Advances in Computer Science for Engineering and Education VI (Lecture Notes on Data Engineering and Communications Technologies, 181) [1st ed. 2023] 3031361172, 9783031361173

Citation preview

Lecture Notes on Data Engineering and Communications Technologies 181

Zhengbing Hu Ivan Dychka Matthew He   Editors

Advances in Computer Science for Engineering and Education VI

Lecture Notes on Data Engineering and Communications Technologies Series Editor Fatos Xhafa, Technical University of Catalonia, Barcelona, Spain

181

The aim of the book series is to present cutting edge engineering approaches to data technologies and communications. It will publish latest advances on the engineering task of building and deploying distributed, scalable and reliable data infrastructures and communication systems. The series will have a prominent applied focus on data technologies and communications with aim to promote the bridging from fundamental research on data science and networking to data engineering and communications that lead to industry products, business knowledge and standardisation. Indexed by SCOPUS, INSPEC, EI Compendex. All books published in the series are submitted for consideration in Web of Science.

Zhengbing Hu · Ivan Dychka · Matthew He Editors

Advances in Computer Science for Engineering and Education VI

Editors Zhengbing Hu Faculty of Applied Mathematics National Technical University of Ukraine “Igor Sikorsky Kiev Polytechnic Institute”, Ukraine Kyiv, Ukraine

Ivan Dychka Faculty of Applied Mathematics National Technical University of Ukraine “Igor Sikorsky Kiev Polytechnic Institute”, Ukraine Kyiv, Ukraine

Matthew He Halmos College of Arts and Sciences Nova Southeastern University Fort Lauderdale, FL, USA

ISSN 2367-4512 ISSN 2367-4520 (electronic) Lecture Notes on Data Engineering and Communications Technologies ISBN 978-3-031-36117-3 ISBN 978-3-031-36118-0 (eBook) https://doi.org/10.1007/978-3-031-36118-0 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

Modern engineering and educational technologies have opened up new opportunities in computer science, such as conducting computer experiments, visualizing objects in detail, remote learning, quickly retrieving information from large databases, and developing artificial intelligence systems. Governments and science and technology communities are increasingly recognizing the significance of computer science and its applications in engineering and education. However, preparing the next generation of professionals to properly use and advance computer science and its applications has become a challenging task for higher education institutions due to the rapid pace of technological advancements and the need for interdisciplinary skills. Therefore, higher education institutions need to focus on developing innovative approaches to teaching and providing opportunities for hands-on learning. Additionally, international cooperation is crucial in facilitating and accelerating the development of solutions to this critical subject. As a result of these factors, the 6th International Conference on Computer Science, Engineering, and Education Applications (ICCSEEA2023) was jointly organized by the National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute,” National Aviation University, Lviv Polytechnic National University, Polish Operational and Systems Society, Warsaw University of Technology, and the International Research Association of Modern Education and Computer Science on March 17–19, 2023, in Warsaw, Poland. The ICCSEEA2023 brings together leading scholars from all around the world to share their findings and discuss outstanding challenges in computer science, engineering, and education applications. Out of all the submissions, the best contributions to the conference were selected by the program committee for inclusion in this book. March 2023

Zhengbing Hu Ivan Dychka Matthew He

Organization

General Chairs Felix Yanovsky Ivan Dychka

Delft University of Technology, Delft, Netherlands National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine

Online conference Organizing Chairs Z. B. Hu Solomiia Fedushko Oleksandra Yeremenko Vadym Mukhin Yuriy Ushenko

National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine Lviv Polytechnic National University, Lviv, Ukraine Kharkiv National University of Radio Electronics, Kharkiv, Ukraine National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine Chernivtsi National University, Chernivtsi, Ukraine

Program Chairs Matthew He Roman Kochan Q. Y. Zhang

Nova Southeastern University, Florida, USA University of Bielsko-Biała, Bielsko-Biala, Poland Wuhan University of Technology, China

Publication Chairs Z. B. Hu Matthew He Ivan Dychka

National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine Nova Southeastern University, USA National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine

viii

Organization

Publicity Chairs Sergiy Gnatyuk Rabah Shboul O. K. Tyshchenko

National Aviation University, Kyiv, Ukraine Al-al-Bayt University, Jordan University of Ostrava, Czech Republic

Program Committee Members Artem Volokyta

Rabah Shboul Krzysztof Kulpa Bo Wang Anatoliy Sachenko Ihor Tereikovskyi

E. Fimmel G. Darvas A. U. Igamberdiev Jacek Misiurewicz X. J. Ma S. C. Qu Ivan Izonin Y. Shi Feng Liu Sergey Petoukhov Z. W. Ye Oleksii K. Tyshchenko C. C. Zhang Andriy Gizun Yevhen Yashchyshyn Vitaly Deibuk

National Technical University of Ukraine “Igor Sikorsky Kyiv - Polytechnic Institute”, Kyiv, Ukraine Al-al-Bayt University, Jordan Warsaw University of Technology, Warsaw, Poland Wuhan University, China Kazimierz Pułaski University of Technology and Humanities in Radom, Poland National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Kyiv, Ukraine Mannheim University of Applied Sciences, Germany Institute Symmetron, Hungary Memorial University of Newfoundland, Canada Warsaw University of Technology, Warsaw, Poland Huazhong University of Science and Technology, China Central China Normal University, China Lviv Polytechnic National University, Lviv, Ukraine Bloomsburg University of Pennsylvania, USA Huazhong Agricultural University, China MERI RAS, Moscow, Russia Hubei University of Technology, China University of Ostrava, Czech Republic Feng Chia University, Taiwan National Aviation University, Kyiv, Ukraine Warsaw University of Technology, Warsaw, Poland Chernivtsi National University, Chernivtsi, Ukraine

Organization

J. Su G. K. Tolokonnikova

ix

Hubei University of Technology, China FNAT VIM of RAS, Moscow, Russia

Conference Organizers and Supporters National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine National Aviation University, Ukraine Lviv Polytechnic National University, Ukraine Polish Operational and Systems Society, Poland Warsaw University of Technology, Poland International Research Association of Modern Education and Computer Science, Hong Kong

Contents

Computer Science for Manage of Natural and Engineering Processes Construction and Operation Analysis of Logistics Sorting Simulation Experimental Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yuexia Tang and Guangping Li

3

Analysis on Ventilation Design of the Negative Pressure Isolation Rooms in Wuhan Leishenshan Hospital . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Xiang Lei, Yilei Liu, Min Xu, Zheng Yang, Qiong Dan, and Xueqin Yan

13

Multi Object Infrared Image Segmentation Based on Multi-level Feature Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Chengquan Liang and Ying Zhang

26

Spear-Man Correlation Data Analysis of Scale of Intelligent Logistics and Investment and Financing of Logistics Technology . . . . . . . . . . . . . . . . . . . . Ni Cheng and Meng Lei

37

Neural Network Model for Laboratory Stand Control System Controller with Parallel Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Peter Kravets, Anatolii Novatskyi, Volodymyr Shymkovych, Antonina Rudakova, Yurii Lebedenko, and Hanna Rudakova Web Application State Management Performance Optimization Methods . . . . . Liubov Oleshchenko and Pavlo Burchak

47

59

Design and Implementation of Signal Acquisition System Based on FPGA and Gigabit Ethernet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Haiyou Wang, Xiaozhe Yang, and Qianxian Bao

75

Design of a Dynamic Probability-Based Automatic Test Paper Generation System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fang Huang, Mingqi Wei, Lianju Su, and Dichen Guan

85

The Extended Fredkin Gates with Reconfiguration in NCT Basis . . . . . . . . . . . . Vitaly Deibuk, Oleksii Dovhaniuk, and Taras Kyryliuk

95

xii

Contents

Applying Data Mining Techniques in People Analytics for Balancing Employees’ Interests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Liana Maznyk, Zoriana Dvulit, Nadiia Seliuchenko, Marian Seliuchenko, and Olena Dragan Object-Oriented Model of Fuzzy Knowledge Representation in Computer Training Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nataliia Shybytska Implementation of Optical Logic Gates Based on Color Filters . . . . . . . . . . . . . . Victor Timchenko, Yuriy Kondratenko, and Vladik Kreinovich

106

116

126

Enhancing Readability in Custom Templates for Displaying Semantically Marked Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Viacheslav Zosimov and Oleksandra Bulgakova

137

Metaheuristic Optimization Algorithms Usage in Recommendation System with User Psychological Portrait Generation . . . . . . . . . . . . . . . . . . . . . . . Mykhailo Vernik and Liubov Oleshchenko

147

Interest Point Detection at Digital Image Based on Averaging of Function . . . . Iryna Yurchuk and Ivan Kosovan Analyzing Ukrainian Media Texts by Means of Support Vector Machines: Aspects of Language and Copyright . . . . . . . . . . . . . . . . . . . . . . . . . . . Maksym Lupei, Oleksandr Mitsa, Vasyl Sharkan, Sabolch Vargha, and Nitsa Lupei Data Mesh as Distributed Data Platform for Large Enterprise Companies . . . . . Yevhenii Vlasiuk and Viktoriia Onyshchenko Using the Strategy of Information Resistance to Improve Content in Virtual Communities Using the Example of the Facebook Social Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mariana Petryk, Oleksandr Marcovets, and Ruslana Pazderska A Modified Method and an Architecture of a Software for a Multivariate Polynomial Regression Building Based on the Results of a Conditional Active Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alexander Pavlov, Maxim Holovchenko, Iryna Mukha, Kateryna Lishchuk, and Valeriia Drozd

163

173

183

193

207

Contents

Research and Development of Bilinear QoS Routing Model over Disjoint Paths with Bandwidth Guarantees in SDN . . . . . . . . . . . . . . . . . . . . Oleksandr Lemeshko, Oleksandra Yeremenko, Maryna Yevdokymenko, and Batoul Sleiman Improved GL-Model of Behavior of Complex Multiprocessor Systems in Failure Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alexei Romankevitch, Kostiantyn Morozov, Vitaliy Romankevich, Daniil Halytskyi, and Zacharioudakis Eleftherios A Method for Evaluating the Efficiency of the Semantic Kernel in the Internet Promotion Channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sergey Orekhov, Andrii Kopp, Dmytro Orlovskyi, and Tetiana Goncharenko Automated Dating of Galaktion Tabidze’s Handwritten Texts . . . . . . . . . . . . . . . Tea Tvalavadze, Ketevan Gigashvili, Esma Mania, and Maksim Iavich Artificial Neural Network and Random Forest Machine Learning Algorithm Based TCSC Controllers for Mitigating Power Problems . . . . . . . . . Niharika Agrawal, Faheem Ahmed Khan, and Mamatha Gowda Child Access Control Based on Age and Personality Traits . . . . . . . . . . . . . . . . . Alguliyev M. Rasim, Fargana J. Abdullayeva, and Sabira S. Ojagverdiyeva Identifying the Application of Process Mining Technique to Visualise and Manage in the Healthcare Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Arezoo Atighehchian, Tahmineh Alidadi, Reyhaneh Rasekh Mohammadi, Farhad Lotfi, and Sima Ajami New About the Principle “Like Begets Like” in Genetics . . . . . . . . . . . . . . . . . . . Sergey V. Petoukhov

xiii

223

236

246

260

269

289

299

309

A Study of Implementing a Blockchain-Based Forensic Model Integration (BBFMI) for IoT Devices in Digital Forensics . . . . . . . . . . . . . . . . . . Chintan Singh, Himanshu Khajuria, and Biswa Prakash Nayak

318

The Method of Determining the Adequacy of Virtual Community’s Themes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Synko Anna

328

A Feature Selection and Regression Refinement Network for Power Facility Detection in Remote Sensing Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yu Zhou, Wenhao Mo, Changyu Cai, Yuntao Yao, and Yanli Zhi

341

xiv

Contents

Development of Arithmetic/Logical Computational Procedures: Computer Aided Digital Filter Design with Bilinear Transformation . . . . . . . . . Abdul Rasak Zubair and Adeolu Johnson Olawale

349

Multidimensional Ranking and Taxonomic Analysis of the Regional Socio-Economic Development in Ukraine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mariana Vdovyn and Larysa Zomchak

361

Advanced Method for Compressing Digital Images as a Part of Video Stream to Pre-processing of UAV Data Before Encryption . . . . . . . . . . . . . . . . . . Zhengbing Hu, Myroslav Ryabyy, Pylyp Prystavka, Sergiy Gnatyuk, Karina Janisz, and Dmytro Proskurin

371

Ternary Description Language and Categorical Systems Theory . . . . . . . . . . . . . G. K. Tolokonnikov

382

Radial Mean LBP (RM-LBP) for Face Recognition . . . . . . . . . . . . . . . . . . . . . . . . Shekhar Karanwal

391

Waste Classification Using Random Forest Classifier with DenseNet201 Deep Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kemal Akyol and Abdulkadir Karacı

401

An Increased Implementation of Generalized Sidelobe Canceller Based on the Expectation - Maximization Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quan Trong The and Sergey Perelygin

410

Pedestrian Re-identification Combined with Triplet Metric Learning and Dual-Domain Filtering in Noisy Scene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wenyu Song, Liyuan Wang, Qi Huang, Shaohuai Yu, and Shurui Wang

419

Performance Analysis of Free Text Keystroke Authentication Using XGBoost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ievgeniia Kuzminykh, Saransh Mathur, and Bogdan Ghita

429

Cyberbullying Detection and Classification in Social Media Texts Using Machine Learning Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Joseph D. Akinyemi, Ayodeji O. J. Ibitoye, Christianah T. Oyewale, and Olufade F. W. Onifade Detection of Potential Mosquito Breeding Sites Using CNN and Fewshot Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gabaalini Ananthajothy, Rudsika Navaratnam, Niluksha Thevarasa, and Maheshi B. Dissanayake

440

450

Contents

Spatial-Temporal Graph Neural Network Based Anomaly Detection . . . . . . . . . Ruoxi Wang, Jun Zhan, and Yun Sun Methods of Topological Organization Synthesis Based on Tree and Dragonfly Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Volodymyr Rusinov, Oleksandr Honcharenko, Artem Volokyta, Heorhii Loutskii, Oleksandr Pustovit, and Artemii Kyrianov Hardware Modified Additive Fibonacci Generators Using Prime Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Volodymyr Maksymovych, Krzysztof Przystupa, Oleh Harasymchuk, Mariia Shabatura, Roman Stakhiv, and Viktor Kuts Data Acquisition System for Monitoring Soil Parameters . . . . . . . . . . . . . . . . . . . Tetiana Fedyshyn, Krzysztof Przystupa, Tetiana Bubela, and Iryna Petrovska Investigation of Drawbacks of the Software Development Artifacts Reuse Approaches based on Semantic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . Olena Chebanyuk

xv

459

472

486

499

514

Perfection of Computer Algorithms and Methods Improving Piezoceramic Artificial Muscles for Flying Insect-Sized Mini Robots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sergey Filimonov, Constantine Bazilo, Olena Filimonova, and Nadiia Filimonova

527

Platform Construction and Structural Design of Passionfruit Picking and Sorting Robot Under the Guidance of TRIZ . . . . . . . . . . . . . . . . . . . . . . . . . . Caigui Huang

539

Development and Application of Ship Lock System Simulation Modules Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yuhong Liu and Qiang Zhou

550

Hydrodynamics of Inhomogeneous Jet-Pulsating Fluidization . . . . . . . . . . . . . . . Bogdan Korniyenko, Yaroslav Kornienko, Serhii Haidai, and Andrii Liubeka Approaches to Improving the Accuracy of Machine Learning Models in Requirements Elicitation Techniques Selection . . . . . . . . . . . . . . . . . . . . . . . . . Denys Gobov and Olga Solovei

560

574

xvi

Contents

Machine Learning Based Function for 5G Working with CPU Threads . . . . . . . Maksim Iavich Analysis of Threat Models for Unmanned Aerial Vehicles from Different Spheres of Life . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hanna Martyniuk, Bagdat Yagaliyeva, Berik Akhmetov, Kayirbek Makulov, and Bakhytzhan Akhmetov

585

595

Identification of Soil Water Potential Sensors Readings in an Irrigation Control System Using Internet-of-Things (IoT): Automatic Tensiometer and Watermark 200SS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Volodymyr Kovalchuk, Oleksandr Voitovich, Pavlo Kovalchuk, and Olena Demchuk

605

Machine Learning Algorithms Comparison for Software Testing Errors Classification Automation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Liubov Oleshchenko

615

Machine Learning for Unmanned Aerial Vehicle Routing on Rough Terrain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ievgen Sidenko, Artem Trukhov, Galyna Kondratenko, Yuriy Zhukov, and Yuriy Kondratenko Improvement of the Dispenser of the Liquid Fertilizer Dosing Control System for Root Feeding of Crops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sergey Filimonov, Dmytro Bacherikov, Constantine Bazilo, and Nadiia Filimonova

626

636

A System of Stress Determination Based on Biomedical Indicators . . . . . . . . . . Lesia Hentosh, Vitalii Savchyn, and Oleksandr Kravchenko

645

Sub-Gigahertz Wireless Sensor Network for Smart Clothes Monitoring . . . . . . Andrii Moshenskyi, Dmitriy Novak, and Liubov Oleshchenko

657

Paradoxical Properties Research of the Pursuit Curve in the Intercepting a Fugitive Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Viktor Legeza and Liubov Oleshchenko

670

Oscillations at Nonlinear Parametric Excitation, Limited Power-Supply and Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alishir A. Alifov and M. G. Farzaliev

682

Fairness Audit and Compositional Analysis in Trusted AI Program . . . . . . . . . . Sergiy Gnatyuk, Pylyp Prystavka, and Serge Dolgikh

690

Contents

xvii

Determination of Brachistochronous Trajectories of Movement of a Material Point in a One-Dimensional Vector Field . . . . . . . . . . . . . . . . . . . . . Viktor Legeza, Oleksandr Neshchadym, and Lyubov Drozdenko

700

A Fuzzy Based Predictive Approach for Soil Classification of Agricultural Land for the Efficient Cultivation and Harvesting . . . . . . . . . . . . Rajalaxmi Hegde and Sandeep Kumar Hegde

709

Sector-Independent Integrated System Architecture for Profiling Hazardous Industrial Wastes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wilson Nwankwo, Charles O. Adetunji, Kingsley E. Ukhurebor, Acheme I. David, Samuel Makinde, Chukwuemeka P. Nwankwo, and Chinecherem Umezuruike Design and Development of a Hybrid Eye and Mobile Controlled Wheelchair Prototype using Haar cascade Classifier: A Proof of Concept . . . . . Yusuf Kola Ahmed, Nasir Ayuba Danmusa, Taofik Ahmed Suleiman, Kafilat Atinuke Salahudeen, Sani Saminu, Abdul Rasak Zubair, and Abdulwasiu Bolakale Adelodun

721

748

A Case-Based Approach in Astrological Prediction of Profession Government Officer or Celebrity Using Machine Learning . . . . . . . . . . . . . . . . . Snehlata Barde, Vijayant Verma, and Apurva Verma

759

Method of Automatic Depersonalization of Databases for Application in Machine Learning Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Artem Volokyta and Vlada Lipska

774

Mathematical Modeling of States of a Complex Landscape System with a Wind Farm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Taras Boyko, Mariia Ruda, Elvira Dzhumelia, Orest Kochan, and Ivan Salamon Development of Distributed System for Electric Personal Transporters Charging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Roman Kochan, Nataliia Kochan, Nataliya Hots, Uliana Kohut, and Volodymyr Kochan Environmental Pollution from Nuclear Power Plants: Modelling for the Khmelnytskyi Nuclear Power Plant (Ukraine) . . . . . . . . . . . . . . . . . . . . . . Igor Vasylkivskyi, Vitalii Ishchenko, Orest Kochan, and Roman Ivakh

788

803

815

xviii

Contents

Location Optimization of Vessel Traffic Service Radar Station in Expansion Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Chuan Huang, Changjian Liu, Jia Tian, Hanbing Sun, and Xiaoqing Zhang

827

Research on the Development Path of Domestic Trade Container Transport in China’s Ports Under the New Situation . . . . . . . . . . . . . . . . . . . . . . . Changjian Liu, Qiqi Zhou, Hongyu Wu, Chuan Huang, and Xunran Yu

845

The Significance and Project Planning of Building a Shipping System in Heilongjiang Province . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Xunran Yu, Lingsan Dong, Changjian Liu, and Hongyu Wu

860

Advances in Technological and Educational Approaches Mindfulness and Academic Achievements of University Students: A Chain Mediation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ying Gao and Yi Chen Evaluation System of College Students’ Key Abilities Development in the Context of “Integration Education of Specialization, Creation and Morality” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Liwei Li, Dalong Liu, and Yanfang Pan

875

885

Exploration and Practice of the Evaluation System of College Teaching Quality Based on MSF Mechanism in OBE Perspective . . . . . . . . . . . . . . . . . . . . Yanzhi Pang and Jianqiu Chen

901

Research on Teaching Application of Course Resources Based on Bloom’s Taxonomy of Educational Objectives . . . . . . . . . . . . . . . . . . . . . . . . . Hao Cai, Jing Kang, Qiang Zhang, and Yue Wang

913

Reform and Practice of Virtual Simulation Practice Course for Logistics Engineering Specialty Based on OBE Concept and School-Enterprise Linkage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yanfang Pan, Dalong Liu, and Liwei Li Application of Google Workspace in Mathematical Training of Future Specialists in the Field of Information Technology . . . . . . . . . . . . . . . . . . . . . . . . Olena Karupu, Tetiana Oleshko, Valeria Pakhnenko, and Anatolii Pashko

927

939

Contents

The Current Situation and Development Trend of the Evaluation of Moral and Ideological Curriculum in China-Visual Analysis Based on CNKI Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yunyue Wu, Yanzhi Pang, and Xiaoli Teng Target Performance of University Logistics Teachers Based on the Combination of AHP and Support Vector Machine in the Context of “Double First-Class” Project Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lei Zhang

xix

950

962

Research and Practice of Outcome-Oriented Innovative Talent Training System of Material Specialty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hu Yang and Qian Cao

972

Development of the Online Library of the University Department to Support Educational Activities: A Conceptual Model . . . . . . . . . . . . . . . . . . . . Nataliia Vovk and Daryna Prychun

984

Research on the Influencing Factors of Students’ General Competency Improvement in Application-Oriented Universities . . . . . . . . . . . . . . . . . . . . . . . . 1000 Jing Zuo and Tongyao Huang Dynamic Adjustment of Talent Training Scenario for Electrical Engineering and Automation Majors Based on Professional Investigation . . . . . 1011 Haoliang Zhu The Effect of Integrating Higher Vocational Aesthetic and Ideological and Political Education Based on FCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1022 Lili Li and Feifei Hu Teaching Reform and Practice of “Sensor and Detection Technology” Course Based on “Specialty and Innovation Integration” and Project-Based Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1033 Deshen Lv and Ying Zhang Improved Method for Grading Bilingual Mathematics Exams Based on Computing with Words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1046 Danylo Tavrov and Liudmyla Kovalchuk-Khymiuk Education System Transition to Fully Online Mode: Possibilities and Opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1057 Md. Asraful Haque, Tauseef Ahmad, and Shoaib Mohd Collaborative Education Mechanism of In-Depth Integration Between Schools and Enterprises Under the Background of the Construction

xx

Contents

of EEE -- Taking the Major of Data Science and Big Data Technology of Nanning University as an Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1070 Zhixiang Lu and Suzhen Qiu Formal and Non-formal Education of Ukraine: Analysis of the Current State and the Role of Digitalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1085 Maryna Nehrey, Nataliia Klymenko, and Inna Kostenko Impact of YouTube Videos in Promoting Learning of Kinematics Graphs in Tanzanian Teachers’ Training Colleges . . . . . . . . . . . . . . . . . . . . . . . . . 1099 Beni Mbwile and Celestin Ntivuguruzwa Logical Relationship Analysis and Architecture Design of Logistics Engineering Curriculum Based on ISM Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 1110 Lijun Jiang, Wanyi Wei, Lei Zhao, and Huasheng Su Exploration of the Application of Project Teaching Method in the Course of Urban Public Transport Planning and Operation Management . . . . . . . . . . . . 1121 Lanfang Zhang, Chun Bao, and Jianqiu Chen Social Engineering Penetration Testing in Higher Education Institutions . . . . . . 1132 Roman Marusenko, Volodymyr Sokolov, and Pavlo Skladannyi Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1149

Computer Science for Manage of Natural and Engineering Processes

Construction and Operation Analysis of Logistics Sorting Simulation Experimental Platform Yuexia Tang(B) and Guangping Li School of Intelligent Manufacturing, Nanning University, Guangxi 530200, China [email protected]

Abstract. The logistics sorting system, which undertakes the function of automatic sorting, is an important equipment for realizing logistics automation. In order to improve the efficiency of the logistics sorting system debugging and reduce the risk of debugging, this paper designs a three-dimensional simulation experimental platform for the logistics sorting system, establishes a three-dimensional virtual simulation model, sets its multiple attributes and compiles the behavior code during operation, takes the programmable controller PLC as the core controller, designs and makes a virtual-reality conversion module, which converts the virtual sensor signal of the model into a real electrical signal, Convert the real PLC electrical signal into the digital information of the virtual model, and realize the real-time interaction between the virtual simulation model and the PLC. The test results show that the designed logistics sorting simulation experimental platform works stably and reliably with good real-time interaction, and effectively solves practical problems in learning and training in related fields. Keywords: Logistics sorting system · Simulation experiment platform · 3D model · Virtual reality conversion · PLC

1 Introduction As an important equipment of modern logistics, logistics sorting system plays an important role in the process of goods warehousing, distribution and transportation. Its increasingly popular application also puts forward higher requirements for professional and technical personnel [1–3]. However, in the study and training in related fields, it was found that the large area of the logistics sorting system, the high cost of training consumables and maintenance, limited the number of sets of training equipment, resulting in short operation time and poor learning effect [4]. At the same time, it is also very risky for beginners to debug directly on the equipment [5–8]. Therefore, the establishment of logistics sorting simulation experimental platform has important practical significance for solving the above problems. At present, most logistics sorting simulation experimental platforms are realized by configuration software. However, most of the practical training carried out on the simulation interface of configuration software development is limited to software design © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 3–12, 2023. https://doi.org/10.1007/978-3-031-36118-0_1

4

Y. Tang and G. Li

and cannot be carried out on hardware design [9, 10]. At the same time, the debugging results cannot be directly applied to the actual equipment. This problem can be effectively solved by using the simulation experiment platform [11–14]. This design takes the logistics sorting system at five crossings as an example, uses UG software to build its three-dimensional model, takes Mitsubishi PLC as the core controller, designs and makes the virtual-reality conversion module, and constructs the logistics sorting simulation experimental platform. The simulation experimental platform can visually verify the circuit and program design, and effectively improve the efficiency and safety of equipment operation and debugging.

2 Design and Three-Dimensional Model Construction of Logistics Sorting Experimental Platform 2.1 Overall Design of Experimental Platform The design of the overall framework is the first step in the design of the experimental platform. As shown in Fig. 1, the logistics sorting simulation experimental platform in this design consists of three parts: the three-dimensional model of the logistics sorting system, the PLC controller and the virtual-reality conversion module.

Fig. 1. Framework of logistics sorting simulation experiment platform

Three-dimensional model is to measure the size of the logistics sorting system according to its entity, use UG software to first establish the three-dimensional model of each component, then assemble the model of each component into a whole according to their motion position, and set various model parameters. The virtual-reality conversion module is to realize the communication between the entity PLC and the virtual model. The module reads the virtual sensor signal of the model and converts it into a real electrical signal to drive the PLC, and converts the electrical signal output by the PLC into a virtual execution signal to drive the actuator action of the model. The experimental platform uses Mitsubishi FX series PLC as the core controller, reads the switch value and sensor signal, and outputs the signal to control the model action. The three parts use serial port mode for two-way communication. 2.2 Construction of Three-Dimensional Model of Logistics Sorting System The function of the logistics sorting platform is to distribute parcels from different regions to different collection bins. The parcels are sent to the identification unit through the main conveyor. If the sorting conditions are met, the baffle in front of the sorting conveyor extends and the parcels are transferred to the corresponding collection box. The platform is mainly composed of one main conveyor belt and four sorting conveyor

Construction and Operation Analysis of Logistics Sorting

5

belts. Use UG software to build the three-dimensional model of each part, and assemble it into a whole according to its relative position to get the overall model of the logistics sorting platform, as shown in Fig. 2.

Fig. 2. Logistics sorting platform model

2.2.1 Model Physical Attribute Setting Physical attribute is to add physical properties to each part based on the geometric model. First of all, the model parts that need to move, such as the baffle of the parcel and sorting conveyor, are rigid bodies. Any geometric object can only accept the influence of external forces if it has added rigid body attributes [15–17]. Secondly, model parts with contact action such as package, main conveyor, sorting conveyor, baffle of sub-conveyor and collecting box are set as collision body. Models without collision bodies will cross each other during simulation, and cannot simulate the real motion of objects. Finally, set the package model as the object source. The setting of the object source allows the model to copy and generate the original model in the set way, so that the package can be automatically generated continuously in the simulation. 2.2.2 Model Motion Attribute Setting The motion attribute is to add motion pairs to the geometric model [18]. Kinematic pair refers to the movable connection between two parts in the model that can contact directly and make relative motion, mainly including hinge pair, sliding pair, cylindrical pair, spiral pair, plane pair, etc. Set the baffle on the sorting conveyor as a hinge pair to realize the rotation function. 2.2.3 Model Electromechanical Object Setting Electromechanical object setting is to add sensor and actuator attributes on the model. The sensor is a kind of detection device, which can convert the information of the object to be measured into electrical signal output according to certain rules [19–21]. Set a camera in front of the main conveyor model to enable it to identify the barcode of the package during the simulation process; A position sensor is set in front of each of the

6

Y. Tang and G. Li

four sorting conveyors to detect the position of the parcel model on the main conveyor [22, 23]. The actuator setting can make the model move, including transmission surface, position control, etc. The transmission surface can make the rigid body on its surface move. Set the main conveyor belt and the four sorting conveyor belts as the transmission surface, so that it can operate according to the set parameters (such as transmission speed, etc.). The position control can make the model complete the specified action, set the position control parameters of the four baffles, and control their rotation speed. 2.2.4 Runtime Behavior Code Through the above settings, the simulation model can achieve most of the actions, but there are some details that need to be programmed separately, such as the automatic generation of packages randomly, the rotation angle of the baffle, etc. Taking package generation as an example, after defining the variables and names of all modules, the operation code is presented below:

if (dtsum > 0.05) { dtsum = 0; if(f.Pin[6]==true || f.Pin[26] == true) time++; if(time>100) { int a1 = r1.Next(99)/20; SB[a1].Active = true; time = 0; } }

3 Design of Virtual-Reality Conversion Module When using ForceControl, MCGS and other configuration software to build a virtual model, the control signal sent by the PLC can control the action of the simulation model, but the virtual sensor signal triggered after the simulation model moves in place cannot return to the PLC, and the interaction between virtual and reality is one-way. Therefore, a virtual-reality conversion module is designed to realize the two-way transmission of information between virtual and reality. 3.1 Hardware Design The core of virtual-reality conversion module hardware mainly includes communication circuit, virtual signal virtualization module and actual signal virtualization module. Considering the cost and real-time of communication, the communication circuit adopts CH340 chip as the communication chip of IO interface circuit board, which can realize the direct conversion between serial port and USB communication. In the

Construction and Operation Analysis of Logistics Sorting

7

virtual signal realization module, the optocoupler is used for electrical isolation. When the virtual sensor is connected, the sensor port in the module will pull the IO port level down after receiving the signal, and the optocoupler output end will be connected to achieve the purpose of converting data signals into electrical signals. There are two kinds of control signals output by PLC: switching value and pulse value. In the actual signal virtualization module, the switch signal is directly read after optoelectronic coupling; The pulse quantity signal is interrupted at the special port, the pulse quantity is read, and the corresponding addition and subtraction calculation is performed to realize the digitalization of the electrical signal. 3.2 Data Conversion Software Design The OPC server can be directly used to communicate between the virtual simulation model and the PLC, but the PLC manufacturers of each major brand have their own OPC server. If the PLC is replaced, the OPC server of the same brand needs to be reinstalled. To solve this problem, a data conversion software is designed for the virtual-reality conversion module. First, in UG software, create runtime behavior code, establish the interface of data conversion software, and map it to the sensor and actuator in the virtual simulation model. Secondly, in order to ensure the continuity of animation and the communication margin, the international standard communication rate of 115200bps is selected as the communication baud rate. On the premise of communication guarantee, using C # language as a development tool, we can control the corresponding virtual sensors and actuators by writing programs and controlling the runtime behavior code. Finally, the data conversion software is embedded into the STEP function of UG software, and the UG software calls the STEP function regularly. During the execution of the STEP function, the data conversion software continuously collects information, reads and writes data to the hardware module, and realizes real-time data exchange without passing through the OPC server, and is not affected by the PLC replacement.

4 Operation of Material Sorting Simulation Experimental Platform After the construction of the logistics sorting simulation experiment platform, the simulation experiment can be carried out. The control requirements of the logistics sorting control system are as follows: after the system is started, packages are transmitted on the main drive belt, their barcode information is identified by the camera, and they are sorted to the corresponding collection box according to the identification results. According to the above control requirements, the I/O address allocation table of the logistics sorting control system is shown in Table 1. The entire logistics sorting equipment control system needs five input ports for camera identification results, representing sorting addresses 1–5 respectively; 4 position detection sensors to detect whether the package has arrived at the sorting conveyor; 8 baffle position detection sensors to detect whether the baffle rotates in place and returns to its original position; With the start and stop buttons, a total of 19 digital input points are required. The logistics sorting control

8

Y. Tang and G. Li

system needs to control the action of four baffles and five drive belts, and the number of digital output points is 9. Considering the 10%–15% I/O point margin, the PLC model is Mitsubishi FX3U -48MT. Some words in Table 1 are abbreviated, for example, Baffle is abbreviated as “B”, Solenoid Valve is expressed as “SV” and Conveyor Belt is indicated as “CB”. Table 1. Input and output port assignment Address X000 X001 X002 X003 X004 X005 X006 X007 X010

Input signal Name Sort Address 1 Sort Address 2 Sort Address 3 Sort Address 4 Sort Address 5 Position 1 detection Position 2 detection Position 3 detection Position 4 detection

Address X011 X012 X013 X014 X015 X016 X017 X020 X021 X022

Name B-1 in place detect B-1 reset detect B-2 in place detect B-2 reset detect B-3 in place detect B-3 reset detect B-4 in place detect B-4 reset detect Start Stop

Address Y000 Y001 Y002 Y003 Y004 Y005 Y006 Y007 Y010

Output signal Name B-1 SV B-2 SV B-3 SV B-4 SV Main CB Sort CB 1 Sort CB 2 Sort CB 3 Sort CB 4

According to the I/O address allocation table, the hardware circuit of the logistics sorting control system is designed, as shown in Fig. 3. The workflow of the logistics sorting control system is shown in Fig. 4. After the system is started, all conveyor belts start to run. Packages are transported on the main conveyor belt. After being identified by the camera one by one, their address information will be stored in five data stacks in the same time period. Before the package runs to the sorting drive belt, it is detected by the position sensor, and the control system compares the current position information of the package with the barcode information stored in the stack. When the comparison results are consistent, the baffle of the current sorting conveyor is opened, the package is received, and the address information of the package in all stacks is deleted. If the comparison results are inconsistent, the address information stored in the current position data stack of the package will be deleted, the package will continue to be transmitted forward, and the information comparison will be performed again before the next sorting conveyor. The control program is completed using the sequential function diagram (SFC) of GX Developer software, which divides a complete control process into several stages. Each stage is responsible for completing different actions. The conversion between different stages is determined by the set conversion conditions. When the conditions are met, the stage transfer is realized. The action of the previous stage ends and the action of the next stage begins. After the control program is written, virtual debugging is conducted on the logistics sorting simulation platform to simulate the process of parcel sorting by region. By observing the working state of the logistics sorting system on the simulation experiment platform, the correctness of the PLC hardware circuit and control program can be verified, and the effectiveness of the logistics sorting simulation experiment can also be verified. Due to the verification of the designed hardware circuit and control program in the simulation experiment, it can be quickly completed in the actual operation of the

Construction and Operation Analysis of Logistics Sorting

9

Fig. 3. Connection between PLC hardware

logistics sorting system. The conveyor belt runs smoothly and the parcel sorting meets the requirements. The actual operation results reflect the application value of the simulation experiment platform, improve the efficiency of programming and debugging, break the limitation of insufficient equipment in training and learning, solve most of the problems that may occur in the actual operation process in advance, and ensure the safety of equipment and personnel.

10

Y. Tang and G. Li

Fig. 4. Flow chart of logistics sorting

5 Conclusion Sorting is an important work in the logistics system, which affects and even determines the function and efficiency of the system to a certain extent. With the progress of science and technology, the technicians in charge of sorting work need to constantly learn and improve their skills and quality. It is an effective way to use the simulation experimental platform of logistics sorting for professional training. The logistics sorting simulation experimental platform realizes the three-dimensional dynamic simulation of virtual and real interaction. The logistics sorting system is modeled, and its geometric model is given physical and motion attributes, and electromechanical objects are added and runtime behavior code is written to obtain a realistic three-dimensional virtual simulation model. The virtual-reality conversion module enables PLC entities to seamlessly interface with virtual simulation models and real-time data interaction. The simulation experiment can carry out hardware design and software design at the same time, find and solve problems such as PLC hardware wiring and control program errors in the process of logistics sorting system debugging in advance, optimize the sorting scheme, and save debugging time. The security of the experiment is improved after the simulation experiment is completed. The logistics sorting simulation platform has certain practical significance for the study of professional courses in colleges and universities and the induction and further training of professional technicians.

Construction and Operation Analysis of Logistics Sorting

11

References 1. Xiang, H., Wang, Y., Gao, Y., et al.: Research on key technology of logistics sorting robot. Lect. Notes Bioinform. (12595), 121–132 (2020) 2. Grambo, P., Mullick, T., Furukawa, T., et al.: Automatic sorting-and-holding for stacking heterogeneous packages in logistic hubs. IFAC-Papers OnLine 52(10), 109–114 (2019) 3. Jouglet, A., Nace, D., Outteryck, C.: Timetabling of sorting slots in a logistic warehouse. Ann. Oper. Res. 239(1), 295–316 (2016) 4. Benotmane, Z., Belalem, G., Neki, A.: A cost measurement system of logistics process. Int. J. Inform. Eng. Electron. Bus. 10(5), 23–29 (2018) 5. Zheng, K., Dai, F., Lian, L.: Virtual debugging of robot grinding system based on NX MCD. Mod. Mach. Tool Autom. Mach. Technol. (12), 57–60+64 (2019). (in Chinese) 6. Qu, Y., Ruan, X., Wang, S.: Construction and design of computer hardware virtual simulation experimental platform. Lab. Res. Explor. 37(12), 116–119 (2018). (in Chinese) 7. Li, X., Lv, L., Sheng, R., et al.: An effective construction method of modular manipulator 3d virtual simulation platform. 3D Res. 9(2), 24 (2018) 8. He, Z., Huang, T.: Exploration of virtual simulation experiment teaching project construction. Experim. Technol. Manage. 35(02), 108–111+116 (2018). (in Chinese) 9. Gjorgjevikj, F., Jakimoski, K.: System monitoring addon analysis in system load simulation. Int. J. Comput. Netw. Inform. Secur. 14(1), 40–51 (2022) 10. Li, Q., Mi, C., Wang, C., Shenyang.: Semi-in-the-loop simulation system of immersive PLC control program based on virtual reality technology. China J. Construct. Mach. 11(01), 41–45 (2013). (in Chinese) 11. Lu, D.: Design of modular automatic production line based on PLC. Labor. Res. Explor. 35(10), 56–59 (2016). (in Chinese) 12. Wang, D., Chen, X., Gao, H.: Application of virtual reality technology in PLC gas-fired boiler training system. J. Syst. Simul. 29(12), 3010–3015 (2017). (in Chinese) 13. Wang, H., Li, H., Cheng, Z., Chen, X.: Design of greenhouse automatic monitoring system based on PLC. Labor. Res. Explor. 36(05), 21–23+71 (2017). (in Chinese) 14. Xiao, Z., Liu, H., Li, B., Rao, X.: Application and research of UG NX in conceptual design of electromechanical products. Mod. Mach. Tools Autom. Process. Technol. 07, 27–30 (2014). (in Chinese) 15. Tang, Y., Guo, X., Zhang, Z., et al.: Parameter optimization method of ADC model based on real-time simulation. Power Autom. Equip. 4(3), 214–218 (2020). (in Chinese) 16. Klucik, S., Taraba, J., Orgon, M., Adamko, D.: The use of PLC technology in broadband services offered to households. Int. J. Inform. Technol. Comput. Sci. 4(4), 1–8 (2012) 17. Afram, R.M., Marie, M.J.: Design and implementation of optimal PID controller using PLC for Al-Tahady ESP. Int. J. Image Graph. Signal Process. 12(5), 1–12 (2020). https://doi.org/ 10.5815/ijigsp.2020.05.01 18. Wang, X., Tang, X., Dong, Z., et al.: Research on rapid development platform of PLC control system. High-tech Commun. (English Version) 27(2), 210–217 (2021) 19. Wu, Y., Wang, Y., Zheng, G., Zhang, J., Li, Y.: Conceptual design and control simulation of electrical integration of CNC lathe loading and unloading machinery based on MCD platform. Mach. Tool Hydraul. 46(15), 99–104 (2018). (in Chinese) 20. Zhang, J., Tan, R., Zhang, Z., Dai, J., Yang, X.: Research on the integration of product conceptual design and detailed design driven by computer-aided innovation technology. J. Mech. Eng. 52(05), 47–57 (2016). (in Chinese) 21. Tang, P., Tian, H.: Research on the layout of national economic mobilization logistics centers. Int. J. Intell. Syst. Appl. 2(1), 44–50 (2010)

12

Y. Tang and G. Li

22. Mohamed, M.: Smart Warehouse management using hybrid architecture of neural network with barcode reader 1D/2D vision technology. Int. J. Intell. Syst. Appl. 11(11), 16–24 (2019) 23. Shieh, J., Zhang, J., Liao, Y., Lin, K.: Applications of barcode images by enhancing the two-dimensional recognition rate. Int. J. Image Graph. Signal Process. 4(7), 26–32 (2012)

Analysis on Ventilation Design of the Negative Pressure Isolation Rooms in Wuhan Leishenshan Hospital Xiang Lei1 , Yilei Liu2 , Min Xu1 , Zheng Yang1 , Qiong Dan1 , and Xueqin Yan1(B) 1 Wuhan Railway Vocational College of Technology, Wuhan 430225, China

[email protected] 2 The University of Hong Kong, Hong Kong, China

Abstract. In response to the outbreak of the novel corona virus pneumonia (COVID-19 pneumonia) in late 2019, many temporary emergency hospitals have been built or rebuilt in some key areas in China. The environment of hospital negative pressure isolation room is closely related to medical workers’ health and safety, while ventilation system has been the major means of environmental control. The design points of Wuhan Leishenshan Hospital (also named as the Thunder Mountain Hospital, which was specifically built to address the COVID-19 pneumonia in February 2020) was briefly introduced and analyzed in the article, as well as some essential elements of the ventilation design to build negative pressure isolation rooms, such as ventilation volume design standard, differential pressure control standard and airflow direction strategy. The measures of differential pressure control are also illustrated in the article. The points in the assay could be used as reference for those engaged in the future designing and construction of infectious disease hospitals. Keywords: COVID-19 pneumonia · negative pressure isolation room · pressure difference · air distribution

1 Introduction In the spring of 2020, corona virus pneumonia flared up widely in the world. According to the statistics, 1,500,830 cases of the virus have been confirmed globally by 9th April 2020. Heavy pressure was imposed to current medical system due to the numerous infected patients. In order to cope with the public health emergency, a large number of temporary medical facilities have been constructed or rebuilt in China. Such construction program is of great social significance with a heavy work burden in limited time. Instructed by the Wuhan government, the author’s team has took over the working mission of HVAC designing as well as epidemic prevention research of the Wuhan Leishenshan Temporary Hospital [1]. The architectural designing core of such infectious disease hospitals is to cut off the infection source and avoid cross-infection among medical staff and patients. Two important factors have to be taken into consideration: firstly, the functional zones and © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 13–25, 2023. https://doi.org/10.1007/978-3-031-36118-0_2

14

X. Lei et al.

flow lines of the hospital should be reasonably arranged to achieve “three zones, two channels”; secondly, the environment of each zone should be well controlled, the crux of which is to control the pressure of each zone to achieve an orderly pressure gradient [2]. A reasonable and effective ventilation system is one of the important means to control the ward’s environment [3]. As the firstly built temporary emergency hospital, Wuhan Leishenshan Hospital is positioned to treat severely ill patients. It could accommodate up to 1500 beds with all the isolation rooms of negative pressure. The box-type board room structure has been adopted, and the overall dimensions (length × width × height) of each unit is 3 × 6 × 2.6 m. 2000 patients of the COVID-19 pneumonia have been adopted and treated up to 19th March 2020. Figure 1 is the Wuhan Leishenshan Hospital.

Fig. 1. The constructing Leishenshan hospital

2 Brief Introduction of Ventilation System Wuhan, where the Leishenshan Hospital is situated, was suffering from a low temperature during the widespread of COVID-19 pneumonia. Heating facilities are of great need in hospitals. Due to the heavy workload as well as limited time, suitable AHU for fresh air system is in shortage. Therefore, the configuration measure, “split air conditioner + centrifugal supply fan + electric heating”, has been adopted in isolation rooms in order to balance indoor ambient temperature. Every air supply and exhaust system serve 4–6 isolation rooms, and each room’s air supply/exhaust branch pipe is equipped with constant air volume valve, ensuring that the wards’ pressure be controlled through the curbing of air supply/exhaust volume differential, thus the miniaturization and simplification of the system could be realized. Figure 2 is the schematic diagram of air supply/exhaust system in Leishenshan’s isolation room zones. The ward is equipped with independent air supply/exhaust system, while the air exhaust system in the bathroom is integrated into the ward’s, where HEPA outlet is installed. A common air supply system is shared between doctor corridor and anteroom, and the doctor corridor enjoys an independent air exhaust system. The ward’s exhaust outlet is 5.4 m above the ground and 20 m horizontal from the supply inlet.

Analysis on Ventilation Design of the Negative Pressure

15

Due to the net height (2.4 m) of the ward, no air pipe is installed indoor. Air supply and return inlet are on one side, when supply inlet is on the top and return inlet is near the floor. Return air terminal was equipped with HEPA inlet. Figure 3 is the negative pressure isolation room of Leishenshan Hospital. Figure 4 is the key factors of ventilation system.

Fig. 2. Schematic diagram of ventilation system

3 Method of Ventilation Design of the Negative Pressure Isolation Rooms 3.1 Isolation Room’s Ventilation Designing Standard and Differential Pressure Control 3.1.1 Air Changes Rate It was emphasized by China National Health Commission in Chinese Clinical Guidance for COVID-19 Pneumonia Diagnosis and Treatment (6th edition) that COVID-19 is transmitted through respiratory droplets and close contact; aerosol transmission is plausible when patients are exposed to high concentration virus-containing aerosols for a long period of time and in a relatively closed environment. Studied by current domestic and foreign researches, short-distance airborne transmission possibly exists in relatively closed environment; airborne transmission is also acknowledged by WHO. Hence, lower the virus concentration of contaminated air in negative pressure isolation room is a crucial method to lower the infectious rate. In the fight against SARS in 2003, it was put forward by domestic experts that the virus-containing air was no longer infectious when diluted above 10000 times [4], showing that virus activity and toxicity decreased when the virus concentration reduced in diluted air. Therefore, the suitable air changes rate has become the essential parameter during isolation room’s design and construction. Many countries have regulated the minimized air changes rate (see Table 1).

16

X. Lei et al.

Fig. 3. Negative pressure isolation room

Fig. 4. Key factors of ventilation system Table 1. Minimized air changes per hour (ACH) Object

Air changes (ACH)

Source

Airborne infectious isolation room

12

ASHRAE Standard 17012

Airborne infectious isolation room

6–12★

The U.S. CDC11

Airborne infectious isolation room

12 and ≥145L/(s·P)

Australia17

Negative pressure isolation room

12

China14

★12 air changes per hour (ACH) new or renovation, 6 ACH existing.

Wells Riley has put forward the noted formula is: P = C/S = 1 − exp(−

Ipqt ) Q

(1)

In the formula, P(C/S) is the infection probability of the most susceptible group where I is each infector’s quanta; p is the expiratory volume of susceptible group (m3 /h); t is

Analysis on Ventilation Design of the Negative Pressure

17

exposure time (h), Q is indoor ventilation volume (m3 /h). The mode is used to predict the probability of airborne respiratory infectious diseases. In a relatively closed environment, airborne transmission possibly exists in a short distance. The basis of isolation room air change rate value could be illustrated by Wells-Riley mode. The infectious rate of medical workers who enter isolation room without protective equipment could be calculated through formula 1. According to the number of SARS cases, supposed that quanta takes the same value of 4680 [5], p equals to 0.3 m3 /h, t equals to 0.5 h; all the patients of COVID-19 virus are requested to wear surgical masks, the infiltrating rate of which could reach above 95%. Hence the quanta of such patients spreading in isolation room could take the value of 234. Figure 5 shows the result. It could be found from Fig. 5 that when ventilation volume is between 0–500 m3 /h, the increase of ventilation volume evidently reduces the infectious risk of susceptible people in isolation room. However, the constant increase contributes little to the decline of infectious rate [6]. If patients did not wear surgical masks, the infectious rate could enjoy a drastic growth. In order to save investment and air-conditioning system operating costs, excessive air changes should not be adopted. When the fresh air volume reaches 500 m3 /h, 12ACH will be correspondent, which is suitable for Leishenshan Hospital. So far, the hospital has been operating for 2 months, the air of which is of high quality due to the well control of the differential pressure with none medical worker infected. Actual testing data of CO2 and PM2.5 in hospital’s isolation room is demonstrated in Figs. 6 and 7.

Fig. 5. Variation in infected persons with different ACHs

During the outbreak of COVID-19 pneumonia, a large amount of temporary negative pressure isolation rooms (TNPIs) have been rebuilt. In order to cope with the volume shortage of fresh air, these isolation, including 19 Fangcang shelter hospitals, have been equipped with portable HEPA. Those devices could provide more than 13ACH air changes rate and remote more than 90% particles larger than 0.3 mm in ten minutes. 24 fan filter units (FFUs) providing 15ACH air changes rate have been used in another hospital. HEAP has been used in FFU which could arrest 99.97% of 0.3 mmm particles, which makes particle concentration 12% less than before. In such TNPI, fresh air volume is around 3–6ACH, and none of the medical workers has got infected. Therefore, it is advised that when air supply unit equipped with HEPA is adopted to dilute contaminated

18

X. Lei et al.

air, the fresh air volume in TNPI should be reduced to 3ACH and total air changes rate should be more than 13ACH.

Fig. 6. CO2 concentration in ward

Fig. 7. PM2.5 in ward

3.1.2 Pressure Difference The pressure difference can be analyzed from four aspects. 1) Purpose of differential pressure control The main non-medical prevention treatment to respiratory infectious disease is to control airborne route and cut off airborne chain, physically separating and isolating infected patients. Reasonable partition is one of the physical separating method. Rooms shall be divided into clean zone, potentially contaminated zone and contaminated zone; independent patient passage and doctor passage could be used to separate patients and medical workers. Although such measures are largely helpful to reduce airborne and contact transmission, virus transmission within the hospital cannot be severed. Crossinfection might be caused without effective control of disorderly air flow, while indoor temperature, outdoor wind, opening or closing the door/window, and medical workers’

Analysis on Ventilation Design of the Negative Pressure

19

passing by might all produce air flow. Air isolation is to maintain an appropriate differential pressure between clean and contaminated zone so as to guarantee the orderly air flow, preventing the spread of pathogenic micro-organisms from contaminated zone. In conclusion, air flow control is essential to cross-infection prevention, and the purpose of differential pressure control is to thoroughly cut off the airborne transmission chain. 2) Standard of differential pressure control Differential pressure control is the crux of isolation room’s design. It might be hard to prevent spread of contaminated air when differential pressure is too low, while energy consumption and system investment might increase if it is too high. Table 2 demonstrates differential pressure setting standard in three countries. It is specifically regulated in China’s standard that no less than 5 pa differential pressure should be maintained between negative pressure isolation room and its adjacent or connected anteroom (or corridor). Differential pressure schematic in various districts is included in the standard. In the U.S, recommendable differential pressure has been 0.25 Pa since 2000 [7]. Scholar Xu Z and his team has studied the relationship between differential pressure and pollutant leakage, discovering that concentration ratio of pollutant leakage only reduce 0.03 when differential pressure changes to −6 Pa from 0, while the ratio remains scarcely unchanged when it turns to −30 Pa from −6 Pa [8]. Scholar Li Y compares ventilation performance of 9 hospitals in Hong Kong. They found that among the 38 tested rooms, differential pressure in 97% rooms complies with the standard −2.5 Pa (all tested rooms have the same average pressure of −7.7 Pa) [9]. Table 2. Smallest differential pressure between adjacent rooms Object

Pressure difference (Pa)

Source

Airborne infectious isolation room

2.5

ASHRAE Standard 17012

Airborne infectious isolation room

2.5

The U.S. CDC11

Airborne infectious isolation room

15

Australia17

Negative pressure isolation room

5

China14

The formality of differential pressure is relevant to building’s air sealing performance and supply/exhaust volume. Given that gap remains unchanged, the greater the differential pressure is, the more the air exhaust volume will be. The relation between open flow and differential pressure could be illustrate in formula below: 1/2  (2) L = 3600μA p /ρ L is air leakage volume (m3 /h); μ is flow coefficient; A is gap width; p is differential pressure between two sides of the gap (Pa); ρ is air density, select the value of 1.2 kg/m3 . In engineering project, differential pressure method is used to calculate air volume infiltrated from door and window gap, the simplified formula is [10]: L = 1.05 AP 1/2

(3)

20

X. Lei et al.

The infiltrated air volume in ward could be calculated by formula (2). When the air supply volume in negative pressure isolation room is fixed, air exhaust volume will be directly influenced by different negative pressure value. Since air infiltration exists among ward and its adjacent anteroom and bathroom, the relation between negative pressure and demanded air supply volume in isolation room is not the functional formula (2) and (3) above. Supposed that the air supply volume is 500 m3 /h (12ACH) in one isolation room in Leishenshan Hospital, the relation between negative pressure and demanded air supply volume under particular condition is shown in Fig. 8. The demanded air supply volume in Fig. 8 is determined by Leishenshan Hospital’s box-type board structure and designed differential pressure between adjacent rooms. Suppose the door gap were 0.0025 m. It could be seen from Fig. 8 that approximate linear relationship exists between ward negative pressure and demanded air exhaust volume. Greater exhaust volume could lead to more investment cost and energy consumption. Apart from that, huge amount of engineering work will extend much more time, which might be impractical to the emergent medical facilities. Therefore, excessive differential pressure is not suitable for temporary hospital. In conclusion, the differential pressure between isolation room, anteroom and corridor shall be no less than −5 Pa in Leishenshan Hospital. According to the actual operating statistics, differential pressure between isolation rooms and anterooms have been maintained between the range of −6 Pa to −10 Pa. Figure 9 shows the differential pressure value on measuring gauge.

Fig. 8. Relations between air exhaust volume and differential pressure

It has been instructed in American CDC Standard [11] and ASHRAE Standard 170 [12] that differential pressure between adjacent rooms shall be no less than 2.5 Pa. The air sealing performance are sometimes not so good in temporary hospitals; according to national current standard, as well as ASHRAE Standard 170, the smallest differential pressure in different zones varies: 5 Pa between isolation room and anteroom; 5 Pa between anteroom and corridor; 2.5 Pa between medical corridor and outside; 2.5 Pa between patient corridor and outside.

Analysis on Ventilation Design of the Negative Pressure

21

Fig. 9. Differential pressure measuring gauge

3) Measures to control differential pressure After ward’s air density has been confirmed, the main measure controlling differential pressure is to control air supply and exhaust volume. The primary task for ventilation system is to build a pressure gradient, making air flow from clean zone to contaminated zone. There are two ways: Variable-frequency control: Supposed that air supply volume remains unchanged, variable-frequency control means to adjust air exhaust volume of the ward according to differential pressure among ward, anteroom and corridor. Constant air volume valve shall be installed on air supply branch pipes [13].

Fig. 10. Constant air volume valve

Constant air volume control: High precision constant air volume valve shall be stalled on every air supply/exhaust branch pipe, which shall be pressure-independent and air volume control deviation shall not be greater than 5%. During the tight construction time, BA system control might not be put into use in temporary hospital. Constant air volume valve could effectively compensate system air volume change produced by high and medium efficiency particulate filter resistance change [14]; hence it has concise procedures and useful function. Figure 10 shows a CAV installed on air supply branch pipe.

22

X. Lei et al.

Fig. 11. Air volume balance in ward

4) Air volume balance in isolation rooms Air volume balance plays a crucial part in differential pressure control. Figure 11 demonstrates an air volume balance schematic in one Leishenshan isolation room. SA represents air supply and EA represents air exhaust. Air supply inlet is installed in isolation room, anteroom and corridor [15]; air exhaust outlet is installed in isolation room and bathroom. Negative pressure in anteroom is realized through negative pressure suction exhaust in isolation room, and, if necessary, 10–20 mm door gap could be left between isolation room and anteroom. In terms of volume difference between air supply and exhaust, it has been instructed in standard that air exhaust volume is 150 m3 /h greater than supply. American CDC standard has raised the volume difference to 210 m3 /h from 85 m3 /h. The volume of air supply and exhaust should be calculated by actual demand in engineering practice.

4 Analysis on Airflow Direction 4.1 Airflow Direction in Ward Zone The directed and organized air flow is one of the main ways to prevent cross-infection in relevant hospital. Airflow direction shall be strictly controlled to make clean air flow from clean zone to potentially contaminated zone and eventually to contaminated zone under the pressure gradient. It could efficiently make pathogen and other contaminated substantiate spread in the smallest range. 4.2 Airflow Distribution Interior to Ward It is required interior the isolation room that the clean airflow shall firstly go through the zone where medical staff pass or work, then the contaminated sources and finally

Analysis on Ventilation Design of the Negative Pressure

23

the exhaust outlet. Figures 12 and 13 show two airflow direction in negative pressure isolation room recommended by American CDC standard as well as ASHRAE Standard 170. Figure 12 is aimed at infectious patients with normal immunity, while Fig. 13 is for susceptible infectious patients. In terms of COVID-19 pneumonia, the airflow direction shown in Fig. 12 is suitable for relevant negative pressure isolation rooms, which is complied with targeted airflow direction principle and adopted by Leishenshan Hospital [16]. As for the location of air supply/exhaust exit, it has been instructed in relative standard that air supply outlet shall be installed on the top of the ward; exhaust air grilles of the ward and clinical room shall be installed on the bottom of the room; the distance between exhaust outlet and the ground shall be no less than 100 mm. The layout of air supply/exhaust exit shall be corresponding to targeted airflow direction principle; air supply outlet shall be installed on the top of the ward, while exhaust air grilles shall be installed near the bed facilitating the emission of contaminated air. Air supply outlet shall be installed above where medical works normally stand, and exhaust air grilles shall be installed below the bed opposite to the inlet. All the standard above aim at protecting medical workers’ safety by practicing targeted airflow principle, making sure that clean air flow past medical zone at first and finally flow outside from patient zone. The location of air supply outlet has not been specifically regulated in ASHRAE Standard 170, as for exhaust outlet, it has been instructed in the standard that the exhaust air grilles in the patient room shall be located directly above the patient bed [17], on the ceiling or on the wall near the head of bed, unless it can be demonstrated that such a location is not practical. Scholar Qian Hua has also argued about the importance of upper air exhaust [18]. During winter, the fine particles exhaled by patient might rise to the ceiling through the hot wind. In order to emit the turbid air on the top of the isolation, exhaust air grilles shall be installed at the lower part of the room and the ceiling above the patient’s head. Electric valve could be installed on the air exhaust branch pipe, which could stop exhausting when medical workers come in and avoid disturbing the targeted airflow. The combination achieves the ventilation type of “partial exhaust, entirely ventilation”. The optimum airflow distribution could not be achieved due to the isolation’s limited storey height, but actual effect of airflow direction in Leishenshan has made isolation’s wind speed and temperature stay at reasonable level. CFD simulation has shown that fresh air orderly flow past medical workers first and then patients.

Fig. 12. Example 1 of ward control for airborne infection isolation

24

X. Lei et al.

Fig. 13. Example 2 of ward control for airborne infection isolation

5 Conclusion The wide spread of COVID-19 pneumonia has been threatening global people’s health. The fight against corona virus pneumonia highlights the importance of infectious disease prevention and, particularly, the significance of medical facility construction for infectious disease. Summarizing relevant experience largely contributes to future’s construction of infectious disease hospitals and protection of medical workers, patients as well as the environment. Key factors in ventilation system design have been analyzed in the article, as well as design and construction difference between temporary and permanent hospitals. Main advises and conclusion are listed below: According to the statistics, in terms of indoor ambient isolation supervision, 12ACH air volume could provide a suitable indoor environment for isolation. As is calculated, 10–15ACH air changes rate could serve as economic and proficient ventilation volume. Differential pressure control is the key factor to controlling isolation. The differential pressure between isolation and adjacent room shall be not less than 5 Pa. Given that temporary building’s air sealing performance might not be so good, and considerable windows are installed along contaminated corridor, negative pressure differential value could be lower to no less than −2.5 Pa. Auxiliary measures shall also be adopted in anteroom to strengthen air isolation. With the fixed building’s configuration and air sealing performance, the differential value is determined by air exhaust volume. Excessive differential pressure needs more air exhaust volume, which will lead to some disadvantages such as much more investment cost, energy consuming and construction burden. The collective effect of anteroom isolation and fresh air diluting makes medical corridor remain relatively clean where virus concentration is approximately 0.0031% of isolation. The air flow direction shall be scientifically determined from both medical workers and patient protection. Acknowledgment. This project is supported by Projects of Hubei Institute for Vocational and Technical Education (ZJGB2022095).

References 1. Odji, E.: Graphic design principles and theories application in rendering aesthetic and functional installations for improved environmental sustainability and development. Int. J. Eng. Manufac. 1(9), 21–37 (2019)

Analysis on Ventilation Design of the Negative Pressure

25

2. Gaomin, D.: Positive crankcase ventilation system. Int. J. Eng. Manufac. 1(1), 13–19 (2011) 3. Dinggen, L., Xiaozhong, L.: Study on gasoline intake model modification based on PCV. Vehicle Eng. 1(1), 31–35 (2009) 4. Jiang, Y., Zhao, B., Li, X., et al.: Investigating a safe ventilation rate for the prevention of indoor SARS transmission: an attempt based on a simulation approach. Build. Simul. 2(4), 281–289 (2009). (in Chinese) 5. Qian, H., Zheng, X.: Prediction of risk of airborne transmitted diseases. J. Univ. 42(3), 468– 472 (2012) 6. Memarzadeh, F., Xu, W.: Role of air changes per hour (ACH) in possible transmission of airborne infections. Build. Simul. 5(1), 15–28 (2011). https://doi.org/10.1007/s12273-0110053-4 7. Marko, H., Anna, R., Pertti, P., et al.: Airborne infection isolation rooms-a review of experimental studies. Indoor Built Environ. 20(2), 584–594 (2011) 8. Xu, Z., Zhang, Y., Wang, Q., et al.: Study on isolation effects of isolation wards. J. HV&AC 36(4), 1–4 (2006) 9. Li, Y., Ching, W.H., Qian, H., et al.: An evaluation of the performance of new SARS isolation wards in nine hospitals in Hong Kong. Indoor Built Environ. 16(1), 400–410 (2007) 10. Technical standard for smoke management systems in buildings: GB51251-2017. Ministry of public security of the People’s Republic of China, Beijing (2015). (in Chinese) 11. ASHRAE. Ventilation of health care facilities: ANSI/ASHRAE/ASHE standard 107–2017. Beijing, Victorian advisory committee on infection control (2017). (in Chinese) 12. Xu, Z., Zhang, Y., Wang, Q., et al.: Isolation principle of isolation wards. J. HV&AC 36(1), 1–7 (2006) 13. Code for design of infectious diseases hospital: GB/T 50849–2014. National health commission of the Peoples Republic of China, Beijing (2015). (in Chinese) 14. SAC/TC319. Requirements of environmental control for hospital negative pressure isolation ward: GB/T 35428-2017[S]. Beijing, National health commission of the Peoples Republic of China (2017). (in Chinese) 15. The design standard of infectious disease emergency medical facilities for novel corona virus (2019-nCoV) infected pneumonia. T/CECS 661-2020. Beijing, China Ippr international Co. Ltd. (2020). (in Chinese) 16. Victorian advisor committee on infection control: guidelines for the classification and design of isolation rooms in health care facilities. Victoria, Victorian advisory committee on infection control (1999) 17. Aganovic, A., Cao, G.: Evaluation of airborne contaminant exposure in a single bed isolation ward equipped with a protected occupied zone ventilation system. Indoor Built Environ. 28(8), 1092–1103 (2019) 18. Qian, H., Li, Y.: Removal of exhaled particles by ventilation and deposition in a multiple airborne infection isolation room. Indoor Air 20(3), 284–297 (2010). (in Chinese)

Multi Object Infrared Image Segmentation Based on Multi-level Feature Fusion Chengquan Liang(B) and Ying Zhang School of Intelligent Manufacturing, Nanning University, Nanning 530200, China [email protected]

Abstract. Aiming at the problem that the segmentation accuracy may be too low when using a single feature to segment infrared image multi objects, a multi object segmentation method based on multi-level feature fusion is proposed in this paper. The entropy feature, contrast feature and gradient feature of infrared image are extracted respectively, and the multi-level features of the extracted infrared image are fused using the parallel weighted feature fusion method to build the multi-level feature fusion space of the infrared image. The multi-level feature fusion space of the infrared image is set as the ergodic space of Mean shift algorithm, and all feature points in the multi-level feature fusion space are subject to mean shift processing, Obtain infrared image multi object segmentation results. This method can use the multi-level features of the extracted infrared image to segment the multi objects of the infrared image, and the multi object segmentation accuracy of the infrared image is high. Keywords: Multi-level feature fusion · Infrared image · Multi object segmentation · Contrast characteristics · Gradient characteristics · Mean shift algorithm

1 Introduction Image segmentation is an important step in image processing [1, 2]. Infrared image object segmentation is often used in object tracking, fault detection and other applications [3, 4]. The infrared image obtained by the infrared imaging technology has the defects of low signal-to-noise ratio and low definition [5]. When the infrared image is affected by light and ambient temperature, the object area has a high similarity to the ambient temperature. When the infrared image background area has a high similarity to the object area, the object segmentation accuracy is affected [6]. Extract the multi-level features contained in the infrared image, improve the defect that the object cannot be accurately presented due to the extraction of a single feature [7–9], and achieve the multi object segmentation of the infrared image through multi-level feature fusion. Multi object segmentation of infrared image is to divide the infrared image into different regions, so that the same region that has been segmented in the infrared image has certain feature similarity [10–12], and different regions have high differences. Infrared image segmentation is an important basis for infrared image analysis and infrared image pattern recognition, and © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 26–36, 2023. https://doi.org/10.1007/978-3-031-36118-0_3

Multi Object Infrared Image Segmentation

27

an important basis for computer vision fields such as infrared image object tracking and object recognition [13]. When infrared image is segmented by multiple objects, contrast features and gradient features are widely used [14]. The infrared image contrast features and gradient features are applied to infrared image multi object segmentation to improve the performance of infrared image multi object segmentation. At present, many researchers have studied image segmentation. Zhang Yishu and others have applied watershed algorithm to laser image segmentation [15]. Experiments have verified that this method can accurately segment inductive laser thermal imaging images, but there is a defect of too much background information in the segmentation results; Zeng Yanyang et al. realized weak light image segmentation by using intercept histogram [16]. This method can effectively segment weak light images without being affected by background interference, but it has the defect of poor real-time segmentation; Ding Yongqian and others carried out adaptive segmentation for narrowband images of spectral images [17]. This method has a high adaptive performance, but it has the defect of small application range. Aiming at the defects of the above research methods when they are applied to image segmentation, the infrared image multi object segmentation method based on multi-level feature fusion is studied. The experimental results show that this method can achieve infrared image multi object segmentation, make full use of the multi-level features of the extracted infrared image, and improve the infrared image object segmentation performance.

2 Multilevel Feature Extraction and Fusion of Infrared Image 2.1 Extraction of Multi-level Features of Infrared Image 2.1.1 Infrared Image Entropy Feature Extraction Entropy feature of infrared image is an important feature to measure texture details of infrared image. When the texture evenness of the infrared image is low, the absolute value of the infrared image entropy feature is large, otherwise the absolute value of the infrared image entropy is small [18]. The gray level co-occurrence matrix method is selected to extract the entropy features of infrared images. If the original infrared image size is M × N , and the infrared image pixel exists in the Sxy area, P(i, j) means that the infrared image sliding pixel block area is. When the pixel gray level is i, the probability of the pixel moving from a fixed position to the target gray j level is P(i, j). See (1) for the calculation formula. gij (1) P(i, j) = (1 + L)2 In the formula, gij and L respectively signify the number of pixels separated from the infrared image and the total number of gray levels. The entropy characteristic expressions of infrared image pixels in different directions are shown in (2). ent(x, y) = −

L L   i

j

P(i, j) logc P(i, j)

(2)

28

C. Liang and Y. Zhang

The mean value expression of infrared image entropy features is shown in (3).  ent(x, y) = − ent(x, y)/4 (3) θ

Traverse all pixels of the infrared image [19], and obtain the entropy characteristic image formula of the infrared image as shown in (4). e1 (x, y) = ent(x, y)

(4)

In the formula, ent(x, y) and θ separately represent the entropy characteristics of the infrared image and the calculation direction of the infrared image pixels. 2.1.2 IR Image Contrast Feature Extraction The gray level co-occurrence matrix is used to extract the contrast features of infrared images. Linear weighting of different components in the infrared image, establish new components of the infrared image, and use the new components to convert the infrared image into a grayscale image [20]. The gray image expression of the infrared image is shown in (5). I1 (x, y) = 0.288 × Rf0 + 0.576 × Gf0

(5)

Where, Rf0 and Gf0 respectively mean the R component and G component of the original infrared image. The contrast value expression of the infrared image pixel in the moving area is visible from (6). con(x, y) =

L L   i

(i − j)2 P(i, j)

(6)

j

Calculate the average contrast of pixels moving in different directions, and set the obtained average contrast con(x, y) to the gray level co-occurrence matrix of pixels (x, y) in the infrared image area Sxy . The matrix c1 (x, y) is established by using the average contrast value of pixels in the infrared image, the contrast feature extraction threshold K0 is set, and the infrared image contrast feature map c2 (x, y) is obtained by c1 (x, y) binary processing. The obtained contrast feature map is the contrast feature extraction result of the infrared image [21]. The contrast feature can reflect the contour of the object area to be segmented in the infrared image. 2.1.3 Infrared Image Gradient Feature Extraction The infrared image has a small imaging range, poor gray level distribution uniformity of the infrared image, and the gray level distribution of the infrared image is prone to extreme distribution [22]. When extracting gradient features of infrared images, it is necessary to reduce the impact of infrared image brightness on infrared images [23, 24].

Multi Object Infrared Image Segmentation

29

Select coverage operator to reduce the influence of infrared image brightness on infrared image. The calculation formula of coverage operator is shown in (7). f2 (x, y) = (1 − η)f0 (x, y) + ηf1 (x, y)

(7)

Where, f0 (x, y) and f1 (x, y) separately represent the original infrared image and the all black image after cross fusion of the original infrared image; η and f2 (x, y) represent the attenuation factor and the infrared image with reduced brightness respectively. The value interval of the attenuation factor μ is [0,1]. The change of the attenuation factor is used to reduce the influence of brightness on the original infrared image. Cross fusion of all black image and infrared image is adopted to reduce the influence of brightness on infrared image, and the infrared image after brightness reduction is obtained and represented by f2 (x, y) [25, 26]. Convert the infrared image f2 (x, y) after the brightness reduction processing from RGB mode to HSI mode, and then complete the brightness reduction processing. The expression of I component of f2 (x, y) is shown in (8). I2 (x, y) = 0.288 × Rf2 + 0.576Gf2 + 0.103Bf2

(8)

Where, Rf2 , Gf2 , and Bf2 represent the R, G, and B components of the infrared image f2 (x, y). Use the I component of the infrared image to obtain the gradient feature image expression of the infrared image. See (9). g(x, y) = |I2 (x, y) − I2 (x + 1, y)| + |I2 (x, y) − I2 (x, y + 1)|

(9)

The gradient feature of the infrared image is extracted by using the difference processing of the horizontal and vertical components of the infrared image through Eq. (9). 2.2 Multi Level Feature Fusion of Infrared Image A parallel weighted feature fusion method is used to fuse the multi-level features of the extracted infrared image. A multi-level feature fusion space is designed for the infrared image, which includes the entropy features, contrast features and gradient features of the extracted infrared image [27–29]. When using parallel feature fusion method to fuse multi-level feature vectors of infrared images, use to represent imaginary units. When the dimensions of the two level feature vectors are different, use the imaginary units to fill zeros behind the feature vectors with lower dimensions, so that the extracted infrared image features have the same dimensions. The feature dimension of infrared image fused by parallel feature fusion method is expressed by max(m, n). The parallel weighted feature fusion method fully considers the influence of the relationship between multi-level features of infrared images on multi object segmentation, provides a basis for multi object segmentation of infrared images, and makes full use of the feature dimension and distinguishability of the extracted multi-level features of infrared images. According to the multi-level features of the extracted infrared image, the feature vector space expression of the fused infrared image is obtained as (10). F = f1 + if2 + jf3

(10)

30

C. Liang and Y. Zhang

Where, i and j are imaginary numbers, f1 , f2 and f3 respectively refer to the entropy feature, contrast feature and gradient feature of the infrared image. Due to the difference in the contribution rate of the extracted infrared image multilevel features to the infrared image multi object segmentation, it is necessary to set different weights w for the extracted infrared image multi-level features. See (11) for the histogram vector expression of multi-level feature fusion of infrared image multi object segmentation. F = w1 f1 + w2 if2 + w3 jf3

(11)

When the units of feature vectors in the same sample space of infrared images are different, the multi-level features of infrared images need to be normalized. See (12) for infrared image multi-level feature normalization processing method. fn = (fo − λ)/(1 + δ)

(12)

In the formula, δ and λ separately imply the variance and mean value of the infrared image feature vector; fn and fo respectively represent the infrared image feature vector that has completed the normalization process and the original infrared image feature vector. The inner product of multi-level feature vectors of infrared images is defined as (X , Y ) = X T Y , both the feature vector X and the feature vector Y are ∈ F, F representing the complex vector of multi-level features of infrared images, T representing the conjugate transpose symbol. In the multi-level feature fusion space of infrared image, the distance expression of two feature vectors is shown in (13).  F1 − F2  = (F1 − F2 )T (F1 − F2 ) (13) It can be seen from Formula (13) that the distance between the two complex vectors in the multi-level feature fusion space is only related to the value of the real part and the value of the imaginary part in the infrared image feature space. The selection order of the real part and the imaginary part of the infrared image does not affect the distance between the complex vectors, which verifies that the infrared image multi-level feature fusion has high stability.

3 Multi Object Segmentation of Infrared Image Based on Mean Shift Algorithm Mean shift algorithm is selected as the algorithm for multi object segmentation of infrared images. This algorithm is a nonparametric kernel density estimation algorithm. Weight coefficients and kernel functions are introduced to achieve multi object segmentation of infrared images using mean shift vectors in the multi-level feature fusion space of infrared images. Mean shift algorithm sets the sample mean calculation formula in the multi-level feature fusion space of infrared image, as shown in formula (14).  xi −x   zw(xi )xi xi ∈Sz K  xzi −x  (14) mz (x) =  w(xi ) xi ∈Sz K z

Multi Object Infrared Image Segmentation

31

Where, K and z purport the kernel function and mean shift bandwidth; w(xi ) represents the weight coefficient. Set the multi-level feature fusion space of infrared image as the traversal space of Mean shift algorithm, and perform mean shift processing on all feature points in the multi-level feature fusion space. All feature points of the infrared image have twodimensional space coordinate feature information. In the multi-level feature fusion space, the mean shift bandwidth is used to implement the mean shift infrared image clustering segmentation, and the multi object segmentation results are converted from the feature space to the infrared image space to obtain the infrared image after the mean shift multi object segmentation. The expression of mean shift vector of infrared image is (15). Mz (x) = mz (x)z − x

(15)

When the mean shift algorithm is used to segment infrared image multi objects, it is necessary to mean shift all feature points in the multi-level feature fusion space to obtain pixels in the multi-dimensional sphere window that meet the feature bandwidth conditions. mz (x) is used to represent the sample mean of the pixel points in the multidimensional sphere obtained through screening, and the mean shift vector Mz (x) is calculated using the sample mean. When the mean shift vector obtained is less than the set threshold, it means that the feature point has drifted to the point with the maximum probability density. The sample mean value obtained at this time is used to replace the value of the completed drift, and the feature point will complete the drift, and the next feature point will drift. When the mean shift vector of the multi-level feature fusion space of the infrared image is not less than the set threshold, assign the sample mean value to the feature point, continue the drift of the feature point until the point with the maximum probability density is obtained, and complete the drift of the feature point. When mean shift is carried out by using multi-level feature fusion results, the feature points in the multi-level feature fusion space of infrared image are regarded as multidimensional vectors. The expression of adding multidimensional kernel function to mean shift vector is shown in (16).  p 2    x s zr   xr 2 c  K    (16) Kzs ,zr (x) = pK  z z  zs zf  zf zs2 zr r f In the formula, xs and xr individually mean the distance between the points in the multi-dimensional sphere and the feature dimension s and r in the multi-level feature fusion space; zs and zr separately represent the spatial bandwidth and spectral bandwidth of mean shift; zf indicates the characteristic bandwidth of the infrared image. Multidimensional Gaussian kernel function is used as the kernel function of infrared image multi object segmentation, and formula (16) is transformed into the expression of formula (17). 







1 − X s zp  − X r zp  Kzs ,zr (x) = √ e zs r e zr r 2π

(17)

32

C. Liang and Y. Zhang

The expression of the multidimensional mean shift sample mean value of the infrared image multi object segmentation obtained according to formula (17) is as follows (18).  s 2  r 2   p p − Xzs zr  − Xzr zr  √1 e e w(xi )xi 2π  2  2  1 − Xz s  − Xz r  s √ e e r w(xi ) 2π

 mzs ,zr (x) =

(18)

In the formula,  2 represents the similarity between the points in the highdimensional sphere obtained through screening and the points in the multi-level feature fusion space in the feature dimensions. The steps for multi object segmentation of infrared image using multi feature mean shift algorithm are as follows. (1) The multi-level features of the extracted infrared image are combined into a multilevel feature fusion space; (2) Set the spatial bandwidth and spectral bandwidth of mean shift, as well as the characteristic bandwidth of infrared image. Mean shift all feature points in the infrared image. Use the current feature point coordinates to preliminarily screen the index of spectral bandwidth points, calculate the spectral characteristics between the points in the multi-level feature fusion space of infrared images and the current feature points, as well as the distance from the third type of features, and obtain the feature vector screening results. The cubic window filtering method is used to obtain the feature points that can be applied to mean shift in the multi-dimensional sphere; (3) Calculate the mean value of infrared image feature samples and the drift amount of infrared image feature samples in the mean shift. Set the error threshold. When the drift between the sample point and the sample point is less than the error threshold, set a new center point to replace the current point, set the center point feature value obtained through calculation in the corresponding position feature space, and obtain the multi-dimensional feature space of the infrared image after the mean shift. In the multi-dimensional feature space of the infrared image with mean shift, the regional growth method is selected to calculate the multi-level feature similarity and determine the growth point. The region growing points of the infrared image are merged, and the multi object segmentation results of the infrared image are obtained after the multi-level feature point traversal of the infrared image is completed.

4 Application of Infrared Image Multi Object Segmentation Method in Fault Diagnosis System In order to verify the effectiveness of the multi object segmentation method of infrared image based on multi-level feature fusion, the method is applied to the fault diagnosis system of an electric power enterprise. The power enterprise uses infrared imaging system to collect many power equipment running in the power system, and the performance of multi object segmentation of infrared image has a great impact on the performance of power equipment fault diagnosis. As an important technology of power equipment fault diagnosis, the multi object segmentation result of infrared image determines the performance of power equipment fault diagnosis, which also has an important impact on the operation of power enterprises.

Multi Object Infrared Image Segmentation

33

The infrared image of the current contactor of the power system is selected as the experimental object. The infrared image of the current contactor collected by the infrared imaging system has low definition, which improves the difficulty of multi object segmentation of the infrared image. The multi-level features of the infrared image of the current contactor are extracted by this method, and the multi-level features of the extracted infrared image are fused. The multi object segmentation of the infrared image of the current contactor is better by using the multi-level feature fusion result of the infrared image. The infrared image of the current contactor of the power system on November 20, 2020 is chosen randomly as the experimental object. The original infrared image of the current contactor collected by the infrared imaging system is shown in Fig. 1. It can be seen from the infrared image of the original current contactor in Fig. 1 that the infrared image of the original current contactor has low definition, which improves the difficulty of multi object segmentation of the infrared image.

Fig. 1. Infrared image of original current contactor

The multi-level features of the infrared image of the current contactor are extracted by the method in this paper, and the multi-level features of the infrared image extracted are fused. The infrared image of the current contactor is segmented by using the multilevel feature fusion result of the infrared image. The result of multi object segmentation is shown in Fig. 2. It can be seen from the experimental results in Fig. 2 that this method can effectively segment multiple objects in the infrared image of the current contactor, and this method can effectively segment the four objects contained in the infrared image of the current contactor. The result of multi object segmentation of infrared image of current contactor in this method can be used as an important basis for fault diagnosis of power equipment and provide a good image basis for fault diagnosis of power system. In order to further verify the multi object segmentation performance of this method for infrared images, this method is used to segment the infrared images of different power equipment in the power system. Statistics of the performance of infrared image multi object segmentation using this method are shown in Table 1. It can be seen from the experimental results in Table 1 that this method can effectively segment multiple objects in the infrared images of different power equipment. Using this method to segment infrared image multi objects, there are only a few wrong segmentation object pixels, and the multi object segmentation accuracy is higher than 99.5%. The experimental results show that the proposed method can effectively use the multi-level features contained in

34

C. Liang and Y. Zhang

Fig. 2. Multi object segmentation result of infrared image

the infrared image to achieve accurate segmentation of multiple objects in the infrared image. Table 1. Infrared Image Multi object Segmentation Results Name of electric equipment

Number of divided objects/piece

Pixels of segmentation object/piece

Actual object pixels/piece

Number of pixels wrongly segmented object/piece

Oil immersed transformer

2

1852

1857

5

Current transformer

3

1678

1680

2

Voltage transformer

2

1984

1988

4

Alternator

3

2052

2055

3

Arrester

4

948

952

4

Motor

3

1085

1090

5

Insulator

2

1385

1381

4

Current limiting reactor

2

1258

1261

3

Circuit breaker

3

1648

1655

7

Isolating switch

7

2354

2361

7

Fuse

5

1852

1858

6

Dice similarity coefficient is selected as the evaluation index to measure the performance of infrared image multi object segmentation. This index is an important index used in image segmentation evaluation and an important index for measuring the similarity between infrared image multi object segmentation results and original infrared image objects. See (19) for the calculation formula of Dice similarity coefficient. Dice =

2|p ∩ q| |p| + |q|

(19)

Multi Object Infrared Image Segmentation

35

In Formula (19), p and q separately represent the pixel value of the infrared image segmentation object and the pixel value of the original infrared image object. The value of Dice similarity coefficient is in the [0,1] range. The higher the Dice similarity coefficient, the better the infrared image multi object segmentation performance. The infrared image multi object obtained by segmentation can completely retain the edge feature details of the object, and solve the defect that the infrared image cannot accurately segment the object due to its low definition.

5 Conclusion Multi object infrared image segmentation is an important content of infrared image analysis and infrared image pattern recognition, and also an important research topic in the field of computer vision. The infrared image contrast feature and gradient feature are applied to infrared image multi object segmentation, which can effectively improve the performance of infrared image multi object segmentation. The multi object segmentation method of infrared image based on multi-level feature fusion is studied. Multilevel features are extracted from infrared image, and the result of multi-level feature fusion is taken as the basis of infrared image multi object segmentation. The experimental results show that the multi object segmentation of infrared image using multi-level feature fusion results has high segmentation accuracy and efficiency. The proposed method can accurately extract multiple objects from the background area of infrared image, and the object information of the infrared image multi object segmentation result is high. The method studied has the advantage of simple calculation, and can be applied to infrared image segmentation occasions that require high segmentation efficiency and high segmentation accuracy.

References 1. Tereikovskyi, I., Hu, Z., Chernyshev, D., et al.: The method of semantic image segmentation using neural networks. Int. J. Image Graph. Signal Proc. 14(6), 1–14 (2022) 2. Kurama, V., Alla, S., Vishnu, K.R.: Image semantic segmentation using deep learning. Int. J. Image Graph. Signal Process. 10(12), 1–10 (2018) 3. Mei, F., Chen, S., Li, Y., et al.: Investigation of infrared image prediction for subsonic exhaust plume. Int. J. Eng. Manufac. 2(3), 46–52 (2012) 4. Mei, F., Chen, S., Jiang, Y., Cai, J.: A preliminary model of infrared image generation for exhaust plume. Int. J. Image Graph. Signal Process. 3(4), 46–52 (2011) 5. Zeng, Y., Xie, G., Zhang, J.: Weak light image segmentation based on intercept histogram and Otsu fusion. Adv. Laser Optoelectron. 58(20), 219–227 (2021). (in Chinese) 6. Chegeni, M.H., Sharbatdar, M.K., Mahjoub, R., Raftari, M.: New supervised learning classifiers for structural damage diagnosis using time series features from a new feature extraction technique. Earthquake Eng. Eng. Vibrat. 21(1), 169–191 (2022) 7. Ding, Y., Xing, Z., Jiang, Y., et al.: Adaptive light intensity segmentation method for narrowband image in spectral index measurement. J. Nanjing Agric. Univ. 43(03), 574–581 (2020). (in Chinese) 8. Anh, N.T.N., Thang, T.N., Solanki, V.K.: Product feature extraction from the descriptions. SpringerBriefs in Applied Sciences and Technology pp. 43–51 (2022)

36

C. Liang and Y. Zhang

9. Long, Y.-H., Chen, Y.-C., Chen, X.-P., Shi, X.-H., Zhou, F.: Test-driven feature extraction of web components. J. Comput. Sci. Technol. (English) 37(2), 389–404 (2022) 10. Zhang, L., Xu, X., Cao, C., et al.: Robot pose estimation method based on image and point cloud fusion of dynamic feature elimination. China Laser 49(06), 58–69 (2022). (in Chinese) 11. Li, C., Lan, H., Wei, X.: Attention based millimeter wave lidar fusion target detection. Comput. Appl. 41(07), 2137–2144 (2021). (in Chinese) 12. Jiang, B., Ma, Y., Wan, J., et al.: Local feature extraction of lidar point cloud based on hemispherical neighborhood. Signal Process. 38(02), 329–339 (2022). (in Chinese) 13. Li, Y., Li, F., Qu, H., et al.: Infrared image segmentation method based on improved artificial bee colony sine cosine optimization. Laser Infrared 51(08), 1076–1080 (2021). (in Chinese) 14. Yao, Y., Cheng, G., Xie, X., et al.: Target detection of optical remote sensing image based on multi-resolution feature fusion. J. Remote Sens. 25(05), 1124–1137 (2021). (in Chinese) 15. Yang, X., Gao, L.: Hyperspectral image classification based on reduced dimension Gabor feature and decision fusion. Comput. Appl. Res. 37(03), 928–931 (2020). (in Chinese) 16. Shen, H., Meng, Q., Liu, Y.: Facial expression recognition based on lightweight convolution network multilayer feature fusion. Adv. Laser Optoelectron. 58(06), 148–155 (2021). (in Chinese) 17. Xi, Z., Yuan, K.: Image super-resolution reconstruction based on residual channel attention and multi-level feature fusion. Progress Laser Optoelectron. 57(04), 262–270 (2020). (in Chinese) 18. Wu, J., Lu, Z., Guan, Y., et al.: SAR target recognition method based on feature fusion of 2D compressed sensing multi projection matrix. Infrared Laser Eng. 50(06), 314–320 (2021). (in Chinese) 19. Wang, F., Sun, H., Wang, Y., et al.: Bilateral filtering enhancement and FCM segmentation of femtosecond laser ablation spot image. Laser Infrared 50(05), 630–633 (2020). (in Chinese) 20. Zhang, Y., Zhang, B., Zhao, Y., et al.: Remote sensing image classification based on dual channel depth dense feature fusion. Laser Technol. 45(01), 73–79 (2021). (in Chinese) 21. Yu, X., Shan, D., Yu, Y., et al.: Hyperspectral image hybrid convolution classification based on multi feature fusion. Infrared Technol. 44(01), 9–20 (2022). (in Chinese) 22. Ma, L., Gou, Y., Lei, T., et al.: Small target detection in remote sensing image based on multi-scale feature fusion. Optoelectron. Eng. 49(04), 49–65 (2022). (in Chinese) 23. Liu, J., Zheng, C., Zhang, L., et al.: Hyperspectral image classification method based on image reconstruction feature fusion. China Laser 48(09), 203–212 (2021). (in Chinese) 24. Zhang, Y., Wang, X., Hou, D., et al.: Image segmentation of inductive laser thermal imaging based on watershed algorithm. Infrared Technol. 43(04), 367–371 (2021). (in Chinese) 25. Xu, X., Xue, D., Qi, G.: Laser thermal imaging image segmentation based on regional differences. Laser J. 43(10), 83–86 (2022). (in Chinese) 26. Xu, W., Jin, G., Miao, Z., Yang, J.: Analysis of on-line transformer fault detection technology based on infrared image. Bonding 49(09), 193–196 (2022). (in Chinese) 27. Lu, Y.: Thermal defect detection method for buildings based on infrared image features. Dalian University of Technology, Dalian (2022). (in Chinese) 28. Islam, S.M.M., Debnath, R.: A Comparative Evaluation of feature extraction and similarity measurement methods for content-based image retrieval. Int. J. Image Graph. Signal Process. 12(6), 19–32 29. Yasaswini, V., Baskaran, S.: An optimization of feature selection for classification using modified bat algorithm. Int. J. Inform. Technol. Comput. Sci. 13(4), 38–46 (2021)

Spear-Man Correlation Data Analysis of Scale of Intelligent Logistics and Investment and Financing of Logistics Technology Ni Cheng1(B) and Meng Lei2 1 Wuhan Railway Vocational College of Technology, Wuhan 430205, Hubei, China

[email protected] 2 Hubei Changjiang Media and Press Group, Wuhan 430079, Hubei, China

Abstract. In recent years, with the development of technologies such as the Internet and artificial intelligence, as well as higher requirements of new retail and intelligent manufacturing, the market scale of intelligent logistics in our country continues to expand. At the same time, the development prospect of intelligent logistics has been favored by social capital of all parties, and the amount of investment and financing in China’s logistics technology is on the rise. Whether there is a correlation between the scale of intelligent logistics and the investment and financing of logistics technology is very important for promoting logistics to enter the smart age and seeking the opportunity of listing in the logistics industry in the future. To solve this problem, This paper firstly collects and collates the data of the scale of intelligent logistics and the amount of investment and financing in logistics technology, and then uses spear man data analysis to explore the correlation between them, finally proposes to promote the diversified development of various subdivided fields of logistics technology, closely grasp the three mainstream fields of investment and financing, and combine the Internet, big data and artificial intelligence to promote the digitized process of logistics. Keywords: Spear-man · Data Analysis · Intelligent logistic · Investment and Financing of Logistics Technology · Artificial Intelligence

1 Introduction Intelligent logistics was first proposed by IBM. In December 2009, the Information Center of China Logistics Technology Association, Hua Xia Internet of Things and editorial Department of Logistics Technology and Application jointly proposed the concept. It refers to the modern logistics mode that realizes the refined, dynamic, and visual management of each link of logistics, improves the intelligent analysis, decision making and automatic operation and execution ability of logistics system, and improves the efficiency of logistics operation through intelligent technology means such as intelligent hardware and software, Internet of Things, big data and so on [1]. According to the data of China Federation of Logistics and Purchasing, logistics data, logistics cloud and logistics equipment are the three major areas that logistics enterprises © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 37–46, 2023. https://doi.org/10.1007/978-3-031-36118-0_4

38

N. Cheng and M. Lei

demand for intelligent logistics at present. From 2015 to 2019, the scale of intelligent logistics in China maintained a double-digit growth rate. In 2019, the market size has reached 500 billion yuan, with a year-on-year growth of 23.1%. Even under the impact of the global epidemic, the scale of intelligent logistics in China reached 584 billion yuan in 2020 and 647.7 billion yuan in 2021, and the market size is expected to exceed one trillion yuan by 2025. In recent years, with the development of technologies such as the Internet of Things and artificial intelligence, as well as higher requirements for logistics in fields such as new retail and intelligent manufacturing, rapidly growing demand is forcing the traditional logistics industry to transform to smart logistics [2]. As a result, the market size of intelligent logistics will continue to expand. On the other hand, the development prospect of intelligent logistics has been favored by social capital of all parties, and the amount of investment and financing in China’s logistics technology is on the rise. In 2018, due to the development of logistics technology towards the direction of refinement, the industry is becoming more mature, and the investment and financing competition of various circuits is becoming increasingly fierce. In this context, although the amount of investment and financing steadily rises, the number of investment and financing cases has decreased significantly. On the other hand, whether the increase of the amount of investment and financing of logistics technology can affect the scale of domestic intelligent logistics is also a question worth discussing. Therefore, it is of great practical significance to clarify the relationship between the scale of intelligent logistics and the investment and financing of logistics technology.

2 Literature Review The research on intelligent logistics at home and abroad mainly focuses on: Professor Wang Zhitai believes that the new generation of information technology and modern management theory are deeply integrated with the Internet and applied to the logistics industry, and finally realize an innovative form of logistics, namely intelligent logistics [3]. The research on the definition and development of intelligent logistics is relatively mature, which is related to enterprise management and talent training. For example, Qi Jianxiu pointed out that the use of automation technology to replace manual operation can improve the comprehensive efficiency of logistics enterprises, and realize the reasonable scheduling and distribution of resources through the effective integration of social resources [4]. Yang Hongyue analyzed the influence on the training of logistics financial talents under the background of intelligent logistics, and proposed the training ideas of logistics financial talents under the “1 + X” certificate system [5]. With the development of informatization, networking and intelligence of smart logistics technology, research on smart logistics mainly focuses on big data analysis, AI technology, blockchain technology, Internet of technology and specific case studies. Dong Haifang uses big data technology to analyze logistics data, thereby improving the efficiency of logistics management [6]. 5G can meet the massive data exchange between management equipment, products, vehicles, and workers, and substantially support the development of smart logistics [7]. Bi Ying proposed the future development trend of logistics industry from the perspective of “intelligent logistics + AI technology” [8]. Based on the characteristics of blockchain technology, Zhou Xinming et al. analyzed its role in the

Spear-Man Correlation Data Analysis of Scale of Intelligent

39

development of intelligent logistics platform [9]. Feng Yayun pointed out that the application of Internet of Things technology changes the traditional management mode and promotes the development of intelligent logistics field to the direction of intelligence and information [10]; Hao Shuchi took the green smart logistics in the Guangdong-Hong Kong Greater Bay Area as the research object to build the evaluation index system of green smart logistics distribution capability [11]. To sum up, the research on intelligent logistics has attracted the full attention and attention of scholars at home and abroad. For example, measuring systems are used to evaluate the costs associated with each logistics process. There are positive implications for the sustainable development of logistics [12]. Most of the research focuses on specific logistics technology fields and case studies, but lacks the overall grasp of logistics technology investment and financing.

3 Research Methods In this paper, the Spearman Correlation Coefficient is used to measure the dependence correlation research between the two variables of Scale of Intelligent Logistics (Y)and Investment and Financing of Logistics Technology (X). The Spearman’s Rank Coefficient of Correlation, or Spearman Correlation Coefficient, is a nonparametric measure of rank correlation. Named after the British statistician Charles Spearman, it is often written as the Greek letter ‘ρ’ (rho). Full version of the calculation formula: 1 n

ρ =  1 n

n      R(xi ) − R(x) · R(yi ) − R(y) i=1

2   2 n  n   1 · R(xi ) − R(x) R(yi ) − R(y) n

i=1

(1)

i=1

Thereinto: • R(x) and R(y) are the positions of x and y, respectively • R(x) and R(y) indicate average ranks, respectively A simpler and easier calculation formula is as follows:  2 6 N i=1 di ρ =1− N (N 2− 1)

(2)

Where: di Represents the difference in the bit value of the i-th data pair. n Total number of observed samples. If Y tends to increase when X increases, the Spearman Correlation Coefficient is positive. If Y tends to decrease as X increases, the Spearman Correlation Coefficient is negative. The Spearman Correlation Coefficient is zero, indicating that there is no tendency for Y to increase as X increases.

40

N. Cheng and M. Lei

3.1 Data Collection Based on the Spearman Correlation Coefficient, this paper selects indicators of the Scale of China’s Intelligent Logistics (Scale of CIL), the Total Annual Revenue of Logistics Industry (TAR of LI), the Amount of Investment and Financing in China’s Logistics Technology (Amount of I&F), and the Investment and Financing Events in China’s logistics technology as indicators (I&F Events). In order to ensure the scientific, authoritative, and objective evaluation results, the statistical data in this paper are from 2015–2021 China Federation of Logistics and Purchasing, National Bureau of Statistics, Statistical bulletins of National Economic and Social development over the years and related websites of China Logistics and Purchasing Website. Some index data are obtained through calculation. Table 1. CIL from 2015 to 2021 (RMB 100 million), TAR of LI (trillion), Amount of I&F (RMB 100 million), I&F Events (number) Year

CIL (RMB 100 million)

TAR of LI (one trillion yuan)

Amount of I&F (RMB 100 million)

I&F Events (number)

2015

2205

7.6

355.1

16

2016

2790

7.9

760.9

143

2017

3375

8.8

1770.3

117

2018

4060

10.1

1345.8

106

2019

4885

10.3

606.4

137

2020

5840

10.5

413.2

90

2021

6477

11.9

1815.7

113

Table 1 shows that from 2015 to 2021, the Scale of China’s Intelligent Logistics (Scale of CIL) and the Total Annual Revenue of Logistics Industry (TAR of LI) had an increasing trend year by year. In 2021, the Scale of China’s Intelligent Logistics (Scale of CIL) exceeded 647.7 billion yuan, hitting a record high. From 2015 to 2019, the Amount of Investment and Financing in China’s Logistics Technology (Amount of I&F) has been on the rise. As shown in Fig. 1, in 2018, due to the development of Logistics Technology in the direction of refinement and the increasingly mature industry, the number of major financing events in the logistics subdivision industry, intelligent logistics continued to stay popular, freight O2O and warehouse logistics gradually received capital attention, and intelligent express cabinets, cold chain logistics, cross-border logistics and logistics drones were favored. Whether there is any correlation between the Scale of China’s Intelligent Logistics (Scale of CIL) and Investment and Financing in China’s Logistics Technology (I&F in CLT) is very important to promote logistics into the age of smart, shape the new pattern of Logistics Technology, and seek the opportunity of future listing in the Logistics Industry.

Spear-Man Correlation Data Analysis of Scale of Intelligent

41

Fig. 1. Investment and Financing situation of each subsector of Logistics Technology in China from 2015 to 2021

3.2 SPSS Determination of Applicable Conditions 3.2.1 Solving Model References [13–15] use Pearson’s Correlation to make statistical measures of the strength of the linear relationship between two random variables, giving only the correlation of X and Y described by the linear equation. Spearman Correlation analysis is used when examining the strength of the monotonic relationship between two variables. References [16–22] Through Spearman Correlation Analysis, we explore the extent to which the two variables tend to become larger or smaller in step, to accurately obtain the correlation between the sampling probability distributions of X and Y. Main steps are listed: 1) Normality test: The variable contains a hierarchical variable, or the variable does not follow a normal distribution or the distribution type is unknown. This is the first condition that satisfies Spearman Correlation Analysis. 2) Monotonicity judgment: There is a monotonic relationship between two variables. This is the second condition that satisfies Spearman Correlation Analysis.

42

N. Cheng and M. Lei

3) Use SPSS to input the data to make the data a function, and the function model can refer to the formula (2) 3.2.2 Normality Test Draw Q-Q charts for Scale of China’s Intelligent Logistics (Scale of CIL) (RMB 100 million) and Investment and Financing in China’s Logistics Technology (I&F in CLT) (RMB 100 million) respectively, as shown in Fig. 2 and Fig. 3.

Fig. 2. Normal Q-Q plot of Scale of CIL

Fig. 3. Normal Q-Q plot of I & F in CLT

The comparison shows that the scattered points of variables deviate from the diagonal distribution more, indicating that the two variables did not obey the normal distribution. Apply spearman conditions. 3.2.3 Monotonically Judgment Monotonic judgment is a meaningful analysis.

Spear-Man Correlation Data Analysis of Scale of Intelligent

43

Fig. 4. Scatter plot of Scale by Investment-f

Draw a scatter plot for Scale of China’s Intelligent Logistics (Scale of CIL) (RMB 100 million) and t Investment and Financing in China’s Logistics Technology (I&F in CLT) (RMB 100 million) (due to the impact of the epidemic, the data in 2019 and 2020 are especially excluded), and the results are shown in Fig. 4. As shown in the Fig. 4, the Scale of CIL (RMB 100 million) and the I & F in CLT (RMB 100 million) increase with the increase, and the two variables show a monotonous relationship. And the relationship between them is positive. In conclusion, the data in this case can be analyzed by Spearman Correlation. 3.2.4 SPSS Calculation of Spearman Correlation Set the Scale of CIL (RMB 100 million) as the analysis item Y, and set the I&F in CLT (RMB 100 million) as X. Use SPSS27.0 statistical software to calculate: Table 2. Correlations Coefficient

Spearman’s rho

Scale

Investment_f

Scale

Investment_f

Correlation Coefficient

1.000

.393

Sig. (2-tailed)

–

.383

N

7

7

Correlation Coefficient

.393

1.000

Sig. (2-tailed)

.383

–

N

7

7

Correlation Analysis was used to study the correlation between the Scale of China’s Intelligent Logistics (Scale of CIL) and Investment and Financing in China’s Logistics Technology (I&F in CLT), and Spearman’s Correlation Coefficient was used to indicate the strength of the correlation.

44

N. Cheng and M. Lei

According to the detailed analysis in Table 2, the correlation value between the Scale of CIL and the amount of I&F in CLT is 0.393, indicating that there is a significant positive correlation between the two. The more investment in Investment and Financing will promote the continuous development of China’s Intelligent Logistics scale. The following suggestions are given based on the Investment and Financing situation of various subdivisions of Logistics Technology in China.

4 Discussion 4.1 The Diversification of Logistics Technology Segments Promotes the Expansion and Development of Intelligent Logistics According to Fig. 1, the investment and financing of logistics technology in 2018 are mainly concentrated in the contract logistics industry, transportation industry, logistics automation and distribution. With the continuous deepening and expansion of China’s intelligent logistics, the subdivided fields of logistics technology have increased to 8 fields in 2019 and 13 fields in 2020. The specific contents include supply chain, crossborder logistics, cold chain logistics, automatic driving, air trunk, intra-city distribution, etc. Diversified development will become an important factor affecting the scale of intelligent logistics. 4.2 Investment and Financing in the Three Major Areas have Become the Mainstream, Promoting the Application and Development of the Intelligent Logistics Industry In 2019, 2020 and 2021, warehousing, transportation and distribution have become the major financing and investment areas of Logistics Technology. The main reasons are as follows: first, the traditional labor force cannot fully meet the warehouse side’s demand for efficiency improvement and refined management; Secondly, in recent years, transportation costs account for more than 50% of the total logistics costs. Reducing transportation costs is an important means to reduce costs in the Logistics Industry. However, logistics costs, which are mainly labor costs and fuel costs, can be compressed. Third, in the terminal distribution, distribution needs are diverse, distribution time conflict, low efficiency and high cost and other problems are prominent. Therefore, investment and financing events in the three areas are highly concerned by the capital. 4.3 Internet of Things, Big Data, and Artificial Intelligence to Promote the Intelligent Logistics The Internet of Things technology connects the virtual network world with real objects, providing part of the data source for Big Data, and Artificial Intelligence technology is the deep learning technology under Big Data. Transportation management platforms, Fintech companies, instant logistics companies, and e-commerce platforms have applied digital technologies in their daily operations and management. With the continuous progress of science and technology and the support of macro policies and economy,

Spear-Man Correlation Data Analysis of Scale of Intelligent

45

Artificial Intelligence will replace human labor and brain power in the warehousing link. In the transportation sector, the application of unmanned driving technology can reduce labor costs and achieve complete autonomous driving. In the delivery sector, drones will become a part of smart delivery in vast rural areas and urban communities.

5 Conclusion With the development of logistics technology in the direction of refinement, the number of major financing events in the logistics segmentation industry and Intelligent Logistics continue to stay hot. Warehousing, transportation, and distribution have become the main areas of Financing and Investment in Logistics Technology. For example, the truck formation technology in the transportation link, automatic sorting technology in the storage link, AGV handling technology, intelligent express cabinet in the distribution link, unmanned distribution technology and so on have gradually attracted the attention of the capital. Logistics Technology helps logistics operations to realize intelligent and efficient operation, and realize the efficient flow of goods from the production place to the consumption place. Therefore, the Investment and Financing of Logistics Technology will shape the new pattern of Logistics Technology and promote the expansion of Intelligent Logistics.

References 1. Qian, H.-M., He, J., Guan, J.: Research on the coupling effect evaluation of Intelligent + Sharing Logistics. China Bus. Market (11), 3–16 (2019). (in Chinese) 2. Wang, T.: Internet of things and big data in smart logistics. In: Proceedings of the 2022 2nd International Conference on Education, Information Management and Service Science (EIMSS 2022), pp. 1265–1271 (2022) 3. Wang, Z.: Smart Logistics Is needed by urbanization. China Bus. Market 3, 4–8 (2014). (in Chinese) 4. Qi, J.: Research on the development status and improvement countermeasures of intelligent logistics integration. Transport. Manager World 20, 50–52 (2022). (in Chinese) 5. Yang, H., Ding, C.: Analysis of problems and countermeasures faced by logistics financial talent training in the context of smart logistics. Logist. Technol. 45(11), 164–166 (2022). (in Chinese) 6. Dong, H.: Discussion on the application mode of smart logistics based on big data. China Logist. Procurem. 20, 72–73 (2022). (in Chinese) 7. Alexandra, L., Chiara, C., Roberto, P., Sergio, C.: 5G in Logistics 4.0: potential applications and challenges. Procedia Comput. Sci. (217), 650–659 (2023) 8. Bi, Y.: Smart logistics + AI technology drives the transformation of the logistics industry. China’s Storage Transport. 08, 142–143 (2020). (in Chinese) 9. Zhou, X., Hu, Y.: Research on the development of intelligent logistics platform based on blockchain technology. Logistics Eng. Manage. 24–26 (2022) 10. Ya, Y.: Research on the application of internet of things in smart logistics management. China Logist. Procurem. (06), 93–94 (2022). (in Chinese) 11. Hao, S.: Research on the development of green and smart logistics distribution in the Guangdong-Hong Kong-Macao greater bay area. Logist. Technol. 45(15), 37–41 (2022). (in Chinese)

46

N. Cheng and M. Lei

12. Benotmane, Z., Belalem, G., Neki, A.: A cost measurement system of logistics process. Int. J. Inform. Eng. Electron. Bus. 10(5), 23–29 (2018) 13. Strelnytskyi, O.O., Svyd, I.V., Obod, I.I., Maltsev, O.S., Zavolodko, G.E.: Optimization of secondary surveillance radar data processing. Int. J. Intell. Syst. Appl. 11(5), 1–8 (2019) 14. Jagadesh, B.N., Srinivasa Rao, K., Satyanarayana, C.: A robust skin colour segmentation using bivariate pearson type iiαα (bivariate beta) mixture model. Int. J. Image Graph. Signal Process. 4(11), 1–8 (2012) 15. Lemmer, H.H.: The allocation of weights in the calculation of batting and bowling performance measures. South African J. Res. Sport Phys. Educ. Recreat. 29(2), 75–85 (2007) 16. Dey, P.K., Ghosh, D.N., Mondal, A.C.: PL team performance analysis: a multi-criteria group decision approach in fuzzy environment. Int. J. Inform. Technol. Comput. Sci. 7(8), 8–15 (2015) 17. Joanna, R., Andrzej, N., Piotr, H., Ewa, D.: Application of Spearman’s method for the analysis of the causes of long-term post-failure downtime of city buses. Appl. Sci. 12(6), 2921 (2022) 18. Jie, T.M.R., He, K.K.: Ultrasound measurements of rectus femoris and locomotor outcomes in patients with spinal cord injury. Life 12(7), 1073–1073 (2022) 19. Irina, K., Asta, P., Ginas, C., Aldona, A., Berita, S., Ramunas, K.: The Impact of achievements in Mathematics on cognitive ability in primary school. Brain Sci. 12(6), 736–736 (2022) 20. Nurul, F.A., Immanuel, M., Nur, H.: The correlation between sedentary lifestyle and physical fitness level in adolescents. Enfermería Clínica 31(5), S668–S671 (2021) 21. Du, H., Sun, L., Chen, S., et al.: Noninvasive prenatal prediction of fetal haplotype with Spearman rank correlation analysis model. Mol. Genet. Genomic Med. 10(8), 1988 (2022) 22. Andréas, H., Alfonso, V.: The Kendall and Spearman rank correlations of the bivariate skew normal distribution. Scand. J. Stat. 49(4), 1669–1698 (2022)

Neural Network Model for Laboratory Stand Control System Controller with Parallel Mechanisms Peter Kravets1 , Anatolii Novatskyi1 , Volodymyr Shymkovych1(B) , Antonina Rudakova1 , Yurii Lebedenko2 , and Hanna Rudakova2 1 National Technical University of Ukraine «Igor Sikorsky Kyiv Polytechnic Institute», 37

Prosp. Peremohy, Kyiv, Ukraine [email protected] 2 Kherson National Technical University, 24 Beryslav Highway, Kherson, Kherson Region, Ukraine

Abstract. The paper considers a frame installation with mechanisms of parallel structure. In it the movement of the tool is carried out in the vertical plane. The movement of the working body is set by the upper control level, each motor is programmed through the local control system, and the sensors monitor their position. To improve the control system, the integration of an automated system using neural networks is proposed. A mathematical model of a frame installation with mechanisms of parallel structure is constructed. The neural network model of this object is constructed with the help of specialized software. This model has the optimal structure for the error of reproduction of the object and the minimum structure. Shows that the smallest error, equal to 0.001236, has a structure with 11 neural networks in the hidden layer, with 1 delayed input and 2 delayed outputs and is the best structure for the selected system. The implementation of neural network control system on hardware the Field-Programmable Gate Array (FPGA) is proposed. This implementation will allow the neural network controller to adapt to undefined system parameters in real time. That made it possible to increase the efficiency of the control of this object. Due to the consideration by the neural network of those parameters that cannot be described mathematically and taken into account at the system design stage. Keywords: Frame installation · manipulator · neural network · parallel structure · control system · FPGA

1 Introduction There are basically two types of manipulators: serial and parallel [1–4]. Serial manipulators are open structures consisting of several links connected in series [1, 2]. Such a manipulator can be efficiently operated throughout the working space. The actuator must basically work and shift the entire manipulator with its links and actuators. Implementing fast and accurate movements with such manipulators is a difficult task. When solving © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 47–58, 2023. https://doi.org/10.1007/978-3-031-36118-0_5

48

P. Kravets et al.

this problem, there are problems of poor rigidity and reduced accuracy. In contrast to serial manipulators, parallel manipulators consist of many circuits with a closed loop [3, 4]. These chains control the end effector together in a parallel structure. They can take a variety of forms. The most common form of parallel manipulators are platform manipulators. Platform manipulators have an architecture similar to the architecture of flight simulators. In such a system, we can distinguish two special links – the base and the moving platform [5, 6]. Parallel manipulators have better positioning accuracy, higher rigidity and higher load capacity. Because in parallel manipulators the total load on the system is distributed between the actuators. The most important advantage of parallel manipulators is the ability to keep all their actuators fixed on the basis of [7, 8]. The moving mass in them can be much higher. This type of manipulator can perform fast movements. But their working spaces are much smaller, which limits the full operation of these preferred features. Perfect control of drives is necessary for fast and exact movements of parallel manipulators. To minimize tracking errors, dynamic forces must be compensated by the controller. For accurate compensation it is necessary to know exactly the parameters of the dynamic model of the manipulator [8, 9]. Closed mechanical circuits make the dynamics of parallel manipulators extremely complex, and their dynamic models nonlinear. When constructing a control system for a parallel manipulator, informational uncertainty in the parameters of the object itself should be taken into account [10, 11]. There are parameters that can be determined, such as mass. And there are a number of parameters that cannot be accurately determined such as strength factors. Because of this, many control methods are ineffective. Therefore, it is necessary to determine the unknown parameters of the system during operation and adapt the control system to them. The application of artificial neural network technology in control systems allows to take into account the uncertainties in the system [12–14, 26, 27]. Neural network control systems are a high-tech direction of control theory and belong to the class of nonlinear dynamic systems. High speed due to parallelization of input information combined with the ability to learn artificial neural networks makes this technology very attractive for creating control devices in automatic systems. Regulators built on the basis of artificial neural networks can be learned in the process and take into account all the uncertainties in the design of the system. Parallel frame installation is a multidimensional control object. Neural networks are a universal tool for modeling multidimensional nonlinear objects and finding solutions to incorrect problems. Using the reinforced learning paradigm, you can teach the neural network to deviate a given position of the platform from the current one. By implementing the entire system on FPGA [15–19], can get an accurate and high-speed control system for the platform of the parallel manipulator. Therefore, the development of a control system for manipulators based on artificial neural networks will increase the efficiency of the parallel manipulator. The aim of the work is to develop neural network model and controller of the control system of the frame installation with mechanisms of parallel structure, which allows to install the platform in a given position. By implementing a neural network controller capable of learning in real time into the control system of this object, we want to achieve an increase in control accuracy

Neural Network Model for Laboratory Stand Control System

49

compared to traditional methods. Due to additional training of the neural network during the operation of the stand, which will allow to take into account those parameters of the object that cannot be taken into account during its mathematical modeling.

2 Design of Frame Installation with Mechanisms of Parallel Structure An example of such a frame installation of a parallel structure is a research stand with two guide rods, developed and constructed at the Kherson National Technical University, which is shown in Fig. 1.

Fig. 1. General view of the installation of the frame structure

The installation consists of a metal frame, which is equipped with two stepper motors, drives, hinges and a working body. A spindle, manipulator and extruder are used as a working body. The movement of the working body is set by the upper level of control, which may consist of a personal computer. Each motor is programmed separately via the local control system. The sensors located on the drives monitor their position and send data to the upper control level. The frame layout equipment uses stepper motors that convert electrical pulses of control signals into angular movements of the rotor (discrete mechanical movements) with its fixation in a given position. Each stepper motor of the installation has the ability to perform precise positioning and speed control. This is well suited for a system that operates at low acceleration and with a relatively constant load. All these systems do not allow detecting information uncertainties in the operation of the equipment. This can lead to accidents. It is advisable to improve the control system of a multi-drive frame layout by integrating an automated control system and functional diagnostics based on FPGA hardware using neural network technology [20–24].

50

P. Kravets et al.

3 Mathematical Model of a Laboratory Stand The movement of the platform with the working tool is caused by movement of separate knots along directing rods. The vertical movement along the axes S1 and S2 is provided by the operation of individual motors. The kinematic system is considered when the movement of the tool is carried out in the vertical plane. The kinematic scheme of a symmetrical frame construction with two guide rods is shown in Fig. 2. The geometric dimensions of the installation elements are set at the stage of design and manufacture. lc1 , l c2 = l – length of beam elements. P – the platform on which the working tool is installed. The distance between the supports is known d /2 = −x01 = x02 , as well as the initial provisions S 01 = 0 and S02 = 0 carriage k 1 i k 2 . The position of the work site, ie the coordinates of its center (xc , zc ) and the angle of deviation of the normal from the vertical axis φ can be determined from the analytical description of the connection of the coordinates of the structure, which is based on the use of the dependences presented below [4, 5].

Fig. 2. Kinematic scheme of frame installation

1. Coordinates (xki , zki ) carriage locations ki can be determined from the ratios: xki = Si · cos α + x0i , zki = Si · sin α + z0i , i = 1, 2

(1)

2. Coordinates (xsi , zsi ) hinge locations shi can be found from the equations of their possible movement in a circle: (xsi − xki )2 + (zsi − zki )2 = lci2 , i = 1, 2 3. Equation of rigid connection of hinges due to the presence of a work platform p: (xs1 − xs2 )2 + (zs1 − zs2 )2 = lp2

(2)

Neural Network Model for Laboratory Stand Control System

51

4. To form a complete system of equations relative to the unknown, you can use the relationship between the angles formed by the sides of the quadrilateral. The maximum possibilities of moving the working body can be estimated on the basis of the solution of the direct problem of kinematics. To find the ranges of change of operating parameters of individual drives, limiting speeds and accelerations, it is necessary to solve the inverse problem of kinematics. A stepper motor is an electric motor, the main feature of which is that its shaft rotates by performing steps moving a fixed number of degrees. This characteristic is obtained due to the internal structure of the engine. It allows you to know the exact angular position of the shaft by counting the steps performed without the need to use a sensor. This feature makes it suitable for a wide range of applications. By switching on one or more phases of the stator, under the action of current flowing in the coil, a magnetic field is generated, and the rotor is aligned with this field. By feeding the different phases in series, the rotor can be rotated by a certain amount to achieve the desired end position. This is the basic principle of operation of a stepper motor. A two-phase hybrid stepper motor is a low-speed hydromagnetic synchronous motor. Due to its stable operation, it can achieve high-precision closed-loop positioning control and input pulse sequences to achieve digital control. Therefore, it is widely used in engineering fields such as robot motion control, antenna scanning and aircraft location control. In high-precision control systems, two-phase hybrid stepper motors are often used as actuators. For example, when establishing satellite-to-satellite communications, custom satellite relay antennas must track targets in real time, and two-phase hybrid stepper motors are often used as antenna drives. The drive of the tracking control system causes the antenna load to indicate the target of the tracking. The two-phase hybrid stepped motor system has a high degree of nonlinearity, which does not contribute to the analysis and design of the engineering control system, so it simplifies the task of obtaining an approximate model of the system and its transfer function. The block diagram of the servo control system with a closed cycle of the two-phase hybrid stepper motor is shown in Fig. 3 [25]. The inner circuit of a closed circuit is a current circuit. The purpose of this circuit is to implement tracking of the winding current of the hybrid stepper motor to a given current, so that the hybrid stepper motor can smoothly output torque under the microstep drive. The outer loop is a position loop, the goal is to track the output load at a given position.

Fig. 3. Block diagram of the closed-loop control system of a hybrid stepper motor

The mathematical model of a two-phase hybrid stepper motor is the basis for the analysis and study of the control system of a two-phase hybrid stepper motor. Because a hybrid stepper motor is a type of highly nonlinear electromechanical device, there are many difficulties in accurately describing and accurately defining nonlinear parameters.

52

P. Kravets et al.

The flux relationship created by the approximate flux of the permanent magnet in the phase winding varies with the position of the rotor according to the sinusoidal law. The influence of the stator current on it is not taken into account, the influence of hysteresis and eddy currents is not taken into account, but only the average permeability of the air gap and the main component, ignoring the mutual inductance between two-phase windings. The transfer function of each stepper motor component is described in the block diagram of the transfer function of the system and is shown in Fig. 4.

Fig. 4. Block diagram of the transfer function of the stepper motor system

4 Block Diagram of a Neural Network Controller To solve the problem of designing a neural network controller, a block diagram based on FPGA is developed, the prototypes of this controller are the tools developed in [15, 16, 22, 28, 29]. In Fig. 5 present a generalized block diagram of a neurocontroller, where as a computing core is used a neurocomputer based on FPGA, which implements an artificial neural network with an algorithm for its training.

Fig. 5. Block diagram of a neurocontroller for controlling a frame installation with mechanisms of parallel structure

Neural Network Model for Laboratory Stand Control System

53

The neurocontroller for controlling a frame installation with mechanisms of parallel structure includes a computing core, means for input of analog and digital information, means for output of analog and digital information, means of data exchange with other computing devices, elements of visualization and operational control. The neurocontroller includes: FPGA – programmable logic integrated circuit; discrete signal input module; discrete signal output module; communication module with control interface; communication modules with position sensors 1 and 2; communication modules with stepper motors 1 and 2; generator. FPGA is a central element of this structure, which implements neural network control algorithms, including the network learning algorithm in the “online” mode. The computing core represented by the FPGA is assigned the role of a “fast” fragment calculator of the general algorithm of the controller associated with neurocomputations; complement the FPGA program on an equal footing, or be a purely technical element of the neural network controller that provides FPGA modes. However, in all cases, the FPGA is designed to implement the neural network elements of the control system, so in the future the main attention is paid to the computer on the FPGA [15, 16]. To model the neural network in MATLAB, system matrices in the form of state spaces are defined. To do this, the transfer function is converted G(s) =

16.47s3 + 1648s2 + 69.81s + 1 0.00005845s5 + 0.6525s4 + 80.94s3 + 1648s2 + 69.81s + 1

to the control object and the necessary coefficients are obtained using Workspace in MATLAB.

5 Construction of a Neural Network System To synthesize the structure of the neural network model of a frame installation with a parallel structure, it is necessary to determine the “external” and “internal” structure of the neural network. The “external” structure of the neural network model is completely determined by the number of inputs and outputs of the object. When choosing the “internal” structure of the neural network model, it is necessary to determine: – the number of hidden layers; – the number of neurons in each hidden layer; – type of activation function for each layer. To solve the problem of choosing a neural network of the minimum structure from the set of possible, [24] describes the technology and software package “MIMO-Plant” for research and evaluation of neural network models of multidimensional control objects for further implementation on FPGA. Their essence is to model all possible variants of “internal” structures within the “external” structure and determine the structure that provides the least standard error, or a neural network of minimal structure whose standard error satisfies the condition of the problem (Fig. 6). Random values in the range from 0 to 1 are fed to the input of the system. The block “RMS” is a requirements management system that calculates the total standard error,

54

P. Kravets et al.

Fig. 6. Block diagram of a neurocontroller for controlling a frame installation with mechanisms of parallel structure

which is mathematically described by the formula:  z  2 i=1 (y − y) , ε= z where z is the number of error values (choose equal to the number of elements of the vector tout from Matlab workspace); y – output model of the object or system; y’is the output of the neural network model of the object or system. The “Display” displays the error values of each of the neural networks. All values of root mean square errors are given in Table 1. The table shows that the smallest error, equal to 0.001236, has a structure with 11 neural networks in the hidden layer, with 1 delayed input and 2 delayed outputs and is the best structure for the selected system. The block diagram of the proposed controller is shown in Fig. 7. The control system works on the basis of a neuro-controller, which generates control signals according to the developed neural network. The movement of the working body is set by the upper level of control, which may consist of a personal computer or controller. The control drivers generate control signals according to the software. The software

Neural Network Model for Laboratory Stand Control System

55

Table 1. The results of the study of neural network models

i_j n 5 6 7 8 9 10 11 12 13 14 15 16

0_1

0_2

0_3

1_1

1_2

1_3

2_1

2_2

2_3

0,251 0,041 0,043 0,039 0,045 0,043 0,053 0,057 0,04 0,075 0,044 0,043

0,247 0,082 0,079 0,088 0,079 0,083 0,074 0,073 0,082 0,087 0,104 0,085

0,251 0,089 0,095 0,094 0,098 0,086 0,091 0,1075 0,107 0,097 0,1075 0,096

0,013 0,245 0,2475 0,147 0,302 0,0051 0,272 0,105 0,015 0,06 0,12 0,125

0,002 0,0024 0,002 0,0025 0,0016 0,0019 0,0012 0,004 0,0036 0,0023 0,0029 0,0024

0,112 0,085 0,09 0,084 0,016 0,089 0,014 0,01 0,021 0,005 0,015 0,024

1,764 0,31 3,167 0,0015 1,053 0,0013 0,152 0,0014 0,6546 0,0027 0,085 0,483

0,259 0,29 1,031 1,728 1,39 2,208 2,211 2,21 0,944 2,203 2,21 1,593

3,96 0,485 0,486 0,484 0,852 1,579 0,486 0,45 0,487 0,662 0,485 0,574

allows you to enter the trajectory of the working body and generates control signals for the two-coordinate machine, which come to the control controllers for the formation and control of actuators - stepper motors. Each motor is programmed separately through the local control system (using separate drivers), and the sensors located on the drives monitor their position and send data to the upper control level [30–32].

Fig. 7. Block diagram of a neurocontroller for controlling a frame installation with mechanisms of parallel structure

The neural network control system of the frame installation with parallel manipulators developed in this work differs from the existing works [6–8, 12] by the introduction of the optimal in complexity and error neural network. This neural network will accurately display a frame installation with parallel manipulators. And the choice of the

56

P. Kravets et al.

minimum neural network architecture and its implementation on FPGA will allow you to control and adapt the controller in real time with high speed.

6 Conclusions The structural scheme of the control system of the frame installation with the mechanisms of the parallel structure based on the neural network controller is developed. A neural network model was built using the MATLAB “MIMO-Plant” package. The root mean square error of neural networks is calculated and the parameters for the most accurate operation of the system are selected. The structure with 11 neural networks in the hidden layer, with 1 delayed input and 2 delayed outputs has the smallest error 0.001236 and is the best structure for the selected system. This model is used as a neural network regulator. The structure of the neural network controller using the FPGA chip has been developed, which will allow managing and adapting to various types of uncertainties in the system in real time with high speed. The controller is responsible for adjusting the control signal by stepper motors depending on the position sensor indicators. In this paper, a control system with two parallel manipulators is modeled. In the following works, control systems with a larger number of manipulators, namely 4, 6, 8, 10 and 12, will be investigated. Neural network models will be selected for them. The prototype of the controller is realized and the experimental data on the workbench are received.

References 1. Wang, M., Ehmann, K.: Error model and accuracy analysis of a six-dof stewart platform. J. Manuf. Sci. Eng. 124(2), 286–295 (2002). https://doi.org/10.1115/1.1445148 2. Patel, Y., George, P.: Parallel manipulators applications—a survey. Mod. Mech. Eng. 2(3), 57–64 (2012). https://doi.org/10.4236/mme.2012.23008 3. Polyvoda, O., Rudakova, H., Kondratieva, I., Rozov, Y., Lebedenko, Y.: Digital acoustic signal processing methods for diagnosing electromechanical systems. In: Lytvynenko, V., Babichev, S., Wójcik, W., Vynokurova, O., Vyshemyrskaya, S., Radetskaya, S. (eds.) ISDMCI 2019. AISC, vol. 1020, pp. 97–109. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-264 74-1_7 4. Kuznetsov, Y.N., Dmitriev, D.A.: Realization of frame-configurations of machine tools with mechanisms parallel structure. J. Techn. Univ. Gabrovo 54, 27–31 (2017) 5. Kondratieva, I.U., Rudakova, H.V., Polyvoda, O.V., Lebedenko, Y.O., Polyvoda, V.V.: Using entropy estimation to detect moving objects. In: 2019 IEEE 5th International Conference Actual Problems of Unmanned Aerial Vehicles Developments (APUAVD), pp. 270–273 (2019). https://doi.org/10.1109/APUAVD47061.2019.8943839 6. Yang, C., Huang, Q., Jiang, H.: PD control with gravity compensation for hydraulic 6-DOF parallel manipulator. Mech. Mach. Theory 45(4), 666–677 (2010). https://doi.org/10.1016/j. mechmachtheory.2009.12.001 7. Rudakova, H., Polyvoda, O., Omelchuk, A.: using recurrent procedures in adaptive control system for identify the model parameters of the moving vessel on the cross slipway. Data 3(4), 1–60 (2018). https://doi.org/10.3390/data3040060

Neural Network Model for Laboratory Stand Control System

57

8. Le, T.D., Kang, H.-J., Suh, Y.-S., Ro, Y.-S.: An online self-gain tuning method using neural networks for nonlinear PD computed torque controller of a 2-dof parallel manipulator. Neurocomputing 116, 53–61 (2013). https://doi.org/10.1016/j.neucom.2012.01.047 9. Rudakova, H., Polvvoda, O., Omelchuk, A.: Using recurrent procedures to identify the parameters of the large-sized object moving process model in real time. In: 2018 IEEE Second International Conference on Data Stream Mining & Processing (DSMP), pp. 247–250 (2018). https://doi.org/10.1109/DSMP.2018.8478506 10. Korniyenko, B., Ladieva, L.: Method of static optimization of the process of granulation of mineral fertilizers in the fluidized bed. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2021. LNDECT, vol. 83, pp. 196–207. Springer, Cham (2021). https://doi.org/10. 1007/978-3-030-80472-5_17 11. Korniyenko, B., Ladieva, L.: Mathematical modeling dynamics of the process dehydration and granulation in the fluidized bed. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2020. AISC, vol. 1247, pp. 18–30. Springer, Cham (2021). https://doi.org/10. 1007/978-3-030-55506-1_2 12. Doan, Q.V., Le, T.D., Le, Q.D., Kang, H.-J.: A neural network–based synchronized computed torque controller for three degree-of-freedom planar parallel manipulators with uncertainties compensation. Int. J. Adv. Rob. Syst. 15(2), 1–13 (2018). https://doi.org/10.1177/172988141 8767307 13. Mugweni, D.T., Harb, H.: Neural networks-based process model and its integration with conventional drum level PID control in a steam boiler plant. Int. J. Eng. Manufac. 11(5), 1–13 (2021). https://doi.org/10.5815/ijem.2021.05.01 14. Rababah, B., Eskicioglu, R.: Distributed intelligence model for iot applications based on neural networks. Int. J. Comput. Netw. Inform. Secur. 13(3), 1–14 (2021). https://doi.org/10. 5815/ijcnis.2021.03.01 15. Shymkovych, V., Telenyk, S., Kravets, P.: Hardware implementation of radial-basis neural networks with Gaussian activation functions on FPGA. Neural Comput. Appl. 33(15), 9467– 9479 (2021). https://doi.org/10.1007/s00521-021-05706-3 16. Kravets, P., Shymkovych, V.: Hardware implementation neural network controller on FPGA for stability ball on the platform. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2019. AISC, vol. 938, pp. 247–256. Springer, Cham (2020). https://doi.org/10.1007/978-3030-16621-2_23 17. Artem, V., Volodymyr, S., Ivan, V., Vladyslav, V.: Research and development of a stereo encoder of a FM-transmitter based on FPGA. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2018. AISC, vol. 754, pp. 92–101. Springer, Cham (2019). https://doi.org/ 10.1007/978-3-319-91008-6_10 18. Opanasenko, V.N., Kryvyi, S.L.: Synthesis of neural-like networks on the basis of conversion of cyclic hamming codes. Cybern. Syst. Anal. 53(4), 627–635 (2017). https://doi.org/10.1007/ s10559-017-9965-z 19. Palagin, A., Opanasenko, V.: The implementation of extended arithmetics on FPGA-based structures. In: 2017 9th IEEE International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications (IDAACS), pp. 1014–1019 (2017). https://doi.org/10.1109/IDAACS.2017.8095239 20. Rayhan Ahmed, Md., Robin, T.I., Shafin, A.A.: Automatic Environmental Sound Recognition (AESR) using convolutional neural network. Int. J. Mod. Educ. Comput. Sci. 12(5), 41–54 (2020). https://doi.org/10.5815/ijmecs.2020.05.04 21. Abdisa, W.T., Harb, H.: A neural network based motor bearing fault diagnosis algorithm and its implementation on programmable logic controller. Int. J. Intell. Syst. Appl. 11(10), 1–14 (2019). https://doi.org/10.5815/ijisa.2019.10.01

58

P. Kravets et al.

22. Shymkovych, V., Samotyy, V., Telenyk, S., Kravets, P., Posvistak, T.: A real time control system for balancing a ball on a platform with FPGA parallel implementation. Techn. Trans. 5, 109–117 (2018). https://doi.org/10.4467/2353737XCT.18.077.8559 23. Awad, M., Zaid-Alkelani, M.: Prediction of water demand using artificial neural networks models and statistical model. Int. J. Intell. Syst. Appl. 11(9), 40–55 (2019). https://doi.org/ 10.5815/ijisa.2019.09.05 24. Kravets, P.I., Lukina, T.I., Shymkovych, V.M., Tkach, I.I.: Development and research the technology of evaluation neural network models MIMO-objects of control. Visnyk NTUU “KPI” Informatics operation and computer systems 57, 144–149 (2012) 25. Xu, W., Jianhong, Y.: Derivation of transmission function model of two-phase hybrid stepping motor. Space Electron. Technol. 3, 50–53 (2011) 26. Vorotyntsev, P., Gordienko, Y., Alienin, O., Rokovyi, O., Stirenko, S.: Satellite image segmentation using deep learning for deforestation detection. In: 2021 IEEE 3rd Ukraine Conference on Electrical and Computer Engineering (UKRCON), pp. 226–231 (2021). https://doi.org/ 10.1109/UKRCON53503.2021.9575783 27. Shibli, M.A., Marques, P.: Artificial intelligent nonlinear auto-regressive external input neural network modeling, design and control of a sea wave electro-mechanical power generating system. Int. J. Intell. Syst. Appl. 11(6), 1–12 (2019). https://doi.org/10.5815/ijisa.2019.06.01 28. Kravets, P., Nevolko, V., Shymkovych, V., Shymkovych, L.: Synthesis of high-speed neurofuzzy-controllers based on FPGA. In: 2020 IEEE 2nd International Conference on Advanced Trends in Information Theory (ATIT), 291–295 (2020). https://doi.org/10.1109/ATIT50783. 2020.9349299 29. Shymkovych, V., Niechkina, V.: The criterion for determining the buffering time of the measuring channel for smoothing the variable changes of the sensor signal. In: 2020 IEEE 7th International Conference on Energy Smart Systems (ESS), pp. 343–346 (2020). https://doi. org/10.1109/ESS50319.2020.9160084 30. Loutskii, H., et al.: Topology synthesis method based on excess de bruijn and dragonfly. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2021. LNDECT, vol. 83, pp. 315–325. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-80472-5_27 31. Loutskii, H., Volokyta, A., Rehida, P., Oleksandr Honcharenko, V., Thinh, D.: Method for synthesis scalable fault-tolerant multi-level topological organizations based on excess code. In: Zhengbing, H., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2020. AISC, vol. 1247, pp. 350–362. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-55506-1_32 32. Shakhovska, N., Montenegro, S., Kryvenchuk, Y., Zakharchuk, M.: The neurocontroller for satellite rotation. Int. J. Intell. Syst. Appl. 11(3), 1 (2019). https://doi.org/10.5815/ijisa.2019. 03.01

Web Application State Management Performance Optimization Methods Liubov Oleshchenko(B) and Pavlo Burchak National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Kyiv 03056, Ukraine [email protected]

Abstract. The use of various libraries often leads to a decrease in the speed of the web application and the complexity of the execution of the program code. The object of this research is the process of storing and managing the data of the client part of the web application, the subject of the research is the software methods of managing the local state of the data of the web application. The research objective is the data processing time reduction of web applications relative to existing software methods. The main idea of the proposed method is to use an atomic approach to the state of the web application data. Having an arbitrary entity, in the general state of the web application, a state fragment is created that is responsible only for this entity. Such a fragment is independent of other state fragments and can only work with the encapsulated entity. Using encapsulation, the configuration of an entity is passed to the React Context API as an object containing data and functions that modify it. The developed software method was compared with popular state management libraries Redux, MobXState-Tree and Recoil. Comparing each of the test scenarios in a percentage ratio, an average decrease in program execution time by 17% was obtained. The analysis of methods was performed using the SonarQube utility. To evaluate the results of the software methods, the Google Chrome browser utility DevTools was used. Based on decreased program execution time it can be stated that proposed optimized software method allows reduce the data processing time and optimize the state management of web applications. The research value is that such a software method can replace existing popular libraries with better proposed solution. Keywords: Performance · web application · optimization · local state of web application data · Redux · MobXState-Tree · Recoil · React

1 Introduction A feature of building web applications to support standard browser functions is that the functions must be performed independently of the client’s operating system. Instead of writing software code for Microsoft Windows, Mac OS X, GNU/Linux and other operating systems, a web application is created once for an arbitrarily chosen platform and deployed on it. Different implementations of HTML, CSS, DOM and other specifications in browsers can cause problems when developing web applications and their © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 59–74, 2023. https://doi.org/10.1007/978-3-031-36118-0_6

60

L. Oleshchenko and P. Burchak

further support. The ability of the user to configure several browser parameters at the same time (for example, font size, colors, disabling scripting support) can prevent the correct operation of the application. The shorter the waiting time, the more efficiently the web application works. A web application is a data store, which can store any type of data in any structure and provides tools to change this data. This is the main task is managing the local state of web application data. The speed of data storage and processing of web applications is also taken into account when ranking web pages, e-commerce web sites also affect the increase in profits along with the decrease in the loading time of the web application page. Therefore, the research of the optimization of the management of the local data state of web applications is an urgent task. A major challenge that arises when developing a client-side web application is managing the internal state of the web application’s data. Currently, there are plenty of solutions exists that help manage the state of application and resolve an issue with storing a lot of application data. Here are some libraries, which are the most popular ones and usually deal with the problem: Redux, MobXState-Tree and Recoil. However, the use of such libraries often leads to a decrease in the speed of the application and the complexity of the code. It is necessary to conduct a study of existing methods to develop software in which the analysis of selected software methods will be performed and it will be done conclusion about their advantages and disadvantages. The major research objective is to find a better solution for state management of modern web application. First of all, it is required to make a research and analysis of popular libraries and based on their performance, propose a new optimized software method of state management. Then proposed method needs to be compared with existing solution, and made a conclusion whether implemented optimization was beneficial. It is necessary to determine the best solutions and, taking into account the advantages of each of the method. 1.1 Analysis of Recent Research and Related Work Research [1] goal is to analyze the most popular state management library Redux and research an alternative to it. In this thesis React Hooks Method was introduced, which showed some significant performance improvements. Given results of the research was measured in a tool to measure performance – Google Chrome Dev Tools. In research [2] Redux library performance was measured and compared to React Context API method. Although, Redux library is easy to use, it shows significant performance issues when working with big data objects. Also, hooks approach was analyzed in detail, proving its scalability. Research [3–14] contains libraries and frameworks for the modern Web Application software development. There was selected some base metrics for libraires to be compared with each other – time of state management method execution for unit of work and suitability for developer to use. Tools that provides that most accurate results, not depending on side-effects was presents, such as Google Chrome Lighthouse and Sonarqube. The existing articles do not describe the methods of optimizing access time to web application data and their changes, so, research of the web application local state data management optimizing is an urgent task.

Web Application State Management Performance Optimization Methods

61

1.2 Analysis of Existing Software Solution Consider the main principles of the web application local state data and what tasks it performs. First of all, such a repository must be a part web application, i.e. cannot be a remote database, the application must be fast and reliable access to data. The web application is not intended for storage large amount of data, it should be only local data that used by a specific user in a specific session. After closing the application, the data is automatically deleted, if not saved on the server. In this case, the use of a database in the system architecture will slow down the application. The state of a web application must span the entire application. That is, each element, or component of the application, has access to the state and possibility to change it. The state in this case should be the same in one time slot for all web application components. It is important that system components be able to exchange state and its updates with each other without directly passing parameters to each other. State should not be passed to components that do not need it. That is, it would be inappropriate to pass state up the tree from the parent component to the lowest-level component, as this creates a lot of extra code. Components that use state are called subscribers, that is, those that subscribe to state updates. The third requirement for the state of a web application is its reactivity. This means that when the state or one of its fields is updated, it is assumed that the subscribing components of the state receive these updates immediately and automatically, without the need to request them. Vuex is a state management library for the Vue.js JavaScript framework. The Angular framework includes its own RxJS library using Observables, while Redux is a common state management library that can be used with any of these frameworks or other libraries, but is very often used with the React library. Redux library designed to manage the local data state of JavaScript programs. Most often used together with React or Angular for building user interfaces. Redux stores the data state of everything application in the tree of objects in the same repository. One state tree makes it easier debugging or testing the application, it allows to save the state of the application data in progress to speed up the development cycle. Store object connects events that record the fact that something happened and reducers that update the state according to these events together. The storage containing the state of the application (application state) provides access to the state using the getState() function; can issue status updates and handles unregistering listeners using subscribe (listener) function. The way to change the state is single out the action, the describing object what happened. This ensures that neither views nor network callbacks will never change state. Instead, they only express intent to do it. All changes are centralized and occur in a clear manner sequence. Since they are simple objects in action, they can be registered, serialized, stored and subsequently reproduced for debugging or testing. Advantages of this method are Typescript language support, using the Redux Toolkit Query extension, availability of additional Redux DevTools for development. The main disadvantages of the method are a large amount of repeated code and lack of solutions that allow third party interception effects. MobX-State-Tree (MST) is a container system state, built on the MobX functional reactive state library. MobX is a state management engine, the MobX-State-Tree provides it structure and general tools needed for specific web application. MST is useful for fast

62

L. Oleshchenko and P. Burchak

scaling code into programs. Compared to Redux, MST offers better performance and much less boilerplate code. MST combines the best features of immutability (transaction, traceability and composition) and approach to management state, based on changeability (openness, joint location, encapsulation). Advantages of this method are the use of several isolated states simplifies scaling, use of a tree structure, does not require writing a large amount of identical code. The main disadvantages of the method are complexity during software debugging, using their own types to describe data that can conflict with Typescript, the need to use the observer function in the necessary system components. Recoil is a state management library for React that allows create a graph of the data flow that flows from atoms (shared state) via selectors to React components. Atoms are units of state to which components can subscribe. Selectors convert this state synchronously or asynchronously. Atoms are units of state that are updated and updated subscriptable. When an atom is updated, each component subscribed to it is re-rendered with the new value. Atoms can be used instead of the state of a local React component. If an atom is used of several components, all of these components have the same state. A selector is a function that takes atoms as input. When these atoms are updated, the function the selector is recreated. Components can subscribe to selectors just like atoms and will then be re-rendered when selectors change. Selectors are used to calculate the received data on based on the condition. This avoids redundant state, as it is minimal the set of states is stored in atoms and everything else is effectively computed as function of this state. Selectors track, what components they need and what state they depend on. Advantages of this method are Typescript language support and availability of tools that support asynchrony. The main disadvantages of the method are compatibility only with the React framework and tendency to duplicate code during web application development. 1.3 Software Requirements The main requirements for the developed software are possibility building a web application to run it in a web browser; the presence of a graphical user interface with the ability to operate with the status data of the web application; the possibility of connecting selected libraries to the software; connecting instruments for measuring time program execution and the ability to connect tools for code quality assessment programs. Software for testing and evaluating the uptime of libraries should be compatible with libraries to manage the local state of web application data. For the development of the software client part, the JavaScript language was chosen for creating web page scripts, which provides an opportunity on the client side to interact with the user, control the browser, and asynchronously exchange data with the server, change the structure and appearance of the web page. To simplify the process of setting up and developing a web application, the React library was chosen due to its compatibility with the selected libraries and the presence of React Dev Tools for measuring the time of program methods and evaluating their work quality. To evaluate the speed of the methods, the Google Chrome web browser for measurement the web application’s operating time for a specified period was used and allows view a diagram of the application’s operating time distribution for specific actions, such as rendering, system operations and program code execution. Taking into account the described requirements, in the software ensuring

Web Application State Management Performance Optimization Methods

63

that each of the methods selected for research should be used in libraries and their work should be demonstrated on the basis of certain data. In a conclusion there are main Software Requirements to implement such software: Google Chrome 95+, React 17+, Node 16+. For the accuracy comparison, the methods should be used in the same conditions, work with the same data and according to their number, the same actions must also be performed on the data: adding data segments, editing, deleting, etc. The data, which will be operated by various methods, should be used for output to the user graphical interface. To estimate the methods operation time operating on the web application data local state, we use a large data volume. The data conform to an array structure with a large number of elements. In quality entities, to simplify the interface, the task entity Todo consists of a field name, a unique identifier and a status. In order to ensure the accuracy of testing methods, an additional, more complex ExtendedTodo data entity is also introduced (Table 1). Table 1. Fields of the ExtendedTodo entity Field name

Data type

id

UNIQUE IDENTIFIER

name

String

status

Boolean

tags

String []

author

{String, Number}

image

Buffer

description

String

createdAt

Date

ExtendedTodo entity fields include more complex data types. The software interface consists of identical segments, each of which is operated by different state management methods. The segment consists of the method name, the elements number in the data array, and the display of the list of tasks. For testing it is possible to manipulate data using the adding elements operations to the array, deleting elements, and editing. In the Todo task list, each task is available to change status and delete. The visual appearance of the software interface is shown in Fig. 1. By clicking on “+” and “×”, respectively, the user can change the task status. Such a function provides an opportunity to test the finding of an element in an array with a large number of elements and its editing. The “Remove” button on each of the elements allows test the removal of the element from the array. The “Add” button to add items to the list for selecting different numbers of items and manipulating data is used. Based on the software requirements, there is a need to connect additional tools to evaluate the quality and methods effectiveness – Google Chrome browser utility DevTools. Chrome DevTools helps the developer edit web pages “on the fly” and quickly diagnose problems. For this research the Performance panel of this utility and tools that allow recording and analyzing the program execution time during the program execution were used.

64

L. Oleshchenko and P. Burchak

Fig. 1. Graphical software interface

After the analysis is completed, the tool provides a report in the time chart form, which allows drawing conclusions about the program operation. All execution time indicators are given in milliseconds. Figure 2 shows the Performance panel of the Google Chrome DevTools utility, which displays the application runtime, the frames per second scale, and the operation distribution diagram. To evaluate the program performance both for the entire time interval and for the allocated time, a diagram of the distribution of operations by time is used. Scripting is JavaScript code execution time, Rendering is time for the browser to display the interface components, Painting is time for the browser to build graphic elements of the program, System are system operations, Idle is idle execution time, Total is total measurement time.

Fig. 2. Performance panel of the Chrome DevTools utility

The Scripting time counter shows how much time was spent executing the JavaScript code. This time will be different when using different methods, comparing it, we

Web Application State Management Performance Optimization Methods

65

can make a conclusion about the greater effectiveness of one or more methods when performing various types of tasks. SonarQube is a code quality testing platform for performing automated reviews with static code analysis for bugs. SonarQube offers review reports on duplicate code, coding standards, unit tests, test code coverage, comments, bugs and security recommendations. With the help of SonarQube, it is possible to analyze the program code using the state management method on a file-by-file basis. The developed software for comparing performance allows to measure the operating time and perform the following operations on the data: adding elements to the array, removing an element from the array, editing an element of the array. The combination of such operations on the data allows exploring the advantages and disadvantages of the methods in different situations. Simple operations on the array are performed in a very short time slot with a small number of elements. Judging by this, it will be appropriate to perform operations on arrays with a large number of elements, for example, thousands and more units. The local data state of the investigated web application before the start of the scripts will be equal to an empty array. In order to correctly test the speed of operation of methods according to the general recommendations for testing the speed of software, it is necessary to carry out work scenarios several times. To calculate the performance index, we will take the arithmetic average of three tests. This approach will avoid possible errors and side effects during the operation of the software. The following speed test scenarios for research are used: • • • • •

adding 1000, 5000, 10000, and 100000 tasks; deleting tasks from an array of 10,000 elements long; changing the status of several list tasks, with a length of 10,000; successive addition, deletion and changes of several elements; use of an extended task data structure for operations with the addition of various elements; • a combination of adding, removing and editing elements in any order with a large number of elements (more than 1000). Having analyzed the scenarios for evaluating the speed of execution of software methods, we can say about the readiness of testing using the developed software and Google DevTools tools. After that, a conclusion will be made about the feasibility of using each of the methods in different situations. The results of this research will be used in the development of a combined optimized method for managing the local state of web application data. 1.4 Objective The object of the research in this article is the reduction in the data processing time of web applications relative to existing software methods.

66

L. Oleshchenko and P. Burchak

2 Research Method 2.1 Evaluation Criteria for Software Method To ensure the accuracy of the estimation, the program code for the research was extended. This approach allows testing the complexity of the code when using a larger number of entities and data. High scalability can indicate that the method with an increase in the amount of software code and expansion of data entities will not lose in the speed of the code. The software methods of the Redux, MobXState-Tree and Recoil libraries were chosen for the performance study. The best result was shown by the Redux library method (Table 2). The MobX library method showed the longest time, as the number of elements to add increases, the difference between MobX and the other methods increases. Table 2. Results of speed analysis when adding elements to the array Number of elements

The time of the operation of adding elements, using the method, ms Redux

MobX-State-Tree

Recoil

1000 elements

239

336

274

5000 elements

1101

1483

1121

10000 elements

2101

3211

2294

To evaluate and compare the performance of state management libraries, different data and scenarios should be used in order to achieve the accuracy of the results and reach an unerring conclusion about the effectiveness of the method. The developed software for comparing the speed of operation allows to perform and measure the operation time of the following data operations: adding elements to the array, removing an element from the array, editing an element of the array. The combination of such operations on the data will allow to accurately investigate the advantages and disadvantages of the methods in different situations. It will be important to say that simple operations, such as those given above, on the array are performed in a very short period of time, with a small number of elements. Judging by this, it will be appropriate to perform operations, by default, on arrays with a large number of elements, for example a thousand and more elements. Let’s single out the following speed test scenarios for research: • • • • •

adding 1000, 5000, 10000, and 100000 tasks; deleting tasks from an array of 10,000 elements long; changing the status of several list tasks, with a length of 10,000; successive addition, deletion and changes of several elements; the use of an extended data structure of the task, for carrying out operations with the addition of a different number of elements;

Web Application State Management Performance Optimization Methods

67

• a combination of adding, deleting and editing elements in any order, with a large number of elements (more than 1000). To ensure reliability and accuracy of our results, it is important to mention conditions in which performance tests will be conducted. This includes both hardware and software setup. Hardware: MacBook Pro, Processor 2 GHz Quad-Core Intel Core i5, Intel Iris Plus Graphics 1536 MB, 16 GB of memory 3733 MHz LPDDR4X. Operation system − macOS Ventura 13.0.1. Software: Google Chrome 108, React v18.0.1, Node v16.13.1. When increasing the number of elements to 100,000, the result of the MobX library was obtained in almost 250 s, which is 3 times slower than analogues. The Redux and Recoil libraries show roughly the same results at ~80 s, which is still quite slow and needs improvement. Analyzing the previous results, we can conclude that on the example of the Redux library, the increase in execution time increases disproportionately with the increase in the number of elements. As can be seen from Fig. 3, the curve of the ratio of the number of added elements to the time for which they were added is similar to the graph of an exponential function. Analyzing the data obtained during the testing of the speed of adding elements, we can conclude that the Redux and Recoil libraries show better results when adding a large number of elements.

Fig. 3. Ratio of the number of added elements to time

Tables 3, 4 show the results of the speed analysis when editing and deleting elements from the array. After analyzing the results of the element editing operation, the methods of the MobX library show better results than when adding a large number of elements. It should also be noted that the Redux library shows much worse results in the operation of editing elements than its counterparts. During the edit operation, it was observed that when trying to pass a new value for the status field that is equal to the current value, the methods change the current value to the new value. But in this case, such editing does not make sense, since the value of the field has not changed and the state has not changed either. In this case, for optimization, it makes sense to add an additional check for the

68

L. Oleshchenko and P. Burchak Table 3. Results of speed analysis when removing elements from an array of length 10000

Number of elements

The time of the operation of adding elements using the method, ms Redux

1 element 10 elements

MobX-State-Tree

Recoil

743

686

723

5260

5245

5177

Table 4. Results of speed analysis during editing three elements from an array of different lengths Number of elements

The time of the operation of adding elements using the method, ms Redux

MobX-State-Tree

Recoil

1000 elements

457

239

289

10000 elements

2151

3211

2011

coincidence of the current state with the new one. This improvement will help prevent redundant operations and significantly improve the execution time of the edit operation. The combined analysis of operations for this research is a combination in any order of the operations of adding, deleting, and editing elements. The following sequence of actions was used: 1. 2. 3. 4. 5.

3 times the addition of 1000 elements; removal of two elements; 5 times adding 1000 elements; editing three arbitrary elements; deletion of one element.

When combining different operations and working with a different number of array elements, the Recoil library method shows the best results in terms of speed compared to its counterparts. Such a result is natural, this method in all previous examples showed effective work in comparison with analogues. Various data structures were used to test the operation of methods for managing the local data state of the web application, and 2 data models, Todo and ExtendedTodo were used. To test the difference, each of the mentioned operations was used in a certain order. For example, the following sequence of actions was used: 1. adding 10,000 elements; 2. removal of two elements; 3. editing two arbitrary elements. In this case, the difference in performance of methods with different data structures fluctuates within the margin of error and does not have a significant impact on the speed of the method. After analyzing the methods of each of the selected libraries for analysis

Web Application State Management Performance Optimization Methods

69

and conducting all the planned test scenarios, the performance results were obtained. Each of the methods has shown its effectiveness in different situations. An important conclusion is the exponential increase in the algorithm’s operating time when adding elements relative to their number. The results showed that the data structure used as an element of the array during the operation of the method is not fundamental. Analyzing the results of the application of combined operations, which included a different number of operations, a conclusion was made about the most efficient operation of the Recoil library method. Bearing in mind that this method performed well in all other performance test scenarios when measuring time, we can conclude that the Recoil library’s method for managing the local state of web application data is the most efficient compared to other libraries. When testing the addition of one hundred thousand elements to the array, each of the methods demonstrated a rather low performance index. 2.2 Proposed Software Method Fundamental the technology of the method, which will allow the developer to have the opportunity to create and configure a system for managing the local data state of a web application, is to use the React Context API technology framework. In the web application using React web application data are passed from top to bottom through parameters, but such usage can be cumbersome for certain data types, such as customizing the user interface theme. The Context API provides a way to share similar values between components without the need for explicit transmission parameter through each level of the tree. To ensure use method throughout the web application in the root file that contains itself calls to all other parts of the web application must be called React Context. The main idea of the method is to use an atomic approach to state. Having an arbitrary entity in the general state of the web application creates a state fragment that is only responsible for that entity. Such a fragment is independent of other state fragments and can only operate with an encapsulated essence. Using encapsulation, in React the Context API is passed the configuration of the entity in the form of a containing object data and functions that change them. Based on the configuration of general state, a fragment is created that will be responsible for such essence. To interact with such a fragment, the method interface provides the hook function. When called, such a function provides access to the data interface and the ability to change it. The program interface can be arbitrary, since at the creation of a state fragment directly indicates its configuration. Consider the features of the proposed software method. 1. Context API is used to provide operation in all files of the web application. 2. Atomic fragment of the state is configured. 3. Using the hook function it is possible to access the local state of the data and change the atomic fragment of the state. Once such a feature has been used in a web application component, that component automatically subscribes to state updates. This means that with any change to the state fragment to which this component is subscribed, the interface component itself will be recalculated and changed using the updated data.

70

L. Oleshchenko and P. Burchak

For a better understanding of how the method works, let’s consider its application for a state fragment example.

interface create = (storeConfig) => useStoreHook;

Syntax (storeConfig) => useStoreHook means that function takes a parameter of type storeConfig and returns a value useStoreHook. Consider the interface of state creation functions:

interface create = (storeConfig) => useStoreHook; interface storeConfig = ( set: (any) => void, get: (void) => any) => (stateConfig:Record ); interface useStoreHook = (selector: (any) => object) => (stateConfig: Record );

Type any means any type, Record is object with key-value pairs, string and any type, respectively. The set and get methods can be called when configuration for further use of field values and for counting certain values in functions. The create method does two things functions: assignment of the initial value of the state and a description of the functions that perform operations on state data. The hook function generated by the useStore method is used as a way to get state data directly for use in the user interface. The main advantage of this approach is the possibility of using the technology of closing and transfer state handler functions. Such approach was inherited from the Recoil library. It is intuitive similar to the standard functions-hooks of the React framework, for which it was this method was developed. After analyzing and comparing popular libraries for managing the state of a web application, the following tasks for optimization were determined: prevention of unnecessary operations, if when editing the state, the current state is equal to the new one; improving the speed of work when adding a large number of elements; possibility of repeated use of similar parts of the code. For better understanding of method flow algorithm diagram is provided (Fig. 4). During the optimized method development, the mentioned disadvantages of analogues were taken into account. The problem of redundant operations is often the reason for the low method efficiency, because after each state change, even if the field values themselves have not changed, all the components that have signed up for the state change will be automatically recalculated. An operation is considered redundant, when attempting to change the state, the initial state and the new state do not differ from each other. To prevent such a problem, before changing the state, we need to check whether the current state is equal to the new state. For this purpose, the possibility of specifying an additional parameter of the useStore hook function for comparing state objects has been added to the implementation of the optimized method.

Web Application State Management Performance Optimization Methods

Fig. 4. Poposed methodology algorithm diagram

71

72

L. Oleshchenko and P. Burchak

ShallowComparison method compares the old and new state before changing it, if the appropriate parameter was specified. This approach is an optimization because the comparison operation is less time-consuming than updating state, including updating state subscriber components. One of the main functional improvements has been added, namely the description of the state change functions during its creation. When creating a state object, the developer must immediately specify the functions that will be used for its operation. This approach allows describe these functions once and use them later in all subscriber components without redefining them. This approach significantly reduces the number of repetitive code lines.

3 Research Results Comparing the popular libraries methods per subject performance, we will conduct similar testing on the developed optimized method. Thanks to the proposed optimization and low by the degree of method dependence on third-party libraries, when adding elements to the array, we get a better result compared to others methods. When adding too many elements to an array the proposed method is only a few seconds ahead of the analogues, which is not a significant improvement with such a number of elements. Let’s compare the operation of library methods with the operation of clearing the state, that is, returning to the default state is an empty array. We get the result of the algorithm time execution 3844 ms with a simple data structure and 3979 ms with a complex one. The use of a complex data structure did not affect the speed of the proposed method. Having analyzed and tested all speed test scenarios and compared the performance indicators of the proposed method with analogues, in all scenarios the proposed method showed better results. It should be especially noted the optimization of the method, which was manifested during the adding operations and editing individual elements of the array. As could be seen from Fig. 5 changing number of elements in the array doesn’t affect method effectiveness in compare with analogues. In 1000 elements selection method is faster than the best of other researched methods by 12%, in 5000 elements selection by 16% and in 10000 elements by 18%. This ensures method stability and ability to scale the store. Comparison of existing popular libraries between each other and proposed method shows which of them performs better in a different condition. For example, MobX library shows fast execution results during operations with small amount of elements or heavy objects, but performs much slowly during a stress test, with a lot of elements. On the other hand, the most popular Redux library, only proving its popularity, showing stable results during multiple tests.

Web Application State Management Performance Optimization Methods

73

Fig. 5. Comparison of performing adding elements operations by the proposed method in comparison with analogues

4 Conclusion The article analyzes the specialized software libraries Redux, MobXState-Tree and Recoil for managing the local state of web application data, according to with the help of which the time of access to data and their changes is optimized, their advantages and disadvantages are considered. A modified method of managing the local state of web application data is proposed for creating applications in the React framework ecosystem, which allows reduce program execution time by 17% on average comparing each of the method test scenarios. As proposed methodology, based on the conducted analysis, performed better than popular existing solutions, in a sense of execution time, it can be said that the research objective was achieved. Analysis of software methods and evaluation of the quality of the scanned code was performed using the SonarQube utility. To analyze and evaluate the results of the considered methods, the Google Chrome browser utility DevTools was used. The research makes some conclusions about use cases of several popular solutions. For instance, lately developed Recoil library, in most situation shows performance even better than other libraries. Also, further development and improvement of optimized method, may become mature enough for commercial usage and outclass current solutions. Further areas of research are the software methods complexity analysis using the Cyclomatic Complexity and Cognitive Complexity metrics, the analysis of each file associated with the use of library tools, testing of the metrics data using the proposed software method. Also, further examination in this area also includes in deeper analysis of other popular solutions and test cases for them. In future on base of received results, if it’s effectiveness will be proved for most cases, the method could be published as library for public usage.

74

L. Oleshchenko and P. Burchak

References 1. Pronina, D., Kyrychenko, I.: Comparison of redux and react hooks methods in terms of performance. In: COLINS-2022: 6th International Conference on Computational Linguistics and Intelligent Systems, May 12–13, 2022, pp. 3–7. Gliwice, Poland (2022) 2. Thanh Le, Comparison of State Management Solutions between Context API and Redux Hook in ReactJS, pp. 14–21 (2021) 3. Atterer, R., Schmidt, A.: Tracking the interaction of users with AJAX applications for usability testing. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 1347–1350 (2007) 4. Zhongsheng, Q.: Test case generation and optimization for user session-based web application testing. J. Comput. 5(11), 1655–1662 (2010) 5. Souders, S.: High-performance web sites. Commun. ACM 51(12), 36–41 (2008) 6. Shivakumar, S., Suresh, P.V.: A survey and analysis of techniques and tools for web performance optimization. J. Inform. Organ. 8(2), 31–57 (2018) 7. Chi, X., Liu, B., Niu, Q., Wu, Q.: Web load balance and cache optimization design based nginx under high-concurrency environment. In: Third International Conference on Digital Manufacturing & Automation, pp. 1029–1032 (2012) 8. Oleshchenko, L., Burchak, P.: Analysis and optimization of methods for storing and processing data of web applications. Applied mathematics and computing. PMK - 2021. Kyiv, November 17–19, pp. 59–63 (2021) 9. Dychka, I., Legeza, V., Oleshchenko, L., Bohutskyi, D.: Method simultaneous using GAN and RNN for generating web page program code from input image. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2020. AISC, vol. 1247, pp. 338–349. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-55506-1_31 10. Wang, Y., Huarui, W., Huang, F.: High-performance concurrent Web application system analysis and research. Comput. Eng. Design Optim. 08, 2976–2981 (2014) 11. Yao, Y., Xia, J.: Analysis and research on the performance optimization of Web application system in high concurrency environment. In: 2016 IEEE Information Technology, Networking, Electronic and Automation Control Conference, pp. 321–326 (2016). https://doi.org/10. 1109/ITNEC.2016.7560374 12. Zagane, M.., Abdi, M..K..: Evaluating and comparing size, complexity and coupling metrics as web applications vulnerabilities predictors. Int. J. Inform. Technol. Comput. Sci. 11(7), 35–42 (2019). https://doi.org/10.5815/ijitcs.2019.07.05 13. Bibi, R., Jannisar, M., Inayet, M.: Quality implication for prognoses success in web applications. IJMECS 8(3), 37–44 (2016). https://doi.org/10.5815/ijmecs.2016.03.05 14. Khanna, M., Chauhan, N., Sharma, D., Toofani, A.: A novel approach for regression testing of web applications. Int. J. Intell. Syst. Appl. 10(2), 55–71 (2018). https://doi.org/10.5815/ ijisa.2018.02.06

Design and Implementation of Signal Acquisition System Based on FPGA and Gigabit Ethernet Haiyou Wang1(B) , Xiaozhe Yang2 , and Qianxian Bao3 1 Beihai Confidential Technology Service Center (Beihai Special Communication Technology

Service Center), Beihai 536000, China [email protected] 2 Faculty of Intelligent Manufacturing, Nanning University, Nanning 530299, China 3 Nanning Branch of China Mobile Guangxi Co., Ltd., Nanning 530022, China

Abstract. In order to meet the demand of high-speed data transmission between PC and FPGA, a signal acquisition system based on FPGA and Gigabit Ethernet is designed and implemented. FPGA is used as the main control platform, SDRAM is used as the external data storage device, Gigabit Ethernet technology is used as the data transmission protocol, and the FPGA is programmed in the Quartus II design software provided by Altera using the hardware description language Verilog HDL, Through socket programming under Linux system and the implementation of UDP_IP_ARP protocol stack on FPGA, and (SOPC Builder) tool, the logic circuits such as microprocessor Nios II, data buffer, program memory, Ethernet controller are integrated in FPGA, and the MAC inside FPGA and the PHY outside the physical layer are sent to PC through RGMII interface for subsequent demodulation, analysis and other work, A programmable system on chip (SOPC) is formed. This design realizes the wireless broadband signal acquisition function with less hardware resources, and has the characteristics of simple circuit and high transmission rate. Keywords: FPGA · Gigabit Ethernet · Programmable system-on-chip

1 Introduction With the development of electronic technology, the system equipment is developing towards miniaturization, integration and networking. FPGA has the characteristics of rich logic and abundant pins, and is widely used in embedded systems of high-speed data processing and communication. The biggest advantage of FPGA is its programmability, that is, users can program the FPGA according to their own needs to achieve user logic, and FPGA can be repeatedly programmed, modified and upgraded without the need to modify the circuit board, only through algorithmic modification and update of the program, which greatly reduces the design work of developers and effectively shortens the development cycle [1–3]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 75–84, 2023. https://doi.org/10.1007/978-3-031-36118-0_7

76

H. Wang et al.

Combining the advantages of FPGA and Ethernet, this paper designs an embedded gigabit Ethernet device based on FPGA. In this design, we use Altera’s high-end FPGA: EP3C10E144C8N chip to build and implement Nios II 32-bit embedded soft-core processor inside, with a working frequency of 185 MHz, and its rich RAM resources can provide high-speed data and code storage for the processor system. A wireless broadband acquisition system is designed around the Nios II processor and the IP core with MAC function, which is in line with the current technological development trend and has certain practical significance.

2 Hardware Design of the System In this paper, the MTV818 of RAON company is selected as the tuner for receiving signals, the EP3C10E144 of Altera company’s Cyclone III series is used as the main control chip, the 88E1111 of Marvell company is used as the physical layer chip of gigabit Ethernet, and the SDRAM chip of ISSI company’s IS42S16160G is used as the external data storage chip [4, 5]. The system inputs the UHF band (frequency range: 470–806 MHz) wireless broadband signal received by the antenna to the tuner MTV818, which performs low-noise filtering, I/Q mixing and low-pass filtering on the input signal, and then performs ADC analog-to-digital conversion and Viterbi demodulation, and then transmits the 8-bit digital signal to the FPGA for further digital signal processing. In FPGA, first build a SOPC (programmable system on chip) with Nios II processor as the core and configured with SDRAM controller, Gigabit Ethernet MAC and other peripherals. SOPC will first receive data and store it in SDRAM, and then read the data stored in SDRAM through FIFO and transmit it to Gigabit Ethernet MAC after packaging the data packet of UDP protocol, and then transmit it to the physical layer chip through RGMII interface. The physical layer chip outputs UDP protocol data to the PC through the Ethernet interface RJ45 interface, and the received data can be further analyzed and processed on the PC. In this paper, a design scheme of wireless broadband signal acquisition system based on Ethernet is proposed by comprehensively using FPGA, SDRAM, Gigabit Ethernet transmission and other technologies. The overall block diagram of the system is shown in Fig. 1.

Fig. 1. Overall block diagram of the system

Design and Implementation of Signal Acquisition System

77

2.1 Tuner MTV818 Design MTV818 is a high-integration system-on-chip receiving system applied to T-DMB, DAB, FM and ISDB-T standards. Tuner MTV818 is a portable mobile TV and baseband single chip solution; Internal integration of functional modules such as DCDC converter and AD analog-to-digital conversion module greatly simplifies peripheral devices [6]. The interface between MTV818 and external communication supports data stream interfaces of SPI, HPI, I2C, EBI2 and MPEG2.The connection diagram of MTV818 and FPGA through SPI interface is shown in Fig. 2.

Fig. 2. MTV818 and FPGA connection diagram

2.2 FPGA Design The system selects the Cyclone III series chip EP3C10E144C8N of Altera Company as the main control platform of the system, which mainly realizes data processing and control of various modules.EP3C10E144C8N support high-speed external memory interface, including DDR, DDR2, DDR3, SDRAM and SRAM.Build an embedded soft-core processor Nios II in FPGA with a three-speed Ethernet MAC core, SDRAM controller, EPCS Flash controller, JTAG debugging interface and self-designed UDP_IP_SOPC of peripheral devices such as ARP protocol module [7–9]. FPGA download circuit and configuration mode are divided into active configuration mode, passive configuration mode and JTAG configuration mode, which are three commonly used configuration modes for FPGA devices. The design selects AS mode and JTAG mode. JTAG and AS pin circuits on FPGA chip are shown in Fig. 3.

Fig. 3. JTAG and AS pin circuits on FPGA chip

78

H. Wang et al.

2.3 SDRAM Module Due to the limited internal storage resources of FPGA, real-time signal processing and transmission are required. The total RAM capacity of the FPGA device selected in this design is only 424 kbits, and the internal storage capacity is small, which cannot meet this requirement. Therefore, the system’s requirements for data cache can be met by adding memory devices. In this design, large capacity data needs to be cached, so the system can meet the requirements of data caching by adding memory devices. Commonly used data memory devices mainly include solid state memory (FLASH), static random access memory (SRAM), synchronous dynamic random access memory (SDRAM) and other devices [10, 11]. However, considering that the price of solid state memory and static random access memory with the same capacity is much higher than that of synchronous dynamic random access memory, this paper selects synchronous dynamic random access memory chip as the external data storage device. At present, there are many SDRAM manufacturers. After comprehensive comparison in various aspects, this paper selects the more common SDRAM chip of ISSI company, the model is IS42S16160G, the total capacity is 256 Mbit, and the working voltage is 3.3 V. The connection diagram between chip and FPGA pin is shown in Fig. 4.

Fig. 4. Design circuit diagram of SDRAM

2.4 Physical Layer Module The PHY chip in this design is the Gigabit Ethernet physical layer chip 88E1111 of Marvell. 88E1111 supports 10/100/1000 Mbps transmission rates. The communication rate of two network interfaces can be automatically negotiated under IEEE 802.3u standard. It supports GMII, RGMII, SGMII, TBI, RTBI and other interfaces. The connection diagram of MAC and PHY is shown in Fig. 5.

Design and Implementation of Signal Acquisition System

79

Fig. 5. Design circuit diagram of PHY chip 88E1111

3 Software Design of the System 3.1 Implementation of UDP_IP_ARP Protocol Stack This paper builds a SOPC system based on Nios II, in which the processor Nios II, Ethernet MAC controller, SDRAM controller, EPCS Flash controller, JTAG debugging interface and other peripherals can be designed using the IP core provided by Altera, but UDP_ IP_ ARP protocol stack must be developed and designed by yourself. The software development flow chart of Nios II is shown in Fig. 6.

Fig. 6. Software development flow chart

This paper uses FPGA hardware to implement UDP_IP_ARP protocol can improve the data transmission rate of the system. The hierarchical model of TCP/IP network is divided into application layer, transport layer, network layer and link layer from top to bottom. Among them, the transport layer and network layer are the focus of this paper. The transport layer is responsible for packaging the data sent from the application layer into packets suitable for sending, and then transmitting them to the network layer for processing. The transport layer mainly includes TCP protocol and UDP protocol [12, 13]. UDP protocol does not have the handshake, confirmation and other mechanisms of

80

H. Wang et al.

TCP protocol. It is a connectionless protocol, which saves the consumption of resources and has the advantages of fast processing speed. Therefore, UDP protocol is selected for the transport layer in this paper. UDP_ IP_ ARP module includes UDP, IP and ARP modules, which can be divided into UDP sending and receiving module, IP sending and receiving module, ARP sending, receiving and caching module according to the independence of each module protocol and the direction of data sending and receiving. The block diagram of the UDP_IP_ARP module is shown in Fig. 7.

Fig. 7. Block diagram of UDP_IP_ARP

3.2 Main Functions of Socket Programming Socket communication can be used for communication between different processes on the same device or between different devices. The principle of socket communication is similar to the principle of “everything is a file” of Unix/Linux. You can use the mode of “open – > read/write write/read – > close” to operate [14]. Therefore, you can regard the socket as a special file and open, read/write and close this special file through some socket functions. The flow chart of socket programming design based on Linux is shown in Fig. 8.

Fig. 8. Socket programming design flow chart

Design and Implementation of Signal Acquisition System

81

4 Simulation Case and Data Analysis 4.1 Implementation and Simulation of UDP_IP_ARP Protocol Stack In the simulation file, we first design the protocol stack to receive an ARP query request, and then the module responds. Then input the length, destination IP address, destination port number, source port number, user data and other information, and observe the MAC frame output of the module. After that, input a MAC frame that is not ARP request to the module, and observe whether the extraction and output of each part of the frame are correct [15, 16]. The above three types of input and output constitute the entire test file. In order to comprehensively test the performance of the module, we have carried out a number of three types of input tests. It is observed that the output of the module meets the expectation. This shows that the module designed in this paper can receive ARP query requests from other devices and give correct responses. At the same time, it can add the data from the application layer first and generate MAC frames to send to the MAC layer. It can also unpack the data from the MAC layer by layer, and finally send it to the application layer [17]. Simulation is such an effective tool that can play an important role in system design [18, 19]. UDP_IP_ARP protocol stack simulation test results are shown in Fig. 9.

Fig. 9. UDP_IP_ARP protocol stack simulation test

4.2 Overall Test of the System The DE1-SoC development board is used as the lower computer. The DE1-SoC development board and the upper computer form a Gigabit Ethernet communication platform

82

H. Wang et al.

through the Gigabit switch. As a server, the PC is used to receive the data transmitted from the development board. As a client, the development board is used to generate data and send data. The communication process of the whole communication system is as follows: first, a 16bit DDS signal is generated on the development board as the source signal of the client, and then a data connection request is sent to the upper computer through socket programming. When the server monitors the connection request, it will make a corresponding response message and send a reply agreeing to the connection request of the client, after receiving the reply from the server, the client can start sending data to the server. Finally, we can observe the transmission rate of the entire gigabit network in real time through third-party software tools. The overall system test schematic diagram is shown in Fig. 10.

Fig. 10. Schematic diagram of overall system test

After connecting the development board, gigabit switch and PC, set the IP address of the development board. In order to achieve the maximum transmission rate after connecting the development board and the PC through the gigabit switch, the IP address of the development board and the PC should be in the same network segment, so the IP address of the development board can be set to 192.168.80.200. Start the control terminal of the development board on the PC, execute the command ifconfig eth0 192.168.80.200, and set the IP address of the development board to 192.168.80.200. To ensure that the IP address of the development board is set successfully, you can check the IP address of the development board by executing the ifconfig command. If the network address is 192.168.80.200 in the execution result, it indicates that the IP address of the development board has been set correctly. Then open the DOS terminal on the PC, execute the command ping 192.168.80.169, and check whether the development board has successfully pinged with the PC through the gigabit switch. After the ping test, the number of data packets sent and received in the network test can be seen from the test results. When the number of data packets sent and received is the same, it indicates that the data transmission has not been lost, indicating that the network of the gigabit network communication system composed of the PC and the development board through the gigabit switch has been successfully connected. The transmission rate of the overall system test is shown in Table 1. After testing, the actual transmission rate of the built system is 630 Mbit/s on average, which is less than the theoretical value of gigabit network, but it can meet the actual application transmission requirements. Table 1 shows the transmission rate of the overall system test.

Design and Implementation of Signal Acquisition System

83

Table 1. Transmission rate of overall system test Sequential

Rate (Mbits / sec)

Time delay (ms)

Datagrams (bit)

1

615

0.007

53419

2

615

0.007

53389

3

617

0.015

53539

4

621

0.006

53949

5

623

0.013

54042

6

624

0.006

54129

7

623

0.013

54117

8

626

0.010

54308

9

625

0.013

54219

10

620

0.014

53829

11

622

0.014

53979

5 Conclusion In this design, firstly, a SOPC (programmable system on chip) with Nios II processor as the core and including peripheral devices such as Gigabit Ethernet MAC is built in FPGA, and the received U-band wireless broadband signal is sent to the tuner chip MTV818. MTV818 performs low-noise filtering on the wireless broadband signal, followed by I/Q mixing and low-pass filtering, and then performs ADC analog-to-digital conversion Viterbi demodulation operation. Finally, the signal is sent to the PC through the RGMII interface between the MAC inside the FPGA and the PHY outside the physical layer for subsequent demodulation and analysis. The design realizes wireless broadband signal acquisition with less hardware resources, and has the characteristics of miniaturization and networking of equipment. Later, it can be upgraded and maintained according to the actual needs, which has a broad application prospect. Acknowledgment. This project is supported by the project of scientific research and basic ability improvement for young and middle-aged teachers of Guangxi universities (2022KY1781).

References 1. Colak, A., Manabe, T., Shibata, Y., et al.: Peak detection implementation for real-time signal analysis based on FPGA. Circ. Syst. 9(10), 148–167 (2018). (in Japan) 2. Dulnik, G., Grzelka, A., Luczak, A.: A Gigabit Ethernet interface with an embedded lossless data encoder on FPGA. Pomiary Automatyka Kontrola 61(7), 364–366 (2015) 3. Windisch, D., Knodel, O., Juckeland, G., et al.: FPGA-based real-time data acquisition for ultrafast x-ray computed tomography. IEEE Trans. Nucl. Sci. 68(12), 2779–2786 (2021) 4. Runqiu, X., Yang, L.: Design and implementation of PCI bus infrared image acquisition system based on FPGA. LCD Disp. 33(09), 772–777 (2018). (in Chinese)

84

H. Wang et al.

5. Jianyu, Z., Jiesi, N., Fuchun, G., Yuchen, J.: Design and analysis of multi-channel signal acquisition circuit based on FPGA. J. Phys: Conf. Ser. 2206(1), 012034 (2022). (in Chinese) 6. Yihe, H., Dongdong, Z., Lei, D.: Design of Gigabit Ethernet data transmission system based on FPGA+MAC+PHY. Sci. Technol. Proj. 19, 275–279 (2014). (in Chinese) 7. Qi, W., Jiangtao, Y., Xihong, M.: Design and implementation of high-speed signal acquisition circuit based on FPGA. Res. Explor. Lab. 34(04), 124–128 (2015). (in Chinese) 8. Mingqing, D., Xiaohu, D., Bao, D.: Design and implementation of a multi-channel data acquisition system. Aeronaut. Comput. Tech. 48(02), 109–112 (2018). (in Chinese) 9. Hao, Z., Haoquan, W., Shilei, R.: Design of portable signal acquisition system based on FPGA and NAND Flash. Appl. Elec. Tech. 44(09), 82–86 (2018). (in Chinese) 10. Wei, W., Qiuyun, Z., Hong, J., et al.: Design and implementation of multiplex acquisition and switching system based on FPGA and TCP/IP. Appl. Elec. Tech. 45(06), 125–129 (2019). (in Chinese) 11. Jinjing, Z., Tao, C., Dong, P., et al.: Design and implementation of 4-DOF manipulator control system based on FPGA. J. Mach. Des. 34(07), 62–66 (2017). (in Chinese) 12. Bin, L., Chunwu, L., Zhiping, H., et al.: Design and implementation of SFI-5 signal source based on FPGA. Opt. Commun. Technol. 38(01), 1–4 (2014). (in Chinese) 13. Wang Xuying, L., Yinghua, Z.L.: Design and implementation of high-speed real-time data acquisition system based on FPGA. J. China Univ. Posts Telecommun. 13(4), 61–66 (2006). (in Chinese) 14. LiLinwei, Jiaoxu, M., Lei, W., et al.: Design and implementation of a data acquisition system based on FPGA and Ethernet. Elec. Des. Eng. 22(07), 1–4 (2014). (in Chinese) 15. Jin, L.: Design and Implementation of FPGA High Speed Data Acquisition and Transmission System Based on Gigabit Ethernet. Beijing University of Posts and Telecommunications, Beijing (2019). (in Chinese) 16. Yuhang, T.: Design and Implementation of Multifunctional Data Acquisition System Based on DSP and Gigabit Ethernet Technology. Huazhong University of Science and Technology, Wuhan (2020). (in Chinese) 17. Xiaoyu, L., Xinyang, H., Peiyan, S.: Design and implementation of rotational speed signal acquisition system based on FPGA. China New Commun. 24(07), 48–50 (2022). (in Chinese) 18. Kalpachka, G.: Computer modeling and simulations of logic circuits. Int. J. Mod. Educ. Comput. Sci. (IJMECS) 8(12), 31–37 (2016) 19. Gjorgjevikj, F., Jakimoski, K.: System monitoring add on analysis in system load simulation. Int. J. Comput. Netw. Inf. Secur. (IJCNIS) 14(1), 40–51 (2022)

Design of a Dynamic Probability-Based Automatic Test Paper Generation System Fang Huang, Mingqi Wei, Lianju Su(B) , and Dichen Guan Global School of Software, Nanning University, Nanning 530022, Guangxi, China [email protected]

Abstract. Our work and life is closely related to computer technology, and computer applications are especially prevalent in the education sector. The traditional manual approach requires teachers to select questions from workbooks and books, compile and edit them, and finally layout the test papers. This process is tedious and time consuming, and it is difficult to ensure the level of knowledge coverage and the repetition rate of questions. This paper takes a dynamic probability-based test paper generation system as the background, follows the standard development model of software engineering, adopts an object-oriented approach to design the system, specifies the functional requirements, Design the system architecture and divide functional modules according to the principles of modularity, independence and abstraction, and realize the detailed design of the main functional modules such as user management, course management, and examination paper management builds a dynamic probability model to realize the automatic formation of papers, and makes the task of assigning questions more scientific. Keywords: Test papers · Automatic generation · Dynamic probability

1 Introduction As an important part of the assessment of the quality of teaching and learning, examinations have traditionally been highly valued by schools. It is a way for students to test their mastery of the knowledge they have learn [1]. For teachers, it is a way of giving feedback on the effectiveness of teaching and learning, and teachers use the quantitative analysis of examination results to make subsequent changes to the curriculum. Traditionally, question papers rely heavily on manual questioning, which can take a lot of time and effort for teachers to consult a large number of books and materials. Once teachers have found the right questions, it takes a lot of time and effort to edit and layout the papers and write the marking scheme. In practice, this can lead to repetition of questions, incomplete coverage of assessment points, and errors and omissions due to manual processing. To this end, the software industry at home and abroad has carried out practical research on the software development of the test paper generation system earlier [2]. Moreover, with the continuous advancement of software and hardware technology and the continuous maturity of development concepts, the software of the test paper generation system is also improving day by day, which is conducive to realizing © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 85–94, 2023. https://doi.org/10.1007/978-3-031-36118-0_8

86

F. Huang et al.

the “separation of examination and teaching”, improving the quality of teaching, and evaluating students’ learning more fairly, which is one of the possible ways for schools to realize the automation, digitalization and information management of examinations [3, 4]. The design of a dynamic probability-based automatic test paper generation system can therefore not only improve the efficiency of teachers’ work in proposing questions, but also ensure the objectivity and scientificity of test papers, which has important significance for improving the quality of teaching.

2 Analysis of Requirements 2.1 Analysis of Functional Requirements Requirements analysis is the first stage of software development, the main task of this stage is to accurately determine “what the target system needs to do in order to solve this problem”, this stage plays a crucial role in the success or failure of software development. The requirements analysis phase typically includes obtaining requirements, analyzing requirements, defining requirements, and validating requirements [5, 6]. Analysis and study of acquisition needs requires consideration of completeness, correctness, reasonableness, feasibility and sufficiency [7]. Designed and developed to meet the urgent needs of current school teachers for test papers, in view of the various problems existing in the management of test questions by many teachers, the test paper generation module is taken as the core research object, and a feasible system is built by modifying and improving the initial system.In order to solve the current problems in the area of question setting and paper assembly, we carried out an analysis of the system’s operational requirements. We started from the actual situation of the system users, took the business process in teachers’ question appointment work as the main reference basis, and carried out requirement analysis by means of offline visits to master the workflow of question paper appointment. The resulting dynamic probability-based test paper generation system provides the following functions: (1) Test bank creation and maintenance. Entry, modification, deletion, query and statistics of test questions, setting of question types, realization of querying test questions based on course and question type. (2) Course management. It includes the addition of new courses, modification of courserelated information and editing relevant knowledge points covered by the courses. (3) Course analysis. Analysis of each test paper by total number of questions, average difficulty, average mark, and by number of questions, difficulty, mark, and knowledge points, with graphical display of analysis results. (4) Paper generation. A variety of ways to assemble papers, either manually by filtering and setting the key examination points, or automatically by setting the course question type and amount of questions or creating a paper template and modifying or adding questions on the basis of the paper template to meet the difficulty requirements, the paper can be saved as a word document for preview and download after generation. (5) User management. The users are mainly administrators and teachers, and the different roles have different permissions. Administrators have all the administrative

Design of a Dynamic Probability-Based Automatic Test Paper Generation System

87

rights of the system, and teachers have only some of the rights of the system. Figure 1 shows the diagram of system use cases.

Fig. 1. Diagram of the system use cases

The system use case diagram describes the management correlation of the administrator and the teacher. The main use cases are analyzed and described below. (1) Course management is to manage the information and knowledge points contained in the course. The table of course management use cases is shown in Table 1. (2) Course analysis is to analyze the number of question banks, test papers, and questions included in the course. The table of use cases for course analysis is shown in Table 2. (3) Question bank management is to manage the basic information of the question bank. The use case table of the question bank management is shown in Table 3. (4) Examination paper management is to manage the examination paper composition and basic information of the questions contained in the system. The table of management use cases is shown in Table 4.

88

F. Huang et al. Table 1. Table of course management use cases

Use case name

Course management

Actor

System administrator / teacher

Use case description

The administrator manages the basic course information data of the system, including the functions of adding, editing, inquiring and deleting the course information; teachers can only view the course information and edit the knowledge information of the course

Prefix condition

System administrator / teacher to login to the system

Post-condition

The system completes the operation

Basic event flow

The system administrator / teacher logs in the system, enters the system home page, select the course management module for functional operation and save

Table 2. Course analysis use case table Use case name

Course analysis

Actor

System administrator / teacher

Use case description

To manage the course information data of the system, through the comprehensive analysis of each course according to the total amount, average difficulty and average score, but also according to the quantity, difficulty, score and knowledge points, and the analysis results will be displayed graphically

Prefix condition

System administrator / teacher to login to the system

Post-condition

The system completes the operation

Basic event flow

The system administrator / teacher logs in the system, enters the system home page, select the course analysis module for functional operation and save

Table 3. Table of question bank management use cases Use case name

Question bank management

Actor

System administrator / teacher

Use case description Manage the basic information and data of the question bank of the system, including the addition, editing, query and deletion of the question bank Prefix condition

System administrator / teacher to login to the system

Post-condition

The system completes the operation

Basic event flow

The system administrator / teacher logs in to the system, enters the main page of the system, and chooses the question bank management module for functional operation and saving

Design of a Dynamic Probability-Based Automatic Test Paper Generation System

89

Table 4. Table of course management use cases Use case name

Examination paper management

Actor

System administrator / teacher

Use case description To manage the paper generation of the system, you can manually screen and set up the key knowledge points, or automatically set up the number of course questions, and the paper can be saved as a word document and downloaded Prefix condition

System administrator / teacher to login to the system

Post-condition

The system completes the operation

Basic event flow

The system administrator / teacher logs in the system, enters the system home page, select the personal setting module for functional operation and save

3 System Design 3.1 System Architecture Design Using a layer-by-layer architecture approach to architecture design, the system takes the front-end UI, view layer, business layer, data layer, database, and operating environment into account in the system architecture, clearly divides the architecture technology and functions of each layer, realizes the interaction of modules at each layer, and builds a complete architecture system for automatic test paper generation, as shown in Fig. 2. 3.2 Global Design It is necessary to use a computer to centralize the management of examination papers and to be able to complete the work of question entry, examination paper enquiry, examination paper generation and examination paper management. The system can improve the low efficiency of traditional manual paper production, use the computer to build a test bank, achieve automatic computer selection of questions to form papers, and give full play to the advantages of computers in information processing. A test paper generation system with a high level of automation can realize the systematization, standardization and automation of test papers [9]. The design of the system is based on a mature software design and development model. A complete set of business functional requirements is analyzed and refined, and specific functional modules and physical realization schemes are planned on the basis of the “high cohesion and low coupling” system design principle. The functional module design of the system is shown in Fig. 3. 3.3 Database Design As a key aspect of database design, a good database will enable the system to run efficiently and safely. Database design will have a direct impact on the efficiency of the system and the effectiveness of the implementation.

90

F. Huang et al.

Fig. 2. Diagram of the system architecture

The system database is designed using the New Orleans method, which organizes, stores and manages data according to a data structure, abstracting the user requirements obtained during the requirements analysis phase into a conceptual model, which is represented by an E-R diagram. The E-R diagram, as a conceptual model, is the basis for both database design by designers and information modelling [10, 11]. In a dynamic probability-based test paper automatic generation system, the entity attributes are more complex. The E-R diagram of the system in general is shown in Fig. 4. (1) A single course can contain multiple knowledge points, and a single knowledge point can exist in multiple courses, so there is a many-to-many relationship between the course and the knowledge points. (2) A course can have multiple papers, but one paper belongs to only one course, so the course and the paper have a one-to-many relationship. (3) One teacher can teach multiple courses, and one course can also be taught by multiple teachers. There is a many-to-many relationship between a teacher and a course. (4) A teacher can create and manage multiple question banks, but a question bank can only be managed by the creator, so there is a one-to-many relationship between the teacher and the question bank.

Design of a Dynamic Probability-Based Automatic Test Paper Generation System

Fig. 3. Diagram of the system architecture

Fig. 4. E-R diagram of the system in general

91

92

F. Huang et al.

4 Critical Techniques for Test Paper Extraction 4.1 Multi-layer Filtering Model Test questions are extracted by filtering all test questions in layers until a paper satisfying the requirements is filtered. A multi-layer filtering model is used to filter all test questions in layers until a test paper satisfying the requirements is generated. The system uses a multi-layer filtering model based on which test questions are extracted, as described below. (1) First level of filtering. Course filtering, which filters out the list of questions for the examination course from the list of all questions; (2) Second level of filtering. Question type filtering, which filters out all questions belonging to different question types from all questions in the current course; (3) Third level of filtering. Grouping filtering, which randomly groups all the questions for that question type and can ensure that different questions are selected in the same conditions; (4) Fourth level of filtering. Question quantity filtering, which filters the number of questions from the current question type. 4.2 Dynamic Probability Models The multi-layer filtering model involves more than one extraction process. The issue of how to extract individuals with relatively high probability from a set containing individuals with different probabilities of being extracted is a problem that needs to be studied, and this is the main focus of the study of dynamic probability models. As a result, the question of how to dynamically adjust the probability of a test being drawn different times from the set of results to be drawn becomes the core of the study of dynamic probability models in extraction algorithms [12, 13]. Based on the timeseries characteristics of the observed data, dynamic probability models are able to use probability distributions to model time-series data, tap the statistical correlation of serial data and use it for predictive analysis. Unlike deterministic dynamic neural networks, such as recurrent neural networks RNN, dynamic probabilistic models can explore the potential uncertainty of the data by modelling the observed data, and can also describe the mutually non-deterministic relationship between the sequential observed data, and the random time-series hidden variables through probability distributions. In solving for the parameters, the dynamic probability model takes into account the uncertainty of the parameters and can obtain the parameter distribution rather than a point in the parameter space [14, 15]. The dynamic filtering model of the system relates to the extraction process several times, and the test paper generation module is the core research object; the questions are queried by course, question type and difficulty factor; the user can randomly draw questions or set the questions manually and extract them to compose the test paper; the total number of questions in the test paper can be adjusted at will, and the questions can even be added or deleted on the basis of the already generated test paper, so that the test paper can be previewed and exported after achieving the best results [16, 17].

Design of a Dynamic Probability-Based Automatic Test Paper Generation System

93

Suppose the question pool of a course is drawn as a set {Q1, Q2,… Qm}, the corresponding decimation probability is denoted as {P1, P2,… Pn}. When drawing for the first time, because the probability of each topic being selected is the same, the system can automatically generate a random number for topic extraction, and after each draw of the topic, the number of draws is counted through the accumulation counter, and the scope of topic extraction is expanded according to the multiple of the number of draws, and the relationship between the number of draws and the expansion multiple is shown in Table 5. Table 5. The relationship between the number of draws and the expansion multiplier Number of draws

0

1

2

…

m–1

Enlarge the multiplier

nm

nm–1

nm–2

…

n

5 Conclusion With the wide application of computer technology, digital campus construction has become the current development trend of education construction, and colleges and universities have begun to use computer technology to assist teaching. For schools, computer operation has become quite widespread. The traditional manual roll-out method has the problems of manpower, time, cumbersome process and low efficiency. How to use computer technology to reduce the light labor of the test taker and realize the scientific and standardized management of the test paper is particularly important. This paper introduces the design of an automatic test paper generation system based on dynamic probability, and takes the actual operation process of question setting as the main line throughout the whole system design process. The system’s functional requirements are outlined by first analyzing the workflow of question assignment, then dividing the functional modules according to the principle of “high cohesion and low coupling”, and designing the system database. In view of the large number of question papers, system used Redis as the cache database to improve the speed of data retrieval, and analyzed the key techniques for extracting question papers. The implementation of the system design can, to a certain extent, reduce the workload of teachers and contribute to the improvement of teaching quality. Acknowledgment. This project is supported by: (1) Nanning University’s first-class undergraduate major cultivation project- software engineering (2020YLZYPY06); (2) The educational training program for 2020’s teaching achievement award in Nanning University-The exploration and practice for application talents training mode based on the combination of curriculum system and practice system (2020JXCGPY09) and (3) The third batch of pilot project for professional certification in Nanning University -Educational certificate of software engineering (ZYRZ06).

94

F. Huang et al.

References 1. Xingrong, L.I.U.: Design and implementation of automatic generation system of test paper based on dynamic probability. University of Electronic Science and Technology of China, Chengdu (2014). (in Chinese) 2. Changli, W.: Design and implementation of the exam paper generation management system. Des. Dev. 3, 120–121 (2011). (in Chinese) 3. Mingqiu, Z., Hongyan, L.: Design and implementation of NET’s online paper-generating system based on ASP. China’s Manag. Informatization 18(22), 149 (2015) (in Chinese) 4. Haoqiang, T.: C++ Programming, (2nd edition). Tsinghua University Press, Beijing, pp. 16–20 (2011) (in Chinese) 5. Rajesh, K., Rakesh, K.: Optimizing requirement analysis by the use of meta-heuristic in search based software engineering. Int. J. Elec. Comput. Eng. (IJECE) 9(5), 4336–4343 (2019) 6. Wang, Z., et al.: A novel data-driven graph-based requirement elicitation framework in the smart product-service system context. Adv. Eng. Inform. 42(C), 100983 (2019) 7. Wang, Z., et al.: A novel requirement analysis approach for periodic control systems. Front. Comput. Sci. 7(2), 214–235 (2013) 8. Xiange, L.: Design and realization of a test paper generation system. Softw. Guide 11(06), 55–56 (2012). (in Chinese) 9. Rui, Z.: Design and realization of a dynamic probability-based automatic test paper generation system. Autom. Instrum. 11, 210–211 (2016). (in Chinese) 10. Tao, Y.: Implementation of test paper library system and web online examination system under WEB. Sci. Technol. Perspect. (03), 145+173 (2016) 11. Bril, R.J., Altmeyer, S., van den Heuvel, M.M.H.P., Davis, R.I., Behnam, M.: Fixed priority scheduling with pre-emption thresholds and cache-related pre-emption delays: integrated analysis and evaluation. Real-Time Syst. 53(4), 403–466 (2017). https://doi.org/10.1007/s11 241-016-9266-z 12. Mian, F.: Study on test question extraction algorithm based on multi-layer filtering and dynamic probability model. J. Xichang College (Natural Science Edition) 33(02), 59–62+124 (2019) (in Chinese) 13. Liancheng, G.: Design of the automatic generative system of examination papers based on ARM. Int. J. Adv. Netw. Monit. Controls 2(4), 110–113 (2017) 14. Guan, L.: Design of the automatic generative system of examination papers based on ARM. Int. J. Adv. Netw. Monit. Controls 2(4), 110–113 (2018) 15. Liu, Z., Liang, R., Pan, X.: Customized online test database systems and automatic generation system for testing papers. Int. J. Educ. Manag. Eng. (IJEME) 2(8), 45–52 (2012) 16. Fu, Y.: Design and implementation of an automatic test paper generation system based on ant colony optimization. In: Proceedings of the 2015 International Conference on Intelligent Systems Research and Mechatronics Engineering, Zhengzhou, China (2015) 17. Jia, Z., Zhang, C., Fang, H.: Department of computer science and engineering North China institute of astronautic engineering, Langfang, China. The research and application of general item bank automatic test paper generation based on improved genetic algorithms. In: Proceedings of 2011 IEEE 2nd International Conference on Computing, Control and Industrial Engineering, pp. 30–34 (2011)

The Extended Fredkin Gates with Reconfiguration in NCT Basis Vitaly Deibuk1(B) , Oleksii Dovhaniuk2 , and Taras Kyryliuk1 1 Yuriy Fedkovych Chernivtsi National University, Chernivtsi 58012, Ukraine

[email protected] 2 School of Computer Science, University College Cork, Cork T12 R229, Ireland

Abstract. The Fredkin gate is a universal reversible logic gate widely used in designing low-power and quantum devices of various complexities. Recently, the gate’s extension was proposed based on its elementary components. It provided variety in functionality and, at the same time, preserved the main features of the gate. This paper offers additional variations of the Fredkin gate, referring to the recent findings. The versatility of logic circuits is particularly beneficial in programmable logic. Therefore, a reconfiguration for the proposed gates has been designed. The device includes 24 newly obtained gates and 8 known extensions. The reconfigurable circuit has a significant advantage in flexibility over the classic gate of 32 extended 3-bit Fredkin gates with a relatively small increase in hardware complexity and delay time. The implementation of the circuit has been tested and verified in Active-HDL, confirming the design’s correctness. Reconfigurable devices synthesized in the work make it possible to create programmable faulttolerant reversible logic circuits effectively. Keywords: Reversible logic · Fredkin gate · NCT gate library · Reconfiguration

1 Introduction The rapid development of computer technologies caused a growing interest in reversible logic for low-power devices, quantum computing, signal processing, computer graphics, bioinformatics, and DNA computing [1, 2]. The processor overheating is a primary reason for researching alternatives to classic computing. According to Landauer’s principle [3], each bit of erased information releases kT ln 2 J of energy, and recent experiments fully confirm it [4]. In contrast to classic computations, reversible computations do not lead to loss of information, which reduces processor overheating and power dissipation [5]. The main goal of the synthesis of binary and quantum reversible devices is to obtain an optimal reversible circuit for a given bijective function. For the functions of several variables, analytical and diagram synthesis methods can produce optimal results [6]. Nevertheless, heuristic evolutionary algorithms, such as genetic, ant colony, and annealing algorithms [7–11], work more efficiently for functions with numerous variables. These approaches utilize the functionally complete reversible NCT gate library of NOT, © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 95–105, 2023. https://doi.org/10.1007/978-3-031-36118-0_9

96

V. Deibuk et al.

controlled-NOT, and Toffoli (double-control NOT) gates. The reversible gates can have positive (black dot), negative (white dot), or mixed control lines that expand their functionality (Fig. 1a–i). The positive control line is activated at a signal of 1, and the negative control line is activated at 0. For multiple controls, activation occurs when all conditions are satisfied. These terms also apply to the Toffoli gates with disjunctive control or OR-Toffoli gates [12].

Fig. 1. Symbolic representation of the NCT basic gates with mixed polarity: (a) NOT, (b) CNOT (Feynman gate), (c) C’NOT, (d) CCNOT (Toffoli gate), (e) C’CNOT, (f) C’C’NOT, (g) ORCCNOT, (h) OR-C’CNOT, (i) OR-C’C’NOT, (j) Fredkin gate on a NCT basis

The functionally complete reversible Fredkin gate (FRG) is of great interest to researchers and engineers since any reversible device can be designed using just this one gate [13]. Besides being functionally complete, the Fredkin gate preserves parity which is critical in quantum computing for simplifying error detection and ensuring fault tolerance. In literature, FRG is also notated as a control SWAP gate for “swapping” the target signals on the output when the control signal equals 1. The classic Fredkin gate contains one control input (X 0 ) and two target inputs (X 1 , X 2 ) satisfying the following conditions: Y0 = X0 ; Y1 = X0 X1 ⊕ X0 X2 ;

(1)

Y2 = X0 X2 ⊕ X0 X1 . A few approaches were proposed to obtain new functionality and, at the same time, maintain the “swap” operation of the gate. First, adding extra control lines creates the generalized Fredkin gate (GFRG) with N controls and two targets [14]. This generalized gate interchanges data on the target lines when all control signals are 1. Second, changing activation conditions for gates with N controls in the way that the “swap” occurs when there is at least one activated control line [12]. Third, similarly to the NCT set, the polarity of controls can be positive (activation on 1), negative (activation on 0), or mixed for multiple controls [19]. A particularly compelling strategy was proposed in [15], where authors consider NCT implementation of the FRG (Fig. 1j), trying different polarity settings on the elementary gates and changing the activation condition for one of the gates. This approach led to the introduction of 8 unique FRG extensions. In the present paper, we analyzed the results of [15] and applied them to introduce 24 new extended Fredkin gates. This paper aims to increase the number of extended Fredkin elements compared to those proposed in [15]. For this purpose, reconfigurable devices have been synthesized,

The Extended Fredkin Gates with Reconfiguration in NCT Basis

97

allowing new reversible fault-tolerant circuit solutions to be created effectively. Such reconfigurable devices for advanced Fredkin gates are an innovation and are offered in the paper for the first time. Combining the FRG extensions into a reconfigurable circuit gives an advantage in flexibility. It can be utilized for assembling reversible encoders, decoders, or more complex devices such as field-programmable gate arrays [18], saving time and resources during production. The rest of this paper is organized as follows. Section 2 explains the main features of the extended Fredkin gates. The reconfigurations for extended Fredkin gates are proposed in Sect. 3 with an analysis of the obtained circuits. Section 4 provides details of the reconfigurable devices described in hardware descriptive language, including simulation and verification of the models. The conclusion of the work is presented in Sect. 5.

2 The Model of Extended Fredkin Gate Reversible logic circuits implement the bijective function f : {0, 1}n → {0, 1}n , where n is the number of input and output binary signals. Function f can be defined as a permutation of n quantities in 2n ways. Consequently, the number of input and output signals (lines) must be the same. A reversible circuit consists of reversible gates and implements the corresponding logical function, where branching at the output (fan-out) and fan-in at the input are prohibited. Reversible logic circuits do not lose information compared to classic logic circuits saving energy and allowing direct and reversible data flow, which are the main advantages. The Fredkin gate is a three-input reversible logic gate that implements the logical function specified by (1). The conditions show that input signal X 0 (control signal) transmits to output Y 0 unchanged. If the control bit is not activated (X 0 = 0), two other input signals, X 1 and X 2 (target signals), transmit to the output unchanged (Y 1 = X 1 , Y 2 = X 2 ). However, if X 0 = 1, the target signals exchange information at the output (Y 1 = X 2 , Y 2 = X 1 ). The specified logical function of the Fredkin gate can be implemented on an NCT basis, as illustrated in Fig. 1j. The gates of the NCT library invert a target signal with (controlled-NOT, Toffoli gate) or without (NOT gate) control. The input signals of the control lines transmit to the output unchanged for CNOT and Toffoli gate, in the same way as for Fredkin gate. The NCT set is also universal. However, unlike FRG, the NCT gates do not preserve parity. Thus circuits built with these gates require additional parity checks to ensure fault tolerance. The FRG modifications proposed in [15] are listed in Table 1. The table contains 16 modified Fredkin gates that implement 8 unique reversible functions. All modifications preserve features of “swapping” target gates on the output and transmitting the control signals without changes. It can be easily confirmed by comparing expressions in column “Function” with conditions (1). Nonetheless, target signals are inverted on the output for some configurations, indicating differences among the extended FRG functions. Additionally, circuits in the four last rows of Table 1 include an OR-Toffoli gate [12] instead of the classic Toffoli gate. This modification of the Toffoli gate has disjunctive control lines schematically represented as upside-down triangles (Fig. 1g– i). The notation is derived from analogy to sign ‘∨’ of the logical OR operation since

98

V. Deibuk et al.

activation on one control line is enough for the OR-Toffoli gate to invert the target signal. An extended Fredkin gate that contains the OR-Toffoli gate is usually called the OR-Fredkin gate. Table 1. Extended Fredkin gates (I) with activation keys [15] 0

1

2

0 0 0

0 0 1

0 1 0

0 1 1

1 1 1

1 1 0

1 0 1

1 0 0

Function

0

)

0

1

0

1

0 2

2

0

2

0 1

1

0

1

0 2

2

0

2

0 1

1

0

1

0 2

2

0

2

0 1

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

2

0 1

2

0

1

0 2

1

0

2

0 1

2

0

1

0 2

1

0

2

0 1

2

0

1

0 2

Circuit

Equivalent Circuit

The Extended Fredkin Gates with Reconfiguration in NCT Basis

99

3 Reconfiguration of Extended Fredkin Gate Reconfiguration is the ability of digital devices to change a functional circuit depending on the key signals applied to the corresponding input bits. A widely used example of a complex reconfigurable device is a field-programmable gate array (FPGA), which allows the programming of various circuits on one chip. FPGAs greatly facilitate and accelerate the development of computing systems [16–18, 21]. The circuit proposed in Fig. 2a activates one of the eight extended Fredkin gates presented in Table 1 by inputting key signals (K 0 , K 1 , K 2 ). The first three CNOT gates change the polarity of the corresponding input signals. The following two Toffoli gates invert the polarity of the target lines concerning the control signal X 0 if K 2 = 1 and do nothing if K 2 = 0. The three gates, outlined with a dashed rectangle, represent the classic Fredkin gate that performs a “swap” on activation. Finally, the last CNOT gate restores the control bit signal (Y 0 = X 0 ).

Fig. 2. Reconfigurable circuits of extended FRGs: (a) RFRG (I), (b) RFRG (II)

The eight extended gates proposed in [15] implement only a part of the possible 32 reversible functions based on the 3-bit Fredkin gate (1). Adding two extra key lines (K 3 , K 4 ) and arranging the NCT gates makes it possible to obtain an extensive reconfigurable circuit RFRG (II) (Fig. 2b) with a similar configuration. It has enough key lines to represent 32 unique variations of extended Fredkin gates (Table 2). RFRG (II) fully incorporates RFRG (I). This attribute is proved by Table 2, which includes all the logic functions of Table 1, offering 24 additional unique logic functions and, thus, new reversible circuits of extended Fredkin gates. Proposed representations of the reconfigurable circuits can be easily adjusted to a specific problem by removing key lines and rearranging the NCT gates on these lines. For instance, a reconfigurable circuit with only the first 16 functions from Table 2 is an RFGR (II) without K 4 and the gate connected to this key line. The key signals are transmitted to the outputs without changes since only the control bits of the reversible gates are connected to them. The hardware complexity of RFRG (I) is the same as that of RFRG (II) and equals three times more than the hardware complexities of the classic FRG, which is insignificant considering the 32-to-1 gates advantage.

100

V. Deibuk et al. Table 2. Activation keys of RFRG (II) 0

1

2

3

4

0 0 0 0 0

1 0 0 0 0

0 1 0 0 0

1 1 0 0 0

0 0 1 0 0

1 0 1 0 0

0 1 1 0 0

1 1 1 0 0

0 0 0 1 0

1 0 0 1 0

0 1 0 1 0

1 1 0 1 0

0 0 1 1 0

1 0 1 1 0

0 1 1 1 0

1 1 1 1 0

Function

0

)

0

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

0

1

2

3

4

0 0 0 0 1

1 0 0 0 1

0 1 0 0 1

1 1 0 0 1

0 0 1 0 1

1 0 1 0 1

0 1 1 0 1

1 1 1 0 1

0 0 0 1 1

1 0 0 1 1

0 1 0 1 1

1 1 0 1 1

0 0 1 1 1

1 0 1 1 1

0 1 1 1 1

1 1 1 1 1

Function

0

)

0

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

1

0

1

0 2

2

0

2

0 1

1

0

2

0 1

2

0

1

0 2

The Extended Fredkin Gates with Reconfiguration in NCT Basis

101

Table 3. Results of the RFRG (II) testing Hex

Gate outputs

Hex

Gate outputs

Hex

Gate outputs

Hex

Gate outputs

00

+X1+X2+X2+X1

02

+X1+X2+X2−X1

01

+X1+X2−X2+X1

03

+X1+X2−X2−X1

10

+X2+X1+X1+X2

12

+X2−X1+X1+X2

11

−X2+X1+X1+X2

13

−X2−X1+X1+X2

08

−X1+X2+X2−X1

0A

−X1+X2+X2+X1

09

−X1+X2−X2−X1

0B

−X1+X2−X2+X1

18

+X2−X1−X1+X2

1A

+X2+X1−X1+X2

19

−X2−X1−X1+X2

1B

−X2+X1−X1+X2

04

+X1−X2−X2+X1

06

+X1−X2−X2−X1

05

+X1−X2+X2+X1

07

+X1−X2+X2−X1

14

−X2+X1+X1−X2

16

−X2−X1+X1−X2

15

+X2+X1+X1−X2

17

+X2−X1+X1−X2

0C

−X1−X2−X2−X1

0E

−X1−X2−X2+X1

0D

−X1−X2+X2−X1

0F

−X1−X2+X2+X1

1C

−X2− X1−X1−X2

1E

−X2+X1−X1−X2

1D

+X2−X1−X1−X2

1F

+X2+X1−X1−X2

The RFRG (I) delay time can be represented as: τI = 2τCNOT + 2τTG + τFRG = 4τCNOT + 3τTG ,

(2)

where τCNOT , τTG , τFRG - delay time of CNOT, Toffoli gate, and classic Fredkin gate, respectively. Since second and third CNOTs operate in parallel, their combined delay time equals the delay of a single CNOT gate. The last two CNOTs also conduct signals in time of one CNOT. Similarly to RFRG (I), the delay time of RFRG (II) becomes: τII = τCNOT + 2τTG + τFRG = 3τCNOT + 3τTG .

(3)

RFRG (II) operates one CNOT gate faster than RFRG (I). At the same time, RFRG (II) allows the implementation of 32 logical functions (Table 2), including 8 logical functions (Table 1), for which RFRG (I) is assigned. Nevertheless, the advantage of RFRG (I) is that it contains only 3 permanent key lines, in contrast to 5 key lines in RFRG (II). Depending on the design task, both reconfigurable circuits can be utilized while considering their advantages and disadvantages.

4 FPGA Implementation A circuit is coded in hardware descriptive language (HDL) to implement it in an FPGA. One of the HDLs is the VHSIC Hardware Description Language (VHDL), which can model the behavior and structure of digital designs at various conceptual levels, from logic primitives to complex independent systems for describing, documenting, and verifying digital circuits [20]. The proposed RFRG (I) and RFRG (II) have been described and tested using VHDL in the Active-HDL 13.0 engineering software. For RFGR (II), the input and output signals have been initialized as binary vectors X [0..2], Ki[0..4] and Y [0..2], Ko[0..4] correspondingly. The architecture of the reconfigurable circuit is presented in Fig. 3. First of all, NCT gates were synthesized in advance and then imported into the design as components (lines 2–5), similar to the previous work [14]. “CNOT” represents the controlled-NOT gate, and component “TG” stands for the Toffoli gate. This approach allows safe modifications in the architecture of the NCT gates without changing anything

102

V. Deibuk et al.

directly in the RFRG’s design. Next, a “port map” operation, which manually links signals to components’ input and output, has been used to connect gates among each other. This way, the links between 9 gates from Fig. 2b have been established (G0 - G8). Gates G5, G6, and G7 represent the classic Fredkin gate when the rest of the gates manage the reconfiguration for the FRG extensions. The number of gates and their order are identical for RFRG (I) and RFRG (II). Therefore the VHDL implementation of RFRG (I) is analogous to RFRG (II), except it contains smaller logical signal vectors Ki and Ko due to fewer key lines in the circuit (Fig. 2a). The VHDL code of both circuits is ready for programming on FPGA boards, such as a 5CSEMA4U23C6N of the Altera Cyclone V, and can be used to synthesize more complex reversible circuits.

Fig. 3. The VHDL implementation of the RFRG (II) architecture

A separate testing program has been created to verify the correctness of the reconfigurable circuits and their VHDL description. The VHDL implementations of RFRG (I) and RFRG (II) are imported as components into the testing code. The best practice is to handle testing circuits as a black box knowing only the input and output data of the devices. The circuit assessments have been carried out to reach the following goals: (1) Ensure that the key and control signals of the circuit are transmitted to the output unchanged. (2) Check whether the key lines are utilized optimally, activating 8 gates for 3 key lines and 32 gates with 5 lines. (3) Confirm that key signals correctly activate the FRG extensions by comparing factual output for RFRG (I) and RFRG (II) with the expected output shown in Tables 1 and 2, respectively. The timing diagram of the RFRG (II) test simulation is presented in Fig. 4. It shows input signals X [0..2], Ki[0..4], output signals Y [0..2], Ko[0..4], and the circuit specifications such as the number of unique gates “unq” and the number of repeating gates

The Extended Fredkin Gates with Reconfiguration in NCT Basis

103

Fig. 4. The timing diagram of extended RFRG (II) in Active-HDL

“rep”. Key signals are represented in hexadecimal values from 00 to 1F, where 00 corresponds to the key K[0..4] = “00000” and 1F to K[0..4] = “11111”, respectively. The full simulation cycle is 514 ns, and the input data changes every two nanoseconds. Figure 4 confirms that the control signal and key signals are transmitted without changes to the outputs (Y 0 = X 0 , Ko = Ki), and the key bits are optimally used as expected since “unq” = 32, “rep” = 0 at the end of the simulation. For RFRG (I), the simulation cycle is almost 4 times shorter due to fewer key lines. The testing results show that RFRG (I) contains 8 unique gates and the output signals of these gates match with the output data of gates from RFGR (II), confirming that RFRG (II) fully contains the functionality of RFRG (I). Table 3 shows the output values of the informational signal lines following the hexadecimal representation of the key signals. Since the control signals are transmitted unchanged (Y 0 = X 0 ), the output signals are represented in the following order Y10 Y20 Y11 Y21 , where Y10 , Y20 stand for target line values on the output while X 0 = 0, and Y11 , Y21 are the values on the output while X 0 = 1. The arithmetic signs ‘+’ and ‘−’ indicate the polarity of the signals. Table 3 shows the correct key-gate activations of the extended Fredkin functions established in Table 2. Based on the simulation results, the proposed RFRG (II) activates 32 unique extended Fredkin gates without repeating activations, as expected. The reconfigurable circuits for reversible extended Fredkin gates have been described in the VHDL code. The programmed models are fully functional and they have been implemented in a 5CSEMA4U23C6N chip of the Altera Cyclone V. This approach allows for the synthesis of new binary logic digital reversible devices in a more compact design and various applications, particularly for message encryption and decryption devices, which will be the subject of our further research.

104

V. Deibuk et al.

5 Summary and Conclusion This paper offers 32 variations of the Fredkin gate containing 8 known gates and 24 new extensions. The analysis of the extended Fredkin gate based on the NCT gate set with mixed polarity was carried out. Two new reconfigurable circuits of extended Fredkin gates have been designed for the obtained reversible gates for the first time. The devices have an advantage in operational flexibility in cost of hardware complexity and delay time. The circuits have been described in HDL code for FPGA implementation. The simulation and evaluation in the Active-HDL environment approve the functional correctness of the HDL models. The circuits can be utilized to design and describe complex reversible devices for low-power and quantum computing. The proposed approach allows synthesizing new extended Fredkin gates with N control lines and using the reconfiguration idea to encrypt and decrypt the messages, which will be the object of future research.

References 1. Chandrika, V.O., Kumar, S.M.: Design and analysis of SRAM cell using reversible logic gates towards smart computing. Supercomput 78(2), 2287–2306 (2022) 2. Shafi, A., Bahar, A.N.: Fredkin circuit in nanoscale: a multilayer approach. Int. J. Inf. Technol. Comp. Sci. (IJITCS) 10(10), 38–43 (2018) 3. Landauer, R.: Irreversibility and heat generation in the computing process. IBM J. Res. Dev. 5(3), 183–191 (1961) 4. Bérut, A., Petrosyan, A., Ciliberto, S.: Information and thermodynamics: experimental verification of Landauer’s erasure principle. Stat. Mech. Theory Exp. 6, P06015 (2015) 5. Bennett, C.H.: Logical reversibility of computation. IBM J. Res. Dev. 17(6), 525–532 (1973) 6. Abdessaied, N., Drechsler, R.: Reversible and Quantum Circuits. Optimization and Complexity Analysis. Springer, Cham (2016) 7. Sasamal, T.N., Gaur, H.M., Singh, A.K., Mohan, A.: Reversible circuit synthesis using evolutionary algorithms. Lect. Notes Electr. Eng. 577, 115–128 (2020) 8. Ghosh, M., Dey, N., Mitra, D., Chakrabarti, A.: A novel quantum algorithm for ant colony optimization. IET Quantum. Commun. 3(1), 13–29 (2022) 9. Shahidi, S.M., Borujeni, E.S.: A new method for reversible circuit synthesis using a simulated annealing algorithm and don’t-cares. Comput. Electron. 20(1), 718–734 (2021) 10. Deibuk, V., Biloshytskyi, A.: Design of a ternary reversible/quantum adder using genetic algorithm. Int. J. Inf. Technol. Comp. Sci. (IJITCS), 7(9), 38–45 (2015) 11. Deibuk, V.G., Yuriychuk, I.M., Lemberski, I.: Fidelity of noisy multiple-control reversible gates. Semicond. Phys. Quantum Elec. Optoelec. 23(4), 385–392 (2020) 12. Moraga, C.: OR-Toffoli and OR-peres reversible gates. In: Yamashita, S., Yokoyama, T. (eds.) Reversible Computation. RC 2021. Lecture Notes in Computer Science, vol. 12805. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-79837-6_17 13. Fredkin, E., Toffoli, T.: Conservative logic. Int. J. Theor. Phys. 21(3–4), 219–253 (1982) 14. Dovhaniuk, O., Deibuk, V.: Synthesis and implementation of reconfigurable reversible generalized Fredkin gate. In: Proceedings of 2021 IEEE 12th International Conference on Electronics and Information Technologies (ELIT), Ukraine, pp. 165–169 (2021) 15. Moraga, C., Hajam, F.Z.: The Fredkin gate in reversible and quantum environments. Facta Univ. Ser.: Electron. Energetics 36(2), 253–266 (2023)

The Extended Fredkin Gates with Reconfiguration in NCT Basis

105

16. Murali Krishna, B., Sri Kavya, K.S., Sai Kumar, P.V.S., Karthik, K., Siva, N.Y.: FPGA implementation of image cryptology using reversible logic gates. Int. J. Adv. Trends in Comput. Sci. Eng. 9(3), 2522–2526 (2020) 17. Rajesh, K., Umamaheswara, R.G.: FPGA implementation of encryption and decryption of a message using optimized reconfigurable reversible gate. Int. J. Recent Technol. Eng. 8(2), 1654–1658 (2019) 18. Pawlowski, M., Szyprowski, Z.: Implementation of reversible gates in FPGA structure. Proc. SPIE 10445, 445–455 (2017) 19. Dovhaniuk, O., Deibuk, V.: CMOS simulation of mixed-polarity generalized Fredkin gates. In: 2022 12th International Conference on Advanced Computer Information Technologies (ACIT), Slovakia, pp. 388–391 (2022) 20. Pedroni, V.A.: Circuit Design with VHDL. The MIT Press, Cambridge (2020) 21. Jamuna, S., Dinesha, P., Shashikala, K.P., Kishore Kumar, K.: Design and implementation of reliable encryption algorithms through soft error mitigation. Int. J. Comp. Netw. Inf. Secur. (IJCNIS) 12(4), 41–50 (2020)

Applying Data Mining Techniques in People Analytics for Balancing Employees’ Interests Liana Maznyk1 , Zoriana Dvulit2(B) , Nadiia Seliuchenko2 , Marian Seliuchenko2 , and Olena Dragan1 1 National University of Food Technologies, Volodymyrska Str. 68, Kyiv 01601, Ukraine 2 Lviv Polytechnic National University, Bandera Str. 12, Lviv 79000, Ukraine

[email protected]

Abstract. In recent years the use of remote or hybrid formats of work organization in enterprises has become the world’s tendency. The introduction of the severe quarantine limitations related to the pandemic of COVID-19 and the avalanchetype character of the digitalization of society activated the changes in the sphere of social-labor relations, where the interests of different groups of workers are asymmetric. Many researchers who studied these phenomena mark the presence of divergences in individual advantages concerning the different formats of a labor organization of employers and hired workers. The dynamic of changes predefines the necessity of further research, which is why this subject acquires greater relevance. Our research aims to substantiate the approaches to balancing the interests of different categories of workers regarding the labor organization formats by exposing disbalances in the conditions of full-scale war in Ukraine. For monitoring, diagnosis, and management of disbalances, we substantiate the use of Data Mining in People Analytics (PA), the necessity of their combination with the classic instruments of study of personnel behavior, the use of active (reactive) formats of study of persons’ opinions in the extreme terms of running a business. The research results correlate with and confirm the world’s tendencies of introducing a hybrid model of labor organization. The novelty of the executed research consists of substantiating the approaches to balancing the interests of the employees regarding different labor organization formats in the conditions of full-scale war in Ukraine. The presented study expands the scope of knowledge in the field of PA, as it takes into account new circumstances that have not existed since the end of World War II. It supplements data on work organization formats obtained by researchers of the impact of the COVID-19 pandemic with new data obtained as a result of a survey of employees and managers of various organizations in Ukraine, where there is a full-scale war. The presented study is one of the first that emphasizes the features of PA during wartime and explains why in such conditions it is necessary to combine Data Mining techniques with classic tools for studying personnel behavior. The authors substantiate possible approaches to balancing the interests of different categories of employees. Keywords: People Analytics (PA) · Data Mining · Employers · Employees · Survey · Disbalance of interest · Formats of labor organization

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 106–115, 2023. https://doi.org/10.1007/978-3-031-36118-0_10

Applying Data Mining Techniques in People Analytics

107

1 Instruction Long-term research conducted by HRM experts and scientists in personnel management proves that the high involvement (loyalty) of personnel directly affects their performance indicators such as the level of labor discipline, quality of performance of production tasks, compliance with deadlines for tasks, etc., which, in turn, affects the overall business performance [1]. Based on the data from the American institute of public opinion Gallup [2], in companies with a high level of personnel engagement the level of workers’ absence in the workplace is 81% lower, the turnover of personnel is 18% lower, and the level of profitability is 23% higher. More than 50 years old research on employee engagement showed that engaged workers achieve better business results than other workers in any industry. The study of personnel engagement problems shows that the most successful collaboration occurs when employees and the organization share common values. The values of the employee determine their behavior. The analysis of the personnel behavior helps to answer the question: «What behavior of worker is effective for the achievement of business goals and helps to react to any challenges faced by the organization?». These and other similar questions are the main subjects of People Analytics (PA) that get increasing interest from practicians and scientists [3–5]. With the introduction of PA, organizations began to actively implement technologies and methods of information collection about workers’ behavior during business hours. The digitalization of the economy, the development of the digital labor market, and the COVID-19 pandemic which acted as the catalyst and activated the transition of the labor organization formats stipulated the necessity of processing large data volumes and the use of Data Mining techniques for PA [6–8]. The results of the use of PA for the last three years show that more workers are predisposed to remote work, and front-rank companies already prepare for it technically and technologically. However, on the whole, there is a significant divergence in the estimations of advantages of different formats of a labor organization of employers and hired workers [9, 10] that predetermines the necessity of social-labor relations harmonization for the corresponding direction. The full-scale war in Ukraine with its negative consequences made adjustments to the processes of functioning of organizations on the territory of Ukraine. Research concerning the peculiarities of the interaction of managers and employees in such realities is limited, therefore the study of people’s behavior in new socio-economic and political conditions is relevant. The presented research targets to study the opinions of workers and leaders of Ukrainian organizations that have experience in different formats of labor not only during the pandemic of COVID-19 but in conditions of the full-scale war in Ukraine, as well as the results of identifying imbalances in their interests and finding possible ways to eliminate them. Hence, the requirement was defined to formulate the following research questions and find answers to them. First is how is the interest of workers and leaders balanced concerning the different formats of labor organization in the conditions of full-scale war with its consequences. Second, what is the organizations’ technical and technological readiness level for remote work? Third, what approaches can organizations use to balance the interests of different categories of employees?

108

L. Maznyk et al.

2 Method of Research 2.1 Methods of Collecting the Opinions of Employers and Employees Regarding Attitudes to Different Formats of Work Organization The composition of the offered research considers the necessity of combining Data Mining for PA technologies with the classic methods of personal behavior study that are especially important in extreme terms (pandemics, military operations, blackouts). Since knowledge about people’s behavior in the conditions of a full-scale war with all its negative consequences is limited, the proposed approach reduces the risks of using exclusively artificial intelligence in People Analytics. It makes it possible to take into account specific situations and the specifics of each person’s reaction to them. Results of previously conducted monitoring confirm the constantly growing asymmetry of employers’ and workers’ answers concerning the advantages and disadvantages of the different formats of labor organization. The presented research was organized in the form of observational data analysis. A sample is formed with the use of the method of representative sampling. The choice of method of observational data analysis is based on the necessity to research the opinions of different employee groups that are included in the general population. The groups are formed based on the qualitative characteristics of “leaders” and “workers (specialists, professionals)”. To answer the research questions we formed two questionnaires that contained 13 questions of the closed type, one for the leaders and one for the workers. The survey envisaged the parallel study of leaders’ and workers’ opinions by an identical list of questions. We collected responses from 101 respondents who participated in the survey and worked in different industries of the economy, where 24 were leaders and 77 were workers (specialists, professionals). The survey took place in Ukraine from November 22 to December 02, 2022. To implement the questionnaires, we used the Google Forms tool and to collect responses we shared the questionnaires on the most popular social networks and messengers. Generalization of the survey results allows for determining the level of coordination of opinions of different groups of respondents. The sample is much smaller than the general population – the number of employees in Ukraine in 2020 was 9 948,1 thousand people, and the corresponding statistical data for 2021–2022 is missing. With a probability of 0,95, we can say that the margin of error of the survey results is 9,75%. Our experimental expectations were to diagnose imbalances in the interests of employees and managers of organizations regarding different formats of work organization. This made it possible to substantiate approaches to balancing their interests. 2.2 Literature Review During the pandemic of COVID-19 and strict quarantine restrictions, when organizations passed to the remote format of work, HR managers faced several problems related to the functions of diagnosis, coordination, supervision, and motivation [11, 12]. It gave the impulse to the development and implementation of innovative and effective methods of employee engagement. The introduction of employee engagement measures based

Applying Data Mining Techniques in People Analytics

109

on digital technologies became especially important for the growth of organizations in crisis [13, 14]. The tendency of the increasing influence of personnel analytics teams, which was observed from 2020 to 2021, was stable in 2022. In 2021 the correlation between analytics team members and total employee headcount changed from 1: 4000 (2020) to 1: 2900. In 2022, this correlation remains at the level of 2021. Leaders of personnel analytics get a more significant influence on the top management of organizations: 21% of leaders report directly to the Chief Human Resources Officer comparatively with 13% in 2021 [15]. In regards to the increasing role of PA, the increasing popularity is acquired by research on the integration of artificial intelligence in managing human resources (HRM) in the context of different functions and practices in organizations. Experts in the field of HRM consider that possession of the relevant digital competencies by the workers in this field is the base condition of forming the modern labor market of HR [16]. The development of PA stipulated the introduction of the systems of HR analytics that are based on the analysis of large arrays of data and apply artificial intelligence to provide the possibility to get practical conclusions (unlike regular reports that give data). Software that developers offer for PA is various: it can be concentrated only on one HR function (payroll management, management of talents, systems of watching of declarants, training and development, production management); offers a single window (HR platforms) that manages all processes from hiring and onboarding to engaging and evaluation of the personnel productivity. To monitor the workers, HR-tech companies offer packages that monitor everything, beginning with keypresses and ending with access to the external disk. The best software offered in the market in 2022 can monitor the visited web pages and used applications, block content and applications, make screenshots of users’ activity, and generate detailed reports [17]. Digitalization of PA allows HR managers to control the workers’ productivity and efficiency in the office and remotely. HR managers can regularly get information about workers that work remotely and, thus, estimate the influence of different formats of labor organization on the productivity, quality, efficiency, psychophysiological state of workers, levels of stress in a team, work difficulties, mutual relations in the team, informal leaders. HR systems that collect anonymized data about employees using questioning or digital tracking (corporate messengers, email, and others), allow forecasting the future, e.g., to forecast personnel turnover or employee behavior. Artificial intelligence gave strong development to the management strategy of talents, as now, in the early stage, it is possible to estimate the potential of the candidate and choose the right moment for rotation and promotion at work. By studying employees’ behavior, an organization can estimate the level of work satisfaction, psychical health, and prosperity of workers, timely reduce their professional burnout, determine the effective or toxic managers, etc. When using PA as an instrument for balancing the interests of leaders and workers, one should take into account its “shadow sides” [18], which are bringing about an illusion of control and reductionism; leading to estimated predictions and self-fulfilling prophecies; fostering path dependencies; impairing transparency and accountability; reducing employees’ autonomy; marginalizing human reasoning; and eroding managerial competence.

110

L. Maznyk et al.

3 Result and Discussion The analysis of the survey’s results showed that most of the polled workers (55,8%) and leaders (50%) represent service businesses. About 24,7% of workers and 16,7% of leaders represent the IT sector. The majority of workers (55,8%) represent organizations with over 100 people, and 41,7% of leaders represent organizations with under 50 people. The survey results showed that almost 45,3% of workers can perform their duties entirely remotely, 34,7% of workers can perform their duties in the hybrid format, and only 20% are not able to work remotely. Moreover, only every third leader gave a positive answer to the possibility of working remotely, 37,5% confirmed they can perform their duties in hybrid mode, and 29,2% of the polled leaders are not able to work remotely. It is necessary to mark that among the polled workers some were in positions with functional duties that require a physical presence in the workplace. Thus, even the influence of extreme circumstances is not a substantial criterion for choosing the format of labor organization. To the question “Does your company regulate the use of different work formats in the regulatory documents?” almost 58% of workers gave a positive answer, 30,3% answered that work formats are not clearly regulated, and 11,8% pointed out the absence of such regulations. Only 45,8% of leaders confirmed the presence of regulation of work formats in normative documents, 37,5% mark the present regulation as not clear enough, and about 17% state the absence of regulation of labor organization formats in their companies. The results testify that leaders have a more serious understanding of approaches to forming internal normative documents concerning corporate measures in extreme terms. To the question “Does your company use the efficiency indexes (KPI, OKR) or other criteria to evaluate your job results?” almost 60% of respondents from the two groups gave positive answers, 37,5% of the polled leaders and 26,6% of workers are sometimes informed about kinds and controls over their professional duties performance and 16,8% of the polled workers expressed no interest in this question. The latter can be explained by the respondents that are confident about adequately implementing their duties. Some respondents (8,3% leaders and 7,8% workers) answered that such questions were not relevant due to the high level of personnel engagement in their organization. According to the given answers, it is possible to mark that not many leaders and workers confirm the absence of information about the kinds and control methods. This is the evidence that Ukraine’s system of social-labor relations is sufficiently transparent and “shadow sides” of PA are minimized. About the work in the office, positive answers gave only 26% of workers, while this format of labor organization chose about 45,8% of leaders. Among the respondents, 55,8% of workers and 37,5% of leaders prefer the hybrid format of work organization. Fractions of workers and leaders that would like to work remotely are close (18,2% and 16,7%) which testifies to the presence of a balance of their interests concerning the organization of the remote work format (Table 1). In the opinion of most workers (52,6%) and leaders (45,8%), the work in the office is characterized by higher productivity in performing their professional duties. Remote work is preferred by about 18,2% of workers and 16,7% of leaders. Interestingly, only 14,5% of workers and 12,5% of leaders consider productivity to be higher in terms of the remote format of work organization. Thus, the disbalance occurs between the

Applying Data Mining Techniques in People Analytics

111

Table 1. The attitude of the employees toward the different formats of labor organization Work format In the office Remote Hybrid

Answers

Groups of employees Leaders, %

Workers, %

Prefer

45,8

26,0

Productivity is higher

45,8

52,6

Prefer

16,7

18,2

Productivity is higher

12,5

14,5

Prefer

37,5

55,8

Productivity is higher

41,7

32,9

answers of both polled groups concerning the advantages and productivity of working in a remote format: a particular part of respondents would like to work remotely despite the understanding that labor productivity will be lower. Fewer leaders (37,5%) prefer the hybrid format, although a more significant part (41,7%) asserts that such a format provides higher labor productivity. The more significant part of workers (55,8%) wants to work in the hybrid format, understanding that the labor productivity will be lower. Thus, leaders estimate labor productivity changes depending on the different work formats. More than half of polled workers consider working in the office more productive, but only a fourth of them prefer this format of work organization. For the management of companies, it must signal the presence of a large disbalance of interests, which is why corporate measures must include specific stimulation instruments. The choice of certain instruments will depend on the results of the PA. Part of the survey questions focused on determining the factors that induce workers or leaders to prefer work in hybrid or remote formats. For this purpose, it was offered to define the priorities of the factors such as saving commuting time and expenses, saving time on complying with business style, and digitalizing the personal space. Summarizing the survey results showed that most leaders considered digitalization a priority in the ground of hybrid or distance formats. For workers, these priorities are saving time and travel expenses. The coordination of respondents’ answers concerning their priorities was evaluated during the selection of remote or hybrid formats of work using the coefficient of mutual conjugating of Cramer. We will mark that the value of this coefficient can change in limits from 0 to 1. The closer the value to 0, the weaker the connection conditions and vice versa. Calculations showed that both polled groups were not homogeneous enough and that higher divergence of thoughts was among the leaders which equals to 0,1145. Concerning workers, the coordination in this group is not far but higher and equals to 0,1511. Thus, such values testify to the weak but noticeable coordination in each investigated group. Answers in the survey concerning existing technical and technological terms for remote work, level of satisfaction of hired workers by these terms, and reporting methods for the executed work allow evaluating the level of technical and technological readiness of organizations to remote or hybrid formats of work. More than 60% of leaders confirm the presence of corresponding technical and technological terms for the remote and hybrid work formats. Only 49,3% of workers

112

L. Maznyk et al.

gave positive answers to this question. It can be explained by (1) comparatively, with the leaders, a more significant amount of workers have experience working in the remote and hybrid formats in the conditions of quarantine limitations, which allows them to estimate present terms more critically in war-time; (2) part of workers having experience in remote or hybrid work demonstrate the belated reaction to the situation of blackout in Ukraine; (3) among the polled leaders, 41,7% represent companies with under 50 employees, most characteristically for the IT sphere, electronic commerce, IT-marketing, where technical and technological terms are more adapted to the modern extreme terms of the functioning of Ukrainian business; (4) leaders perceive corresponding challenges more strategically and comprehensively approach to the evaluation of the created terms, which envisages not only the use of software facilities for remote and hybrid formats of work but also the necessity of the use of alternative methods of energy supply. Regarding the level of satisfaction by the technical and technological terms for the realization of the remote and hybrid formats of work, fractions of answers of leaders and workers almost coincided, i.e., disbalance in this question is absent. The majority of respondents are totally or partly satisfied with the created terms. Answers of respondents to the question «What method do they use for evaluating the executed work»? were divided into 43,4% of workers and 45,8% of leaders reporting about the executed work by electronic reports; every third worker (30,3%) and every eighth leader (12,5%) reported using specialized software. This testifies that organizations use modern technologies for information collection about the workers. The results of the survey we conducted correlate with the conclusions obtained in other analogical surveys. However, such survey plans did not envisage the parallel study of leaders’ and workers’ opinions with an identical list of questions. A Survey of leaders’ opinions of leading companies of Ukraine, conducted from July to October 2022 by KPMG Ukraine [19], showed that 63% of business leaders of Ukrainian companies and 48% of leaders of the world companies marked that hybrid work positively influenced on investments in technology of workplace. At the same time, 5% of leaders in the world said that a hybrid format negatively influenced workers’ maintenance. In contrast, business leaders in Ukraine admitted vice versa, that the possibility to work from home or the office had positively influenced the maintenance of talents. However, in the opinion of Ukrainian leaders, hybrid work negatively influences workers’ productivity (42%) and worsens their morale state (37%). In the world, these indexes are far below the level of 13% and 12%, accordingly. Without regard to that, the leaders of Ukrainian companies in the future think about the work in a hybrid format (88%) that exceeds the expectation of world leaders (28%) three times. Only 5% of the Ukrainian leaders speak about the obligatoriness of traditional work in the office. At the same time, leaders in Ukraine (5%) and in the world (7%) do not plan to transform the working environment to the format of fully remote collaboration. The results of our research practically coincided with the survey of KPMG Ukraine [11] regarding the leaders’ that give the advantage to the hybrid format of work organization. However, a contradiction occurs concerning the results of the influence of such a format on labor productivity. From our data, in the opinion of 41,7% of leaders, the hybrid format of labor organization assists the increase of labor productivity, and according to the results of KPMG Ukraine, 42% of leaders think that such format will

Applying Data Mining Techniques in People Analytics

113

negatively influence labor productivity. Such divergences in the results of the survey can be explained by different times of their realization: KPMG Ukraine was conducted during July - October 2022, when the destruction of power infrastructure of Ukraine was not yet there; our research is undertaken in November - December 2022, when periodic partial and complete blackouts began to occur in the context of war. Thus, we took into account the change of terms of doing business in Ukrainian realities by using the instruments of Data Mining. According to research [10], 95% of employers want workers to be in the office. At the existence of choice, 99% of workers expressed the desire to work remotely their whole life even in part-time jobs. However, Ukrainian realities showed quite another structure of such priority distribution in the choice of different work formats. About 45,8% of leaders and 26% of workers entirely give the advantage to working in the office, and among the supporters of remote and hybrid formats are 54,2% of leaders and 74% of workers. According to our research, such distribution of ideas does not conflict with the survey [19] results, where two-thirds of people want to work at home, and 80% of workers think about the remote format as the competitive edge of the enterprise at the labor market. According to AT&T research [20], the hybrid model of work in the USA was 42% in 2021, with the prospect of increasing to 81% in 2024. This format is the prevailing model of the organization of the working process, and most business leaders consider this as a new standard. In Ukraine, according to our research, 79,2% of leaders and 82,4% of workers have the experience to work in remote or hybrid formats. These numbers for 2022 in Ukraine correspond to the forecasts of researchers for the USA in 2024. Despite all advantages of hybrid and remote formats of work for the different groups of hired workers, some problems take place such as formal and informal communication in teams, work and personal life balance, level of trust in the team; inoperative reactions to current problems, etc. To balance the interests of the hired workers, we recommend using such approaches: (i) combination of Data Mining, People Analytics and classical instruments of study of personnel behavior; (ii) the use of active (reactive) forms of studying people’s opinions in the extreme terms of running a business; (iii) integration of the module of behavior analytics in any automated HRM system that is used in the enterprise; (iv) providing a transparent process of collecting information about workers using a personal electronic cabinet for every worker, where such individual information can be accessed by each worker in read-only mode as well as by the leader with the corresponding access level; (v) development of specific technical, technological, and organizational measures based on diagnostics of the hired workers attitude to implementing functional duties in the different formats of labor organization.

4 Summary and Conclusion In the presented research, the disbalances of interests of separate categories of workers concerning the different formats of labor organization are exposed and approaches to harmonizing corresponding social-labor relations in the conditions of full-scale war in Ukraine. Our results showed imbalances that occured between the answers of both polled groups concerning the advantages and productivity of working in a remote format: a particular part of respondents would like to work remotely despite the understanding that

114

L. Maznyk et al.

labor productivity will be lower. Fewer leaders and workers prefer the hybrid format, although a more significant part asserts that such a format provides higher labor productivity. The majority of leaders confirm the presence of corresponding technical and technological terms for the remote and hybrid work formats. Less than half of workers gave positive answers to this question. The presented study expands the scope of knowledge in the field of PA, as it takes into account new circumstances that have not existed since the end of World War II. Modern PA systems, for objective reasons, cannot use historical data on the behavior of employees in the conditions of a full-scale war in Ukraine. Therefore, to identify imbalances in the interests of employees, it is recommended to combine Data Mining, People Analytics and classical instruments of study of personnel behavior. The results of the study complement the already available data on the attitude of employees to various formats of work organization during the COVID-19 pandemic. Surveying employees through the distribution of Google Forms does not provide an opportunity to ensure the representativeness of the sample. However, in the difficult conditions of the war, with its help, the most acute problems in the field of labor organization can be identified. HR managers can further use the information obtained during the formation of a questionnaire for a regular survey within a specific organization and for making management decisions based on Data Mining techniques. We expect that the results of the conducted research will contribute to the growth of scientific interest in the field of application of Data Mining in People Analytics in the conditions of global challenges of society (digital economy, pandemics, climate problems, armed conflicts, full-scale war in Ukraine, world food crisis, etc.).

References 1. Govender, P., Sukdeo, N., Ramdass, K.: Employee engagement and the impact on productivity in a South African FMCG company. In: Proceedings of the 7th North American International Conference on Industrial Engineering and Operations Management, Orlando, pp. 205–220 (2022) 2. Website Gallup. https://cutt.ly/p0KNlta 3. Huselid, M.: The science and practice of workforce analytics: introduction to the HRM special issue. Hum. Resour. Manage. 57(3), 679–684 (2018) 4. Allaham, M.: Bibliometric analysis of HR analytics literature. Elec. J. Soc. Sci. 21(83), 1147–1169 (2022) 5. McCartney, S., Fu, N.: Promise versus reality: a systematic review of the ongoing debates in people analytics. J. Organ. Effect. People Perform. 9(2), 281–311 (2022) 6. Dahlbom, P., Siikanen, N., Sajasalo, P., Jarvenpää, M.: Big data and HR analytics in the digital era. Balt. J. Manag. 15(1), 120–138 (2019) 7. Chamorro-Premuzic, T., Bailie, I.: Tech is transforming people analytics. Is that a good thing? Harvard Business Review (2020). https://cutt.ly/Z0KB23S 8. Repaso, J., Capariño, E., Hermogenes, M., Perez, J.: Determining factors resulting to employee attrition using data mining techniques. Int. J. Educ. Manag. Eng. 12(3), 22–29 (2022) 9. Website Moneyzine. Martynas Pupkevicius. Revealing remote work statistics & facts for 2022. https://moneyzine.com/careers-resources/remote-work-statistics 10. Website KPMG. https://cutt.ly/E0KNyvc

Applying Data Mining Techniques in People Analytics

115

11. Hussain, M., Mirza, T., Hassan, M.: Impact of COVID-19 pandemic on the human behavior. Int. J. Educ. Manag. Eng. 10(5), 35–61 (2020) 12. Salehin, I., Tamim, D.S., Mohammad, T.I., Rayhan, I., Fatema, N.K.: Impact on human mental behavior after pass through a long time home quarantine using machine learning. Int. J. Educ. Manag. Eng. 11(1), 41–50 (2021) 13. Nisha Chanana, S.: Employee engagement practices during COVID-19 lockdown. J. Publ. Affairs (2020) 14. Jack Linchuan Qiu: Humanizing the posthuman: digital labour, food delivery, and openings for the new human during the pandemic. Int. J. Cult. Stud. 25(3–4), 445–461 (2022) 15. Ferrar, J., Verghese, N., González. N.: Impacting Business Value: Leading Companies in People Analytics. shorturl.at/uwD57 16. Shpak, N., Maznyk, L., Dvulit, Z., Seliuchenko, N., Dragan, O., Doroshkevych, K.: Influence of digital technologies on the labor market of HR specialists. In: Proceedings of the CEUR Workshop Proceedings, vol. 3171, pp. 1475–1487 (2022) 17. The Best Employee Monitoring Software for 2022. Business News Daily. https://www.bus inessnewsdaily.com/11143-best-employee-monitoring-software.html 18. Giermindl, L.M., Strich, F., Christ, O., Leicht-Deobald, U., Redzepi, A.: The dark sides of people analytics: reviewing the perils for organisations and employees. Eur. J. Inf. Syst. 31(3), 410–435 (2022) 19. Global Workplace Analytics. Costs and Benefits. https://globalworkplaceanalytics.com/res ources/costs-benefits 20. Is corporate America ready for The Future of Work? AT&T. shorturl.at/oqAG5

Object-Oriented Model of Fuzzy Knowledge Representation in Computer Training Systems Nataliia Shybytska(B) National Aviation University, Kyiv 03058, Ukraine [email protected]

Abstract. The situation in the world with covid and other historical events has accelerated the transition of the educational process to training remotely. That’s why the task of developing models for the representation of didactic knowledge in a form suitable for computer processing is currently acquiring scientific and practical significance. The article shows one of the aspects of optimization of the learning environment by formalizing the process of representation and obtaining knowledge in computer-based training systems based on an object-oriented model of the representation of didactic knowledge. It is proposed to optimize the learning process by eliminating redundancy (procedural method), optimizing the execution of actions at a time (composition), and combining procedures to obtain one result (generalization). The object-oriented model of fuzzy knowledge representation in computer training systems is based on combining elementary semantic units into classes and identifying the corresponding relations between them by specifying a fuzzy membership function of the student’s knowledge to the system’s knowledge, obtained during the test. Thus, the proposed principles make it possible to create computer training systems with a flexible learning scenario by correcting the learning trajectory taking into account the results of intermediate testing. Keywords: Knowledge representation · computer training systems · object-oriented technology · knowledge model · fuzzy set · fuzzy membership function

1 Introduction Long-term didactic experience and the situation in the world with covid-19 have revealed a lot of disadvantages of the traditional classroom learning model [1–3]. The training process should be highly individual in speed and form, and take place in the location and time most suitable for a student. The creation of intelligent learning systems - computer training courses with a flexible learning strategy and with multimedia elements of visualization, improves the efficiency of the learning process [2–4]. One of the aspects of the activation of the learning environment is the formalization of the process of presenting and acquiring knowledge in computer-based training systems.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 116–125, 2023. https://doi.org/10.1007/978-3-031-36118-0_11

Object-Oriented Model of Fuzzy Knowledge Representation

117

The development of information technologies and the growth of the productivity of computer systems have predetermined the improvement of structural forms of presentation of information, which are characterized by a high level of formalization and abstraction. Thus, the task of developing models for the representation of didactic knowledge in a form suitable for computer processing is currently acquiring scientific and practical importance [5–7]. This article proposed the object-oriented model of fuzzy knowledge representation in computer training systems, where the student’s knowledge is a subset of the knowledge of the system with some fuzzy membership function. This allows you to create of fully-connected and dynamically customizable computer-based training systems with a flexible learning scenario by correcting the learning trajectory using fuzzy function obtained as a result of intermediate test control students’ knowledge.

2 The Task Analysis of Knowledge Representation Models The representation of knowledge as a methodology for modeling and formalization of conceptual knowledge, focused on computer processing, is one of the most important sections of knowledge engineering [8–13]. This is due to the fact that the representation of knowledge ultimately determines the characteristics of computer training systems and depends on the nature and complexity of the tasks. Let us analyze the existing knowledge representation models used to build computer training systems. In a broad sense, the learner model is understood as knowledge about the learner used to organize the learning process. The learner mode a set of accurately presented facts about the student, which describe various aspects of his condition: level of knowledge, personal characteristics, professional qualities, etc. Computer training systems are developing in following directions [12]: 1) Creation of learner (student) models - is a structured representation of a learner’s knowledge, misunderstandings, and difficulties, is constructed from learner data usually gathered by a system through the learner’s interaction with a training system. 2) Creation of information structure subject area for obtaining a knowledge representation model which can be the basis of a computer-based training system. 2.1 Student Models The student (learner) model is the core of a computer-based training system. For each user of the system, the student model maintains up-to-date information about his learning goals, path, experience, student knowledge, misunderstandings, mistakes, etc. It stores the history of his interactions with the system. Thus, student model includes: objectives of a current training session; a target level of knowledge; a user’s current knowledge level; a current training program; a system working parameters and options set by the user and/or automatically detected by the system; technical characteristics of the current client computer. The most famous student models are: overlay learner model, difference model, perturbative model. The easiest model to implement is the overlay student model [15]. It is based on the assumption that the knowledge of the student and the knowledge of the system have a

118

N. Shybytska

similar structure, while the knowledge of the student is a subset of the knowledge of the system. A numeric attribute is added to each topic to indicate the extent to which the learner understands the material on that topic. The value of this attribute will be determined during the student’s survey. Student difference model is the model of the student, in the construction of which the system analyzes the answers of the student and compares them with the knowledge that is embedded in the system and used by the expert when solving similar problems. The differences between this knowledge form the basis of the user model. This model allows taking into account not only the lack of knowledge of the student, but also their incorrect use. Perturbation model of the student is the model of the student, which is built on the assumption that the knowledge of the student and the knowledge of the system may not partially coincide. In this case, an important prerequisite for constructing such a model is the identification of the reasons for the discrepancy, since without defining discrepancies, the learner model will be too fuzzy. 2.2 Knowledge Representation Models Let’s consider the most widely-used knowledge representation models: logical, production, semantic networks, frame and ontologies models [7, 13, 14]. The logical model is used to represent knowledge in the system of first-order predicate logic and formulate conclusions using syllogism. Such a model makes it possible to implement a system of formally accurate definitions and conclusions. To represent knowledge using predicate logic, it is necessary to choose constants that define objects in a given area, functional and predicate symbols that define functional dependence and relations between objects, and build logical formulas based on them. In the case when the choice of objects is difficult, you should choose a different model of knowledge. In production models, knowledge is represented by a set of rules of the form “IF THAT ”. These models are of two types: with direct and inverse conclusions and include three components: a rule base consisting of a set of products (rules of inference); a database containing many facts; interpreter to obtain a logical conclusion based on this knowledge. The rule base and the database form the knowledge base, and the interpreter corresponds to the inference mechanism. Direct output systems convert the contents of the database from the original to the target. However, such a conclusion is characterized by a large amount of data, as well as estimates of the state tree that are not directly related to the conclusion. With the help of rules, a AND/OR tree is constructed in the production system with inverse conclusions, linking facts and conclusions. The logical conclusion is an assessment of this tree based on facts. Parts of the tree that are not relevant to the conclusion are not considered in this case. Thus, production systems, along with positive properties - the simplicity of creating and understanding individual rules and the simplicity of the inference mechanism - also have disadvantages: low processing efficiency, the complexity of the mutual relations of rules. The next model for the representation of knowledge is semantic networks, representing the knowledge system in the form of a holistic image of the network, the nodes of which correspond to concepts and objects, and arcs - to the relations between objects.

Object-Oriented Model of Fuzzy Knowledge Representation

119

Initially, the semantic network was invented as a model for representing the structure of long-term memory in psychology. The hierarchical structure of concepts shows the relation of inclusion of concepts using relationship predicates: IS-A (inheritance of attributes in the hierarchical knowledge system) and PART-OF (“part-whole” relation). The advantage of the representation of knowledge by semantic networks is the simplicity of the description model based on the relationship between elements. However, with the increase in network size, the search time increases significantly. The methods of knowledge representation based on semantic network models have found application in natural language understanding systems, in particular, at the stage of structural analysis of the recognition of parts of speech, as well as highlighting keywords. The next step in the formalization of the representation of knowledge was the creation of the Minsky frame theory [16]. A frame is a data structure designed to represent some standard situations. The frame representation model is a hierarchical structure of frames, each of which, in turn, consists of an arbitrary number of slots represented by a certain data structure, which is briefly described as a triplet “object-attribute-value”. Frame systems are used not only in diagnostic expert systems, but also in the field of pattern recognition, natural language processing, and knowledge processing [16–18]. Ontology is a modern technique of knowledge representation that is grounded on a conceptualization. Conceptualization is an abstract, simplified view of some selected part of the world, containing the objects, concepts, and other entities that are assumed to exist in some area of interest and the relationships that hold among them [19, 20].

3 Object-Oriented Model of Knowledge Representation The object-oriented model of presentation data is based on the concept of an object, which is a combination of data and methods for processing them. In object-oriented programming (OOP) environments, any object is defined by two characteristics - state and behavior. The state (attributes) of an object is determined by the current values of its parameters. Behavior (event) determines the ways of changing the object’s own state and its interaction with others [21, 22]. The developed graphical interface of modern software allows you to process objects in the form of graphic images, which allows you to optimize not only the process of obtaining knowledge but also to simulate the process of training skills. A perspective direction of using computer technologies in the learning process is the integration of computer capabilities and various methods of transmitting audiovisual information, which is made available using multimedia tools that allow you to create and use 3D graphics, animation, and sound. At the same time, formalization of many different methods of assembling and analyzing information about the course of the learning process and the preparedness level of students becomes available. The task of modeling and optimizing the process of managing the information flow in the training system, taking into account the individual level of training of the student, becomes relevant. The didactic principle of the formation of the scientific content of education can be formulated as follows: structuring the scientific content to indivisible (unable to be divided or separated) elements of knowledge within the discipline in order

120

N. Shybytska

to identify interdisciplinary relationships and manage the process of declaring knowledge [23]. From a pedagogical viewpoint, setting goals for discipline training is to identify and logically strictly define the skill system that a learner has to master to solve professional problems. The objective learning function defines a system of skills, for the formation of which it is necessary to organize the process of mastering the elements xi ∈ X of the knowledge set X of the system. To organize a learning process management system, firstly need to define goals skills, then form the scientific content of the training, which is transformed into a skill system. To determine specific goals, it is necessary to structure the learning process and identify typical learning activities. As a result of this, the activity can be represented in the form of simpler actions and set to an adaptive level to the learner or to choose a higher level and to raise the speed of education. Convolution is the merging of elementary actions into a new action of a higher level. The typical learning actions can be performed: simultaneously, sequentially, continuously, discretely, be combined into a linear, branching or cyclic structure. Let us imagine the scientific content of training in the form of a hierarchical model (Fig. 1).

Fig. 1. The hierarchical structure of the discipline

The knowledge in this model is described in a modular fashion. The structure of the modules is determined by the relations of concepts typical of each module. The top-level xi modules can be sections and topics of the studied subject area. At the lower level, the concepts xn of the discipline under study are considered, while xi , xn are the modules of the upper and lower levels, respectively, vi are the relations between the modules. Optimization of the processes of knowledge formation is carried out on the basis of compilation mechanisms: – The procedure method (Proceduralization) consists of the allocation of typical actions and the formation of procedures with the replacement of variables in the originally used universal rules by some special values suitable for solving the problem.

Object-Oriented Model of Fuzzy Knowledge Representation

121

– The composition is a combination of independent rules and their transformation into one general. – The mechanism of a knowledge compilation allows increasing the speed of assimilation and processing of information by the student while reducing the load on long-term memory, the learning process becomes more adaptive. The stage of didactic knowledge coordination is carried out using three mechanisms: 1) Strengthening the rules is used when comparing them taking into account frequently used rules and is achieved through the organization of repetitions and visual-auditory effects. 2) Separation or specialization is used to create a new rule based on the information received in a previous attempt to apply known rules. 3) The mechanism of generalization allows you to expand the area of the rules and ensure their application in a wider context. Thus, the optimization of the education process in computer-based training systems can be achieved using mechanisms of eliminating redundancy of didactic information (proceduralization), a mechanism that allows performing actions at a time (composition), and a mechanism for combining procedures that allow getting one result - (generalization). The mechanism of skill learning is that of compilation, which is a complex process consisting of proceduralization and composition.

4 Method of Describing Student’s Knowledge as Fuzzy Subset of System’s Knowledge To manage the learning process and organize feedback from the student to the computerbased learning system, the object-oriented model of presenting didactic knowledge is proposed and shown in Fig. 2. This model is based on the principles of encapsulation, inheritance, and polymorphism allows you to build a hierarchical model of the subject area by combining objects into classes and identifying the corresponding relations between them. Subsets of test indicators ti ∈ T are determined at the stage of test control of students’ knowledge and form classes at the levels of hierarchy in the knowledge system with elements xi ∈ X . Today, there are no standards for estimating the processes of learning information; therefore, it is important to structure the scientific content of education to elementary semantic units - information objects. The task of creating semantic units of knowledge on the topic of lectures and a system of didactic knowledge is solved by the teacher subjectively. We do not have an objective definition of a knowledge unit. A knowledge unit is something that brings meaning to particular data. The types of attributes of elements of these sets and test indicators are shown in Fig. 3. For creating computer learning systems, it must be really important to understand that each information object could represent a different level of detail (or abstraction) with different levels of information detail inside the structure.

122

N. Shybytska

Fig. 2. The structure model of knowledge representation

Fig. 3. Types of relations between elements

Concerning the system of content units of knowledge, which corresponds to the principle of information and logical completeness, test tasks are developed. Each question of the test task is of the following types: – conceptual questions “who, what” and “define the concept; – conceptual and analytical questions “define the analogy”; – conceptual and semantic tasks in order to determine the semantic equivalence or proximity of concepts and situations; – factual questions “what - where - when” to determine the location of the object S and the time t of its existence. Let the studied object (phenomenon or situation) S1 be connected through the system-forming operator ψ1 with the object S at time t, therefore, we can write: S(t)ψS1(t).., Sn(t);

(1)

– functional (target or causal) questions “for what - why” - to study the characteristic changes of an object in the chain of cause-and-effect relationships: S → S1(t − 2) → S1(t − 1) → S1(t) → S(t + 1) → . . . ;

(2)

– structural questions “of what” (or what parts the object consists of) - to analyze the appearance of the object S1, its structure and its components S11, S12, . . . , S1n,

Object-Oriented Model of Fuzzy Knowledge Representation

123

which, using the operator χ , form a whole: Sχ S1, S2, . . . , Sn;

(3)

– problem - operational tasks in order to solve the problem; – quantitative questions “how much - how much” - to identify knowledge about the quantitative properties of objects and their application under various conditions and situations. To assess the degree of matching of knowledge or skills of a group of student’s subset Y ⊂ X to the required knowledge subset Z ⊂ X , the concept of fuzzy function μZ (xi ) inheritance is proposed. The fuzzy function μZ (xi ) inheritance describes the process of accumulation by the student of a subset of Y ⊂ X didactic knowledge or skills and evaluates the contribution of xi elements from the set of X - descendants to the formation of the element of set Z of knowledge - the parent by the formula: μZ (xi ) =

N 

μy (xi ) · ξV (xi ),

(4)

i=1

where xV (xi ) is the relative weight of the binary inheritance relation in a pair of elements (xi , zi ), calculated by the formula: μV (xi ) . ξV (xi ) = N i=1 μV (xi )

(5)

Let’s look at these formulas in more detail. For a fixed element z of the set of knowledge - the ancestor will expertly set the structure of the inheritance relation in the form of the function: v = H (xi , zi ), where i = 1, N .

(6)

We will assume that the set of v relations between the elements xi , zi knowledge of neighboring levels is represented by the weights of the values of the function μZ (xi ), given expertly at the stage of structuring knowledge and building a tree of goals (skills). The values of the coefficients μZ (xi ) for elements xi → zi ∈ Z inheritance relations are given as follows: – μZ (xi ) = 0, if the parent-child relationship xi → zi ∈ Z is not defined; – 0 < μZ (xi ) 30 places and events through the created data administration system for “Entertainment Planner” recommendation system. Table 1. Accuracy comparison of classic method for providing recommendations and MOA combinations of the proposed method for sampling ~6,000 recommended events Method

Accuracy

Classic (MLP + BP)

76.68%

Proposed (MLP + MOA)

84.61%

Proposed (MLP + PSO)

44.27%

Proposed (MLP + GWO)

67.81%

Proposed (MLP + WOA)

65.52%

Proposed (MLP + GWO/WOA)

70.34%

The classic method implements a neuro-collaborative filter model using multilayer perceptron technology with back propagation method, the sample results showed high results compared to standard, non-personalized systems, such as the classic implementation of the collaborative filter, the subject filter. The accuracy of the method is 76.68%. The proposed method implements a neuro-collaborative filter model with optimization modification using MOA set, which, in turn, use the vector technology of neural

160

M. Vernik and L. Oleshchenko

network decomposition. In this case a multilayer perceptron is used instead of classical stochastic gradient descent. Metaheuristic optimization methods have a fairly fast convergence and the possibility of getting out of the local traps of the function. The proposed method uses a set of 12 MOA algorithms, the most popular of which are PSO, GWO, WOA and the hybrid GWO/WOA algorithm. These algorithms have a competent capacity compared to most other modified algorithms. The result of the proposed method is increased accuracy in contrast to the classical implementation, which is 84.61%, which is more accurate by almost 6% compared to the classical implementation. The proposed implementation of the neuro-collaborative filter using PSO [21] algorithm shows an accuracy result of 44.27%, which is quite small compared to the above implementation methods of the proposed method and compared to the classical implementation. Although this method uses personalized recommendations with metaheuristic optimization anyway does not give a high accuracy rating due to the simplicity and weakness of the algorithm on complex functions. The proposed implementation of a neuro-collaborative filter using the Gray Wolf metaheuristic optimization (GWO [20]), which, in turn, is one of the very popular metaheuristic algorithms of recent years. The algorithm itself has demonstrated considerable effectiveness in solving a number of optimization problems in various areas of constructive engineering. In this method, the accuracy result is 67.81%, which differs from the classical one by about 9%. This result has the right to live with the use of various types of modifications of the method, which can provide greater accuracy to the proposed method, such as the implementation of the hybrid Gray Wolf and Blue Whales algorithms (GWO/WOA [22]), which, in turn, confirmed the hypothesis of an increase in accuracy when using modifications to the GWO [20] algorithm up to 70.34%. The reliability and accuracy of the results have been provided by using a dataset of 6,000 real events, as a proof of concept of the proposed method, on the basis of a representative sample an 8% increase in accuracy was obtained from the classical method MLP + BP 76.68% to proposed MLP + MOA 84.61%. Disadvantage of using the proposed software method is time costs, since the neural network is decomposed into a vector of weights and displacement weights, and processing and superimposition of activation functions and other matrix operations takes place over it, which in general have an O(N^3) time estimate, which significantly slows down the processing process on large neural networks.

4 Summary and Conclusion For the first time, it found further development, the superposition of methods proposed by the authors “MLP + MOA” improves accuracy for providing recommendations to users of the recommender system, which is better compared to existing software solutions in the recommendation systems field. The model for learning and optimizing neural networks is based on metaheuristic optimization algorithms, which, in turn, have advantages in solving specific tasks compared to accurate optimization methods. In particular, an advantage in the speed of convergence and the advantage of being able to avoid local traps (depending on the task, these can be both minima and maxima). Disadvantage of using the proposed software method is time costs for data processing.

Metaheuristic Optimization Algorithms Usage in Recommendation System

161

The proposed software method uses as a basis a neuro-collaborative filter with metaheuristic optimization methods, which ensures fast convergence and finding a way out of local traps of the function, thanks to which the result of improving the accuracy of the user’s selection of events by 6–8% when using the full set of MOA algorithms compared to available methods. Based on the received data, which, in turn, allows improving the accuracy of user analysis, the proposed method provides a reduction in the complexity of calculations and provides greater accuracy than the classical implementation. The modular structure of the proposed software, the modules of which implement the above-mentioned algorithms, greatly facilitates the implementation of the software product as a whole. In the future, based on the obtained results, can be created universal approach of a self-developing recommendation system with super positioning optimization algorithms. Proposed method can change the classical realization of recommender systems development, as well as apply new technologies to optimize neural networks. The analyzed evaluation results and comparison of the efficiency of similar solutions can be used for various applied problems, where a choice of possible options is implemented.

References 1. Ghazanfar, M.A., Prügel-Bennett, A., Szedmak, S.: Kernel-mapping recommender system algorithms. Inf. Sci. 208, 81–104 (2012). https://doi.org/10.1016/j.ins.2012.04.012. CiteSeerX 10.1.1.701.7729 2. Aggarwal, C.C.: Recommender Systems: The Textbook. Springer, Cham (2016). https://doi. org/10.1007/978-3-319-29659-3. ISBN 9783319296579 3. Brusilovsky, P.: The Adaptive Web, p. 325 (2007). ISBN 978-3-540-72078-2 4. Wang, D., Liang, Y., Xu, D., Feng, X., Guan, R.: A content-based recommender system for computer science publications. Knowl.-Based Syst. 157, 1–9 (2018). https://doi.org/10.1016/ j.knosys.2018.05.001 5. Blanda, S.: Online Recommender Systems – How Does a Website Know What I Want? American Mathematical Society. Accessed 31 Oct 2016 6. Feng, X.Y., et al.: The deep learning-based recommender system “Pubmender” for choosing a biomedical publication venue: development and validation study. J. Med. Internet Res. 21(5), e12957 (2019) 7. Chen, M., Beutel, A., Covington, P., Jain, S., Belletti, F., Chi, E.: Top-K Off-Policy Correction for a REINFORCE Recommender System (2018). arXiv:1812.02353 [cs.LG] 8. Yifei, M., Narayanaswamy, B., Haibin, L., Hao, D.: Temporal-contextual recommendation in real-time. In: KDD 2020: Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 2291–2299. Association for Computing Machinery (2020) 9. Hidasi, B., Karatzoglou, A.: Recurrent neural networks with top-k gains for session-based recommendations. In: Proceedings of the 27th ACM International Conference on Information and Knowledge Management, CIKM 2018, pp. 843–852. Association for Computing Machinery, Torino (2018). https://doi.org/10.1145/3269206.3271761. arXiv:1706.03847. ISBN 978-1-4503-6014-2. S2CID 1159769 10. Kang, W.-C., McAuley, J.: Self-Attentive Sequential Recommendation (2018). arXiv:1808. 09781 [cs.IR]

162

M. Vernik and L. Oleshchenko

11. Li, J., Ren, P., Chen, Z., Ren, Z., Lian, T, Ma, J.: Neural attentive session-based recommendation. In: Proceedings of the 2017 ACM on Conference on Information and Knowledge Management, CIKM 2017, pp. 1419–1428. Association for Computing Machinery, Singapore (2017). https://doi.org/10.1145/3132847.3132926. arXiv:1711.04725. ISBN 978-1-4503-4918-5. S2CID 21066930 12. Liu, Q., Zeng, Y., Mokhosi, R., Zhang, H.: STAMP: short-term attention/memory priority model for session-based recommendation. In: Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD 2018, pp. 1831–1839. ACM, London (2018) 13. Xin, X., Karatzoglou, A., Arapakis, I., Jose, J.: Self-Supervised Reinforcement Learning for Recommender Systems (2020). arXiv:2006.05779 [cs.LG] 14. Ie, E., et al.: SlateQ: a tractable decomposition for reinforcement learning with recommendation sets. In: Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence (IJCAI-19), pp. 2592–2599 (2019) 15. Zou, L., Xia, L., Ding, Z., Song, J., Liu, W., Yin, D.: Reinforcement learning to optimize long-term user engagement in recommender systems. In: KDD 2019: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD 2019, pp. 2810–2818. https://doi.org/10.1145/3292500.3330668. arXiv:1902.05570. ISBN 9781450362016. S2CID 62903207 16. Lakiotaki, K., Matsatsinis, N.F., Tsoukias, A.: Multicriteria user modeling in recommender systems. IEEE Intell. Syst. 26(2), 64–76 (2011). https://doi.org/10.1109/mis.2011.33. S2CID 16752808. CiteSeerX 10.1.1.476.6726 17. Yong, G., Hui, X., Alexander, T., Keli, X., Marco, G., Pazzani, M.J.: An energy-efficient mobile recommender system (PDF). In: Proceedings of the 16th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 899–908. ACM, New York (2010). Accessed 17 Nov 2011 18. Pimenidis, E., Polatidis, N., Mouratidis, H.: Mobile recommender systems: identifying the major concepts. J. Inf. Sci. 45(3), 387–397 (2018). https://doi.org/10.1177/016555151879 2213. arXiv:1805.02276. S2CID 19209845 19. Gomez-Uribe, C.A., Hunt, N.: The Netflix recommender system. ACM Trans. Manag. Inf. Syst. 6(4), 1–19 (2015). https://doi.org/10.1145/2843948 20. Mirjalili, S., Mirjalili, S.M., Lewis, A.: Grey wolf optimizer. Adv. Eng. Softw. 69, 46–61 (2014). https://doi.org/10.1016/j.advengsoft.2013.12.007. ISSN 0965-9978 21. Trivedi, I.N., Jangir, P., Kumar, A., Jangir, N., Totlani, R.: A novel hybrid PSO–WOA algorithm for global numerical functions optimization. In: Bhatia, S.K., Mishra, K.K., Tiwari, S., Singh, V.K. (eds.) Advances in Computer and Computational Sciences. AISC, vol. 554, pp. 53–60. Springer, Singapore (2018). https://doi.org/10.1007/978-981-10-3773-3_6 22. Mirjalili, S., Lewis, A.: The whale optimization algorithm. Adv. Eng. Softw. 95, 51–67 (2016). https://doi.org/10.1016/j.advengsoft.2016.01.008. ISSN 0965-9978 23. Jadhav, S.L., Mali, M.P.: Pre-recommendation clustering and review based approach for collaborative filtering based movie recommendation. Int. J. Inf. Technol. Comput. Sci. (IJITCS) 8(7), 72–80 (2016). https://doi.org/10.5815/ijitcs.2016.07.10 24. Raghavendra, C.K., Srikantaiah, K.C., Venugopal, K.R.: Personalized recommendation systems (PRES): a comprehensive study and research issues. IJMECS 10(10), 11–21 (2018). https://doi.org/10.5815/ijmecs.2018.10.02 25. Neysiani, B.S., Soltani, N., Mofidi, R., Nadimi-Shahraki, M.H.: Improving performance of association rule-based collaborative filtering recommendation systems using genetic algorithm. IJITCS 11(2), 48–55 (2019). https://doi.org/10.5815/ijitcs.2019.02.06 26. Soto, C., Jackson, J.: Five-factor model of personality. In: Dunn, D.S. (ed.) Oxford Bibliographies in Psychology. Oxford, New York (2020). https://doi.org/10.1093/obo/978019982 8340-0120

Interest Point Detection at Digital Image Based on Averaging of Function Iryna Yurchuk(B) and Ivan Kosovan Software Systems and Technologies Department, Taras Shevchenko National University of Kyiv, Kyiv, Ukraine [email protected]

Abstract. Digital image processing is a fundamental task of computer vision. It can be realized with different purposes, in particular, images can be compared, certain segments are tracked, image quality changes, etc. There are many methods and approaches for processing, but methods that can make inferences about an image based on individual pixels, which are the individual card of an image, are valuable and useful. In particular, one of these concepts is interest pixels (points). The search for them is based on detectors at all. In this paper, the mathematical basis with further its application into the algorithm to research a pixel as an interest point of a digital image is obtained. It is based on the conditions of topological persistence of averaging of a function of two variables. Also, the authors implement this algorithm, test it on grey-scaled images, and compare it with the Harris detector. Keywords: Parameter-dependent integral · Interest point · Digital image · Averaging · Topological persistence

1 Introduction Nowadays, there are some points of view on conditions to pixel has to satisfy to be an interest point [1]. In general, it means any point in the image for which the signal changes two-dimensionally. It can be X − corner (T − corner, Y − corner, etc.), a black dot on white, an ending of a branch, etc. Also, there are three categories of methods to detect an interest point. They are based on contour (extract contours that have to be approximated, search for maximal curvature, etc., and search for intersection points), intensity (compute a measure that distinguishes a point from others based on calculations in many directions of discrete analogs of derivatives) or parametric model (fit a parametric intensity model to the signal). Each of them has its advantages and disadvantages, which are taken into account when applied in a certain subject area. For example, the parametric model is limited to specific types of interest points by the detected features sample such that each of them is closest to the observed signal. Purpose of work – to propose the mathematical basis and its algorithmic realization for the detection of the point of interest based on the intensity model with providing © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 163–172, 2023. https://doi.org/10.1007/978-3-031-36118-0_15

164

I. Yurchuk and I. Kosovan

approximated manifold with the further possibility of its persistent analysis. It facilitates universality due to a combination of both intensity and contour methods of interest point detection. The aim of research – algorithm of detection of an interest point as a point that is persistent (stable) up to a translation and a scaling, taking into account the possibility of changing the value of its neighborhood, and detected interest point has to be distinct, stable, unique, interpreted and local. Major research objectives: • Formulate mathematical statements that provide the conditions for the persistence of a point up to a translation and a scaling. • Based on the obtained results, propose an algorithm for finding a point of interest. • Compare the obtained algorithm with the existing ready-made solution.

2 Literature Review In the field of digital image processing, the averaging method is quite common. In addition, this method can be found by studying the processing of local pixels of one. For example, averaging the variance of neighboring pixels is used to increase the performance [2], and pixel-by-pixel averaging of variations of multiple images has increased the accuracy of predictions to change the color of the pixel in the original image [3]. The averaging of equalization and threshold made it possible to develop an algorithm that automatically corrects the white balance and increases the color contrast. The color range is not overstretched, and the nature of the original image does not change [4]. Averaging is also used to reduce noise, which greatly improves the quality of digital images. As an example, weighted averaging with overcomplete orthogonal transformation fights noise at the level of cutting-edge algorithms [5]. Another example is using the average values of neighboring non-noise pixels to detect and replace noisy pixels in the original image. This approach has shown high efficiency and provides low computational complexity, as it is implemented using iterative processes [6]. In modern realities, it is obvious that portable devices are an integral part of human life. However, such devices have limited memory and computing power, so they must use modern filtering methods, such as a spatial averaging filter [7]. The method of analysis and comparison of templates is also used to solve the problem of resource intensity. It can be used to recognize a particular person’s face in a digital image with several different faces. For example, there is an approach that uses the averaging of half of the face, namely: divides the face in half, determines the features, and recognizes them in other images [8]. This method not only showed greater recognition accuracy but also improved resource efficiency. It should also be noted that sufficient conditions have been derived for the topological stability of the averaging of piecewise smooth functions f : R → R with a finite number of extremes with respect to discrete measures with finite supports [9]. At the current stage of development of digital image processing, there are many algorithms to define the points of interest. As an example, we can take the use of the Dilation Operator [10]. First, the image is divided pixel by pixel into equidistant grids. As a stopping metric, the SpeedThreshold is used, which is calculated as critical values

Interest Point Detection at Digital Image

165

of two manually set speeds. The stopping points and their neighborhoods are the areas of further investigation - we select the area where the number of stopping points is greatest. First, we binarize this area to get rid of unnecessary noise. Using the Dilation operator, we reduce the studied area to the minimum value, which will be the point of interest. Kumar and Pang [11] noted that a pixel of interest can be determined by examining its neighbors, namely, gray-scale grades. However, newer scientific work, such as that by Zhao et al. [12], departs from this principle and uses a multiscale Gabor filter to extract information from pixels in completely different directions. Information about the different scales of grayscale change is used to build an “entropy map” of each pixel. The pixels of interest are extracted directly from the map by entropy normalization. In this way, the work has paved the way for the study of pixels of interest not only for the contour itself but also for the grayscale, as well as increasing the robustness to noise and reducing the time-consuming execution with the right selection of the scale. The authors single out this method for finding pixels of interest as being more suitable for applications such as screening monitoring, sensing, and working with medical images. In addition to working directly with images, local feature search is also used for video files. This is possible because a video is structured as a set of individual images. For example, Li et al. [13] describe a method for finding Spatio-Temporal Interest Points in video files, which is based on calculations of curvature and gradient points. The process of finding points of interest is divided into three steps: intensity calculation using a response function, obtaining candidate points by non-maximal suppression, and imposing constraints.

3 Background Let consider points A1 (x1 , y1 , z1 ), A2 (x2 , y2 , z2 ), A3 (x3 , y3 , z3 ) and O(0, 0, 0) such −−→ −−→ −−→ that for any i, i = 1, 3, a value zi > 0, OA1 , OA2 and OA3 are pairwise non-collinear    and non-coplanar. Also O(0, 0, 0) ∈ Int A1 A2 A3 , where A1 , A2 and A3 are projections of A1 , A2 and A3 onto the xOy coordinate plane, respectively. We define 2-simplex formed by A1 , A2 , A3 and O. It is easy to write the equations of planes α, β and γ such that A1 , A2 , O ∈ α, A2 , A3 , O ∈ β and A1 , A3 , O ∈ γ by using the coordinates of corresponding points. So, we obtain α : a1 x + b1 y + c1 z = 0, β : a2 x + b2 y + c2 z = 0 and γ : a3 x + b3 y + c3 z = 0. Let define a function f (x, y) by the following equalities: ⎧  ⎨ a1 x + b1 y, if (x, y) ∈ D1 ; f (x, y) = a2 x + b2 y, if (x, y) ∈ D2 ; (1) ⎩  a3 x + b3 y, if (x, y) ∈ D3 , where ai = − acii , bi = − bcii and Di = prxOy Ai OAi+1 where A4 = A1 for i = 1, 3. → ni such that It is easy to see that ci = 0. Since it is z− coordinate of a vector −   j   i  k  x   yi  − → → = 0. ni ⊥Ai OAi+1 , we can write − ni =  xi yi zi  and  i xi+1 yi+1  x  i+1 yi+1 zi+1

166

I. Yurchuk and I. Kosovan

By D we denote a closed domain that contains a projection of a triangle A1 A2 A3 onto the xOy coordinate plane. At a finite set of points {Pi (xi ; yi )}m i=1 of D we define a function μ(x, y) such that μ(xi , yi ) = pi , where pi ∈ R+ . m P  ∪ P  ∪ P  , P  ⊂ D1 , P  ⊂ D2 and Let assume that    {Pi (xi ; yi)} i=1 = P =        P ⊂ D3 , also P = m1 , P = m2 and P  = m3 , where m1 + m2 + m3 = m. We can renumber the points Pi : the first m1 points belong to D1 , next m2 points belong to D2 and the last m3 to D3 . Let f (x, y) : R×R → R and α > 0 be a number. Then a function fα (x, y) : R×R → R can be defined by the following formula: ¨ f (x + αt, y + αs)μ(t, s)dtds, (2) fα (x, y) = D

where D is a closed domain, D ⊂ R × R. ˜ We have to remark that if f (x, y) is defined into D, then fα (x, y) is defined into D, ˜ ⊃ D. D Two continuous functions f (x, y) : R × R → R and g(x, y) : R × R → R are called to be topologically equivalent if there exist the preserving orientation homeomorphisms h1 : R × R → R × R and h2 : R → R such that h2 ◦ f = g ◦ h1 . A function f (x, y) is topologically persistent up to averaging to a function μ if there exists ε > 0 such that for all α ∈ (0, ε) the functions f (x, y) and fα (x, y) are topologically equivalent. Let μ(t, s) be a discrete probabilistic measure with a finite support P1 , P2 , · · · , Pm at D. Set μ(Pi ) = μ(xi , yi ) = pi , where pi ∈ R+ and m that i=1 pi = 1. We assume ∞ f : D → R belongs to a class C 1 excepting a set of points S = Si (xi∗ ; yi∗ ) i=1 which ∂f (x,y) ∂f (x,y) i is zero measurable and there exist finite limits Li1 = lim ∂x , R1 = lim ∂x , ∗ ∗ Li2 =

lim ∗

y→yi −0

∂f (x,y) ∂y ,

Ri2 =

lim ∗

y→yi +0

∂f (x,y) ∂y .

x→xi −0

x→xi +0

We remind that a point M0 (x0 , y0 ) is a local minimum (maximum) of function f (x, y) if there exists its neighborhood U (M0 ) such that for any point (x, y), M ∈ U (M0 ), the inequality f (M ) > f (M0 ) is true. For formulating the conditions of function to be topologically persistent up to averaging we use Theorem 5.2., see [9]. (x,y) R1 = lim ∂f ∂x (L2 = x→0+0

∂f (x,y) ∂f (x,y) lim lim ∂y , R2 = x→0+0 ∂y ) are finite and Xj  = 0, where Xj = L1 p1 + . . . + pj + x→0−0





R1 pj+1 + . . . + pm (Xj = L2 p1 + . . . + pj +R2 pj+1 + . . . + pm ), j = 1, . . . , m−1. Then f (x, y) at (0;0) is topologically persistent up to averaging to μ. We only provide the main argument for the proof and omit the technical details. Since ∂f (x,y) ∂x can be considered as a function of one variable, we can apply Theorem 5.2., which contains conditions to be topologically persistent up to averaging to μ for a function of one variable.

Theorem. Let assume that both of limits L1 = lim

x→0−0

∂f (x,y) ∂x ,

Corollary. If f (x, y) is piecewise linear, defined by the formula (1) and values X = L1 pk + R1 pk  and Y = L2 pk  + R2 pk  , where k, k  , k  , k  ∈ {1, 2, 3} and correspond

Interest Point Detection at Digital Image

167

to Di , are defined correctly. Then X and Y have the same substance as Xj in the Theorem and exhaust the whole set of Xj . We have to remark before using Theorem for digital image processing in Sect. 4: • any pixel has integer coordinates, all xi and yi are integer and zi is also integer as its intensity and belongs to the interval [0; 255]; • any pixel has to be translated along the two axes at the origin as a function of two variables.

4 Algorithm An interest point in a digital image is a pixel such that its neighborhood is a blob, namely a region that is associated with at least one local extremum (either a maximum or a minimum), see [14]. In this paper, the authors consider the following conditions on a pixel to be an interest point of an image: • The interest point position is accurate in the image domain; • It is persistent up to local and global deformations in the image domain: according to definitions of fα (x, y) and topologically persistent up to averaging, term stability includes a translation and a scaling; • An interest point should include an attribute of scale so that the interest point can be obtained from the image even when the scale changes. We have to remark that the last condition can be expanded to a concept of multi scaling, see [12, 15]. Let consider the algorithm of researching a pixel as an interest point in a digital image in denotations and terms of the previous section. A point O coincides with a researched pixel X . It is easy to hold a condition to be origin by using a translation. Points A1 (x1 , y1 , z1 ), A2 (x2 , y2 , z2 ) and A3 (x3 , y3 , z3 ) are pixels that belong to a neighborhood of X . These four points have to be checked on conditions of the possibility to construct 2-simplex on them. Then it needs to be checked whether a point O belongs to the internal area of the projection of 2-simplex on the xOy – plane. It is quite simple: first, we read the points, then we look for the vector product of the paired vectors, calculate the sum of the products and check whether it is equal to 3 or −3. After obtaining the required construction, we proceed to the stage of integration with a parameter. In the course of the work, we have identified 16 possible arrangements of a triangle in which one of the vertices coincides with the origin of coordinates in the plane. Then, we have to check the stability criterion theorem: 1. Obtain the coefficients at the variables in the three equations of the planes as parts of 2-simplex, see the formula (1); 2. Define the function μ(xi , yi ) = pi , i = 1, 2, 3 by using pixels that belong to part Di i of a neighborhood of O. Let {Ks }N s=1 be a set of pixels that belong to Di . Then, we

168

I. Yurchuk and I. Kosovan

set for a grey-scaled image: Ni pi =

s=1 I (Ks )

255 Ni

,

(3)

where I (Ks ) is an intensity of Ks and i = 1, 2, 3; 3. Calculate values X and Y , according to the corollary of the Theorem; 4. Check whether either X = 0 or Y = 0. If the statement is true, pixel O is stable. The final check is to check for linearity preservation. The planes obtained by integration are passed through the construction check again. We define those points that passed all the described steps as points of interest and plot them on the image. A parameter α at formula (2) can be interpreted as a learning rate. As usually in machine learning it must be manually chosen for each given type of digital image (landscape, portrait, MRI, etc.) and may vary greatly depending on it.

Fig. 1. Algorithm block diagram: the first condition is followed from pairwise non-collinear and non-coplanar of vectors which are the basis of the construction; f (x, y) and fα (x, y) are defined by formulas (1) and (2), respectively

In Fig. 1, there is a representation of the algorithm. It has a starting point in terms of the image to be processed. The processing consists of the alternate selection of three random neighbors of each pixel of the image, followed by checking the possibility of constructing a 2-simplex on the given points (see the first condition in the block diagram), as well as calculating the averages. The second condition is checking of the Corollary. If the conditions of the Corollary are satisfied, then a point is an interest point. At the output of the algorithm, we have an image with selected points of interest. We have to remark that the last condition checks the preservation of 2-simplex.

Interest Point Detection at Digital Image

169

5 Illustrative Examples and Testing Certain processing results were obtained for images of size 45 × 45. For the first test, a handwritten symbol similar to the “+” sign was chosen. It can be seen that the number of points found is greater than the Harris detector (Fig. 2). We remind that the Harris detector extracts corners (a corner is a junction of two edges, where an edge is a sudden change in image brightness) and infers interest point. It is a first-order derivative-based interest point detection. According to the definition of interest point, the stability of constructed 2-simplex in obtained algorithm can be compared with the Harris detector correctly.

Fig. 2. The first comparison of the work of the algorithms. On the left is the author’s one, and on the right is the Harris detector

Next, another handwritten symbol, “∀”, was chosen. One pass over the image showed the same result as in the first test (Fig. 3). Much more interesting is the picture after several passes (Fig. 4). You can see that almost the whole figure is highlighted as pixels of interest. From this, we can conclude that more research needs to be done in the field of use as a contour detector.

Fig. 3. The second comparison of the work of the algorithms. On the left is the author’s one, and on the right is the Harris detector

On the larger image, we got a slightly worse result because of the number of points found (Fig. 5). This can be explained by the fact that the Harris method examines the image through a sliding window, while our method goes through each pixel individually. Let use a visual inspection to summarize distinct, stable, unique, interpreted and local properties of interest points: the more passes of the algorithm, the less distinct points are; the stability does not depend on passes and is defined by α as a parameter; the more passes of the algorithm, the less unique points are; the interpretability and locality are satisfied and do not depend on passes.

170

I. Yurchuk and I. Kosovan

Fig. 4. Several passes of the algorithm over the image with the results saved. On the left is the author’s one, and on the right is the Harris detector

The listed properties can be both advantages and disadvantages depending on the field of application. Obviously, if there are many small items (coins, etc.) in the scene, then it is necessary to reduce the uniqueness, which can be done with a larger number of passes. Let summarize the obtained results: 1) Compared to the Harris detector, the obtained algorithm finds more interest points. Several passes of the algorithm over the image enable to increase of interest points set. 2) During the testing, there are no false positives, in particular, background points not specified as interest points. To compare the performance of the algorithms, let summarize the obtained values in the table (Table 1). Table 1. Algorithm performance comparison Algorithm

Image size 45 × 45

255 × 255

1001 × 1001

The suggested algorithm

4.931a

13.392a

52.844a

Harris detector

0.022a

0.123a

0.482a

Fig. 5. Algorithm passing over the larger image. On the left is the author’s one, and on the right is the Harris detector

Interest Point Detection at Digital Image

171

6 Summary and Conclusion The detection interest point algorithm based on topological persistence up to averaging to a two-variable function includes the possibility of constructing an approximation manifold in the neighborhood of the point using a 2-simplex, whose stability is respective up to the translation and scaling transformations. The basis of the algorithm is a criterion determined by partial derivatives. The author’s algorithm was tested on grey-scaled images and compared with the Harris detector as a detection interest point algorithm based on the intensity model. The most advantage is a large enough amount of detected points. The most disadvantage of it is the long processing time. In further research, the digital image processing based on the obtained algorithm will be adapted to video files with real-time processing. It needs additional research possibility to increase the number of considered neighbors (pixels) on which 2-simplex is constructed. In addition, the calculation of scaling coefficients which are realized as average values of the intensity of pixels in a neighborhood can be improved.

References 1. Jing, J., Gao, T., Zhang, W., Gao, Y., Sun, C.: Image Feature Information Extraction for Interest Point Detection: A Comprehensive Review. ArXiv, abs/2106.07929 (2021) 2. Sritapanu, S., Amornraksa, T.: Adaptive partial image watermarking based on the averaged variance of neighbor pixels. In: Proceedings of the International Symposium on Intelligent Signal Processing & Communication Systems, pp. 1–4. IEEE, Chengdu (2010). https://doi. org/10.1109/ISPACS.2010.5704661 3. Peungpenich, A., Amornraksa, T.: An improving method for image watermarking using image averaging and tuned pixels prediction. In: Proceedings of the 10th International Symposium on Communications & Information Technologies, pp. 755–760. IEEE, Tokyo (2010). https:// doi.org/10.1109/ISCIT.2010.5665090 4. Tai, S.-C., Liao, T.-W., Chang, Y.-Y., Yeh, C.-P.: Automatic White Balance algorithm through the average equalization and threshold. In: Proceedings of the 8th International Conference on Information Science & Digital Content Technology (ICIDT 2012), pp. 571–576. IEEE, Jeju (2012) 5. Xu, J., Lin, F.: Hybrid multi-resolution analysis and weighted averaging of overcomplete orthogonal transform scheme for digital image denoising. In: Proceedings of the International Conference on Multimedia & Expo Workshops (ICMEW), pp. 1–6. IEEE, Chengdu (2014). https://doi.org/10.1109/ICMEW.2014.6890575 6. Konieczka, A., Balcerek, J., D˛abrowski, A.: Method of adaptive pixel averaging for impulse noise reduction in digital images. In: Proceedings of the Baltic URSI Symposium (URSI), pp. 221–224. IEEE, Poznan (2018). https://doi.org/10.23919/URSI.2018.8406738 7. Tsai, C.-M., Yeh, Z.-M.: Fast spatial averaging filter. In: Proceedings of the International Symposium on Computer, Consumer & Control, pp. 153–156. IEEE, Taichung (2012). https:// doi.org/10.1109/IS3C.2012.47 8. Arora, S., Chawla, S.: An intensified approach to face recognition through average half face. In: IEEE India Conference (INDICON), pp. 1–6. IEEE, New Delhi (2015). https://doi.org/ 10.1109/INDICON.2015.7443802 9. Maksymenko, S., Marunkevych, O.: Topological stability of the averages of functions. Ukr. Math. J. 68(5), 707–717 (2016)

172

I. Yurchuk and I. Kosovan

10. Qing, W., Ding, C., Fu, K., Ren, W.: Interest point detection method based on dilation operation. Acta Armamentarii 38(10), 2041–2047 (2017) 11. Kumar, A., Pang, G.K.H.: Defect detection in textured materials using optimized filters. IEEE Trans. Syst. Man Cybern. B Cybern. 32(5), 553–570 (2002) 12. Zhao, Z., Li, B., Chen, L., Xin, M., Gao, F., Zhao, Q.: Interest point detection method based on multi-scale Gabor filters. IET Image Proc. 13(12), 2098–2105 (2019) 13. Li, Y., Xia, R., Huang, Q., Xie, W., Li, X.: Survey of spatio-temporal interest point detection algorithms in video. IEEE Access 5, 10323–10331 (2017). https://doi.org/10.1109/ACCESS. 2017.2712789 14. Lindeberg, T.: Detecting salient blob-like image structures and their scales with a scale-space primal sketch: A method for focus-of-attention. Int. J. Comput. Vis. 11(3), 283–318 (1993) 15. Satyabrata, S., Santosh, K.N., Tanushree, M.: Digital image texture classification and detection using Radon transform. Int. J. Image Graph. Sig. Process. (12), 38–48 (2013)

Analyzing Ukrainian Media Texts by Means of Support Vector Machines: Aspects of Language and Copyright Maksym Lupei1(B) , Oleksandr Mitsa2 , Vasyl Sharkan3 , Sabolch Vargha2 , and Nitsa Lupei4 1 Department of Intelligent Information Technologies, Institute of Cybernetics,

National Academy of Sciences of Ukraine, Kyiv, Ukraine [email protected] 2 Department of Informative and Operating Systems and Technologies, Uzhhorod National University, Uzhhorod, Ukraine 3 Department of Journalism, Uzhhorod National University, Uzhhorod, Ukraine 4 Department of Theory and History of State and Law, National Academy of Management, Kyiv, Ukraine

Abstract. The paper focuses on the possibilities of studying Ukrainian online media publications using the support vector machine. Through a database including about 95,000 texts of three Ukrainian online media («Censor.NET», «Ukrainska Pravda», «Zakarpattya Online») within the period from September 2021 to August 2022 there were traced possibilities for identifying the authorship of a media text (belonging to certain media). It was established that the accuracy of determining whether a text belongs to online media (regional, national) using SVM is significantly influenced by the presence in published texts keywords included in the names of the media: if these words remain in the dataset, the accuracy of identification of the media through the text fragment is 0.979, if these words are removed, the accuracy drops to 0.930. This means that the high frequency of use of certain words (primarily related to the name of the publication, which is typical for the texts of Ukrainian online media) significantly affects the results of the experiment. A more detailed analysis of language elements employing SVM makes it possible to conclude that the identification of the media occurs primarily on the basis of linguistic units of the lexical level. Therefore, for the analysis of texts written in the Ukrainian language, which has an inflectional system, the stemming procedure for variable words is essential. The study also considered the legal aspects of authorship determination using the proposed machine learning method and highlighted new legal challenges when authorship determination is associated with automated journalism, that is, the creation of texts using artificial intelligence. Keywords: Machine learning · support vector machine · definition of authorship · legal regulation · copyright protection · artificial intelligence · mass media

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 173–182, 2023. https://doi.org/10.1007/978-3-031-36118-0_16

174

M. Lupei et al.

1 Introduction In today’s world, machine learning is intended to solve tasks related to large volumes of data (e.g. with large text corpora or all the words of a certain language in general) and for which there is no clear algorithm (e.g. the imitation of the work of the human brain, which has so far not yet been studied fully). Trying to solve such tasks with the help of algorithms, which are hard-coded into the program’s source code, would make it too complex to both design and execute such a program even using a very powerful computer. It is in situations like these that machine learning methods are necessary. In connection with the rapid development of artificial intelligence, the limits of its application are expanding. For instance, in the field of journalism, when creating various types of content and texts with the use of artificial intelligence, new legal discussions arise, since the perception of missing authorship can also lead to the perception of the disappearance of responsibility [1]. Modern studies of robotics in media news [2] distinguish three areas in which such robotics is used: content creation, news gathering (gathering, sorting and verifying information from sources) and news distribution (personalized news and advertising). Journalistic materials are special from a copyright perspective, as some news items are not protected by copyright law. To illustrate, the Berne Convention for the Protection of Literary and Artistic Works (1979) does not protect «news of the day or various facts which are of the nature of press information only» [3]. The same applies in Ukrainian national legislation. The reason for this is to prevent the violations of copyright law to monopolize information. The mere nature of information or news cannot be considered «creative» and thus protected by copyright law. However, there are other legal tools that allow stakeholders to protect their investment in news production, such as protection against unfair competition or unjust enrichment. The development of systems for automated journalism can also be protected as a know-how by the trade secret. The issue of copyright protection for works created with the help of computers appeared in legal literature as early as the 1960s [4]. Since then, quite a large amount of literature has explored this issue from different points of view. There currently exist two legislative models in the world: a) special protection of computer-generated works (adopted, e. g. by Great Britain), b) requirement for a person’s original creative activity [5]. Copyright protection has traditionally proven itself as a valuable regulatory tool. However, when revolutionary technologies are challenged, unprecedented questions arise. Given the global and pervasive impact of artificial intelligence on our society, legislation now more than ever needs to be more supportive of ethical behavior. Ethical design and ethical use of artificial intelligence systems are some of the priorities of the European Union. To remain competitive with the rest of the world, the EU aims to ensure user trust in artificial intelligence (AI) systems by guaranteeing that these systems developed in the EU provide their users with the safeguards for protection of human rights. In particular, «the EU seeks to remain true to its cultural preferences and highest standards of protection against social risks posed by artificial intelligence, inter alia, those affecting privacy, data protection and discrimination rules, unlike other more lenient jurisdictions» [6].

Analyzing Ukrainian Media Texts by Means of Support Vector Machines

175

Thus, in April 2021, the European Commission suggested new rules and actions aimed at turning Europe into a global center for reliable artificial intelligence. The combination of the first ever AI legislative framework and the new coordinated plan with the member states is meant to guarantee safety and fundamental human rights while enhancing the expansion of AI, investment and innovation across the European Union [7]. New technology regulations will complement this approach by adapting safety rules to increase user trust in a new, versatile generation of products. Since copyright is not supposed to apply to materials that are of the nature of ordinary press information, in many cases the question of authorship of specific publications that belong to the genre of informative notes is not as relevant as it is with respect to authored materials of other genres. However, there is a separate category consisting of publications whose sources are the materials of news agencies, the use of which requires special permissions. For example, in many Ukrainian publications, particularly, on one of the most popular Ukrainian news websites «Ukrainska Pravda» («The Ukrainian Truth») and «Censor.NET», there is a warning about the prohibition of copying, reprinting information containing links to the agencies «Interfax-Ukraine», «Ukrainski Novyny» («The Ukrainian News»). To detect the unauthorized use of such materials, special technologies are required, and the usual identification of the sameness of text fragments may not be sufficient in this case. The purpose of this study is to analyze the possibilities of SVM in determining the belonging of a Ukrainian publication to certain media based on language analysis and taking into account the legal aspects of authorship. To realize the goal, the following tasks must be completed: 1) create a dataset of Ukrainian media publications, which will allow us to draw conclusions about the reliability of using SVM to identify the text belonging to a certain media; 2) to determine with the help of a linear SVM model the accuracy of identification of text belonging to one or another media; 3) find out the prospects of using the proposed approach to solve the problem of determining the authorship of media publications.

2 Review of the Literature Machine learning is effective for solving tasks related to analyzing data from the Internet – publications from social networks, articles in scientific journals, publications on news portals, in particular, with regard to determining their authorship [8]. Authorship identification is important in order to preserve intellectual property rights, prevent theft, and ensure that each article is referenced to its specific author. This enables institutions to provide identification data about the author [9]. It is also worth highlighting that the use of machine learning is relevant with regard to recognizing fake news, which is used in politics for various kinds of propaganda, in social networks such as Facebook, Twitter etc. [10]. Authorship identification programs have been recently developed in many fields, such as cybercrime law [11], opinion mining (also known as sentiment analysis) [12], etc. Artificial intelligence is also part of cryptographic detection, signature detection and intrusion detection. The biggest challenge of authorship identification

176

M. Lupei et al.

is the selection of the most important features representing the author’s writing style. The possibility to single out the most important characteristics can ensure an accurate determination of authorship. The approaches to authorship identification can combine accumulated knowledge from the theory of image recognition, mathematical statistics and probability theory, neural networks, cluster analysis, Markov chains, and others. It should be noted that Support Vector Machines (SVM) gives good results when solving the problem of determining authorship. This is a popular and powerful tool in the field of machine learning. It is a type of supervised learning algorithm that can be used for both classification and regression problems. Support Vector Machines (SVM) was effective for sentiment analysis in social networks [13]. In the following research [14], SVM was used with a linear kernel using the One-Vrest strategy to classify blogs, tweets, and other documents. In the research work [15] a novel up-selling and appetency prediction scheme is presented based on a support vector machine (SVM) algorithm using linear and polynomial kernel functions. In the paper [16] new combined classification method is proposed using a bagging classifier in conjunction with a support vector machine as the base learner. It should be noted that the choice of the classifier for the bagging is problem-dependent. E.g., it is possible to employ neural networks with bithreshold activations [17] instead of SVM for some identification task. Sometimes this approach can provide better performance [18].

3 Dataset of Publications for the Analysis Articles from three Ukrainian digital media were used for the analysis: «Ukrainska Pravda» (https://www.pravda.com.ua/), «Censor.NET» (https://censor.net/) and «Zakarpattya Online» (https://zakarpattya.net.ua/). About 95,000 publications of these media for the period from September 2021 to August 2022 were stored in the local database (all publications were presented on the sites within this period): • about 37,000 publications from «Ukrainska Pravda», • about 50,000 publications from «Censor.NET», • about 8,000 publications from «Zakarpattya Online». To compare the results, there were selected two media of the all-Ukrainian level («Ukrainska Pravda», «Censor.NET») which use the materials of the information agencies «Interfax-Ukraine», «Ukrainski Novyny» (in total, the publications included in the dataset contain about 4,500 references to these news agencies), and one regional media («Zakarpattya Online») which does not use materials of news agencies as a direct source. Since the vast majority of publications of these Internet media are published without specifying the authorship and authors are different persons, yet publications are often made on the basis of press releases, it is difficult (and sometimes impossible) to trace the peculiarities of the language style of the publications with the help of traditional analysis. In this regard, the data obtained with the help of SVM can be interesting for a deeper linguistic analysis – relating to the peculiarities of the language of certain media, as well as for analysis from the copyright point of view – concerning the use of news agencies as a source of materials.

Analyzing Ukrainian Media Texts by Means of Support Vector Machines

177

The language of some of these media has already been the subject matter of the study conducted by linguists. Most of the publications deal with the functioning of the Ukrainian language units in the «Ukrainska Pravda» media, this topic has been systematically studied by researcher M. Navalna (publication [19] and others), certain aspects have been studied by Yu. Makarets [20], N. Polishchuk [21]. Various aspects of the functioning of the language of the Ukrainian mass media were analyzed in the collective monograph edited by M. Navalna [22]. These studies were conducted using traditional linguistic methods. The language of «Censor.NET», «Zakarpattya Online» has not yet been in the research spotlight, inter alia, in a comparative aspect. The study of the Ukrainian media «Ukrainska Pravda» and «Politeka» was carried out employing SVM through publication [23].

4 Study Methods SVMs work by finding a hyperplane in a high-dimensional space that maximally separates different classes. This hyperplane is called the maximal margin hyperplane. The distance between the hyperplane and the closest data points of each class is called the margin. The purpose of SVM is to find a hyperplane with the maximum margin. The support vector classification (SVC) is the type of support vector machine (SVM) which is employed for classification tasks. SVC aims to find a hyperplane in a highdimensional space that maximally separates different classes. Once the hyperplane is found, new data points can be classified by assigning them to the class in which they are closest to the hyperplane. One of the key advantages of SVC is its ability to process non-linearly partitioned data. This is achieved by using kernel functions that map the data points into a higher dimensional space where they can be linearly separable. The common kernel functions include linear, polynomial and radial basis function (RBF). The choice of kernel function can significantly affect SVC performance, so it is important to carefully consider which kernel to use. Another advantage of SVC is its noise resistance. Because it relies on only a subset of data points, called support vectors, to define the hyperplane, it is less sensitive to the presence of outliers. This makes SVC a good choice for tasks where the data may be noisy or have some degree of uncertainty. In spite of its advantages, SVC has some shortcomings. One shortcoming is that it can be sensitive to the choice of parameters such as the cost parameter C and the kernel function. It is of importance to carefully set these parameters in order to obtain good performance. Moreover, SVC can be time-consuming, especially when working with large datasets. Overall, SVC is a powerful and widely used machine learning tool that can be applied to a variety of classification tasks. Its ability to process non-linearly separable data and its noise resistance make it a perfect choice for many programs. However, it is essential to address its restrictions and to thoroughly set its parameters to achieve good performance.

178

M. Lupei et al.

The goal of SVC is to find a hyperplane in high-dimensional space which maximally separates different classes. Mathematically, it can be represented as follows: wT x + b = 0,

(1)

where w is the vector of the normal to a hyperplane, x is a data point, b is the bias term. The bias term determines the position of the hyperplane along the normal vector. To classify a new data point, we can use the following formula: y = sign(wT x + b),

(2)

where y is the predicted class label (1 or -−1). If the result of the equation is positive, the data point is classified as the one that belongs to class 1. If the result is negative, the data point is classified as the one that belongs to class −1. To find the hyperplane of the maximum field, we need to maximize the distance between the hyperplane and the nearest data points from any class. This distance is called the margin. Mathematically, the margin can be represented as: margin =

2 , ||w||

(3)

where ||w|| is the norm of the vector of the normal w. In order to maximize the margin, we need to minimize the norm of the normal vector w. This can be achieved through the following optimization problem: 1 minimize ||w||2 , 2

(4)

yi (wT xi + b) >= 1 for all i,

(5)

where yi is the true label of the data point class x i . This optimization problem can be solved by means of different methods such as quadratic programming or gradient descent. Once the optimal values of w and b are found, the maximum field hyperplane can be used to classify new data points. The main parameters and toolsets employed in the computational experiments are listed in Table 1. Table 1. Basic parameters and toolsets Sample

Characters in a text

Converter

Model

95000

0, b = c − z1 . zn − z1 zn − z1

(9)

We substitute az + b for the input variable x in (6). Then, r 

θj (az + b)j =

j=0

r 

γj z j .

(10)

j=0

Having set x1 = c, xj = azj + b, j = 2, n,

(11)

we get x1 = c < x2 < . . . < xn = d ,

r  j=0

θj (xi )j =

r 

γj (zi )j .

(12)

j=0

Remark 6. It is shown in [18] that the coefficients θj and γj , j = 0, r, are interrelated through a nondegenerate system of linear equations with an upper triangular matrix of constraints. This result allowed us [18, 19] to estimate the coefficients θj , j = 0, r, at arbitrary numbers c < d in the UPR problem using only a single set of NOPFs (7).

A Modified Method and an Architecture of a Software

213

3.2 Modified Algorithmic Procedures of the Method for Estimating the Coefficients at Nonlinear Terms of an MPR (1) We illustrate the meaning of the outlined below modified algorithmic procedures using the following example: Y (x1 , x2 , x3 ) = θ0 + θ1 x1 + θ2 x2 + θ3 x3 + θ4 x1 x2 + θ5 x1 x22 +θ6 x14 x2 + θ7 x1 x2 x3 + θ8 x1 x22 x32 + E.

(13)

According to (8), r = 4. We find NOPFs Qj (z), j = 0, 4, with a given accuracy [14] for z1 < z2 < . . . < zn , n > 4. 3.3 Modified Algorithmic Procedure of the First Sub-algorithm of Sect. 2 The modification in the first sub-algorithm of Sect. 2 consists in replacement of the segment [c, d ] with



cxF ...xF , dxF ...xF , il ∈ (14) / j1 , . . . , j m−1 j1

jm−1

j1

jm−1

where we change only the input variable xil in the active experiment, and for each of its steps we choose a nonlinear term of the MPR (1) that has the input variable in the largest degree. For the example (13), the first sub-algorithm implements the first step for the nonlinear term θ6 x14 x2 . The corresponding virtual UPR has the form Y (z) = θ0 + θ1 (az + b) + θ2 x2F1 + θ3 x3F1 + θ4 (az + b)x2F1  2 +θ5 (az + b) x2F1 +θ6 (az + b)4 x2F1 + θ7 (az + b)x2F1 x3F1  2  2 x3F1 + E, +θ8 (az + b) x2F1

(15)

the coefficients a and b are found by (9) for c = cxF1 xF1 , d = dxF1 xF1 . The values of the 2

3

input variable x1 in the active experiment for the real UPR

2

3

  2  y x1i , x2F1 , x3F1 = θ0 + θ1 x1i + θ2 x2F1 + θ3 x3F1 + θ4 x1i x2F1 + θ5 x1i x2F1  2  2 4 F1 x3F1 + εi +θ6 x1i x2 + θ7 x1i x2F1 x3F1 + θ8 x1i x2F1 4 = θ∗0 + θ∗1 x1i + θ∗4 x1i + εi

(16)

are found according to (11). In (16), εi is the realization of the random variable E in the i-th test. Depending on the upper estimate of Var(E) = σ 2 , to find the estimates

214

A. Pavlov et al.

∗



θj , j ≥ 2, with a given variance, the active experiment can be repeated [14]. That is, the input variable value sequence is x1 , . . . , xn , x1 , . . . , xn , …, x1 , . . . , xn and  the results of the active experiment have the form xi → yj+i , i = 1, n, j = 0, p − 1 . The results of a real active experiment are used to design a virtual active experiment based on the input values of the virtual input variable z: z1 , . . . , zn , …, z1 , . . . , zn for the virtual UPR  r j j=0 γj (zi ) + E (15). Taking [14] into account, the experiment is implemented as 

 yi + yn+i + · · · + y(p−1)n+i zi → , i = 1, n . p

(17)



In this case, the variance of estimates γj , j = 0, 4, is reduced by p times compared to   the estimates obtained as the result of the virtual active experiment zi → yi , i = 1, n . The estimate of the coefficient θ6 based on the results of the virtual active experiment   − is equal to θ6 = γ4 / a4 x2F1 . Due to [19], the estimates of the coefficients θ∗j , j =0, 4, obtained using the NOPF built directly from the results of the real active experiment, and those obtained from the virtual active experiment are the same. The second step of the first sub-algorithm is implemented for the nonlinear term θ8 x1 x22 x32 of the MPR (13) where the variable in the active experiment is the input variable x2 . The input variables x1 and x3 take fixed values. At the second step of the first sub-algorithm, we estimate the coefficients for the input variable in the second degree of two UPRs (for different fixed values of the input variables x1 , x3 , i.e., for x1F1 x3F1 , x1F2 x3F2 ) and form a nondegenerate system of two equations, where the variables are the estimates of the coefficients θ5 and θ8 . 



Remark 7. To avoid solving the system of two equations, we could set the variable x3 in the second step and perform the third step of the first sub-algorithm for the term θ5 x1 x22 of the MPR (13). 3.4 Modified Algorithmic Procedure of the Second Sub-algorithm of Sect. 2 The l-th step of the second sub-algorithm is performed for a subsequent term of the MPR (1) which coefficient was not evaluated in the previous steps of the second sub-algorithm and which contains the maximum number of input variables only in the first degree. Remark 8. Such a choice of a nonlinear term of the MPR (1) leads to the fact that the estimate of its coefficient is the solution of a single equation with a single variable. Let this nonlinear term have the form b1...1 i1 ...it

tl  l

xij .

(18)

j=1

In this case, the conditional active experiment is implemented  under

the following 

restrictions. If xjp i = xjFp , i = 1, n, p = 1, m − tl , i1 , . . . , itl ∩ j1 , . . . , jm−tl =

A Modified Method and an Architecture of a Software

215

  ∅, i1 , . . . , itl ∪ j1 , . . . , jm−tl = {1, . . . , m}, then the input variables can take arbitrary values within the ranges

ip ip . (19) xip i ∈ cxF ...xF , dxF ...xF j1

jm−t

j1

l

jm−t

l

The ranges (19) may have an empty intersection. In this case, we replace each input variable xip with xip = aip z + bip , p = 1, tl , xip i = aip zi + bip , i = 1, n

(20)

where aip , bip are found according to (9) where i

c = cxpF ...xF j1

jm−t

i

l

, d = dxpF ...xF j1

jm−t

,

(21)

l

and the real experiment for the UPR has the form   xi1 i , . . . , xitl i , xj1 i = xjF1 , . . . , xjm−tl i = xjFm−t → yi , i = 1, n . l

(22)

Remark 9. In a repeated active experiment with p repetitions, the sequence of input values (22) is repeated p times. We determine the value of p by the restriction on the variance of estimate of the UPR’s coefficient btl at the variable z to the power of tl .   1...1 1 btl + const bi1 ...it = tl (23) l j=1 aip 



where the known value of the constant in (23) is not zero if some nonlinear terms of (1) containing at least one input variable in a degree not lower than two, were transformed in the UPR into terms at z to the power of tl . Remark 10. For the example (13), the UPR at the first step of the second sub-algorithm has the form Y (z) = θ0 + θ1 (a1 z + b1 ) + θ2 (a2 z + b2 ) + θ3 (a3 z + b3 ) + θ4 (a1 z + b1 )(a2 z + b2 ) + θ5 (a1 z + b1 )(a2 z + b2 )2 + θ6 (a1 z + b1 )4 (a2 z + b2 ) + θ7 (a1 z + b1 )(a2 z + b2 )(a3 z + b3 ) + θ8 (a1 z + b1 )(a2 z + b2 )2 (a3 z + b3 )2 + E,

(24) 





the constant in (23) is uniquely determined by the parameters θ5 , θ6 , θ8 , ai , bi , i = 1, 3. Remark 11. The number of steps of the second sub-algorithm given by the number of different terms of an MPR (1) in the form (18) can be reduced by evaluating more than one coefficient of the MPR (1) at some steps by building more than one UPR. Thus, the implementation of the modified sub-algorithms of paragraphs 3.3, 3.4 is guaranteed to allow estimation, at a limited number of steps, of the coefficients at nonlinear terms of an MPR given by a redundant representation for the case when restrictions (3)–(5) are met. This conclusion is based on the fact that the use of the result [18] (see formulas (6)–(11)) in the modified sub-algorithms reduces their theoretical substantiation to the theoretical substantiation (given in [14]) of the sub-algorithms of paragraphs 2.1, 2.2. This is clearly illustrated by the example (13) for which the modified sub-algorithms of paragraphs 3.3, 3.4 are applied in Sect. 3.

216

A. Pavlov et al.

4 Variances of Estimates of MLR Coefficients in the Repeated Experiment As shown in [14], when the estimates for nonlinear terms of an MPR (1) are found sufficiently accurately and the terms with sufficiently small values of coefficient estimates are excluded, the problem of building an MPR given by a redundant representation (1) is reduced to building an MLR given by a redundant representation. We proposed [14] to solve this problem using one of the well-known universal methods. This section presents a theoretical result that can be useful for the implementation of any method of estimating the coefficients of an MLR given by a redundant representation using the general LSM procedure. Suppose we have an MLR in the form Y (x) = θ x + E

(25)

where θ = (θ0 , θ1 , . . . , θm ) , x = (1, x1 , . . . , xm ) , xj , j = 1, m, are input deterministic variables, E is a random variable with ME = 0, Var(E) = σ2 < ∞. The active  experiment has the form xi → yi , i = 1, n where yi = x i θ + εi ,

(26)

εi is the realization of E in the i-th test, xi = (1, x1i , . . . , xmi ) is the vector of input variables in the i-th test. We can consider yi , εi to be realizations of independent virtual random variables Yi , Ei , i = 1, n where Ei are independent virtual copies of E. Then Yi = x i θ + Ei , i = 1, n, or in vector form Y = Aθ + E

(27)

where E = (E1 , . . . , En ) , Y = (Y1 , . . . , Yn ) . Then MY = Aθ, the vector of estimates   θ = θ0 , θ1 , . . . , θm obtained by the LSM has the form









 −1 θ = A A A Y , M θ = θ.





(28)

Remark 11. In the MPR (1) building problem, the vector Y is replaced by the vector. Y ∗ = (Y1 − f (x1 ), . . . , Yn − f (xn ))

(29)

where f (xi ) is the known value of the nonlinear part of the MPR considering the found coefficient estimates for its nonlinear terms and excluding the nonlinear terms that are considered redundant. Suppose that we conduct a repeated active experiment   xi → yi , i = 1, n, xi → yn+i , i = 1, n, . . . , xi → yn(p−1)+i , i = 1, n , −

i.e., we have conducted np tests in which the sequence of input variables xi , i =1, n is repeated p times. Then the following statement holds.

A Modified Method and an Architecture of a Software

217

Statement 1. The Following is True: 

1. The vector θ of estimates of unknown coefficients θ0 , θ1 , . . . , θm obtained by the LSM has the form ⎛ ⎞ p  −1  1 θ = A A A ⎝ Y j⎠ (30) p 

j=1

where the random vectors Y j , j = 2, p, are virtual independent copies of the virtual random vector Y (27) and Y 1 = Y . Realizations of their components are the numbers y(j−1)n+i , i = 1, n, j = 1, p. 2. The estimate of the random vector θ (30) is unbiased, and the variance of its components is p times less than that of the estimates vector (28). 

Proof. Formula (30) follows from the vector equality ⎛

⎛ ⎞⎞−1 ⎛ 1⎞ p   A   Y 1   −1   j ⎜   ⎜ .. ⎟⎟   ⎜ .. ⎟ A A A = · · · A · · · A A A Y . ⎝ ⎝ . ⎠⎠ ⎝ . ⎠ p j=1 p Y A

(31)



The unbiasedness of the random vector θ (30) follows from the equality  −1  −1 M A A A Y = M A A A Y j = θ, j = 1, p,

(32)

and the property of its components from theequalities   of variances −1  1 p  −1 1 p  j Var A A A p j=1 Y = p2 j=1 Var A A A Y j ,  Var

  −1 −1  A A A Y j = Var A A A Y , j = 1, p.

(33) 

Corollary 1. Knowing the variance of the components of the estimates vector θ (28), we can find such the number of repetitions p of the active experiment that the largest variance of the estimates θ0 , . . . , θm has an acceptable value. 





Remark 13. The variances of the components of the vector θ (28) are easily found because its components Yi are virtual independent random variables, and Var(Yi ) = σ2 , i = 1, n. Corollary 2. We do not need to additionally compute the matrix inverse in the general formula of the LSM at arbitrary p, by virtue of (31). We have used a partial case of the presented theoretical result in the method of estimating the linear terms of an MPR given by a redundant representation [21]. The software architecture presented in Sect. 5 includes algorithmic procedures for estimating all the coefficients of an MPR given by a redundant representation. It can be considered an original model of the software we are now designing. We publish it for the first time in this form.

218

A. Pavlov et al.

5 Software Architecture Figure 1 shows the architecture of a software that implements the mathematical methods for estimating the values of unknown coefficients of an MPR (its nonlinear and linear parts) described in this paper. It is presented in the form of a universal library with a high-level component-based software architecture. In this library, we use a set of independent components for an MPR coefficients estimation. The components can be used to estimate unknown coefficients of other types of regressions, e.g., an MLR. The library itself is implemented as a separate component, but of a higher level, with the possibility of embedding in a third-party software. We briefly describe below the components that make up the library: • MPR is the main component, in fact the library itself, that implements the mathematical apparatus for an MPR coefficients estimation described in this work. • RA_to_rational: provides the input data of the MPR problem in the form of rational fractions. This is necessary to ensure the calculation accuracy (eliminate rounding errors during calculations). • RA_redundant_represent: builds a redundant representation for an MPR building problem. • RA_calculation_MPR_evalt: finds estimates at nonlinear MPR terms implementing modified versions of the first and second sub-algorithms. • MLR: finds coefficient estimates for linear parts of an MPR. • RA_multivariate_linear: implements the LSM for finding estimates during multivariate linear regression analysis. To increase its performance, we use parallel calculations during the matrix multiplication and inversion operations in the general formula of the LSM [20]. • RA_cluster_analysis: significantly reduces the number of partial representations by splitting all the coefficients into two classes. The set of partial representations definitely contains the sought solution. • RA_search_structure_MLR: finds the sought structure of the MLR (in interaction with RA_multivariate_linear component). To increase its performance, we use parallel calculations in the part of estimating the coefficients from the partial representations of MLR and further calculations of the LSM [20]. • RA_calculation_epsilon_evalt: finds estimates of the realizations of E. • RA_chi_square: contains a modified χ 2 criterion to determine the reliability degree of the found estimates for an MLR. • UPR: finds estimates of a UPR coefficients (is auxiliary for finding the coefficient estimates for the nonlinear part of an MPR). • RA_Forsythe_uniform: implements an algorithm for finding the coefficients of NOPFs of a given degree with the required accuracy based on the use of symbolic calculations in a segment with a uniform distribution of a deterministic variable. • RA_detx: finds the value of a deterministic variable x for the given starting and ending points of an active experiment. • RA_detx_real_to_virtual: finds the value of a deterministic variable x for the given starting and ending points of a real active experiment based on the previously conducted virtual experiment.

A Modified Method and an Architecture of a Software

219

Fig. 1. The component-based software architecture of the MPR library

• RA_Forsythe_polynomial: uses NOPFs to implement the method for estimating the coefficients at nonlinear terms of a UPR using the output data of an active experiment.

220

A. Pavlov et al.

• RA_Forsythe_virtual_to_real: implements the method of a UPR coefficients estimation based on the estimates’ values for nonlinear terms of a virtual UPR. • RA_to_double: provides the output data of the MPR problem in the form of real numbers. • RA_error_handler is a component responsible for handling various types of errors.

6 Conclusions 1. We have considered the method of efficient coefficient estimation for nonlinear terms of a multivariate polynomial regression given by a redundant representation. The method reduces the problem solution to the sequential building of univariate polynomial regressions and the solution of nondegenerate linear equations. We have extended the method to apply it to the case when the ranges of values of the input variables have an empty intersection and depend in the general case on the fixed values of the input variables in all tests of the active experiment. Thus, the described modified method of building a multivariate polynomial regression given by a redundant representation significantly extends the scope of its application, in particular, for medical diagnostic systems. 2. We have given the theoretical properties of estimates of multivariate linear regression coefficients based on the results of a repeated active experiment which allowed us to implement restrictions on the value of their variances and not repeatedly invert the matrix in the general formula of the least squares method at the number of repetitions not less than two. We have used a partial case of this result in the original method of building a multivariate linear regression described in [21]. 3. We have shown an original architecture of the software that efficiently and systematically implements the proposed method for estimating nonlinear coefficients of a multivariate polynomial regression given by a redundant representation, as well as the method [21] for estimating the coefficients at its linear terms. 4. It is appropriate actually to carry out statistical studies of the integrated method of building an MPR given by a redundant description that includes the above-mentioned modified method of estimating the coefficients at nonlinear terms of an MPR given by a redundant description, as well as the method [21] of estimating the coefficients at its linear terms.

References 1. Yu, L.: Using negative binomial regression analysis to predict software faults: a study of Apache Ant. Int. J. Inf. Technol. Comput. Sci. (IJITCS) 4(8), 63–70 (2012). https://doi.org/ 10.5815/ijitcs.2012.08.08 2. Shahrel, M.Z., Mutalib, S., Abdul-Rahman, S.: PriceCop–price monitor and prediction using linear regression and LSVM-ABC methods for e-commerce platform. Int. J. Inf. Eng. Electron. Bus. (IJIEEB) 13(1), 1–14 (2021). https://doi.org/10.5815/ijieeb.2021.01.01 3. Satter, A., Ibtehaz, N.: A regression based sensor data prediction technique to analyze data trustworthiness in cyber-physical system. Int. J. Inf. Eng. Electron. Bus. (IJIEEB) 10(3), 15–22 (2018). https://doi.org/10.5815/ijieeb.2018.03.03

A Modified Method and an Architecture of a Software

221

4. Isabona, J., Ojuh, D.O.: Machine learning based on kernel function controlled Gaussian process regression method for in-depth extrapolative analysis of Covid-19 daily cases drift rates. Int. J. Math. Sci. Comput. (IJMSC) 7(2), 14–23 (2021). https://doi.org/10.5815/ijmsc. 2021.02.02 5. Sinha, P.: Multivariate polynomial regression in data mining: methodology, problems and solutions. Int. J. Sci. Eng. Res. 4(12), 962–965 (2013) 6. Kalivas, J.H.: Interrelationships of multivariate regression methods using eigenvector basis sets. J. Chemom. 13(2), 111–132 (1999). https://doi.org/10.1002/(SICI)1099-128X(199903/ 04)13:2%3c111::AID-CEM532%3e3.0.CO;2-N 7. Ortiz-Herrero, L., Maguregui, M.I., Bartolomé, L.: Multivariate (O)PLS regression methods in forensic dating. TrAC Trends Anal. Chem. 141, 116278 (2021). https://doi.org/10.1016/j. trac.2021.116278 8. Guo, G., Niu, G., Shi, Q., Lin, Q., Tian, D., Duan, Y.: Multi-element quantitative analysis of soils by laser induced breakdown spectroscopy (LIBS) coupled with univariate and multivariate regression methods. Anal. Methods 11(23), 3006–3013 (2019). https://doi.org/10.1039/ C9AY00890J 9. Nastenko, E.A, Pavlov, V.A, Boyko, A.L., Nosovets, O.K: Mnogokriterialnyi algoritm shagovoi regressii (Multi-criterion step-regression algorithm). Biomedychna inzheneriya i tekhnolohiya 3, 48–53 (2020). https://doi.org/10.20535/2617-8974.2020.3.195661 (in Russian) 10. Sergeev, V.V., Kopenkov, V.N., Chernov, A.V.: Comparative analysis of function approximation methods in image processing tasks. Comput. Opt. 26, 119–122 (2004). (in Russian) 11. Babatunde, G., Emmanuel, A.A., Oluwaseun, O.R., Bunmi, O.B., Precious, A.E.: Impact of climatic change on agricultural product yield using k-means and multiple linear regressions. Int. J. Educ. Manag. Eng. (IJEME) 9(3), 16–26 (2019). https://doi.org/10.5815/ijeme.2019. 03.02 12. Bolshakov, A.A., Karimov, R.N.: Metody obrabotki mnogomernykh dannykh i vremennykh riadov (Methods of multidimensional data and time series processing). Goriachaia liniia– Telekom, Moscow (2007). (in Russian) 13. Zgurovsky, M.Z., Pavlov, A.A.: The Four-Level Model of Planning and Decision Making. In: Zgurovsky, M.Z., Pavlov, A.A. (eds.) Combinatorial Optimization Problems in Planning and Decision Making. SSDC, vol. 173, pp. 347–406. Springer, Cham (2019). https://doi.org/10. 1007/978-3-319-98977-8_8 14. Pavlov, A., Holovchenko, M., Mukha, I., Lishchuk, K.: Mathematics and software for building nonlinear polynomial regressions using estimates for univariate polynomial regressions coefficients with a given (small) variance. In: Hu Z., Dychka I., Petoukhov S., He M. (eds) Advances in Computer Science for Engineering and Education V. ICCSEEA 2022. Lecture Notes on Data Engineering and Communications Technologies, vol. 134, pp. 288–303 (2022). https://doi.org/10.1007/978-3-031-04812-8_25 15. Vaccari, D.A., Wang, H.K.: Multivariate polynomial regression for identification of chaotic time series. Math. Comput. Model. Dyn. Syst. 13(4), 395–412 (2007). https://doi.org/10. 1080/13873950600883691 16. Ivahnenko, A.G.: Modelirovanie Slojnyh Sistem. Informacionnyi Podhod (Complex Systems Modeling. Informational Approach). Vyshcha shkola, Kyiv (1987). (in Russian) 17. Jackson, E.C., Hughes, J.A., Daley, M.: On the generalizability of linear and non-linear region of interest-based multivariate regression models for fMRI data. IEEE Conf. Comput. Intell. Bioinform. Comput. Biol. (CIBCB) 2018, 1–8 (2018). https://doi.org/10.1109/CIBCB.2018. 8404973

222

A. Pavlov et al.

18. Pavlov, A.A.: Estimating with a given accuracy of the coefficients at nonlinear terms of univariate polynomial regression using a small number of tests in an arbitrary limited active experiment. Bulletin of Natl. Tech. Univ. “KhPI”. Seri.: Syst. Anal. Control Inf. Technol. 2(6), 3–7 (2021). https://doi.org/10.20998/2079-0023.2021.02.01 19. Pavlov, A.A., Holovchenko, M.N., Drozd, V.V.: Construction of a multivariate polynomial given by a redundant description in stochastic and deterministic formulations using an active experiment. Bull. Natl. Tech. Univ. “KhPI”. Ser.: Syst. Anal. Control Inf. Technol. 1(7), 3–8 (2022). https://doi.org/10.20998/2079-0023.2022.01.01 20. Pavlov, A.A., Holovchenko, M.N., Drozd, V.V., Revych, M.M.: Doslidzhennya efektyvnosti metodu pobudovy bahatovymirnoyi liniynoyi rehresiyi, zadanoyi nadlyshkovym opysom (Doclidenn efektivnocti metody pobydovi bagatovimipno| linino| pegpeci|, zadano| nadlixkovim opicom; Study of the efficiency of a method of building a multivariate linear regression given by a redundant representation). In: Materialy Vseukrayins’koyi naukovo-praktychnoyi konferentsiyi molodykh vchenykh ta studentiv “Inzheneriya prohramnoho zabezpechennya i peredovi informatsiyni tekhnolohiyi” (SoftTech-2022), Kyiv, 22– 26 May and 22–25 October 2022, pp. 10–13. NTUU “KPI”, Kyiv (2022). https://drive.goo gle.com/file/d/1CP9EaBTT_rJAXsINbanSVGnP2jkg9FJ0/view. Accessed 30 Dec 2022. (in Ukrainian) 21. Pavlov, A.A., Holovchenko, M.N.: Modified method of constructing a multivariate linear regression given by a redundant description. Bull. Natl. Tech. Univ. “KhPI”. Ser.: Syst. Anal. Control Inf. Technol. 2(8), 3–8 (2022). https://doi.org/10.20998/2079-0023.2022.02.01

Research and Development of Bilinear QoS Routing Model over Disjoint Paths with Bandwidth Guarantees in SDN Oleksandr Lemeshko , Oleksandra Yeremenko(B) and Batoul Sleiman

, Maryna Yevdokymenko ,

Kharkiv National University of Radio Electronics, Nauky Ave. 14, Kharkiv 61166, Ukraine {oleksandr.lemeshko.ua,oleksandra.yeremenko.ua, maryna.yevdokymenko}@ieee.org

Abstract. The work proposes developing a bilinear Quality of Service routing model over disjoint paths with bandwidth guarantees in the Software-Defined Network data plane. The advantage of the solution is formulating the routing problem as optimization under bilinear conditions aimed at bandwidth guarantee over the set of disjoint paths (multipath). Solving the routing problem allows obtaining disjoint routes by Boolean variables calculation responsible for the link belonging to the collection of network paths. Also, the model introduces conditions of the disjoint routes connectivity and the intersection absence. The model’s novelty is the bilinear conditions for a given Quality of Service level guarantee. Fulfilling such bilinear requirements provides that the total bandwidth of the disjoint routing multipath will not be lower than the demanded level. Also, the optimality criterion is defined by the Quality of Service requirements. A numerical study confirmed the improved model’s workability in providing a set level of obtained disjoint paths bandwidth. Keywords: Multipath routing · Disjoint Paths · Bandwidth · Load Balancing · QoS · Software-Defined Network

1 Introduction The significant interest in Software-Defined Networks (SDN) is due to many reasons and advantages compared to traditional networks. The use of softwarized networking approaches makes it possible to make traffic management more flexible and use optimization techniques for different purposes [1–3]. The practical application of SDN solutions is directly related to increasing the efficiency of providing a given level of Quality of Service (QoS) [4]. Since the control plane is separated from the data plane, containing network equipment, the so-called router’s programming based on table-based forwarding becomes possible [1–5]. Due to various reasons and requirements for Quality of Service, fault tolerance, and security, different types of SDN architectures are used [3–8]. Figure 1 represents the primary differences between conventional (traditional) networks and SDN [1, 3]. At the © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 223–235, 2023. https://doi.org/10.1007/978-3-031-36118-0_20

224

O. Lemeshko et al.

same time, one should consider the existing deployment of communication networks, which include both SDN devices and traditional network equipment. In this way, hybrid SDNs are organized [3, 5]. It is noted that appropriate routing tools are actively used to ensure the Quality of Service, network security, and fault tolerance under established requirements [9– 12]. One of the well-known directions is disjoint routing solutions [13]. Therefore, the presented work is devoted to improving the mathematical model of routing by disjoint paths, which can serve as a basis for QoS routing protocols in the SDN data plane. The paper structure is as follows: 1. Section 2 is devoted to related work analysis in the field of disjoint routes calculation and its application to QoS routing in SDN. 2. Section 3 describes the mathematical model of disjoint paths calculation that provides multipath guaranteed bandwidth. 3. Section 4 contains the numerical research results of the developed model. 4. Section 5 discusses the proposed QoS routing model. 5. Section 6 concludes the work and suggests further development directions in advanced performance-based disjoint QoS routing.

Fig. 1. Difference between traditional networks and SDN [1, 3]

Research and Development of Bilinear QoS

225

2 Related Work Analysis The analysis [10–24] shows that multipath routing over disjoint paths is an effective solution for better resource allocation, scalability, network resilience, and security. Moreover, such an approach for multipath routing is often utilized for Quality of Service and Quality of Experience improvement. Consider in more detail the results of some recent research in this area. Thus, in [14] the need to develop search algorithms for multiple disjoint paths is substantiated. Furthermore, the corresponding algorithms must satisfy the demands of computational complexity to achieve scalability. Having such a set of routes adds more value to the multipath. In addition, it is noted that a set consisting of only two disjoint paths is usually used. This approach is widely used in fault-tolerant routing, where one route is used as the primary and the other as a backup. Thus, in [14] an efficient algorithm for calculating multiple disjoint paths with acceptable computational cost and ensuring scalability is proposed and investigated. Moreover, in [15], the authors propose One-Shot Multiple Disjoint Path Discovery Protocol (1S-MDP) that can provide both node-disjoint or link-disjoint paths. The widespread use and application of disjoint paths routing are found in Mobile Ad-hoc Networks (MANETs). The specifics of this type of network impose certain limitations when developing routing algorithms. First, you should consider the dynamics of the MANET topology, error-less data broadcasting, and the need to ensure a high level of fault tolerance and security. For example, in [16], a link-disjoint multipath routing method was proposed to choose the shortest path from multiple paths in MANET. Moreover, the simulation results proved the possibility of using and efficiency in the traffic load of the proposed method in a dynamic environment. Also, many current research works relate to improving the Quality of Service during QoS routing and implementing the strategy of disjoint multipath routing [17–22], including applying the Traffic Engineering concept principles [19]. Additionally, in [20], the authors consider adaptive multipath routing over both shortest and non-shortest disjoint paths. While in [22], a novel multipath transport scheme for real-time multimedia using disjoint multipath and segment routing in Software-Defined Networks is proposed. It is noted that satisfactory Quality of Experience (QoE) level currently remains an urgent task. Thus, in the solution [22], the SDN controller centrally calculates multiple disjoint paths meeting bandwidth requirements and load balancing in the network. In this case, sub-flows are transmitted over disjoint paths to reduce the end-to-end delay and improve QoE. Therefore, the basic model proposed in [10] is further used to modify and develop an improved model of disjoint multipath routing to improve the overall Quality of Service level by guaranteeing bandwidth for multipath flow transmission in application to SDN. In contrast to the solutions presented in previous studies, primary attention is paid to the task of disjoint paths set calculation in this work. The load balancing processes in the SDN data plane are not explicitly considered, but the multipath routing strategy is used. The problem of calculating disjoint paths in the proposed mathematical model is presented in an optimization form. Mixed integer linear and nonlinear programming methods are used to solve the optimization problem formulated in work.

226

O. Lemeshko et al.

3 Mathematical Model of Disjoint Paths Calculation that Provides Multipath Guaranteed Bandwidth Work [10] presents a model we will modify further and calculate the maximum number of disjoint paths and their bandwidth. In the modified model, the network structure and its functional parameters we described as follows: G = (R, E)

sk

  graph presenting the network structure, in which R = Ri ; i = 1, m is a set of   vertices simulating routers, and E = Ei,j ; i, j = 1, m; i  = j is a set of arcs representing links; source node;

dk

destination node;

K

set of flows for transmitting in the network (k ∈ K);

k ai,j

control variables, each of which determines whether the link Ei,j ∈ E belongs to

ϕi,j

link Ei,j ∈ E capacity (measured in packets per second, pps);

Mk

integer variable characterizing the number of disjoint paths;

wi,j

weighting coefficients related to the capacity of a link Ei,j ∈ E

the set of calculated disjoint paths for transmission of the k th flow;

As a result of solving the problem of calculating the maximum number of disjoint k , paths for transmission of the k th flow, it is necessary to calculate the set of variables ai,j where the set of calculated disjoint paths corresponds to maximum multipath bandwidth. k have the constraints of type: The routing variables ai,j k ai,j ∈ {0; 1}.

(1)

In addition, the following conditions for every distinct pair of source and destination nodes must be fulfilled [10]:  k ai,j = Mk ; k ∈ K, Ri = sk ; (2) j:Ei,j ∈E



k aj,i = Mk ; k ∈ K, Ri = dk

(3)

j:Ej,i ∈E

At the same time, for the transit nodes in the network (Ri = sk , dk ), the following restrictions are introduced in the basic model [10]: ⎧  k ai,j ≤ 1, k ∈ K; ⎪ ⎪ ⎪ ⎪ ⎪ j:E ∈E i,j ⎪ ⎪  ⎪ ⎨ k aj,i ≤ 1, k ∈ K; (4) ⎪ j:Ej,i ∈E ⎪ ⎪ ⎪   ⎪ k k ⎪ ⎪ ai,j − aj,i = 0, k ∈ K. ⎪ ⎩ j:Ei,j ∈E

j:Ej,i ∈E

Research and Development of Bilinear QoS

227

The physical meaning of the first condition in the system (4) is that no more than one path can outgo from the transit node Ri . Fulfillment of the second condition in the system (4) must ensure that the transit node Ri cannot belong to more than one path from the calculated set. Implementing the third condition from the system (4) is responsible for the fact that the path can outgo from the transit router Ri only if it has income previously to this node. For the telecommunication network in general, the implementation of the restrictions system (4) must guarantee compliance with the following conditions: 1. Involved communication links ensure the connectivity of each specific calculated path. 2. Calculated paths will not intersect, i.e., they can share only source sk and destination dk nodes. However, in any case, the following condition takes place: Mk ≥ 1.

(5)

In the general case, the range of available values Mk directly depends on the network topology, the network nodes’ connectivity level, and the degree of the graph G vertices, which simulate the source and destination routers. The basic mathematical model (1)–(5) can also be modified in QoS routing to provide maximum or specified bandwidth using the calculated set of disjoint paths. Consequently, within the frame of the basic model, it is necessary to introduce additional conditions to ensure a given Quality of Service level relative to bandwidth. Therefore, we denote by k the minimum bound value for the bandwidth of any of the disjoint paths set used βpath to transmit the kth packet flow. Then the following condition can be introduced into the framework of the routing model (by analogy with [10]): k k k ϕi,j + W (1 − ai,j ) ≥ βpath , ai,j

(6)

where the weighting coefficients W take values higher than the maximum bandwidth of the communication network links, fulfillment of condition (6) ensures that each route belonging to the set of disjoint paths calculated for the kth flow has a bandwidth of not k . less than βpath Next, we should choose the following objective function J for maximization as the optimality criterion for the QoS-routing over disjoint paths problem solution:  k k J = cM Mk + cβ βpath − cv vi,j ai,j , (7) Ei,j ∈E

where weighting coefficients cM , cβ , and cv determine the importance of each component in expression (7). Introducing the first term in the objective function (7) is associated with maximizing the number of disjoint paths used. The second term is responsible for maximizing the lower bound bandwidth of the calculated paths. If we limit ourselves to using only these two terms, then, on the one hand, the least efficient path will have a bandwidth equal to k . However, such a solution may not always contribute to the inclusion in the value βpath the calculated set of the highest bandwidth routes. Therefore, the novelty of the proposed

228

O. Lemeshko et al.

model is the use of the third term in criterion (7), which is introduced by analogy with the metrics of the OSPF and EIGRP protocols to include in the calculated disjoint paths high-bandwidth links. Thus, it is proposed that in the objective function (7) the weighting coefficients vi,j , taking into account the bandwidth ϕi,j of the corresponding communication link Ei,j ∈ E, are defined as: vi,j = 10/ϕi,j .

(8)

It is experimentally established that to implement QoS routing with the calculation of the maximum number of disjoint paths with maximum bandwidth, the weighting coefficients in expression (7) must meet the following condition: M >> cβ >> cv .

(9)

The peculiarity of using the proposed mathematical model (1)–(9) is that it focuses on choosing paths in the network that do not intersect and provide maximum bandwidth. Increasing the bandwidth allocated to a particular flow allows for improving average delay and packet loss. However, in practice, to save network resources when servicing the traffic of network users, there are requirements for the boundary value of the network bandwidth available for use by a particular flow. Let us denote, for example, for the kth packet flow this boundary value through β k . Therefore, in the model (1)–(9), it is proposed to introduce the following condition: k Mk βpath ≥ βk .

(10)

Therefore, in the general case, the left part of inequality (10) is a bilinear form k of two types of control variables Mk and βpath that characterize the lower bound of bandwidth, which in total provides the use of the calculated paths. The bound is lowest, as each of these disjoint paths, according to conditions (6), has a capacity not lower but k . Fulfillment of condition (10), depending on the form of the selected may be higher βpath optimality criterion, can be achieved either by increasing the number of involved disjoint k . routes Mk or by raising the boundary value relative to their minimum bandwidth βpath Given the use of the optimality criterion of routing solutions (7), the priority of increasing the control variables that are included in the bilinear form in (10) will be determined by the hierarchy of values of weights in (7). In the case of introducing conditions (10), the optimization problem with criterion (7) and the supplemented set of constraints (1)–(6), (8)–(10) will belong to the class of problems of mixed-integer nonlinear programming (MINLP). However, for example, if the number of paths to be calculated is known in advance, i.e., in criterion (7), the first term also becomes a constant, condition (10) becomes linear, and the optimization problem remains of the MILP class.

Research and Development of Bilinear QoS

229

4 Numerical Research Consider several numerical examples for calculating disjoint paths and providing a given bandwidth on the network structure presented in Fig. 2. Communication link gaps for this network structure (Fig. 2) indicate their bandwidth (pps). Additionally, Table 1 shows the set of available paths and their bandwidth. Then Table 2 shows the calculations to determine for one flow transmitted in the network from the first to the ninth router (Fig. 2) a set of disjoint paths and guarantees the lower bound of total bandwidth. Table 2 demonstrates possible solutions to the QoS-routing problem using disjoint paths that provide a given bandwidth. Depending on the ratio of weighting coefficients M, cβ , and cv in (7), the use of the model (1)–(6), (10) allows for obtaining a different set of paths. If condition (9) was satisfied, then the fulfillment of condition (10) was achieved, as a rule, based on increasing the number of involved disjoint routes Mk . However, as shown in Table 2, it is better to use the network resource economically to satisfy the constraint M >> cβ >> cv . Then condition (10) was fulfilled based on raising the k ) of the calculated paths. If β k boundary value relative to the minimum bandwidth (βpath is equal to 1200 pps, then in Table 2, only the solutions at Mk = 3 (third row) satisfy (Fig. 3).

Fig. 2. Network structure for numerical research.

The introduction of the third term in the objective function (7) under cv = 0 allowed for including more productive communication links in the set of calculated paths. Such a result was accompanied by an increase in the total bandwidth of disjoint paths used in QoS routing (Table 2, Fig. 3).

230

O. Lemeshko et al. Table 1. Characteristics of available disjoint paths for the network structure (Fig. 2).

Path #

Designation

The set of links that form the path

Path Bandwidth, pps

1

L1

{E 1,2 , E 2,6 , E 6,9 }

400

2

L2

{E 1,2 , E 2,9 }

400

3

L3

{E 1,3 , E 3,6 , E 6,9 }

440

4

L4

{E 1,3 , E 3,7 , E 7,9 }

500

5

L5

{E 1,4 , E 4,7 , E 7,9 }

300

6

L6

{E 1,4 , E 4,8 , E 8,9 }

820

7

L7

{E 1,5 , E 5,9 }

250

8

L8

{E 1,5 , E 5,8 , E 8,9 }

550

Table 2. The results of calculations to determine the set of disjoint paths and provide guarantees for their lower bound of total bandwidth (k = 1). β k , pps

1000

Mk

k , βpath pps

The set of paths Total that are calculated bandwidth, using (7) pps

The set of paths Total that are calculated bandwidth, using (7) pps

cv = 0

cv  = 0

4

250

{L 2 , L 3 , L 5 , L 7 }

1390

{L 2 , L 4 , L 6 , L 7 }

1970

2

500

{L 4 , L 8 }

1000

{L 4 , L 6 }

1220

3

400

{L 2 , L 4 , L 8 }

1450

{L 2 , L 4 , L 6 }

1720

Figure 4 presents the result of solving the QoS routing problem over disjoint paths with bandwidth guarantees in SDN when two paths were calculated to ensure the required network bandwidth (β k is equal to 1000 pps), namely, {L 4 , L 6 } (Table 2, second row) with total bandwidth 1220 pps. Figure 5 shows the case when ensuring the required network bandwidth (β k = 1000 pps) using the QoS routing model over disjoint paths with bandwidth guarantees in SDN (1)–(10) using (7) under cv = 0 four paths were calculated {L 2 , L 3 , L 5 , L 7 } (Table 2, first row) with total bandwidth 1390 1/s. The increase in the number of disjoint paths used can be dictated, for example, by the requirements of network security policies [14].

Research and Development of Bilinear QoS

231

Fig. 3. The set of paths {L 2 , L 4 , L 6 } calculated using (7) under cv  = 0 with total bandwidth 1720 pps (Mk = 3).

Fig. 4. The set of paths {L 4 , L 6 } calculated using (7) under cv  = 0 with total bandwidth of 1220 pps (Mk = 2).

232

O. Lemeshko et al.

Fig. 5. The set of paths {L 2 , L 3 , L 5 , L 7 } calculated using (7) under cv = 0 with total bandwidth of 1390 pps (Mk = 4).

5 Discussion The routing problem solution allows getting disjoint routes by calculating a set of Boolean-type variables. Each determines the link belonging to the collection of calculated network routes. Also, the model includes the connectivity conditions of the obtained paths and the intersection absence. The model’s novelty is the bilinear conditions for ensuring a demanded level of QoS. Conditions fulfillment ensures that the calculated disjoint paths’ total bandwidth will not be lower than the given level. Therefore, in the mathematical model framework, the technical task of QoS routing in the SDN data plane is formulated as an optimization problem with the optimality criterion depending on QoS requirements. Additionally, considering the routing variables constraints, the optimization problem belongs to the mixed integer nonlinear programming class.

6 Conclusion The work proposes developing a bilinear QoS routing model over disjoint paths with bandwidth guarantees. A numerical study of QoS routing applying the proposed model on a set of examples proved the improved model’s workability and effectiveness in providing a demanded level of obtained disjoint paths bandwidth. The basic mathematical routing model (1)–(6) is proposed for further improvement. Thus, improved model (1)–(10) allows for calculating the corresponding multipath, namely, a set of disjoint paths under Quality of Service conditions. A graph in the model

Research and Development of Bilinear QoS

233

(1)–(6) describes the network structure where the vertices set represents network routers, and the arcs set represents communication links. The application of the routing model proposed in the work is, first of all, softwarized telecommunication networks (SDN data plane) where the functions of route calculation are assigned to SDN controllers. The linear nature of the expressions that form the basis of the developed QoS routing model, criteria, and limitations of the formulated optimization problem regarding the calculation of disjoint paths should contribute to the acceptable computational complexity. Therefore, their technological implementation in SDN could be the basis of the algorithmic and software background of promising QoS routing protocols. We see further development of an advanced performance-based disjoint path QoS routing model in implementing fault-tolerant and secure routing. Moreover, complementary protection schemes will support additional QoS indicators such as bandwidth, average delay, packet loss probability, and network security parameters, which are crucial when transmitting mission-critical applications’ traffic.

References 1. Ahmad, S., Mir, A.H.: SDN interfaces: protocols, taxonomy and challenges. Int. J. Wirel. Microwave Technol. (IJWMT) 12(2), 11–32 (2022). https://doi.org/10.5815/ijwmt.2022. 02.02 2. Barabash, O., Kravchenko, Y., Mukhin, V., Kornaga, Y., Leshchenko, O.: Optimization of parameters at SDN technologie networks. Int. J. Intell. Syst. Appl. 9(9), 1–9 (2017). https:// doi.org/10.5815/ijisa.2017.09.01 3. Al Mtawa, Y., Haque, A., Lutfiyya, H.: Migrating from legacy to software defined networks: a network reliability perspective. IEEE Trans. Reliab. 70(4), 1525–1541 (2021). https://doi. org/10.1109/TR.2021.3066526 4. Islam, N., Rahman, H., Nasir, M.K.: A comprehensive analysis of QoS in wired and wireless SDN based on mobile IP. Int. J. Comput. Netw. Inf. Secur. 13(5), 18–28 (2021). https://doi. org/10.5815/ijcnis.2021.05.02 5. Rohitaksha, K., Rajendra, A.B., Mohan, J.: Routing in hybrid software defined networking using hop-count approach. Int. J. Wirel. Microwave Technol. (IJWMT) 12(3), 54–60 (2022). https://doi.org/10.5815/ijwmt.2022.03.04 6. Strelkovskaya, I., Solovskaya, I., Makoganiuk, A.: Predicting self-similar traffic using cubic B-splines. In: 2019 3rd International Conference on Advanced Information and Communications Technologies (AICT) Proceedings, pp. 153–156. IEEE (2019). https://doi.org/10.1109/ AIACT.2019.8847761 7. Bielievtsov, S., Ruban, I., Smelyakov, K., Sumtsov, D.: Network technology for transmission of visual information. selected papers of the XVIII international scientific and practical conference “information technologies and security” (ITS 2018). In: CEUR Workshop Proceedings, vol. 2318, pp. 160–175 (2018). https://ceur-ws.org/Vol-2318/paper14.pdf 8. Smelyakov, K., Smelyakov, S., Chupryna, A.: Adaptive edge detection models and algorithms. In: Mashtalir, V., Ruban, I., Levashenko, V. (eds.) Advances in Spatio-Temporal Segmentation of Visual Data. Studies in Computational Intelligence, vol. 876, pp. 1–51. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-35480-0_1

234

O. Lemeshko et al.

9. Lemeshko, O., Hu, Z., Shapovalova, A., Yeremenko, O., Yevdokymenko, M.: Research of the influence of compromise probability in secure based traffic engineering model in SDN. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) Advances in Computer Science for Engineering and Education IV. ICCSEEA 2021. Lecture Notes on Data Engineering and Communications Technologies, vol. 83, pp. 47–55. Springer, Cham (2021). https://doi.org/10.1007/978-3-03080472-5_5 10. Lemeshko, O., Yeremenko, O., Yevdokymenko, M., Sleiman, B.: System of solutions the maximum number of disjoint paths computation under quality of service and security parameters. In: Ilchenko, M., Uryvsky, L., Globa, L. (eds.) Advances in Information and Communication Technology and Systems. MCT 2019. Lecture Notes in Networks and Systems, vol. 152, pp. 191–205. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-58359-0_10 11. Lemeshko, O., Yeremenko, O., Yevdokymenko, M., Shapovalova, A., Hailan, A.M., Mersni, A.: Cyber resilience approach based on traffic engineering fast reroute with policing. In: 2019 10th IEEE International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications (IDAACS) Proceedings, vol. 1, pp. 117–122. IEEE (2019). https://doi.org/10.1109/IDAACS.2019.8924294 12. Lemeshko, O., Yevdokymenko, M., Yeremenko, O., Hailan, A.M., Segeˇc, P., Papán, J.: Design of the fast reroute QoS protection scheme for bandwidth and probability of packet loss in software-defined WAN. In: 2019 IEEE 15th International Conference on the Experience of Designing and Application of CAD Systems (CADSM), pp. 1–5. IEEE (2019). https://doi. org/10.1109/CADSM.2019.8779321 13. Gomes, T., Jorge, L., Girão-Silva, R., Yallouz, J., Babarczi, P., Rak, J.: Fundamental schemes to determine disjoint paths for multiple failure scenarios. In: Rak, J., Hutchison, D. (eds.) Guide to Disaster-Resilient Communication Networks. Computer Communications and Networks, pp. 429–453. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-446857_17 14. Lopez-Pajares, D., Rojas, E., Carral, J.A., Martinez-Yelmo, I., Alvarez-Horcajo, J.: The disjoint multipath challenge: multiple disjoint paths guaranteeing scalability. IEEE Access 9, 74422–74436 (2021). https://doi.org/10.1109/ACCESS.2021.3080931 15. Lopez-Pajares, D., Alvarez-Horcajo, J., Rojas, E., Carral, J.A., Martinez-Yelmo, I.: One-shot multiple disjoint path discovery protocol (1S-MDP). IEEE Communications Letters 24(8), 1660–1663 (2020). https://doi.org/10.1109/LCOMM.2020.2990885 16. Robinson, Y.H., Julie, E.G., Saravanan, K., Kumar, R., Abdel-Basset, M., Thong, P.H.: Linkdisjoint multipath routing for network traffic overload handling in mobile ad-hoc networks. IEEE Access 7, 143312–143323 (2019). https://doi.org/10.1109/ACCESS.2019.2943145 17. Sreeram, K., Unnisa, A. and Poornima, V.: QoS aware multi-constrained node disjoint multipath routing for wireless sensor networks. In: 2019 5th International Conference on Advanced Computing & Communication Systems (ICACCS) Proceedings, pp. 382–385. IEEE (2019). https://doi.org/10.1109/ICACCS.2019.8728475 18. Sarma, H.K.D., Dutta, M.P., Dutta, M.P.: A quality of service aware routing protocol for mesh networks based on congestion prediction. In: 2019 International Conference on Information Technology (ICIT) Proceedings, pp. 430–435. IEEE (2019). https://doi.org/10.1109/ICIT48 102.2019.00082 19. Zhang, C., et al.: Scalable traffic engineering for higher throughput in heavily-loaded software defined networks. In: NOMS 2020–2020 IEEE/IFIP Network Operations and Management Symposium Proceedings, pp. 1–7. IEEE (2020). https://doi.org/10.1109/NOMS47738.2020. 9110259 20. Besta, M., et al.: High-performance routing with multipathing and path diversity in ethernet and hpc networks. IEEE Trans. Parallel Distrib. Syst. 32(4), 943–959 (2020). https://doi.org/ 10.1109/TPDS.2020.3035761

Research and Development of Bilinear QoS

235

21. Hou, A., Wu, C.Q., Zuo, L., Zhang, X., Wang, T., Fang, D.: Bandwidth scheduling for big data transfer with two variable node-disjoint paths. J. Commun. Netw. 22(2), 130–144 (2020). https://doi.org/10.1109/JCN.2020.000004 22. Zhang, W., Lei, W., Zhang, S.: A multipath transport scheme for real-time multimedia services based on software-defined networking and segment routing. IEEE Access 8, 93962–93977 (2020). https://doi.org/10.1109/ACCESS.2020.2994346 23. Zeng, G., Zhan, Y., Pan, X.: Failure-tolerant and low-latency telecommand in megaconstellations: the redundant multi-path routing. IEEE Access 9, 34975–34985 (2021). https:// doi.org/10.1109/ACCESS.2021.3061736 24. Kaneko, K., Van Nguyen, S., Binh, H.T.T.: Pairwise disjoint paths routing in Tori. IEEE Access 8, 192206–192217 (2020). https://doi.org/10.1109/ACCESS.2020.3032684

Improved GL-Model of Behavior of Complex Multiprocessor Systems in Failure Flow Alexei Romankevitch1 , Kostiantyn Morozov1(B) , Vitaliy Romankevich1 , Daniil Halytskyi1 , and Zacharioudakis Eleftherios2 1 National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”,

37 Peremohy Avenue, Kyiv 03056, Ukraine [email protected], [email protected] 2 Neapolis University Paphos, 2 Danais Avenue, 8042 Paphos, Cyprus

Abstract. GL-models can be used as behavioral models of fault-tolerant multiprocessor systems in the flow of failures, to evaluate their reliability parameters by conducting statistical experiments. Estimation of the reliability parameters of fault-tolerant multiprocessor systems is an extremely complex and important theoretical and practical problem. To build models of complex systems, in particular those consisting of several subsystems, simpler basic GL-models can be combined. To build models of systems, between processors of different subsystems of which task exchange is possible for the purpose of restoring performance, a so-called dual model can be applied, built by replacing the edge functions of the original GL-model with corresponding dual functions. In addition, the work describes a special auxiliary model that reflects the limitation of the volume of tasks that can be transferred between subsystems. Keywords: GL-model · fault-tolerant multiprocessor system · dual function

1 Introduction In modern world, automation is becoming more and more important. This trend has several reasons. Yes, on the one hand, it frees a person from monotonous and uncreative work. On the other hand, it allows you to reduce the probability of errors caused by the human factor, as well as bypass the physical limitations of the human body, such as reaction speed, amount of calculation capabilities, etc. To ensure automatic and automated functioning of various systems and complexes, so-called control systems (CS) are used [1, 2]. Today such systems are often created based on microprocessors (MP). However, if one MP is usually enough to build a simple CS, then more complex and responsible systems may require the use of many processors. This especially applies to the so-called mission critical systems (MCS) [3], the failure of which can lead to significant material losses, threaten people’s lives and health, etc. In particular, such systems often include energy facilities (nuclear power plants, thermal power plants, hydroelectric power plants), transport systems (aircraft, trains, railway hubs), large industrial enterprises, space vehicles and systems, complex military © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 236–245, 2023. https://doi.org/10.1007/978-3-031-36118-0_21

Improved GL-Model of Behavior of Complex Multiprocessor Systems

237

equipment. Recently, this is beginning to apply to more commonly used objects, in particular, such as automatically piloted cars. The control system for such objects should usually be very reliable. In addition, such systems often must perform a large number of calculations per unit of time. Therefore, the developer of such an CS must solve the problem of ensuring both a high level of its reliability and a high level of its productivity. Both problems can be solved in the case of building a CS based on the so-called faulttolerant multiprocessor systems (FTMS) [4–6]. FTMS can continue full functioning in the event of failure of a certain number or subset of its processors. Such systems may include processors of various types with a complex system of connections between them. The FTMS developer faces the task of assessing the reliability parameters of the developed system [7, 8], in particular, the probability of its trouble-free operation. The calculation of these values can occur in different ways, which can be divided into two categories [9]. The first category includes approaches based on the construction of certain analytical expressions for the sought parameters of system reliability. This allows you to get their values with great accuracy. However, for complex and heterogeneous systems, the formation of such expressions becomes a difficult task. For each of the systems, it may be necessary to develop a separate specialized method. That is, approaches from this category are usually non-universal [10–14]. Approaches from the second category make it possible to evaluate the reliability parameters of the system with a given accuracy, by performing statistical experiments with models of its behavior in the flow of failures [15]. Such a model takes as input the socalled system state vector and returns a certain value characterizing the state of the system as a whole, which corresponds to it (for example, 1 – the system is operational and 0 – the system has failed). The system state vector is a vector whose elements correspond to the state of each of the system’s processors (for example, 1 – the processor is operational, 0 – the processor has failed). By applying the methods of mathematical statistics to the values obtained with the help of such models on sets of random system state vectors, an estimate of its reliability parameters can be obtained with a certain accuracy. Accuracy can be increased, in particular, by increasing the number of analyzed vectors, and/or by using specialized procedures for their formation.

2 GL-Models and Their Properties The so-called GL-models [16], which combine the properties of graphs and Boolean functions, can be used as behavioral models of FTMS in the flow of failures. Such a model is an undirected graph, each edge of which corresponds to a certain Boolean edge function that depends on all or some of the elements of the system state vector. If the edge function takes a zero value on a certain vector, the corresponding edge is removed from the graph. The connectivity of the graph corresponds to the state of the system: if the graph remains connected, the system is operational, and if not, it fails. A number of methods can be used to build GL-models of different types of systems. In particular, the method described in [17] allows building models based on cyclic graphs. In addition, these models have a number of additional properties.

238

A. Romankevitch et al.

A system that remains operational when any m of its n processors fail and fails in the event of a failure of a larger number of processors (any combination of them), as well as the GL-model corresponding to it, are called basic and denoted by K(m, n). Model K(m, n), built according to [17], will have exactly ϕ(m, n) edges, where ϕ(m, n) = n − m + 1. In addition, on vectors with l zeros it will lose exactly ψ(m, l) edges, where  0, if l < m ψ(m, l) = . l − m + 1, if l ≥ m

(1)

(2)

Such models are called MLE-models (minimum of lost edges), because on vectors with m + 1 zeros, they always lose a minimum sufficient number of edges, in other words, 2 edges. Unfortunately, real systems are not always basic. In particular, this may apply to so called multi-module systems consisting of several subsystems. In some cases, processors of one subsystem can perform part of the tasks of another subsystem. Methods of building GL-models for some of these subsystems are proposed in [18, 19]. However, for cases of more complex interaction between subsystems, these methods may not be effective enough.

3 Dual GL-Model Consider a multimodule system in which the processors of one subsystem can be used to perform the tasks of others. Let the subsystem contain n processors and be resistant to failure of m of them. In other words, n – m capable processors are enough to perform tasks of the subsystem. The rest of the processors can be used to perform tasks of other subsystems. To build a GL-model of the system, it will be useful to be able to obtain a vector containing exactly as many ones as the processors of the subsystem can be involved in solving external problems. For the aforementioned subsystem, this number can be calculated using the formula:  0, if l ≥ m μ(m, l) = . (3) m − l, if l < m Let the model K  (m, n) contain edge functions that are dual with respect to the functions of the MLE-model K(m, n). We will call such a model a dual model. Using de Morgan’s rules, it is easy to show that the expressions of the edge functions of the dual model can be obtained from the expressions of the edge functions of the basic model by replacing conjunctions with disjunctions, and disjunctions with conjunctions. It is obvious that in the dual model K  (m, n) on vectors with l ones exactly ψ(m, l) functions will take a value equal to one. Suppose that in a basic m-fault-tolerant system consisting of n processors, l of them failed. Accordingly, the number of working processors is n − l. In the dual model K’(n −

Improved GL-Model of Behavior of Complex Multiprocessor Systems

239

m + 1, n), which accepts as an input the system state vector of exactly ψ(n − m + 1, n – l) functions will take a value equal to one. According to (2) and (3) and taking into account that n, m and l are integers, we can conclude that ψ(n − m + 1, n – l) ≡ μ(m, l). Thus, the vector consisting of the values of the edge functions of the dual model K’(n − m + 1, n) will contain exactly μ(m, l) ones. In addition, according to (1), such a vector will have a length equal to m. We will call such a model a redundancy model and, to simplify notation, denote it as R(m, n), i.e., R(m, n)  K  (n − m + 1, n)

(4)

4 Use of the Dual GL-Model 4.1 Model of a Pair of a “Donor” - “Recipient” Subsystems Consider the situation when “redundant” processors of one subsystem can be used to restore the performance of another subsystem. That is, we can say that the first subsystem is a “donor”, and the second – a “recipient”. Let the first of the subsystems contain n1 processors and be resistant to the failure of any m1 of them, and the second – n2 processors and be resistant to the failure of m2 of them. At the same time, the “excess” resources of the first subsystem can be used to solve part of the problems of the second subsystem, if its own resources are not enough. That is, the first subsystem acts as a “donor” and the second as a “recipient”. For the first subsystem, it is enough to build the usual basic model K(m1 , n1 ), which will reflect its behavior in the flow of failures. In addition, let’s build a dual redundancy model R(m1 , n1 ), which accepts as input the state vector of the first subsystem, and with the help of which the auxiliary vector w, length m1 , will be obtained. By combining this vector (by concatenation) with the state vector of the second subsystem, we will get a vector (denoted as v) containing as many ones as there are processors available to solve the problems of the second subsystem. Its length will be equal to m1 + n2 . Recall that the second subsystem is resistant to the failure of no more than m2 of n2 of its own processors, i.e., to solve its problems, no less than n2 − m2 of operational processors are required. Considering the processors that can be provided by the first subsystem (the number of which can be up to m1 ), we can say that the system is resistant to the failure of no more than m1 + m2 of m1 + n2 processors in total, that is, it will remain operational exactly as long as in the vector v there will be no more than m1 + m2 zeros. Thus, the second subsystem will correspond to the model K(m1 + m2 , m1 + n2 ), which receives the vector v described above as an input. The model of the entire system can subsequently be built on the basis of the models of its subsystems obtained above, in particular, by the methods proposed in [19]. 4.2 Any Number of “Donor” and “Recipient” Subsystems In general, the number of subsystems corresponding to the behavior of the “donor”, as well as the number of subsystems corresponding to the behavior of the “recipient”, can be arbitrary.

240

A. Romankevitch et al.

In the case of several “donors”, all their redundant processors can be used to restore the functionality of the “recipient” subsystems. Let the system have s “donor” subsystems, each of which contains ni processors and is resistant to the failure of mi of them (i = 1, 2, …, s). For each of them, we will construct dual redundancy models R(mi , ni ), with the help of which the vectors wi will be obtained. Let’s combine these vectors by concatenation and get the vector w. This vector, which will have a length of m = m1 + m2 +… + ms , will contain exactly as many ones as there are reserve processors in all “donor” subsystems in general. Further, this vector can be used similarly to the case of one “donor” system. In the case of multiple “recipient” subsystems, each of which can use available donor processors, the approach described in [18] can be applied. In this case, the available reserved processors of the “donor” subsystems can be considered as a sliding redundancy. Accordingly, the vector w will correspond to the part of the state vector that characterizes the processors of the sliding redundancy in [18]. 4.3 Model of a “Symmetric” Pair of “Donor”-“Recipient” Subsystems The case was considered above when one subsystem can contribute to the restoration of the performance of the second subsystem. In turn, it is possible to organize the system in such a way that the second subsystem can also contribute to the restoration of the first subsystem. That is, depending on the situation, each of the subsystems can become both a “donor” and a “recipient”. In this case, to build a model of the first subsystem, it is enough, by analogy with the previous sections, to build a dual redundancy model R(m2 , n2 ), which accepts the state vector of the second subsystem as an input. With the help of this model, an auxiliary vector of length m2 will be obtained. By combining this vector (by concatenation) with the state vector of the first subsystem, we will get a vector containing as many ones as the number of processors that can be used to solve the problems of the first subsystem. Its length will be equal to m2 + n1 . The first subsystem will correspond to the model K(m2 + m1 , m2 + n1 ), which accepts this vector as an input. The model of the second subsystem is built in the same way as in the previous case. Note that for both dual models, parts of state vectors containing elements corresponding to processors that are contained only in one or another subsystem are used. Indeed, these models reflect the amount of “excess” resources directly in the subsystems, without taking into account the resources that can be obtained from other subsystems. We also note that in a simple case (in particular, in the case considered in this section), the behavior of such a system corresponds to the basic model K(m1 + m2 , n1 + n2 ). However, the proposed approach can be used for more complex cases that cannot be reduced to basic models, in particular, when there are more than two systems and the relationships between them are more complex. For example, some of the subsystems cannot act as “donors” for others, or the number of processors that can participate in restoring the subsystem’s performance is limited (which is discussed in the next section). Also, the proposed model can be useful in cases where it is necessary to obtain separate models for each of the subsystems, and not only the entire system.

Improved GL-Model of Behavior of Complex Multiprocessor Systems

241

4.4 Limiting the Amount of the “Donor” Processors In a real system, the possibilities of using “excess” resources of one subsystem to restore the operability of another may be limited, for example, by bus bandwidth. In this case, no more than a certain number of available “redundant” processors can be used. This feature should also be reflected in the model. To do this, we will solve the following problem. Let there be a vector of length n containing l ones. We need to get a vector of length h, h < n, containing k = min(l, h) ones. It is obvious that the vector considered above contains n − l zeros. The K(n − h + 1, n) model that accepts it as an input will lose ψ(n − h + 1, n − l) edges, i.e. exactly the same number of its edge functions will take zero value. The total number of edge functions of this model according to (1): n − (n − h + 1) + 1 = h.

(5)

Thus, the vector consisting of the values of the edge functions of this model will have a length of h, will contain ψ(n − h + 1, n − l) zeros and, accordingly, ρ(l) = h − ψ(n − h + 1, n − l) ones. So, according to (2):  h, if l > h − 1 . (6) ρ(l) = l, if l ≤ h − 1 Taking into account that l and h are integers, ρ(l) ≡ min(l, h). Thus, in the event that the number of processors of the subsystem that can be used to solve external problems is limited to some number h, it is sufficient to apply the auxiliary vector obtained using the dual model to the model K(n − h + 1, n), where n is the length of this vector. The vector built on the basis of the values of the edge functions of such a model will contain the number of ones equal to the number of “redundant” processors of the subsystem, taking into account the above limitation. This vector can be used in the same way as the vector obtained by the auxiliary dual model was used in the previous cases. Thus, in the event that the number of processors of the subsystem that can be used to solve external problems is limited to some number h, it is sufficient to apply the auxiliary vector obtained using the dual model to the model K(n − h + 1, n), where n is the length of this vector. The vector built based on the values of the edge functions of such a model will contain the number of ones equal to the number of “redundant” processors of the subsystem, taking into account the above limitation. This vector can be used in the same way as the vector obtained by the auxiliary dual model was used in the previous cases. We will also call such a model a constraint model and, to simplify notation, denote it as L(h, n), i.e., L(h, n)  K(n − h + 1, n)

(7)

5 Examples 5.1 An Example of a Model of a Pair of “Donor”-“Recipient” Subsystems Let the system consist of two subsystems. The first subsystem contains 10 processors and is tolerant to the failure of no more than 5 of them, and the second subsystem contains 7 processors and is tolerant to the failure of no more than 3 of them. At the same time,

242

A. Romankevitch et al.

“redundant” processors of the first subsystem can be used to restore the performance of the second. The structure of this system is presented on Fig. 1 (A).

Fig. 1. System structure: A – a pair of “donor”-“recipient” subsystems, B – a “symmetric” pair of “donor”-“recipient” subsystems, C – a system with a limit on the number of “donor” processors

Let’s build models of each of the subsystems. Let the states of the processors of the first subsystem correspond to the variables x1 , x2 , . . . , x10 , and the variables x11 , x12 , . . . , x17 of the second. The first subsystem will correspond to the model K 1 (5, 10) with 6 edge functions, which we denote by f11 , f21 , . . . , f61 . We will also construct an auxiliary dual redundancy model R1 (5, 10) = K1 (6, 10) with 5 edge functions, which we denote by g11 , g21 , . . . , g51 . We will form the auxiliary vector v1 , which will contain the values of the elements of the system state vector that correspond to the states of the processors of the second subsystem, as well as the values of the edge functions of the auxiliary model:   v1 = x11 , x12 , x13 , x14 , x15 , x16 , x17 , g11 , g21 , g31 , g41 , g51 . The second subsystem will correspond to the model K 2 (8, 12), which accepts v1 as an input vector. This model will have 5 edge functions. In order to save space, the cumbersome expressions of the edge functions are not given [17]. Structure of the model is shown on Fig. 2 (A). 5.2 An Example of a Model of a “Symmetric” Pair of “Donor”-“Recipient” Subsystems Consider a system similar to the one considered in the previous section, but in which the second subsystem can also use its “excess” resources to restore the performance of the first subsystem. The structure of the system is presented on Fig. 1 (B). Note that the model of the second subsystem will be exactly the same as in the previous example, so we will limit ourselves to building the model of the first subsystem. Let’s build an auxiliary dual redundancy model R2 (3, 7) = K2 (5, 7) , which will contain 3 edge functions: g12 , g22 , g32 . Let’s form the auxiliary vector v2 , which will contain the values of the elements of the system state vector that correspond to the states of the processors of the first subsystem, as well as the values of the edge functions of the auxiliary model:   v2 = x1 , x2 , x3 , x4 , x5 , x6 , x7 , x8 , x9 , x10 , g12 , g22 , g32 . The first subsystem will correspond to the model K 1 (8, 13), which accepts the vector v2 as an input vector. This model will contain 6 edge features. In order to save space, the cumbersome expressions of the edge functions are not given [17]. Structure of the model is depicted at Fig. 2 (B).

Improved GL-Model of Behavior of Complex Multiprocessor Systems

243

Fig. 2. System models: A – a pair of “donor”-“recipient” subsystems, B – a “symmetric” pair of “donor”-“recipient” subsystems, C – a system with a limit on the number of “donor” processors

5.3 An Example of Limiting the Number of “Donor” Processors For example, consider a system similar to the one considered in Sect. 5.1, but in which the number of “redundant” processors of the first subsystem that can be used to restore the performance of the second subsystem is limited to 3. The structure of the system is presented on Fig. 1 (C). Note that in this case only the model of the second subsystem will be changed. At the same time, the auxiliary dual redundancy model will remain the same as in the previous example. Let’s build an auxiliary constraint model that will limit the number of “redundant“ processors This model L(3, 5) = K1 (3, 5) , which will accept the input vector   1 1used. 1 1 1 u1 = g1 , g2 , g3 , g4 , g5 will contain three edge functions: h11 , h12 , h13 .

244

A. Romankevitch et al.

Let’s form an auxiliary vector u2 , which will contain the values of the elements of the system state vector that correspond to the states of the processors of the second subsystem, as well as the values of the edge functions  of the auxiliary model formed above: u2 = x11 , x12 , x13 , x14 , x15 , x16 , x17 , h11 , h12 , h13 . The second subsystem will correspond to the model K 2 (6, 10), which accepts u2 as an input vector. This model will have 5 edge functions. In order to save space, cumbersome expressions of edge functions are not given [17]. The structure of the model is shown in Fig. 2 (C).

6 Summary and Conclusion The work is devoted to the issue of building GL-models of complex multi-module systems, in which it is possible to transfer some of the tasks of one subsystem to another in order to restore operational efficiency. It is proposed to use the so-called dual model, the edge functions of which are formed from the edge functions of the original GLmodel by replacing conjunction operations with disjunction operations and vice versa. It is shown that such a model can be used as a so-called redundancy model and indicate the number of processors that can be used to restore the performance of other subsystems. It is also shown the possibility of introducing into the model a limitation on the volume of tasks that can be transferred between subsystems, which is achieved by introducing an additional so-called constraint model. Subsystems of the “donor”-“recipient” type are considered, in which one of the subsystems (the so-called “donor”) can use its processors to restore the performance of another (the so-called “recipient”), as well as their variations (with an arbitrary number of subsystems of each type, “symmetric”, with restrictions). For each case, methods of constructing a system model based on a combination of dual and ordinary GL-models are proposed. The application of the proposed approaches for models of some of the considered types is shown using specific examples based on the experiments carried out by the authors.

References 1. Yalcinkaya, E., Maffei, A., Onori, M.: Application of attribute based access control model for industrial control systems. IJCNIS 2, 12–21 (2017) 2. Gokhale, S., Dalvi, A., Siddavatam, I.: Industrial control systems honeypot: a formal analysis of conpot. IJCNIS 6, 44–56 (2020) 3. Drozd, O.V., Rucinski, A., Zashcholkin, K.V., Drozd, M.O., Sulima, Y.Y.: Improving FPGA components of critical systems based on natural version redundancy. Appl. Asp. Inf. Technol. 4(2), 168–177 (2021) 4. Yizhong, F., Yuting, L., Tuo, H., et al.: Design of missile incremental adaptive fault tolerant control system. J. Beijing Univ. Aeronaut. Astronaut. 48(5), 920–928 (2022). (in Chinese) 5. Abbaspour, A., Mokhtari, S., Sargolzaei, A., Yen, K.K.: A survey on active fault-tolerant control systems. Electron. 9(9), 1513 (2020) 6. Arfat, Y., Eassa, F.E.: A survey on fault tolerant multi agent system. IJITCS 9, 39–48 (2016)

Improved GL-Model of Behavior of Complex Multiprocessor Systems

245

7. Wason, R., Soni, A.K., Qasim Rafiq, M.: Estimating software reliability by monitoring software execution through OpCode. IJITCS 7(9), 23–30 (2015) 8. Thomas, M.O., Rad, B.B.: Reliability evaluation metrics for internet of things, car tracking system: a review. IJITCS 9(2), 1–10 (2017) 9. Kuo, W., Zuo, M.J.: Optimal Reliability Modeling: Principles and Applications, p. 560. Willey, Hoboken (2003) 10. Belfore, L.A.: An O(n(log2(n))2) algorithm for computing the reliability of k-out-of-n:G & k-to-l-out-of-n:G systems. IEEE Trans. Reliab. R-44, 132–136 (1995) 11. Rushdi, A.M.: Utilization of symmetric switching function in the computation of k-out-of-n system reliability Microelectron. Reliab. 26(5), 973–987 (1986) 12. Chang, G.J., Cui, L., Hwang, F.K.: Reliabilities for (n, f, k) systems. Stat. Probab. Lett. 43, 237–242 (1999) 13. Sfakianakis, M., Kounias, S., Hillaris, A.: Reliability of a consecutive k-out-of-r-from-n: F system. IEEE Trans. Reliab. 41(3), 442–447 (1992) 14. Hwang, F.K.: Simplified reliabilities for consecutive k-out-of-n systems. SIAM J. Algebraic Discrete Methods 7(2), 258–264 (1986) 15. Lee, W.S., Grosh, D.L, Tillman, F.A., Lie, C.H.: Fault tree analysis, methods and applications: a review. IEEE Trans. Reliab. 34(3), 194–203 (1985) 16. Romankevich, A., Feseniuk, A., Maidaniuk, I., Romankevich, V.: Fault-tolerant multiprocessor systems reliability estimation using statistical experiments with GL-models. Adv. Intell. Syst. Comput. 754, 186–193 (2019) 17. Romankevich, V.A., Potapova, E.R., Hedayatollah, B., Nazarenko, V.V.: GL-model of behavior of fault-tolerant multiprocessor systems with a minimum number of lost edges. Visnyk NTUU KPI Inform. Control Comput. 45, 93–100 (2006). (In Russian) 18. Romankevich, A., Romankevich, V., Morozov, K.: On one GL-model of a system with a sliding redundancy. Radio Electron. Comput. Syst. 5, 333–336 (2013). (in Russian) 19. Romankevich, A.M., Morozov, K.V., Romankevich, V.A.: Graph-logic models of hierarchical fault-tolerant multiprocessor systems. IJCSNS 19(7), 151–156 (2019)

A Method for Evaluating the Efficiency of the Semantic Kernel in the Internet Promotion Channel Sergey Orekhov(B) , Andrii Kopp, Dmytro Orlovskyi, and Tetiana Goncharenko Intelligent Information Technologies and Software Engineering Department, National Technical University Kharkiv Polytechnic Institute, Kharkiv 61002, Ukraine [email protected]

Abstract. Over the past twelve years, search engine optimization has become a classic technology for product promotion. Its tools have also become classics, such as Google AdWords advertising, HTML code optimization and building the semantic core of a WEB resource. However, recently they no longer give significant results, and their efficiency drops. We completed several WEB projects in 2020– 2021, the results of which show a catastrophic drop in the efficiency of classic SEO tools by at least 50%. The main reason for this is that the online market is constantly changing due to competition. In our study, we propose to look for ways to improve efficiency based on the reengineering of existing tools. To implement the reengineering procedure, it is necessary to evaluate the effectiveness of existing SEO tools. Therefore, in this paper, we focused on the problem of evaluating the effectiveness of the semantic kernel of digital content as the most common approach in SEO. It is based on the method of analyzing the current situation and its automatic evaluation. By the situation, we mean a set of WEB documents that contain a semantic kernel. Therefore, the evaluation of the efficiency of the kernel must be performed both from the point of view of the document and from the point of view of the keyword. Further, such an assessment should also include a measure of the quality of the semantic kernel from the position of the search server. Such a server offers two evaluation tools. The first is WEB statistics, that is, how many visits this kernel attracted. Another way is the structure of the kernel itself, that is, what keywords are included in its composition. And the third component of the assessment is the budget or financial resources that need to be spent on the synthesis and implementation of the kernel on the Internet. Based on the results of this modeling, the article shows a real example (WEB project in the American online services market), in which this evaluation procedure was applied. Keywords: Semantic Kernel · Virtual promotion · Search engine optimization · Customer journey map

1 Introduction For five years, we have been researching the effectiveness of classic SEO tools, in particular, methods for synthesizing semantic kernels. Completed WEB projects concerned the market of Ukraine and the USA. The results obtained show a catastrophic decrease in © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 246–259, 2023. https://doi.org/10.1007/978-3-031-36118-0_22

A Method for Evaluating the Efficiency of the Semantic Kernel

247

the effectiveness of methods for implementing the semantic kernel. For example, a WEB project for the wood products market in Ukraine showed that a simple implementation of a semantic core in parallel with code optimization did not give a single online order for half a year. These techniques increased the attendance of the WEB resource, but without result. At the moment, there is no alternative to SEO. Therefore, the only way to improve the efficiency of product promotion on the Internet is to reengineer classic SEO techniques. It is known that reengineering includes three stages: assessment, change and planning. In this paper, we have focused our attention on the assessment stage. As an object of study, we have chosen the process of introducing the semantic core of digital content. Our goal is to create a methodology for evaluating the effectiveness of the semantic core of digital content. Consider a review of scientific publications on this issue and start with the concept of digital content.

2 Related Works Nowadays, digital content has become a real product that is actively bought and sold. Digital content describes real goods on the Internet and exists in parallel with them. Digital content performs several tasks. All of them are divided into two groups depending on the two roles played by the end user on the Internet. The first role of the user is the seller of digital content, and the second role is the buyer. In both roles, digital content is an intermediary. In turn, according to the classical theory of marketing, digital content plays the role of a message in the marketing channel [1–4]. That is, it is a message between the buyer and the seller. Such a message informs the buyer that a digital product or service is available to satisfy a particular need. The reason for the creation of both digital content and digital marketing was a significant change in the scheme of selling goods and services. Earlier we lived in the conditions of a post-industrial society. The schema of business was “goods – money – goods” and it existed because there was a limitation in the quantity of goods. However, with the advent of globalization and the concept of open markets, the goods are in abundance, and now it is difficult to find a buyer for these numerous goods and services. Thus, the main problem was finding a solvent buyer. That is, a new paradigm was formed: attract – earn – spend. Under new conditions, a digital product (content) becomes a combination of three elements: an audience, a communication channel (promotion) and a target message (digital content). The following restrictions are imposed on digital content by the Internet: – First, the digital description of the product is a set of keywords that are scanned by the search engine. The server forms answers according to a certain algorithm (Page Rank). Therefore, an important factor is the first position in the answers of a search server, for example, Google [5]. – Secondly, the format of the digital product description becomes the HTML language. It allows you to select keywords that the search server responds to. – Third, the search server registers the number of links to the web page where our digital content is hosted.

248

S. Orekhov et al.

– Fourthly, end users work massively in social networks. They have a technological connection with the search server and built-in search algorithms. – Fifth, mobile applications allow you to connect to social networks. To manage digital content a new business process scheme called virtual promotion was proposed (Fig. 1) [6, 7]. The analysis of the new process of virtual promotion shows that the digital content moving in the channel was called the semantic kernel (SK) [8, 9]. This is a new object for research. Let us consider its properties. The first step in the implementation of virtual promotion is the synthesis of the semantic kernel. The success of virtual promotion depends on its effectiveness. The works [8, 9] define the semantic kernel as a set of keywords, phrases and morphological forms that most accurately characterize the type of activity, goods and services offered for sale. For the first time, the term “semantic kernel” was voiced in the methodology of search optimization [10]. Here, too, the semantic core is defined as a set of keywords that should be included in digital content to improve the search engine side of a web page. However, in the methodology of search optimization, the semantic kernel is considered as a duplicate of a potential request from a user in the form of a request to a search server. In other words, the semantic kernel is a targeted set of keywords that tells the search server what page to display to the user in response to his query. Moreover, the query must also contain a similar set of keywords. That is, the semantic kernel is a potential request from a user on the Internet.

Fig. 1. The process of virtual promotion

The research conducted in [11] expands the concept of the semantic kernel. Here, based on many years of research into various forms of interaction between users and the search server, it is proposed to consider the semantic core as a concise description of a product or need circulating on the Internet. If you correctly create a core and place it in the right points (nodes) of the Internet, you can create a larger flow of customers for the enterprise. In addition, it has already been proven that digital content hides marketing

A Method for Evaluating the Efficiency of the Semantic Kernel

249

information that is useful not only for the buyer, but also for the seller and his competitors [12]. In this way, SJ begins to play a much larger role than just copies of the user’s request in the network. It becomes a message in the format of an HTML document, which can potentially be responded to by a buyer on the Internet. The effectiveness of the entire promotion process on the Internet becomes dependent on SK. As a result, the task of evaluating the effectiveness of the semantic kernel, as a concise annotation of knowledge about the product and the need it covers, becomes relevant. In which case is it necessary to solve this problem? First, the performance evaluation model allows you to manage the semantic core, that is, to change it if its performance falls or is insufficient. This is shown in work [13], where the assessment of the quality of the SI is the basis for solving the management problem. The second case, where the semantic core efficiency model is appropriate, is contextual advertising [3, 4, 14]. Here, the effect of the kernel on the advertising message is analyzed: it attracts the required number of buyers of the product. The third case, where the efficiency of the kernel is analyzed, is the thematic search or thematic traffic on the Internet [15]. Here SK is embedded in the WEB page and the quality of the core depends on whether the page will be in the TOP or not. In this case, the terms quality and effectiveness of the SK should be considered as synonyms. Thus, the assessment of effectiveness is the central essence of various channels of promotion of goods on the Internet which is confirmed by the data (Fig. 1). To determine the quality or effectiveness of a message in a marketing channel, the following indicators are identified that affect its use [16]: – – – – – –

message display frequency; demand or the number of effective actions of the buyer in response to the message; competitiveness of the request, where there is a message; rejection rate (the number of visitors who did not want to view the message); seasonality of the request, which affects the display of the message; geographical dependence of the request, where the message is displayed.

As can be seen from the indicators listed above, they all describe the functioning of the message in the promotion channel. Let us assume that SK is embedded in the message in the promotion channel. Then we will consider the metrics that are already available for evaluating the message in the channel. The classic definition in the marketing theory says that the goal of a product promotion is to increase the efficiency of demand and sales [1]. The content of promotion consists in communication, that is, to inform potential customers about the product, its packaging, brand, advertising exhibitions, demonstrations, relation to the enterprise, etc. As a result, it is clear that the effectiveness of sales depends on the effectiveness or attractiveness of the message about the product that reaches potential customers (buyers). That is, the effectiveness of promotion is determined by the effectiveness of communication in the communication channel between the company and buyers. Currently, the main key performance indicator that can give an answer about the effectiveness of a message on the Internet will be traffic [3]. It is the number of visits to a WEB page with a semantic kernel or the number of visits to a WEB site through an

250

S. Orekhov et al.

advertising message with a semantic kernel that forms the number of unique users per unit of time. Based on the “traffic” metric, the first metric is formed - the traffic indicator. It determines the conditional efficiency of the channel and message [3]: Efficiency =

traffic2 . budget

(1)

Another classic indicator of the effectiveness of the promotion channel is the return on investment indicator [3]: ROI =

profit − exp enses . exp enses

(2)

The next important marketing funnel metric is conversion: Conversion =

buyers . visitors

(3)

That is, it reflects the achievement of the goal - to attract the required number of buyers of the product. In parallel with this, other indicators can be found in the literature, for example: the number of refusals, the cost of attracting one buyer, the average check, the number of repeated visits. But all the above indicators are only metrics, unfortunately, no stable efficiency criterion has been defined among them. The modern theory of Internet promotion is based only on the metric approach. Thus, there is an urgent task of evaluating the effectiveness of the semantic kernel in the promotion channel created on the Internet. Let us consider its statement.

3 Problem Statement To formulate the research problem, it is necessary to analyze the requirements for functioning and quality from the end user. In other words, it is required to understand how, when and with what result the user will apply this technique for evaluating the semantic kernel. To do this, we will construct a diagram of use cases (Fig. 2) in the UML language [17, 18]. The evaluation process will be realized thanks to the implementation of two roles: the developer and the user. It can be seen that the product promotion cycle must be managed (activated and stopped). It is also necessary to adjust the digital content, SY and a set of metrics during the entire time of its execution. It is also clear that management is based on evaluations. It is clear that the promotion cycle is generally unlimited [6] because competitors also have every chance to use either search engine optimization techniques based on the semantic kernel or virtual promotion. As a result, there is a situation of competition of semantic cores in the promotion channel on the Internet. Moreover, the victory is won by SK with the highest efficiency, which means that the company itself wins over its competitors. Thus, as a result of the

A Method for Evaluating the Efficiency of the Semantic Kernel

251

Fig. 2. Use cases of evaluation process

high efficiency of SK, there will be a maximum of calls and online payments. Now it is possible to formulate the statement of the research problem. According to the definition of SK, its essence is a set of keywords contained in an HTML document. Therefore, efficiency should also be determined on the basis of keywords and document evaluation. The classic metric for evaluating the quality of a keyword is the C-value metric:  not included log2 |a|f (a),  (4) C − value = log2 |a|(f (a) − P(T1 a ) b∈Ta f (a)), included where a is the keyword, |a| is the length of the keyword, f (a) is the frequency of occurrence of the keyword a,  Ta is the set of elements that include the word a, P(Ta ) is the number of elements in Ta , f (a) is the sum of frequency elements b ∈ Ta , including a. That is, a is part of the sentence b. This metric (4) is based on the fact that the keyword exists within the text corpus. A text corpus is a set of HTML documents describing a given digital content. That is, the input information for evaluation is a text corpus containing a semantic kernel. We will assume that the text corpus is a set of documents, and each document, in turn, is a set of sentences and the sentences themselves contain key words. Thus, the first assessment of effectiveness will be on the side of the quality of the semantic kernel, that is, on the side of the validity of the semantic kernel in terms of the structure and composition of keywords. In this case, metric (4) is the most suitable. Thus, the task of evaluating the effectiveness of the semantic kernel consists in stepby-step evaluation first of HTML documents from the text corpus, and then of keywords from sentences. Then our evaluation should be carried out in two stages. First, you need

252

S. Orekhov et al.

to evaluate the documents and choose the highest quality document. Then, if on the basis of this document, it is possible to get a high-quality composition of keywords which are the cores with the maximum quality that will be used when performing search optimization or virtual promotion procedures [6]. The other side of the evaluation will be the marketing side, for example, metric (1) and WEB statistics – metric (2). Then, the task of evaluating the effectiveness of the SK can be formulated as follows: through the evaluation of the corpus of HTML documents, the digital content that should be applied to the synthesis of the semantic kernel was identified, and then the conditional effectiveness and conversion of the semantic kernel in the promotion channel on the Internet were evaluated.

4 Proposed Approach The paper proposes the following scheme for solving the problem of evaluating the effectiveness of the semantic kernel (Fig. 3). The scheme is presented as an action diagram in UML. The solution diagram (Fig. 3) shows two cycles of data processing. In the first, we operate on a corpus of documents, and in the second, we evaluate keywords from the highest quality document. Based on the algorithm Page Rank [19], we assume that a collocation describes an ideal Internet query that will lead a potential buyer to the desired product. That is, collocation describes the ideal semantic kernel of a product on the Internet. Then the presence of collocation in the document indicates the high efficiency of this digital content and, accordingly, the semantic kernel in the promotion channel. Therefore, initially, we propose to select documents with the highest quality of keywords, and then check whether it is possible to build at least a third-order collocation based on the selected documents. Why collocations are of the third order? In work [12], it was shown that in order to build an ideal query on the Internet, a user needs a query that consists of three keywords. Moreover, each word answers one of the questions asked. Only such a query structure leads to the maximum efficiency of Page Rank algorithm [5]. Therefore, if a collocation or an ideal query is hidden in our document, then Page Rank algorithm will work with maximum efficiency. Let us consider the SK model for building an evaluation method. Taking into account the proposal to evaluate the SK from three sides: document and keyword evaluation, conditional efficiency evaluation, and conversion evaluation, let us first consider the modeling of the “document-keyword” based on the text model shown in [20]. We will assume that at the input we have a text corpus of documents D, presented as HTML text: D = {s1 , ..., sL },

(5)

where si , i = 1, L is a sentence ending with a comma or other end-of-sentence sign. Each sentence si is a combination of three sets: a set of keywords to describe the main

A Method for Evaluating the Efficiency of the Semantic Kernel

253

Fig. 3. Schema of the solution

content, a set of sentence-ending characters, and a set of keywords that expresses the general meaning [20, 21]: si = K, G, E i = 1, L

(6)

where K are keywords that form the product description on the Internet, G are keywords of general content, E are end-of-sentence symbols. Next, we introduce the operation of metric evaluation of documents: M : D → DM . M is a set of metrics to evaluate a document D. Let’s introduce the normalization operation N, when we exclude the set E, and N. transform the other keywords to the infinitive: N : DM → DM In work [12], when the user’s request should include the answer to three questions (what, where, when) to describe the product. Then trigram models can be used to describe

254

S. Orekhov et al.

the semantic kernel, as shown in [21]. After that, if the kernel has the form of a trigram model, then its efficiency is maximum. Thus, if the kernel has a trigram in its composition, then the efficiency is the highest. The MI measure and its modifications are the most often used to identify terms as collocations: MI = log2

f (a, b)N , f (a)f (b)

(7)

where N is the number of words in the text document. f (a, b) is the frequency of cooccurrence of words a and b, which evaluates the degree of dependence of the occurrence of two words in the corpus on each other. f (a), f (b) – frequency of occurrence of words a and b separately from each other. If the detected two-word collocations are considered as a whole, then with the help of the mentioned measures, longer word combinations (three-word, four-word, etc.) can be recognized in the text, which makes it possible to extract long terms with an arbitrary syntactic structure using statistical criteria. Thus, two mechanisms are established. The first is to find the correct document with a semantic kernel, and the second is to detect collocations. If they exist, then interaction with them is more effective than others.

5 Proposed Method Let us describe verbally the effectiveness of the evaluation method. As input information, we have a text corpus with document D, which is presented in the form (5). First, we conduct an evaluation operation of documents in the corpus. Next, the normalization operation must be applied to the set of keywords in selected document. This set will be the source of semantic kernels. That is, all non-essential symbols and keywords are removed from the texts of each document where the SK is hidden. Next, we form a set of keywords for each document where SK is hidden. Then, using the metric (7), we check which order the collocations have in the selected documents. If there are collocations of the second and third order in the documents, then the effectiveness of such a system will be the highest from the point of view of the ideology of linguistic analysis of texts [12]. The verbal description of the method coincides with the scheme for solving the research problem shown in Fig. 3. Thus, evaluating the document and finding the second- and third-order collocations will be the first evaluation cycle. Next, it is proposed to make an assessment from the side of marketing theory and WEB statistics. Then, if the semantic kernel passes this evaluation, it has the highest level of efficiency. To implement the second evaluation cycle, we suggest using metrics (1)–(4). The specified algorithm makes it possible to evaluate SK both from the side of the theory of texts, and from the side of the theory of marketing and WEB statistics. In addition, our evaluation is based on both the document evaluation and the SK. This is an important point, because the search server operates with the concept of document and keyword, and the user builds his queries on the basis of collocations and WEB statistics.

A Method for Evaluating the Efficiency of the Semantic Kernel

255

Each buyer analyzes the number of visits to the page of the product that he wants to buy, as well as the number of reviews, i.e. refusals and so on. The advantage of this method is that it is the implementation of the first stage of reengineering of the classic SEO tool – the semantic kernel. At the same time, it takes into account both the aspect of evaluating digital content from where the kernel is extracted, and the aspect of the influence of the search engine on this SEO tool. Let us consider an example of using the specified evaluation method to understand the effectiveness of the real semantic kernel in the process of product promotion on the Internet.

6 Experiment To test the method of evaluating the effectiveness, the paper considered a WEB project to promote online services in the USA: www.celestialtiming.com. Currently, the project is fully operational and is constantly online. According to the solution scheme (Fig. 3), at the first stage we analyze the text corpus of our WEB project. The corpus includes N documents. To do this, all documents must first be uploaded and evaluated. The following document metrics M = {A, R, S, Q, Z} were selected for evaluation [20, 21]: Verbs , N

(8)

Verbs , Nouns + Adjs + Pr onouns

(9)

Nouns + Pr onouns , Verbs + Adjs

(10)

Adjs + Advs , Nouns + Verbs

(11)

Pr epositions . Sentences

(12)

A= R=

S=

Q=

Z=

Based on metrics (8)–(12), the quality of each document in the corpus was calculated (Table 1). The results show that the semantic kernel of this web project exists on the “contacts” page. It is necessary to check now whether it is possible to extract the kernel from the content of the “contacts” page. An example of content is presented in Table 2. We have found that in this digital content there are collocations of the second order (Fig. 4). This means that the semantic kernels that can be synthesized will include collocations of the second order. It is also seen that it is potentially possible to obtain collocations of the third order. Now we need to compare how the kernels behave in terms of WEB statistics. To do this, we use the data of the WEB service – Google Analytics (Fig. 5).

256

S. Orekhov et al. Table 1. Evaluating the documents in a text corpus

No

Link

A

R

S

Q

Z

Sum

1

home

1.336

0.265

0.127

0.347

0.417

2.491

2

celebrity

1.196

0.339

0.138

0.415

2.125

4.213

3

myself

0.778

0.455

0.152

0.455

2.0

3.838

4

another

0.636

0.5

0.182

0.5

2.0

3.818

5

saved

0.83

0.451

0.152

0.403

1.609

3.445

6

applications

1.071

0.287

0.13

0.477

2.764

4.728

7

free

1.263

0.407

0.125

0.303

3.4

5.499

8

contact

1.5

0.469

0.057

0.111

7.667

9.803

9

account

0.667

0.455

0.15

0.471

2.25

3.991

Table 2. The fragment of digital content of tested WEB project No Link 1

Content

contact CelestialTiming is the product of international collaboration with a goal of providing site users with an accurate and easy tool to bring them in touch with their natural rhythms of the universe and enable them to make more effective personal and business decisions Members of the CelestialTiming team have extensive experience and advanced degrees in the fields of physics, engineering, computer science, psychology, and education. They have developed professional psychological and astrological software, taught in colleges and universities, and published books and articles on a variety of topics

Fig. 4. Searching second order collocations

There are three spikes on the visit chart. They correspond to three variants of the semantic kernel that have been launched on the Internet. Thanks to our evaluation method, three points of decrease in the efficiency of the kernel were identified.

A Method for Evaluating the Efficiency of the Semantic Kernel

257

Fig. 5. WEB statistics of tested WEB project

Three variants of the semantic kernel were successively launched on the Internet. However, all these variants were synthesized on the basis of second-order collocations. This is confirmed by the table data (Table 2). Also, these data show that it is difficult to build third-order collocations for this case of digital content. This means that the efficiency of such cores is low. Therefore, if we take into account the effect of aging of nuclei, which was described in [8], then such kernels are unreliable and quickly age. This fact for this WEB project is also confirmed by WEB statistics (Fig. 5). Therefore, we recommended the owners of this WEB project to change the digital content so that it is dominated by collocations of the third and fourth order. What kind of collocations these should be can be seen from the data in Fig. 4. To ensure the accuracy and reliability of the results obtained, three WEB projects were analyzed in this work. Two projects prove the need to develop this assessment method. The third project, as the most complete and versatile, is shown above. All projects initially used the classical method of semantic kernel synthesis, which led to long inefficient cycles of product promotion on the Internet. But after applying a new method for evaluating the efficiency of the kernel, noticeable improvements were obtained. The accuracy and validity of the new method is proven by the fact that all three projects were carried out for different markets and market segments, both in terms of geography and in terms of products to be promoted.

7 Conclusion Among the main tools of search engine optimization on the Internet, work with semantic kernels occupies a central place. However, it is not enough to simply synthesize the kernel based on WEB services. The state and efficiency of the synthesized kernel must be constantly monitored. The fact is that we discovered the effect of aging of the semantic kernel [20, 21]. It lies in the fact that the kernel repeats the product life cycle model in the Internet according to the classical marketing theory. Therefore, the problem becomes determining at what stage of this life cycle the kernel is. An example of the kernel life cycle is perfectly visible in Fig. 5. Therefore, to reveal changes in the cycle we need to check the efficiency of the kernel. The paper proposes an approach that is based on evaluating the efficiency of the kernel both from the standpoint of the Page Rank algorithm and from the standpoint of WEB statistics.

258

S. Orekhov et al.

First, we check whether the synthesized kernel contains collocations. It is the presence of a third-order collocation that gives the maximum effect from the kernel. If, as in our test, since only second-order collocations are stable in the project, such kernels are characterized by the minimum lifetime in the Internet. This fact is also confirmed in Fig. 5. The second part of the assessment method relies on life cycle analysis through WEB statistics. Therefore, our article considers a generalized method for evaluating the effectiveness of the kernel in order to identify points of change in the life cycle. This is an urgent task, the solution of which will increase the effectiveness of search engine optimization. Based on the results obtained, the following conclusions can be drawn: 1) The practical application of the new evaluation method lies in the fact that it can be used to recognize the moment when the semantic kernel changes either to a new one or to a changed one. 2) The implementation of this method helps the content manager to understand the relationship between the life cycle of the semantic core, WEB document and WEB statistics. Until now no such relationship has been found in scientific publications. And since competition on the Internet is high, without noticing the fact that competitors use third-order collocations, we risk to lose in this market segment. 3) The development of a new evaluation method opens up prospects for applying the theory of situational management. This is a smart control class. Therefore, it is advisable to implement this control method in automatic mode. Then you can manage the process of promotion on the Internet using artificial intelligence systems [9]. The purpose of our further work is the software implementation of a special component of the semantic kernel. Its creation is planned on the platform NodeJS. Such a component will work as a machine learning system that will accumulate information about successful kernels in order to automatically generate new ones. At the same time, without a method for evaluating the effectiveness of the core, which works at the level of a keyword, document, and metric, it is not possible.

References 1. Kotler, P., Keller, K.L.: Marketing Management. Prentice Hall, USA (2012). 812 p. 2. Sihare, S.R.: Image-based digital marketing. Int. J. Inf. Eng. Electron. Bus. 5, 10–17 (2017) 3. Hashimova, K.K.: Analysis method of internet advertising-marketing information’s dynamic changes. Int. J. Inf. Eng. Electron. Bus. 5, 28–33 (2017) 4. Hashimova, K.K.: Development of an effective method of data collection for advertising and marketing on the internet. Int. J. Math. Sci. Comput. 3, 1–11 (2021) 5. Ziakis, C., Vlachopoulou, M., Kyrkoudis, T., Karagkiozidou, M.: Important factors for improving Google search rank. Future Internet 11(32), 3–14 (2019) 6. Orekhov, S.: Analysis of virtual promotion of a product. In: Hu, Z., Petoukhov, S., Yanovsky, F., He, M. (eds.) ISEM 2021. LNNS, vol. 463, pp. 3–13. Springer, Cham (2022). https://doi. org/10.1007/978-3-031-03877-8_1 7. Orekhov, S., Malyhon, H.: Metrics of virtual promotion of a product. Bulletin of the National Technical University “KhPI”. Series: System analysis, control and information technology. NTU «KPI», Kharkiv, vol. 2, no. 6, pp. 23–26 (2021)

A Method for Evaluating the Efficiency of the Semantic Kernel

259

8. Orekhov, S., Malyhon, H.: Method for synthesizing the semantic kernel of web content. In: CEUR-WS, vol. 3171, pp. 127–137 (2022) 9. Orekhov, S., Malyhon, H.: Method of solving the problem of situational control of semantic kernel of web content. Comput. Syst. Inf. Technol. 2, 6–13 (2022) 10. Clarke, A.: SEO 2016. Learn search optimization with smart internet marketing strategies. DigitalBook.Guru, USA (2016). 140 p. 11. Orekhov, S.: An alternative way to search engine optimization. Inf. Technol. Sci. Eng. Technol. Educ. Health (2022). Abstracts of reports of the XXX international scientific and practical conference Micro-CAD-2022 (19–21 October 2022). Kharkiv. Ukraine: NTU «KPI», 2022. – P. 804 12. Orekhov, S., Malyhon, H., Liutenko, I., Goncharenko, T.: Using internet news flows as marketing data component. In: CEUR Workshop Proceedings, vol. 2604, pp. 358–373 (2020) 13. Godlevsky, M., Orekhov, S., Orekhova, E.: Theoretical fundamentals of search engine optimization based on machine learning. In: CEUR Workshop Proceedings, no. 1844, pp. 23–32 (2017) 14. Mezzi, M., Benblidia, N.: Study of context modelling criteria in information retrieval. Int. J. Inf. Technol. Comput. Sci. 3, 28–39 (2017) 15. Pawade, D.Y.: Analyzing the impact of search engine optimization techniques on web development using experiential and collaborative learning techniques. Int. J. Mod. Educ. Comput. Sci. 2, 1–10 (2021) 16. Bendle, N.T., Farris, P.W., Pfeifer, P.E., Reibstein, D.J.: Marketing Metrics. The Manager’s Guide to Measuring Marketing Performance, 3rd edn. Pearson Education, Inc., London (2016). 456 p. 17. Zaretska, I., Kulankhina, O., Mykhailenko, H., Butenko, T.: Consistency of UML design. Int. J. Inf. Technol. Comput. Sci. 9, 47–56 (2018) 18. Pathak, N.: An empirical perspective of roundtrip engineering for the development of secure web application using UML 2.0. Int. J. Intell. Syst. Appl. 5, 43–54 (2017) 19. Srivastava, A.K., Garg, R., Mishra, P.K.: Discussion on damping factor value in PageRank computation. Int. J. Intell. Syst. Appl. 9, 19–28 (2017) 20. Orekhov, S.: Advanced method of synthesis of semantic Kernel of E-content. In: CEUR-WS, vol. 3312, pp. 87–97 (2022) 21. Orekhov, S., Malyhon, H., Goncharenko, T.: Mathematical model of semantic Kernel of WEB site. In: CEUR Workshop Proceedings, vol. 2917, pp. 273–282 (2021)

Automated Dating of Galaktion Tabidze’s Handwritten Texts Tea Tvalavadze1 , Ketevan Gigashvili2 , Esma Mania3 , and Maksim Iavich4(B) 1 Giorgi Leonidze State Museum of Georgian Literature, 8 Gia Chanturia Street, 0108 Tbilisi,

Georgia 2 Iakob Gogebashvili Telavi State University, 1 Georgian University Street, 2202 Telavi, Georgia 3 Korneli Kekelidze Georgian National Centre of Manuscripts, 1/3 Merab Aleksidze Street,

1093 Tbilisi, Georgia 4 School of Technology, Caucasus University, 1 Paata Saakadze Street, 0102 Tbilisi, Georgia

[email protected]

Abstract. The study of Georgian history greatly involves the fundamental investigation of the handwritten texts. The scientists use different methods for this investigation, but they face different problems. It is related to the fact that the time of the text creation of manuscript in the most cases differs from the date of the manuscript creation and the identification of each of them needs special approach. The automatic dating approaches of the manuscripts is now very common in practice, and it involves the work of computer scientists. The paper studies the approaches of the dating the manuscripts and offers the novel methodology for dating Georgian ones using neural networks. The authors offer the neural network and train it with the existing data. The system is trained using the classified writings of Galaktion Tabidze, which is the greatest Georgian poet of XX century. The system is needed for dating his manuscripts, which are not dated yet. It is very important for the study of Georgian history. The paper also offers the trained model, which can classify Georgian handwritings of Galaktion Tabidze. The received system has the accuracy score, which is considered good in comparison with the related research in this field. The offered approach can be used to date the handwritings of different authors. Keywords: Automatic dating approach · Neural network · Automated handwriting dating

1 Instruction Georgia has 16-century-old written heritage and as far as any kind of historical research, be it social, political, cultural etc. is based on written sources, our awareness of Georgian history greatly depends on the fundamental investigation of the handwritten texts. One of the most basic aspects of the researches is dating of the manuscript sources and scholars use different methods for this purpose. The time of the text creation often differs from the date of the manuscript creation and identification of each of them needs special approach. e.g. dating of the text based on the information extracted from its contents cannot always © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 260–268, 2023. https://doi.org/10.1007/978-3-031-36118-0_23

Automated Dating of Galaktion Tabidze’s Handwritten Texts

261

be applied to the manuscript dating. On the other hand, in cases when manuscripts are the first original copies of the texts, the dates of the text and the manuscript are the same. Dating of such handwritings also means the text datation and it is very important. When we identify the time of the text creation, we specify at least the upper time limit of all the happenings described in it. It is one of the most challenging issues of historical studies as without knowing the time of the events we can’t appoint them to their particular historical context or define their exact place in the biographical timeline. While working on the archival materials of different writers and historical figures we have often come to the dead end with the dating of undated holographs. It generally happens when the contents of the texts do not contain any information that could be appointed to the particular period of time. In such cases, the only way to solve the problem is the manuscript analysis. In order to identify specific features of one and the same author’s handwriting according to the years we need to analyze a great number of his/her manuscripts belonging to each chronological year (paradigmatic row). Georgian manuscript repositories are full of undated documents that contain important information but as they are undated, they can’t be appointed to the particular period of time and can’t be relied on in the scientific researches. Of course, we would be glad to date each of them, but the automated dating is based on statistics and therefore the great amount of samples in the datasets that should be investigated (including training sets, validation sets and testing sets) is very important. In order to use this method for manuscript dating we had to choose such an author whose archive could provide enough material needed for such researches. The largest Georgian manuscript repositories: Historical Archives of Georgia, Korneli Kekelidze National Institute of Manuscripts and Giorgi Leonidze National Museum of Georgian Literature include a great number of personal archives among which one of the richest is the archive of the XX century greatest Georgian poet Galaktion Tabidze. He had been writing poems for more than half a century and at the same time was very productive. His archive includes dated manuscripts for quite a long period from 1908 to 1959 and there is a great number of undated manuscripts too. The main objective of the research is to offer the methodology of the dating of Georgian handwritings, to offer the corresponding model and to train it according the dated samples. The most existing solutions offer the manual dating, which is very time consuming. Therefore, the automated solution must be offered. In the papers we offer the automated dating solution using Neural Network. The main limitation of our approach is that the accuracy score is not high enough, but is still considered good in the comparison with the related researches, but in the different languages. In the paper the authors try to achieve the maximal possible accuracy score.

2 Related Works Dating of the literary pieces is extremely important for literary studies as it often helps us to identify the main idea of the text. e.g. dating of Galaktion Tabidze’s poem “Let the banners wave on high” proved that it couldn’t be dedicated to the 1917 October revolution in Russia as it was written before it. It became clear that the idea of its connection with the revolution was promoted by the soviet ideologized critics. It was supposed that the first variant copy of Galaktion Tabidze’s another poem “Aspindza” was dedicated to a

262

T. Tvalavadze et al.

woman (in some researches to Meri Shervashidze and in the others – to Olia Okudjava) and the feelings represented in it were romantic but dating of the text made it clear that the poem was patriotic from the very first variant copies and was dedicated to the Georgians’ victory over the Ottoman Empire in the Battle of Aspindza [1, 2]. Nowadays the use of digital technologies has greatly enhanced the possibilities of textual and manuscript studies. Using the grapheme-based method, scientists of Groningen University have dated Dead Sea Scrolls. Computer technologies and software enable automatic dating of handwritten manuscripts but, of course, there should be the data, chosen by scholars as the basis for such an analysis. Gustave Flaubert’s undated manuscripts have been dated with an accuracy of about 62% and as the scholars say, “incorporation of newer methods of handwriting analysis can improve the accuracy of automatic datation” [3, 4]. Samson Ullmann used the grapheme-based method to date Matthew Arnold’s manuscript of the poem “Dover Beach”. He considers that there should be at least three thousand words for each year obtained from dated manuscripts and at least two lines for each letter of the alphabet to get accurate results. Tara Gilliam in her research “Writer identification in medieval and modern handwriting” suggested “three novel methodological approaches to the analysis of writer identification data” [5–7]. Miroslava Bozekova made experiments on 100 images from 40 different writers and achieved 96.5% accuracy in writer verification. Manoj Kumar Sharm and Vanshika Chanderiya created a writer identification framework which measures “the hand pressures during script writing using identical grapheme and writing strokes and then generates the pressure descriptors”. They came to the conclusion that this method “gives the encouraging results and average success rate of the descriptors with the distinct alphabets is 98%”. Kalthoum Adam, Somaya Al-Maadeed and Younes Akbari suggest a novel approach that improves classical dating methods by applying feature-level hierarchical fusion” and shows that “that applying a hierarchical fusion based on subsets of multi features in the KERTAS dataset can obtain promising results” [8–11]. There are lots of other projects, publications and novelties that prove effectiveness of digital technologies in these kinds of researches. There are some new approaches of dating the data using different machine learning techniques [12–15].

3 Automated Handwriting Dating Georgia is making its first steps in this direction and has to do a lot to catch up with the processes. Automated handwriting dating as a method has never been used by our scientists before but being aware of those very interesting projects carried out in different countries for the author identification and handwriting dating we always desired to practice similar approaches on Georgian data. These two aspects of research are based on the same principles and both of them need voluminous databases with lots of highresolution binarized images (manuscript images should be cleaned of the background noise). For the authorship identification scholars use algorithms and training sets to train the network investigate holographs of a great number of authors and by means of feature vectors, codebooks and self-organizing feature maps (SOFM) identify as many betweenwriter variations or so called writer-informative features in them as possible. The more discriminative these classifiers are the higher is the verification ratio of the results. After

Automated Dating of Galaktion Tabidze’s Handwritten Texts

263

having trained the system on this data scholars run in the system the query sample – an anonymous handwriting in order to discover which are its closest-matching samples, or which are the manuscripts that have the greatest resemblance. As the manuscripts of the training sets are authorized, finding the resemblance between the anonymous manuscript with any of the images from the training set means that the author is identified. For the automated dating of manuscripts scholars use different training sets – handwritings belonging to one and the same person and arranged according to the years. They use special algorithms to make computer identify and group their classifiers. These features are also mentioned as within-writer variations and date-informative features. Then scholars run in the system the query sample – an undated handwriting to find out manuscripts of which year in the paradigmatic row resemble it most of all (“nearest-neighbour-based verification system”). The received answer will be the date of the undated manuscript. In both automated authorship investigation and automated dating researches that are based on handwriting studies there are two main approaches that mean two different objects of investigation. The first one is graphematic (based on allograph studies) and the other – textural. Both of the methods are applied to make the computer identify differences and similarities and detect the differentiating features but in the first case, the object of the investigation is: how different writers (writer-specific information) or one and the same writer in different years (date-specific information) write graphemes. The similarity and the difference both are based on the study of the grapheme forms. In the other case scholars do not consider graphemes, text contents and even which alphabet is used. They are oriented only at the text-independent statistics, which they call “textural”. These are: “spacing between graphemes, words and lines, margins, proportions of the grapheme parts, ligatures, links, ties, slant, deviation from the baseline, the relation between height and width of letters, paper and ink specifications, height of the text line, hand pressure, pace of writing, roundness, curvature, contour-direction, Contour-hinge, Direction co-occurrence, horizontal run length, vertical run length, white run length, thickness of the ink trace etc. Georgian alphabet consists of 33 letters and is very specific. Our handwritings have never been analysed, structured and clustered for automated processing. We have never used manuscript segmentation, feature extraction or pattern recognition techniques. Therefore, at this stage graphematic approach would be much more complicated and take a lot of time. So, in order to reach the higher level of performance we decided to apply textural method but at the same time to prepare some foundation for the usage of the graphematic method in future. Every new project of this kind besides carrying out particular investigation and receiving particular results supports the further development of the Humanities as the algorithms and the databases created for these researches can be used for a great number of other investigations as well.

4 Offered Methodology The image classification problem is considered to be rather complicated problem in computer vision. It is considered to identify the presence of different visual structures in the uploaded image. The algorithm of image classification has to analyze the needed

264

T. Tvalavadze et al.

numerical properties of the different image features and afterwards has to classify them into different categories. Many different classification methods, such as artificial fuzzysets, expert systems and neural networks can be used to classify the images, all of them have different advantages and disadvantages, but their main problem is the low accuracy score. Deep convolutional neural network (CNN) is considered to be the modern and efficient approach of image classification, it can obtain rather good results when solving complicated machine learning problems such as classification [16–18]. In deep learning, CNN is considered as a class of artificial neural network. It is mostly used to analyze the visual images. CNNs can be mentioned as Space Invariant Artificial Neural Networks or Shift Invariant networks, they are based on the architecture, which is shared-weight of the convolution kernels or filters that process input features and output the responses, which are known as feature maps [19, 20]. Therefore, to date our image samples we decided to create the algorithm using image classification, by means of creating Deep Neural Network model. As the label, we have used the year of handwriting. We have used Keras library in order to implement deep convolutional neural network algorithm. We had the labeled images of 54 different years. Each image we have saved with the resolution 300 dpi. The whole size of the images dataset was 16 GB. We have divided our dataset into the validation and test datasets, the best result we have got for 75% for training and 25% for validation. We have processed the following procedures: Image Scaling: The images we have resized to the size of: img_height = 180 and img_width = 180. It must be mentioned that this scaling was got after the experiments, the algorithm could classify the writings. It is illustrated on Fig. 1. Mean/Standard Deviation of the Images: We considered the ‘mean image’, which we got by calculating the mean values for every pixel in our train datasets. It gave us the opportunity to understand the underlying structure in input datasets. For the concrete images we chose to augment the information with the perturbed images, because we considered the corresponding structure inappropriate. Data Normalization: Data normalization is obligatory in order to ensure that all input pixels have the similar distribution of data. It makes the algorithm to process the information much faster during the training procedure. We normalized our data by means of subtracting the mean values from every pixel and afterwards divided the obtained result by the values of the standard deviation. This gave us opportunity to get the curve similar to a Gaussian one, which is centered at zero. For the input images the pixel values must be positive, therefore we scaled the normalized data in 0–255 range. It must be also mentioned that the writer was also drawing on the handwritings and he was writing on the handwritings with the already printed text. It caused the reduction of the accuracy score; therefore, we cleaned the images in Photoshop before applying them to the input. As our model has faced the overfitting, when the accuracy was reaching 52–58% the validation accuracy began to decrease.

Automated Dating of Galaktion Tabidze’s Handwritten Texts

265

Fig. 1. Scaled images.

Therefore, we performed the following procedures: Data Augmentation: We used the pre-processing technique, it means that we augmented the input data with the perturbed versions of the input data. Rotation gave us very good results to mitigate the overfitting. By means of this, we exposed CNN at the much more variations of the data. It is rather interesting techniques because in this case it is much more complicated for CNN to recognize the not needed characteristics in data. As it was mentioned above, we had 54 folders of data, after the literature review we have decided to keep accuracy score not less than 74%, as it the good result in the related works, which consider the hand writing classification, but for the different languages. We decided to add the folders to training and validation data sets one by one. If the adding concrete folders caused the accuracy decrease, we continued to clean the image using Photoshop. Finally, we succeeded to add 50 folders. That folders which contained the small quantity of images in them decreased the accuracy score and caused the overfitting. The proposed gives us the possibility to achieve the main objective of the research, to date Georgian handwritings automatically using artificial intelligence.

266

T. Tvalavadze et al.

5 Experiments We have trained our model and the validation accuracy score 74.8%. We used Google Collab for it. The attributes of the final model are shown on the Fig. 2.

Fig. 2. Model attributes

We have tested our model and it worked fine in the most cases, but it mixed the years with the previous and next ones, which can be considered logical, because the write could write the manuscript in the beginning or in the end of the year. In the experiment we used 100 not dated handwriting. We dated the using our model and using the manual dating. We have got the same output is 78 handwritings, which confirms our accuracy score.

6 Conclusion and Future Work Our research shows the automated dating method of handwritten texts. Our research show that the best was for this is to use neural networks. We offer the methodology of training the system using neural network with the corresponding pre-processing procedures and techniques. As the result we have designed our model and made the experiments on

Automated Dating of Galaktion Tabidze’s Handwritten Texts

267

Galaktion Tabidze’s handwritten texts, written in Georgian. Our system has the accuracy score 74.8%, which is good in comparison with the related research, but still is low to be used in practice. In our future work, we will work on improving the accuracy score and we think to combine manual and automated techniques. The results of the work advance the process of the study of Georgian history. It must be mentioned that mostly the existing approaches are manual and time consuming. The offered approach can date the handwritings much more efficiently. The offered methodology can also be using to date the handwritings of different writes, which will work with much higher accuracy. Acknowledgment. This work was supported by Shota Rustaveli National Science Foundation of Georgia under grant [No. FR-21-7997] – Graphematic research and methodology of dating manuscripts.

References 1. Ninidze, M.: Creative history of Gelaktion Tabidze’s poem (once more about “Aspindza”), Galaktionology, IX, Tbilisi, pp. 265–266 (1922) 2. Dhali, M.A., Jansen, C.N., de Wit, J.W., Schomaker, L.: Feature-extraction methods for historical manuscript dating based on writing style development. Pattern Recognit. Lett. 131, 413–420 (2020) 3. Garain, U., Parui, S.K., Paquet, T., Heutte, L.: Machine dating of manuscripts written by an individual. J. Electron. Imaging 17(1) (2008) 4. Ullmann, S.O.A.: Dating through calligraphy: the example of “Dover Beach.” Stud. Bibliogr. 26, 20–25 (1973) 5. Gilliam, T.: Writer identification in medieval and modern handwriting, p. 160. University of York Department of Computer Science, September 2011 6. Bozekova, M.: Comparison of handwritings, physics and informatics, pp. 7–8. Comenius University Bratislava, Slovak Republic (2008) 7. Sharma, M.K., Chanderiya, V.: Writer identification using graphemes, Department of Computer and Communication Engineering, School of Computer Science & Information Technology, Manipal University Jaipur, p. 14. (2019) 8. Adam, K., Al-Maadeed, S., Akbari, Y.: Hierarchical fusion using subsets of multi-features for historical arabic manuscript dating. J. Imaging 8(60), 10 (2022). https://doi.org/10.3390/ jimaging8030060 9. Ninidze, M.: Modern Research Technologies and the Electronic Scholarly Edition, Tbilisi, p. 10 (2016) 10. Bulacu, M., Schomaker, L.: Automatic handwriting identification on medieval documents. In: Proceedings of 14th International Conference on Image Analysis and Processing (ICIAP 2007), p. 279. IEEE Computer Society (2007) 11. Tvalavadze, T.: Grapheme-based method of handwriting dating and its use in archival studies 12. van der Lubbe, M.F.J.A., Vaidyanathan, A., de Wit, M., et al.: A non-invasive, automated diagnosis of Menière’s disease using radiomics and machine learning on conventional magnetic resonance imaging: a multicentric, case-controlled feasibility study. Radiol. Med. 127, 72–82 (2022). https://doi.org/10.1007/s11547-021-01425-w 13. Goswami, A., et al.: Change detection in remote sensing image data comparing algebraic and machine learning methods. Electronics 11, 431 (2022). https://doi.org/10.3390/electronics1 1030431

268

T. Tvalavadze et al.

14. Yadav, S.S., Jadhav, S.M.: Deep convolutional neural network based medical image classification for disease diagnosis. J. Big Data 6(1), 1–18 (2019). https://doi.org/10.1186/s40537019-0276-2 15. Deep Learning: Challenges, Methods, Benchmarks, and Opportunities. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 13, 3735–3756 (2020). https://doi.org/10.1109/JSTARS.2020. 3005403 16. Bocu, R., Iavich, M.: Real-time intrusion detection and prevention system for 5G and beyond software-defined networks. Symmetry 15, 110 (2023). https://doi.org/10.3390/sym15010110 17. Bocu, R., Bocu, D., Iavich, M.: An extended review concerning the relevance of deep learning and privacy techniques for data-driven soft sensors. Sensors 23, 294 (2023). https://doi.org/ 10.3390/s23010294 18. Sarvamangala, D.R., Kulkarni, R.V.: Convolutional neural networks in medical image understanding: a survey. Evol. Intell. 15(1), 1–22 (2021). https://doi.org/10.1007/s12065-020-005 40-3 19. Maduranga, M.W.P., Nandasena, D.: Mobile-based skin disease diagnosis system using convolutional neural networks (CNN). Int. J. Image Graph. Signal Process. (IJIGSP) 14(3), 47–57 (2022). https://doi.org/10.5815/ijigsp.2022.03.05 20. Zaman, S., Rabiul Islam, S.Md.: Classification of FNIRS using Wigner-ville distribution and CNN. Int. J. Image Graph. Signal Process. (IJIGSP) 13(5), 1–13 (2021). https://doi.org/10. 5815/ijigsp.2021.05.01

Artificial Neural Network and Random Forest Machine Learning Algorithm Based TCSC Controllers for Mitigating Power Problems Niharika Agrawal1(B) , Faheem Ahmed Khan1 , and Mamatha Gowda2 1 Department of Electrical and Electronics Engineering, Ghousia College of Engineering,

Ramanagaram, Karnataka 562 159, India [email protected], [email protected] 2 Department of Artificial Intelligence and Data Science, BGS College of Engineering and Technology, Mahalakshmi Puram, Bengaluru, Karnataka 560 086, India [email protected]

Abstract. The demand for electrical energy is rising every year. There is a need for more generation and transmission facilities which require huge investment in the erection of new power stations and lines. The conventional sources of energy are coal, oil, fossil fuel, and nuclear energy-based power plants which are fast depleting. The coal-based plants release different gases like nitrogen dioxide, sulfur dioxide, particulate matter (PM), mercury, and other substances which are very harmful to human life as these gases pollute the atmosphere. The erection of new power plants is costly and affects the environemnt too. The existing transmission line was overloaded which created stability problems in the system. Fixed capacitors (FC) were used in the transmission lines to meet this problem but there were problems like series resonance, wear and tear and slow response. Here comes the role of Power -Electronics based FACTS device Thyristor Controlled Series Capacitor (TCSC) to provide variable compensation and to mitigate these power problems and challenges without the need for investment in the construction of new lines. Traditionally Proportional Integral (PI) controllers have been used for power flow control by TCSC. In the present work the Artificial Neural Network (ANN) technique and Random Forest Machine Learning Algorithm (RFMLA) which comes under the umbrella of Artificial Intelligence have been applied to enhance the power flow transfer capacity of the Kanpur Ballabhgarh transmission line. Power obtained at the receiving end is more with an ANN-based controller. Power obtained is furthermore by the use of RFMLA. The existing power system is utilized more effectively, the efficiency of the system is enhanced, the losses are reduced with modern technology enabled/based TCSC. Thus the objective of meeting the shortage of power without harming the environment is achieved with ANN and RFMLA-based TCSCs. The TCSC based on ANN and RFMLA is successful in meeting various other power challenges and problems also like stability, power quality, low voltage profile of the system. The results of power flow are checked with MATLAB simulation. Keywords: oscillations · power flow · power system · stability · TCSC

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 269–288, 2023. https://doi.org/10.1007/978-3-031-36118-0_24

270

N. Agrawal et al.

1 Introduction The objective of Sustainable Development Goals (SDGs) is the development of prosperity and peace for people and the planet now and also into the future. SDGs were formulated in 2015 by the United Nations General Assembly (UNGA). There are 17 SDGs and for each SDGs there are 8 to 12 targets and between one and four indicators for each target. The indicators are for measuring the progress towards reaching the targets. From the 17 different goals, the goal 7 is to provide an affordable, reliable sustainable and modern energy to all. Some of the targets to achieve this goal are to ensure the availability of affordable, reliable, and modern energy services by 2030, to increase the share of renewable energy by 2030, to make the global rate of energy efficiency improvement rate double in 2030, to use advanced and cleaner fossil fuel technology. In order to meet the above goal and target of SDGs the solution is to meet the rising demand of power through efficiently using the existing generation and transmission system without deteriorating/harming the environment. The installation of new plants and lines for meeting the rising demand is costly and impact the environment negatively. Fixed series capacitors were used for improving the power capacity but there were problems of wear and tear, slow response and sub synchronous resonance leading to damage of shaft. This challenge/limitation is overcome in this paper by the installation of a Series FACTS device -TCSC based on Machine Learning Algorithms (MLAs) which are ANN and RFMLA. The MLAs have extensive computational capabilities to solve complex Power System Engineering problems and produce accurate and fast results. The simulation results showed an improvement in transmission line power transfer capacity with MLA based TCSCs. The accuracy of decision tree, ANN and RFMLA is checked and the results are shown.

Fig. 1. Affordable and clean energy

The objective of the proposed work which is meeting the goal of providing clean and affordable energy to all is achieved successfully. Increased power flow by MLA based TCSC will help in the technical and economical development of country. The TCSC behaviour has been discussed for the two operating modes of TCSC and the results are

Artificial Neural Network and Random Forest Machine Learning Algorithm

271

analysed. The different errors are plotted for the inductive and capacitve modes of TCSC. Figure 1 shows the goal 7 and Fig. 2 shows the TCSC.

Fig. 2. TCSC (hitachy energy)

2 Literature Review Various techniques have been used to control the power flow by TCSC. The PI controller was implemnted in the system for controlling the power flow. The simulation diagram consisted of various MATLAB blocks like AC Source, transmision line, three phase load, measurement block. A switch was provided for changing the different modes of TCSC. The results showed an increase in power transfer. For the generation of appropriate firing pulses SR flip flop was used [1]. To reduce the voltage sag using TCSC PWM Generator Pulse Controller was developed. The programmable voltage source was used to vary the voltage with time. The testing with TCSC was done with connections at different locations. The different locations were testing at sending end, testing at middle end, and testing at receiving end of the transmission line, with varying lengths of the transmission line i.e., 100 km, 200 km, 300 km and 400 km long. The power transfer capacity was found to be better at the sending end as compared to other ends [2]. The Pulse Generator block was implemented in the system for giving the firing pulses for analysing the thyristor, capacitor current, capacitor voltage. For proper working of TCSC in capacitive and inductive mode appropriate pulses were generated. In the inductive mode the TCSC acted as a controlled inductor. Power flow decreased in inductive mode and increased in capacitive mode [4]. The efficiency and voltage regulation of the transmission line was improved on using TCSC. The results of power flow improvement with different values of firing angle were plotted. The power was increased, the stability and damping profile of the system was improved using a decentralised synergetic TCSC Controller [3]. The TCSC was implemented using Arduino uno and the laboratory model was prepared. The improvement in power flow, voltage regulation was found when the TCSC system was implemented in series with the pi section of transmission line [5].The TCSC was used in multimachine bus system for improving stability and power system oscillations. The TCSC is effective in fault current limiting. TCSC is also used to mitigate the various power quality issues

272

N. Agrawal et al.

and problems like voltage sag and swell which were created due to disturbances or faults [6].The closed loop model of TCSC was found to be effective than open loop model for power flow control. The TCSC can be used in constant current and constant power mode. The simulation was done without TCSC and with TCSC and various waveforms for power, impedance and firing angle were plotted for both the modes of TCSC [7]. FACTS devices are used to control power flow in the system by governing the various parameters like impedance, voltage, and phase angle. The TCSC lead lag structure and impedance firing angle relation was nonlinear and the nonlinear parameter/element was given to the neural network using ANN because of non-linear activation layers. The TCSC was used as a fault current limiter by taking care of impedance value to large inductive range depending on the design of TCSC. Using TCSC in the system there was enhancement of voltage sag at the time of disturbances [8]. The split TCSC is used for meeting the small change in power demand using many modules of TCSC with different or same values of inductor and capacitor. There is possibility of giving different firing angles with split TCSC. Many reactance characteristic curves are now possible and fine tuning of line reactance is done. The dimension of triggering matrix now is 92 * 92 which gives a huge range of triggering choices increasing the flexibility of the system [12]. Due to faults, there were losses to the company, losses to the consumers. The correct location and repair of faults was significant. For the different values of distances, the simulation of fault in different modes of TCSC was done [9]. The result showed the plot of Fault section indicator FSI to locate the exact position of fault in the line. The identification of faults at the time of power swing was analysed [10]. A robust and unique fault detection algorithm was proposed. The two FACTS devices TCSC and Static Var Compensator (SVC) were used for the control of power flow and improvement of voltage profile. The control parameters of the devices were estimated based on ANN. The objective of the study was the enhancement of available transfer capacity with FACTS devices based on ANN [11].

3 Test System The Kanpur Ballabhgarh transmission line in India and its Project Data is taken as for testing the PI, ANN and RFMLA techniques [12]. The compensation chosen was 75%. The inductance of the line is 1.044 mH /km. Total inductance of line of length 400 km is 0.4176 H. Total reactance of line 131.1929 ohms. For 75% series compensation the value of TCSC capacitor is 32.3503 µF and TCSC inductor is 0.07828 H. The resonance factor is 2 and resonance angle is 45 degrees. The resonance region is between 35 to 55 degrees. The ratio of TCR reactor inductance and the capacitance of series capacitor in this work is 0.25 which is in the desired range of 0.1 to 0.3. For 90% compensation the values of TCSC Capacitor and Inductor are 26.9722 µF and 0.09400 H respectively. This is 400 kV and 400 km long line. In this work the firing angle in capacitive mode of TCSC is chosen. The TCSC can be operated in blocked, bypassed and vernier mode. The vernier modes are the capacitive and the inductive mode. There is a resonance region between the inductive and capacitive vernier regions. In the present system the TCSC is worked in the capacitive mode to increase the power flow capacity of the system.

Artificial Neural Network and Random Forest Machine Learning Algorithm

273

4 Proposed Methodology and Simulation Diagram 4.1 Proposed Methodology In the present work the power transfer capacity of the system is enhanced using Artificial Intelligence based techniques like ANN and RFMLA. The Fig. 3. Shows the four main types of AI approaches. The major branches were developed and modified by Kalogirou, Dounis, Smolensky. Applications of AI are increasing in all domains at a fast rate because AI has the remarkable capacity to mimic the intelligence of human brain. The first category of AI that is Symbolic AI was based on symbols and has wide applications in process engineering. ANN are very popular and special machine learning algorithms that learn from the data and provide output/predictions. The heuristic methods solve the problem faster and in an efficient manner [13, 14].The entire solution space is properly searched to find the best solution. Some studies require Hybrid method that combine ANN and Fuzzy Logic. In the present work the TCSC is connected between source and infinite bus with transmission line. The different types of controllers selected can be PI, ANN and RFMLA. The PI is the traditional controller used in power systems. The key contribution of the paper is the use of ANN and RFMLA controllers to improve the power flow. ANN and RFMLA based controllers are very fast and are excellent computing systems. Figure 4 shows the methodology of power flow control.

Fig. 3. The major branches of AI

4.2 Simulation Diagram Without TCSC in the System This simulation diagram consists of different blocks in MATLAB like source block, the load block, the transmission line which is represented by the three phase series RLC branch block, RMS, Scope, display, VI measurement, Power measurement, from and go to blocks. The scope block is used to plot the variation of voltage and the variation of active power in the line. Then the simulation diagrams are developed including TCSC with different controller techniques like the traditional PI, the ANN method and the novel RFMLA. The results for power flow enhancement using different methods are tabulated and the waveforms are plotted. Fig. 5 shows the system without TCSC.

274

N. Agrawal et al.

Fig. 4. Flow chart of Methodology

Fig. 5. System without TCSC

4.3 Simulation Diagram for System with PI, ANN, RFMLA Based TCSC The simulation diagram with PI, ANN and RFMLA based TCSC controller is shown in Fig. 6. The systen consists of transmission line, TCSC block, source, load, control system and firing unit, active and reactive power blocks. The internal details of control subsystem are different for the three types of controller. The TCSC block is added in this system. The TCSC is bypassed for the first 0.5 s using the circuit breaker[20].The simulation is run and the different waveforms are seen on various scopes. The details of the simulation diagram is discussed in Sect. 4.3.1.

Artificial Neural Network and Random Forest Machine Learning Algorithm

275

Fig. 6. System with PI, ANN, RFMLA based TCSC

4.3.1 System with PI Based TCSC Controller The traditional controllers used in the power system are P, PI, PD and PID controllers. In the present work the tuning of firing angles of SCR is done first by PI controller. By properly tuning the controller the impedance error is reduced. The impedance errror is the difference betweeen the measured impedance and the reference impedance given to the system. Properly tuning the system is important to improve the performance and stablity. The simulation system consists of a PI controller block, the impedance measurementt block, firing unit subsystem. In the firing unit subsystem block there are blocks for firing unit of all the three phases A,B and C. There are three PLLs in the firing circuit subsystem which are synchronised with the line current.The PI controller adjust the firing angle of TCSC properly according to the change in impedance. There is a toggle switch for the change of modes which are the manual, inductive and the capacitive modes. In the present work the capacitive mode is used at firing angle of 78 degrees for the PI based TCSC controller. Fig. 7 shows the PI based Controller.

Fig. 7. PI based Controller.

276

N. Agrawal et al.

4.3.2 System with ANN Based TCSC Soft computing techniques have huge computational and mathematical capabilites to solve complex problems. The conventional PI controller is implemented here using (Artficial Intelligence) AI technique. The ANN is an excellent AI based tool for solving Power problems. The ANN model receives the input, process it on the basis of activation function and produces output. It is inspired by human brain and try to simulate the neural network so that decisions can be made in a human like manner [17]. In the present paper the ANN TCSC simulation model is designed with ANN based on feed-forward backpropagation network architecture. The input to the ANN controller is the impedance and the output is the appropriate/predicted firing pulses to the TCSC. The controller is designed for both modes of TCSC which are the inductive and the capacitive modes. The initialization of weights, biases and the training of network is done properly after taking the appropriate values of various parameters like error goal, the maximum number of epochs and the learning rate. The training function used in the present work is trainlm. It uses the Levenberg-Marquardt optimization and updates the values of weights and biases in the system. There are many benefits of using the trainlm training function such as it is fast, powerful and it does not require heavy memory as required by other algorithms. Figure 8 shows the ANN model. Proper selection of weights and input output function/transfer function affect the working of ANN. The transfer functions here is tansig and purelin functions. The MATLAB command newff is used in the program which is used to initialize the network and creating a network object. The random number generator is used by newff and the initial weights for the network are generated. The parameters of trainlm are used to train the network. This ANN model trains the data of input power and angle. The final value of firing angle is generated based on the values of various ANN parameters fed into the system[18, 19].The ANN TCSC in the capacitive mode finally improved the system’s power transfer capacity. The Fig. 9 shows the manual alpha control, the ANN controls (Custom designed Capacitive and Inductive Controls) for the two operating modes of TCSC.

Fig. 8. The ANN Model

Artificial Neural Network and Random Forest Machine Learning Algorithm

277

Fig. 9. ANN based Controller

4.3.3 System with Random Forest Machine Learning Algorithm RFMLA is very popular and having high computational capabilities supervised machine learning method for classification and regression problems. The different types of classification algorithms are the decision tree, random forest-nearest neighbour, Naïve Bayes. As the name suggests RFMLA randomly creates a forest with trees. More the number of trees in the forest the more robust and accurate results are. Multiple decision trees are created here from the dataset and then they are combined together to get and more stable and accurate prediction of results. This is called ensemble learning in RFMLA. Here the predictions are done on the combined results of individual sample models Earlier decision tree method was used which worked effectively with the trained data, but this method captured the results of the data which is then reflected in the test data. Hence the results are influenced by the trained data of decision tree method. As against this is the random forest method, the training of data set is done by the bagging method. This method is used to decrease the variations in the predictions. It combined the results from many decision trees built on different samples of the given data set. The entire data set is divided into a different random subset and then these random subsets are used to split the tree. Hence the drawbacks of the decision tree method are removed here. The representation of Random Forest by Breiman, 2001; Geurts et al., 2006 is as given in Fig. 10 [15]. The internal detail of the system is shown in Fig. 11 below:

Fig. 10. Random Forest Machine Learning Algorithm (RFMLA) [15]

A call function is created.The data of power and angle is loaded in the workspace from mydata. Mat file. The variable X is chosen for power in MW and variable Y is for the output firing angle in degrees. The values of power are chosen according to the

278

N. Agrawal et al.

Fig. 11. RFMLA based Controller

TCSC characteristics that power flow decrease in inductive vernier region and increase in capacitive vernier region such that Power in MW = [390 380 360 340 320 300 590 580 570 560] and Firing angle in degrees = [0 10 20 30 40 50 60 70 80 90]. This input data is loaded into the system and the classification method of RFMLA is performed. The entire data set is used for training the classification trees based on bagged ensemble learning method. The tree bagger class is used in this function to create a bag of decision trees. The Baggedensemle. Trees stored the 60-by-1 cell vector which are trained classification trees. Using tree bagger, the trees are grown deeper in size. Hence it is used in the work. The number of bags is chosen for bootstrapping is 60. The different samples are taken from the data and many decision trees are prepared. As the classification is used in the above function hence the majority of the votes is taken for decision. The variable X and variable Y was converted into a [30 × 4] matrix and [30 × 1] matrix of power and angle respectively to plot the errors. The Num Predictors and Num Predictors to Sample are 4 and 2 respectively. The minimum leaf size by default is 1. The in-bag fraction, and the sample with replacement are kept both 1. The compute OOB Prediction command was kept 1 to show the error of out of bag. The class names obtained after optimization are all the angles in the variable Y. The firing angle by RFMLA is found to be 60 degrees. This firing angle is assigned to the simulation file of the system with TCSC using the MATLAB function. This function used EVALIN and ASSIGNIN commands and the data from the workspace. A for loop is created inside the function which sets the appropriate firing pulses to the TCSC system which finally improved the power transfer capacity to the maximum by RFMLA[15, 16].

5 Results and Discussion Table 1 shows the power obtained with PI, ANN and RFMLA based TCSC. The power is maximum with RFMLA. The Power with ANN and RFMLA is more than that of traditional PI based TCSC. This is the contribution of this paper. RFMLA properly tuned the system based on input power and firing angle data and the line’s capacity is significantly enhanced. From the Table 2.the highest accuracy is found with RFMLA. This work is further extended for checking the errors with RFMLA based TCSC for the two operating modes of TCSC. Two test cases are generated by RFMLA for building the decision trees and various errors are plotted or two cases.

Artificial Neural Network and Random Forest Machine Learning Algorithm

279

Table 1. Power and Voltage (Phase to Ground) at receiving end of the line S.No

Controller/Device

Power (MW)

Voltage (kV)

1

Without TCSC

339.44

212.70

2

With PI based TCSC

380.04

225.10

3

With ANN based TCSC

382.43

225.50

4

With RFMLA based TCSC

386.00

227.00

Table 2. Accuracy of different algorithms S.No

Method

Accuracy (%)

1

Decision Tree

94.20

2

Artificial Neural Network(ANN)

95.00

3

Random Forest Machine Learning Algorithm

97.04

5.1 Case 1: Test Case for Capacitive Mode of TCSC 5.1.1 Building Decision Tree The decision tree is a popular tool in machine learning for making decisions and is commonly used in operational research and other studies The Random Forest creates many decision trees and then reaches to the solution based on input data The Pseudo code for the decision tree by Breiman et al., 1984 is shown in Fig. 12 [15, 16].

Fig. 12. Decision Tree [15]

For the capacitve region of TCSC the power value chosen is 590 MW and the desired firing angle is predicted for this power using predict (Bagged Ensemble, [590]). The prediction is based on the trained input data of power and angle. The result shown by the decision tree based on RFMLA is 60 degrees which lies in the capacitive range of this TCSC. The following decision trees for classification are generated by MATLAB simulation and are shown in Figs. 13a and b.

280

N. Agrawal et al.

Fig. 13. a. Decision tree (DT) of type 1 b. Decision Tree (DT) of type 2

Many decision trees are created by the RFMLA and the desired result for firing angle is generated by MATLAB. The classification method is chosen by the tree bagger. For splitting two variables are selected at random. The decision tree is plotted for getting outcomes of all the possible choices of angles for the given input of power demand. The node of the tree is used for classification and a condition is created on the features of the input to divide the data set according to the conditions chosen. The first decision tree shown in Fig. 13a is in the form of algorithm and the second is shown in Fig. 13b is in the form of a tree by RFMLA. 5.1.2 Plotting Various Errors On applying bootstrap aggregation two independent sets of bags which are in the bag and out of the bag are created. In the bootstrap sample using sampling with replacement method the data chosen is “in-the-bag”. The out-of-bag set of data is that data which is not selected in the sampling process from the original set of data. Using Random Forest many bootstrap samples and out of bag (OOB) sets are created. For each of the tree the observations which are out of bag are stored. The pseudo code for algorithm for OOB error consists of initialising the data set for training purpose. The random forest is created with k number of trees. The random forest is trained using the data set. The bootstrapping is done for each of the decision trees created by random forest. The predictions are done on data set (Figs. 14, 15, 16, 17).

Fig. 14. MSE with no of trees grown

Artificial Neural Network and Random Forest Machine Learning Algorithm

281

Fig. 15. Out of the Bag (OOB) Classification error (CE)

Fig. 16. Out of the Bag (OOB) MSE with trees

Fig. 17. Out of the Bag (OOB) feature importance

Then the errors are calculated using the difference between the actual value and predicted value. The MSE give how far the classifications are from the true values. It is given by the following equation: 

yt = mean(y|t), i(t) =

 2 1 Lt y − yt x,y∈ Nt 

The error is plotted for out of the bag (OOB) observations and the number of trees grown using OOB Error Bagged Ensemble. From the results of above figures, it is seen that all the errors which are mean square error (MSE), out of the bag (OOB) classification error (CE) and out of the bag (OOB) mean square error (MSE) are decreasing with the number of trees grown. The MSE is plotted for different leaf sizes. The minimum leaf size by default is 1. The error was calculated for leaf sizes 5, 6, 7, 8 also. The error is minimum in leaf size 5 and the optimal leaf size found by algorithm is also 5. If the size of leaf is increased, it will lead to the problem of overfitting. The out of bag feature importance is calculated using the command OOBPermutedVarDeltaError. The power

282

N. Agrawal et al.

values are included in the variable X which is a matrix of dimension 30 by 4. In all the 4 columns of X variable the values of power flow corresponding to angles from 0 degree to 90 degree is filled. The out of bag feature is found to be column no 2 and 4 of the variable X which indicated that the values of power in that column of dataset is out of bag. The columns 2 and 4 of the matrix of variable X is having values which is not in range and match with the data of columns in 1 and 3. The data in column no 1 and column no 3 have high probabilities than the column no 2 and column no 4 for getting the appropriate values of power flow. 5.2 Case 2: Test Case for Inductive Mode of TCSC 5.2.1 Building Decision Tree: For predicting the angle for power flow 340 MW using predict (Bagged Ensemble, [340]) the angle predicted was 30 degrees by RFOA which lies in the inductive range of this TCSC. The following decision tree was generated by running the algorithm on simulation in MATLAB (Figs. 18, 19).

Fig. 18. Decision tree type 1

Fig. 19. Decision Tree type 2

For the inductive region also the treebagger is used which provided the ensemble with 60 bagged decision trees. The tree bagger was created with the method of classification. The number of predictors were 4 and the number of predictors to sample was 2. This decision tree is created for 340 MW power. Firing angle obtained from the decision tree

Artificial Neural Network and Random Forest Machine Learning Algorithm

283

corresponding to this power is 30 degrees. This is passed to the simulation file using the MATLAB function. 5.2.2 Plotting Various Errors

Fig. 20. MSE with grown trees

Fig. 21. Out of the Bag (OOB) Classification error (CE)

The compute OOB prediction was set to 1 for calculating the error. The plot of various errors like MSE, out of the bag (OOB) classification (CE) error and out of the bag (OOB) mean square error (MSE) with trees is done for the TCSC in inductive mode also. From the various Figs. 20, 21, 22 it is seen that all these errors are decreasing with the number of grown trees. The plot of out of bag feature importance is shown in Fig. 23. The out of the (OOB) bag feature result showed that the column no 3 and 4 of the data set have identical features and that of column 1 and column 2 dataset have small difference.

284

N. Agrawal et al.

Fig. 22. Out of the Bag (OOB) MSE with trees

Fig. 23. Out of Bag feature importance with feature number

5.3 Plot of Variation of Power Without and with Different Controllers Figure 24 shows the plot of variation of power without using any controller and then Figs. 25, 26, 27 shows the power with all the three different types of controllers.During the first 0.5 s the TCSC is bypassed. At 0.5 s the TCSC starts working and the transfer of power is increased. There is more power transfer with ANN method than with PI method. The transfer of power is maximum with RFMLA based controller.

Fig. 24. Power without any Controller

Artificial Neural Network and Random Forest Machine Learning Algorithm

285

Fig. 25. Power with PI based Controller

Fig. 26. Power with ANN based Controller

Fig. 27. Power with RFMLA based Controller

5.4 Plot of Variation in Voltage at the Receving End Without TCSC and with TCSC Based on Three Types of Controllers. The receiving end voltage is plotted for the system without any controller and then the system with different controllers. The TCSC is bypassed for the first 0.5 s. The voltage by ANN is better than the PI based TCSC.The voltage is maximum with RFOA based TCSC controller which shows that RFMLA is an excellent machine learning algorithm (Figs. 28, 29, 30, 31).

286

N. Agrawal et al.

Fig. 28. Receiving End Voltage without TCSC

Fig. 29. Receiving End Voltage with PI controller

Fig. 30. Receiving end Voltage with ANN based Controller

Fig. 31. Receiving End Voltage with RFMLA based controller

Artificial Neural Network and Random Forest Machine Learning Algorithm

287

6 Conclusion and Future Work Normally PI controllers are used for power flow enhancement by TCSC. In the proposed research work the use of AI based techniques which are the ANN, and the Random Forest Machine Learning Algorithm is done for power flow enhancement and voltage profile improvement. These ANN and RFMLA are intelligent algorithms and are very fast with excellent computational capabilities. From the results of Table 1 it is seen that using these AI techniques the power and voltage obtained at the receiving end are increased. These AI based techniques properly tuned the system and improved the stability of the system also by increasing the power transfer. The values of power and voltage at the receiving end are obtained more by ANN method than that received by traditional PI controller. These values of power and voltage are further enhanced by RFMLA based model. In this way the power is utilized more efficiently and effectively without the need for investment in new lines and stations. The Power system is now robust due to integration with AI enable technology is and capable of meeting the power challenges due to faults and other disturbances. The equation showing the relationship between the impedance and firing angle of thyristor is non-linear and the ANN has very good ability to model this non-linear and complex relationship. It is successfully applied to the present system. As the number of trees are higher in random forest there is higher accuracy in the results with RFMLA than the decision tree method. The results with RFMLA agrees with the results of [15]. The decision trees are plotted for both the modes of TCSC. The different types of errors are also plotted with the number of trees and the error is found decreasing in both the operating modes of TCSC. In this way ANN and RFMLA based controllers are successfully applied in the present work to meet the various power problems and challenges. The work can be extended on different systems with different loading or operating conditions and the accuracy and performance of the controller can be checked for those conditions. The work on load flow analysis and THD can be done. Thus, TCSC based on ANN and RFMLA is found to be a versatile device for meeting the challenge/objective of various power problems, providing clean, affordable energy to all and for the economic and all-round development of the country. Acknowledgement. The authors would like to pay a vote of thanks to Electrical and Electronics Engineering Department, Principal of Ghousia College of Engineering for providing warm support.

References 1. Arora, K., Agarwal, S.K., Kumar, N., Vir, D.: Analysis of power flow control in power system model using TCSC. Int. J. Eng. Res. Appl. 3(3), 821–826 (2013) 2. Kamel, S., Abokrisha, M., Selim, A., Jurado, F.: Power flow control of power systems based on a simple TCSC model. Ain Shams Eng. J. 12, 2781–2788 (2021) 3. Fathollahi, A., Kargar, A., Derakhshandeh, S. Y.: Enhancement of power system transient stability and voltage regulation performance with decentralized synergetic TCSC controller. Int. J. Electr. Power Energy Syst. 1–11 (2021)

288

N. Agrawal et al.

4. Hemeida, A.M., Hamada, M.M., Mobarak, Y.A., El-Bahnasawy, A., Ashmawy, M.G., Senjyu, T.: TCSC with auxiliary controls-based voltage and reactive power controls on grid power system. Ain Shams Eng. J. 11, 587–609 (2020) 5. Ibrahim, D.K., Abo-Hamad, G.M., Zahab, E.E. D.M.A., Zobaa, A.F.: Comprehensive analysis of the impact of the TCSC on distance relays in interconnected transmission networks 8, 228315–228325 (2020) 6. Fasihipour, H., Seyedtabaii, S.: Fault detection and faulty phase(s) identification in TCSC compensated transmission lines. IET Gener. Transm. Distrib. 14(6), 1042–1050 (2020) 7. Cephas, O., Pandya, A.: Analysis of fault location in a transmission line with and without TCSC. In: International Conference on Advances in Power Generation from Renewable Energy Sources, pp. 1–12 (2019) 8. Pandiyan, M.K., Idhayaselvi, V.A., Danalakshmi, D., Sheela, A.: Online estimation of control parameters of FACTS devices for ATC enhancement using artificial neural network. In: IOP Conference Series: Materials Science and Engineering, IOP Publishing, pp. 1–11 (2021) 9. Ruhi, Z.M., Jahan, S., Uddin, J.: A hybrid signal decomposition technique for transfer learning based industrial fault diagnosis. Ann. Emerg. Technol. Comput. 5(4), 37–53 (2022) 10. Dodangeh, M., Ghaffarzadeh, N.: An intelligent machine learning-based protection of ac microgrids using dynamic mode decomposition. Iranian J. Electr. Electron. Eng. 4, 1–9 (2022) 11. Gandotra, R., Pal, K.: FACTS technology: a comprehensive review on FACTS optimal placement and application in power system. Iranian J. Electr. Electron. Eng. 18(3), 1–14 (2022) 12. Meikandasivam, S., Vijayakumar, D., Nema, R.K., Jain, S.K.: Investigation of split TCSC on Kanpur-Ballabhgarh transmission system. Int. J. Electr. Comput. Eng. 8, 76–83 (2014) 13. Liao, M., Yao, Y.: Applications of artificial intelligence-based modelling for bioenergy systems: a review. Wiley On Line Library (2021) 14. Keshkeh, K., Jantan, A., Alieyan, K.: A machine learning classification approach to detect TLS-based malware using entropy-based flow set features. J. ICT 21(3), 279–313 (2022) 15. Akash, N.S., Rouf, S., Jahan, S., Chowdhury, A., Uddin, J.: Botnet detection in IoT devices using random forest classifier with independent component analysis. J. ICT 21(2), 201–232 (2022) 16. Niva Mohapatra, K., Shreya, A.C.: Optimization of the random forest algorithm. In: Borah, S., Balas, V.E., Polkowski, Z. (eds.) Advances in Data Science and Management: Proceedings of ICDSM 2019, pp. 201–208. Springer Singapore, Singapore (2020). https://doi.org/10.1007/ 978-981-15-0978-0_19 17. Santra, A., Dutta, A.: A comprehensive review of machine learning techniques for predicting the outbreak of Covid-19 cases. Int. J. Intell. Syst. Appl. 14(3), 40–53 (2022) 18. Panda, D., Mukhopadhyay, S., Bachhar, A.K., Roy, M.: Multi criteria decision making based approach to assist marketers for targeting BoPs regarding packaging influenced purchase during Covid-19. Int. J. Intel. Syst. Appl. 14(3), 1–17 (2022) 19. Hagshenas, N., Mojarad, M., Arfaeinia, H.: A fuzzy approach to fault tolerant in cloud using the checkpoint migration technique. Int. J. Intel. Syst. Appl. 14(3), 18–26 (2022) 20. Jovcic, D., Pillai, G.N.: Analytical modeling of TCSC dynamics. IEEE Trans. Power Delivery 20(2), 1097–1104 (2005)

Child Access Control Based on Age and Personality Traits Alguliyev M. Rasim, Fargana J. Abdullayeva, and Sabira S. Ojagverdiyeva(B) Institute of Information Technology, Baku, Azerbaijan [email protected]

Abstract. Exposure to harmful information on the Internet and constant use of digital devices (computer, phone, tablet, etc.) harm children’s psychology and health. The use of methods that control access to the Internet and filter web content considered harmful to children is an effective means of solving this problem. The article proposes a method to control access to the Internet, taking into account several personality traits of the user-child (age, eye diseases, heart diseases, neurological and psychological conditions, etc.) using Mamdani-based fuzzy logic inference system. The values of the input parameters of this system are described by five linguistic parameters, which are “very low”, “low”, “medium”, “high” and “very high” and with a triangular membership function. This approach is focused on the individual user and the main advantage is that parameters are used in vector form. This research is significant for the use of parents, guardians and those responsible for children. Keywords: fuzzy logic extraction system · malicious information · Mamdani model · screen time · children’s personality traits

1 Instruction Modern children use digital devices from a younger age, and digital technology has become a key component of their lives [1]. Using smartphones and tablets has already become an easy activity for children. According to research, 97% of families have at least one smartphone, 75% have a tablet or a computer, and 44% of young children have a personal tablet and a computer [2]. Content posted on the web pages affects children’s intellectual level, emotional state, mental and physical development, personality formation, and in short, psychology. These factors require a serious concern for children’s online safety and finding effective solutions to protect them in cyberspace [3, 4]. There are current works on this feature, but in these cases one or two privacy features are used to block harmful internet content. (age, voice, etc.) The approach we propose is multi-criteria: age, psychological state, heart diseases, eye diseases, etc. includes such personality traits. Some studies suggest an approach to protecting children’s use of digital devices by implementing built-in safety and privacy controls, such as screen time or parental © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 289–298, 2023. https://doi.org/10.1007/978-3-031-36118-0_25

290

A. M. Rasim et al.

controls. Here are some age-appropriate safety precautions for children when using smart devices like the iPhone [4]. Cyber grooming is a threat that can cause more damage to children’s psychology. [5] proposes a model that classifies cyber grooming attacks based on artificial intelligence to organize online protection of children. In [6], children’s access to harmful content is prevented by creating age-restricted profiles, and a method based on machine learning that automatically detects their age and filters out any suitable content by applying suitable restriction programs is proposed. Another point here is related to the proper regulation of the time children and adolescents spend in front of the computer. In many cases, users get certain diseases by spending a lot of time in front of a computer [7]. It should be noted that a standard approach to protecting children from the Internet or digital devices, a set of official rules, etc. are not developed. In this case, experts offer a variety of options to monitor the time a child spends on the computer. The purpose of the article is to prevent children from being exposed to harmful information while using Internet resources and to organize screen time control. User-child health, psychology, etc. when certain problems arise, the time spent in front of the monitor is limited and decisions are made by applying a control system based on Mamdani.This method is applied to each child individually, and as a result, decisions are made according to the child’s situation. 1.1 Related Work In order to protect minors from harmful web content and age-inappropriate information on websites, Youtube and other virtual spaces, methods that detect web content and implement web filtering are used. Systems that control users’ access to web content are based more on biometric methods. These systems use many personality traits: age, face recognition, gender recognition, keystroke dynamics, voice recognition, etc. uses various functions such as [8, 9]. The methods used to detect web pages that are considered harmful content to children are based on the use of “blacklists”. In the field of computer science, “blacklists” are a mechanism that controls access. [10] suggests a hybrid approach using a blacklisted URL database and the Naive Bayes algorithm. By blocking a link to a malicious website, only positive results are stored in the list and presented to the user. This content is considered accessible only for children aged 5–12. Considering that pornographic information (images, audio, video, text containing pornographic content) is not relevant for child audience, using the data sanitization method, the contents of the malicious multimedia are cleaned and presented to the user [11]. The literature has developed a new Deep Neuron Network architecture called ChildNet, which filters out harmful image content that is age-inappropriate for children [12]. Another study uses machine learning methods to detect erotic texts on social media. Minors are prevented from exposure to the content in the form of obscene, aggressive, erotic or rude comments on the Internet [13]. Based on the fact that more and more unwanted and harmful information is spread among the users of the social network environment, a unified system for monitoring the harmful effects of social networks is introduced [14].

Child Access Control Based on Age and Personality Traits

291

Some approaches offer methods to protect children online using biometric technologies. When using biometric methods, the privacy of the individual is more reliably ensured. The security condition applied here requires the security of sensitive information belonging to individuals, services or means. Such security systems provide access only to authorized persons, the access to the children’s profiles by other persons, as well as criminals (sexual exploiters, cybercriminals, etc.) is restricted. Access control is a key element of security systems that is responsible for assessing whether a user is allowed to run a particular program. The literature [15] introduces an access control system called BiometricAccessFilter. This system restricts the access of malicious people to the virtual world of children and adolescents and is important in preventing violence or aggression against them and in ensuring security. Here the age of the user is determined by voice recognition by the system. A lot of research is being done with IoT applications on child protection issues [16]. A GSM-based security system is proposed. Another approach is child safety through wearables [17]. An integrated Care4Student in-system notification system is offered, which includes a server-based as well as a mobile-based module for security [18]. According to some of the approaches mentioned above, it can be concluded that existing research designed to protect children from harmful content and habits acquired from the Internet has taken into account mostly one biometric trait of the users. However, in our proposed work, screen time is set for each user, taking into account several individual characteristics of children and adolescents (age, heart and eye disease, psychological condition) and their access to information is controlled.

2 The Architecture of the Proposed System Parental control programs can prevent children from accessing content that is not suitable for their age, manage the applications they use, check their log files, virtual friends they connect with, view their messages, and more. is used for. The architecture of the system providing individual protection of children is described below (Fig. 1).

Fig. 1. Children’s access control system on the Internet

292

A. M. Rasim et al.

According to the given architecture, the information about the child’s age, health and psychological condition is considered as a criterion. The system is installed by the parent on the child’s personal computer, and the process is performed sequentially when the child wants to access the computer. The elements that make up this system are described below. Identification: the process of recognition of a child or adolescent to enter the system. Features vector: includes information about the child’s personal traits (name, surname, age, health and psychological condition). Database: this database contains information about the diseases and psychological state of children from an early age. Fuzzy assessment: in this section, the knowledge base is formed by determining the conditions for access to the content, taking into account the age, health and psychological condition of the child, and decisions are made based on this knowledge. Verification of signs: according to the age of the child, his access to Child content or Adult content is determined. As shown in Fig. 1, if a child is under the age of 18, he or she is allowed to access Child content or a screen time is set. Otherwise, i.e. if the child is older than 18 years, his access to Adult content is determined.

3 Fuzzy Logic Extraction System The expert’s knowledge and experience are used in building the system. Rules based on the IF-THEN condition are given between the input and output data of the process [19]. The information processed in the fuzzy logic extraction model consists of two parts. The first part defines the membership functions of the input and output variables. The second part is based on rules. Fuzzy variable is a set in the form < α, X , A >, here we have: α is the name of the fuzzy variable; X is the domain of the function; A is a fuzzy set in the space X. The knowledge base is covered by general knowledge related to this domain, with the conditions and results of the problem. FLES generally consists of 3 blocks (Fig. 2) [19]:

Fig. 2. Fuzzy logic extraction process

1. Fuzzification block. Converts fuzzy input data into fuzzy data (creates a fuzzy subset). 2. Fuzzy extract block.

Child Access Control Based on Age and Personality Traits

293

3. Defuzzification block. Converts fuzzy output values to non-fuzzy values. The structure of Mamdani algorithm consists of several successive stages. At the same time, each subsequent step takes the values obtained in the previous step as input. Here, the Rules consist of conditions and conclusions, which create fuzzy expressions. 3.1 Application of Mamdani Type Fuzzy Logic Extraction System In the proposed work, a fuzzy logic extraction algorithm of the Mamdani type is used. In this model, the evaluation stages consist of the following steps [19]: Database of Rules. It is designed to formally represent the empirical knowledge of experts in a particular subject area in the form of fuzzy extraction rules. This stage is the initial stage, a set of rules where a specific weight factor is assigned to each sub-result. In this study, a knowledge base on the topic is established first to determine security and screen time. Fuzzification. Evaluation is carried out on the basis of 6 rules in the implementation of this task and each of these rules depends on whether the entry belongs to a different linguistic fuzzy set. In the process of fuzzification, regardless of what the input variables describe, the output variables are rated in the range 0 to 1. This step shows the basic parameters needed to determine security. The dimensions of each parameter are given in linguistic terms and converted to the corresponding fuzzy number. The research uses the triangular membership function. The triangular membership function is defined by 3 parameters {a, b, c} and is written as in the (1) formula: ⎫ ⎧ 0, x≤a ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ x − a ⎪ ⎪ ⎪ , a ≤ x ≤ b⎪ ⎬ ⎨ b−a (1) f (x : a, b, c) = c−x ⎪ ⎪ ⎪ ⎪ , b ≤ x ≤ c ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ c−b ⎪ ⎪ ⎪ ⎪ ⎭ ⎩ 0, c≤x Establishing Fuzzy Rules. Fuzzy rules are defined using the conditional operators If and Then. The result of the If-Then rule is determined by a fuzzy set. Condition 1: If x = A, then y = B and z = C. Here x and y are linguistic variables, and A and B are linguistic values defined by fuzzy sets. The values of the input parameters of the fuzzy logic extraction system are described by five fuzzy sets “very low”, “low”, “medium”, “high”, and “very high”. Three parameters are taken as input parameters: the user’s age, health and psychological condition. These fuzzy sets define the form and state of the membership function and are described in Table 1, Table 2 shows the values of the Internet content extraction block. Extraction. At this stage, decisions are made on the basis of fuzzy rules. The output parameters are calculated for the set rules. Some of these rules are structured as follows: 1. If (child’s age is 5, very low) and (health condition is medium) and (psychological condition is medium) then childcontent can be accessed no more than 10 min.

294

A. M. Rasim et al. Table 1. Values of the fuzzy triangular number

Output parameters

Linguistic values

Interval values for the fuzzy triangular number

Adult content (no access)

Very low

[−0.4 0 0.2474]

Child content (no more than 10 min)

Low

[0.0127 0.2447 0.491]

Child content (no more than 20 min)

Medium

[0.251 0.504 0.742]

Child content (no more than 1.5 h)

High

[0.497 0.718 0.996]

Adult content (accessible)

Very high

[0.747 0.9987 1.05]

Table 2. Linguistic values of the Internet content extraction block Linguistic values

Child’s age

Very low

4 years old

Low

6 years old

Medium

12 years old

High

16 years old

Very high

18 years old

2. If (child’s age is 4, very low) and (health condition is medium) and (psychological condition is medium) then childcontent cannot be accessed. 3. If (child’s age is 6, low) and (health condition is medium) and (psychological condition is medium) then childcontent can be accessed for 15 min. 4. If (child’s age is 12, medium) and (health condition is low) and (psychological condition is low) then childcontent cannot be accessed. 5. If (child’s age is 14, medium) and (health condition is low) and (psychological condition is medium) then childnet can be accessed for 20 min. 6. If (child’s age is 16, high) and (health condition is low) and (psychological condition is low) then childnet cannot be accessed. Aggregation. This step uses the values of the functions obtained in the fuzzification stage. For each of the fuzzy extraction system rules, the conditions of the fuzzy extraction system and the situation of the validity of the conditions are calculated. The maximum or sum of the output of the rules assigned is calculated to check the accuracy. Defuzzification. In this step, the fuzzy number is converted to a non-fuzzy value using the defuzzification method. Here, a decision algorithm is implemented that selects the best non-fuzzy value based on a fuzzy set and the value of the i-th output variable is calculated as in formula (2) using the central gravity method. Here, the i-th output variable is included in Ei (i = 1...n). max x · μi (x)dx max (2) yi = min min μi (x)dx

Child Access Control Based on Age and Personality Traits

295

Here yi is the value obtained as a result of defuzzification, μi is a membership function corresponding to Ei fuzzy set. min, max are the boundaries of the fuzzy set. 3.2 Results The presentation of graphic data in the proposed expert system is processed in the fuzzy extraction system of the Matlab program. As one of the results of the experiment, a surface model based on certain dependencies is presented in the following style. Here, the child’s access to harmful content is visualized, taking into account parameters suitable for the child’s age and neurological condition (Fig. 3).

Fig. 3. Dependence graph of children by age and neurological condition

In the figure below, the child’s access to harmful content is visualized, taking into account the parameters appropriate to the child’s age and neurological condition. 3.3 Comparison of the Proposed Method with the AHP Method Comparison of the proposed method with AHP (Analytic Hierarchy Process) can be done as follows. In the approach proposed using the AHP method, the type of information is determined using only the age category. In the model below, this idea is depicted and its decomposition is given [20] (Fig. 4). Here, the purpose of the analysis is described in the first layer, the set of criteria in the second layer, and the alternatives to be selected in the third layer. As can be seen from the calculations made according to Table 3. Ranking of children’s access to Internet content is determined based on age category.

296

A. M. Rasim et al.

Fig. 4. Description of the decision problem in terms of criteria and alternatives Table 3. Alternative selection matrix Alternatives

Children Children Children Children Users Weightage Score under 6

over 6

over 12

over 12

over 18

Harmless information

0.47

0.47

0.37

0.10

0.12

Training information

0.37

0.38

0.16

0.18

0.31

0.35

0.28

0.24

0.23

Entertaininginformation 0.22

0.22

0.18

0.40

0.39

0.21

0.25

News

0.11

0.09

0.10

0.11

0.11

0.11

0.10

Harmful information

0.03

0.03

0.04

0.03

0.09

0.08

0.04

As can be seen from the table, when using this method, information for children is determined by age. However, in our proposed Mamdani method, the number of personal traits is more.

4 Conclusion and Discussion Experiments showed that different surface models were obtained. The various parameters of these surface models are interdependent and visually different. As the parameters we choose change, so does the convexity of the model we build. In today’s age, it is necessary for children and adolescents to have computer skills and develop digital skills. But protecting their health is more important. 1) The article proposes an approach to protect children from harmful content, to ensure proper use of digital devices without harming their health and psychology, using a Mamdani-type fuzzy logic extraction system. 2) If-then rules are defined in this approach. A fuzzy output method was developed in the Matlab software package.

Child Access Control Based on Age and Personality Traits

297

3) The results obtained in the output of the fuzzy logic inference system were satisfactory.

References 1. Konca, A.S.: Digital technology usage of young children: screen time and families. Early Childhood Educ. J. 50(7), 1097–1108 (2022) 2. Rideout V., Robb M.B:. The common sense census: media use by kids age zero to eight 2020, 2020 (2020) 3. Alguliyev, R., Ojagverdieva, S.: Conceptual model of national intellectucal system for children safety in internet environment. Int. J. Comput. Network Inf. Secur. 10(3), 40–47 (2019) 4. Siddiqui, Z., Zeeshan, N.: A survey on cybersecurity challenges and awareness for children of all ages. In: 2020 International Conference on Computing, Electronics & Communications Engineering (iCCECE), Southend, UK, 131–136 (2020) 5. Isaza, G., Muñoz, F., Castillo, L., Buitrago, F.: Classifying cybergrooming for child online protection using hybrid machine learning model. Neurocomputing 484, 250–259 (2022) 6. Pulfrey, J., Hossain, M.S.: Zoom gesture analysis for age-inappropriate internet content filtering. Expert Syst. Appl. 1(199), 116869 (2022) 7. Stajduhar, A., Ganel, T., Avidan, G., Rosenbaum, R.S., Freud, E.: Face masks disrupt holistic processing and face perception in school-age children. Cognitive Res. Principles Implications 7(1), 1–10 (2022) 8. Safavi, S., Russell, M., Janˇcoviˇc, P.: Automatic speaker, age-group and gender identification from children’s speech. Comput. Speech Lang. 50, 141–156 (2018) 9. Roy, S., Sinha, D., Roy, U.: Identifying age group and gender based on activities on touchscreen. Int. J. Biometrics 14(1), 61–82 (2022) 10. Kamarudin, A.N.A., Ranaivo-Malançon, B.: Simple internet filtering access for kids using Naïve Bayes and blacklisted URLs. In: International Knowledge Conference 2015, Kuching, Sarawak, Malaysia, pp. 1–8 (2015) 11. Alguliyev, R.M., Abdullayeva, F.J., Ojagverdiyeva, S.S.: Protecting children on the internet using deep generative adversarial networks. Int. J. Comput. Syst. Eng. 6(2), 84–90 (2020) 12. Alguliyev, R.M., Abdullayeva, F.J., Ojagverdiyeva, S.S.: Image-based malicious Internet content filtering method for child protection. J. Inf. Secur. Appl. 65, 103123 (2022) 13. Barrientos, G.M., Alaiz-Rodríguez, R., González-Castro, V., Parnell, A.C.: Machine learning techniques for the detection of inappropriate erotic content in text. Int. J. Comput. Intell. Syst. 13(1), 591–603 (2020) 14. Kotenko, I., Saenko, I.., Chechulin, A.., Desnitsky, V.., Vitkova, L.., Pronoza, A..: Monitoring and counteraction to malicious influences in the information space of social networks. In: Staab, S., Koltsova, O., Ignatov, D.I. (eds.) Social Informatics: 10th International Conference, SocInfo 2018, St. Petersburg, Russia, September 25-28, 2018, Proceedings, Part II, pp. 159–167. Springer International Publishing, Cham (2018). https://doi.org/10.1007/9783-030-01159-8_15 15. Ilyas, M., Fournier, R., Othmani, A., Nait-Ali, A.: BiometricAccessFilter: a web control access system based on human auditory perception for children protection. Electronics 9(2), 361 (2020) 16. Deva, S.V., Akashe, S.: Implementation of GSM based security system with IOT applications. Int. J. Comput. Network Inf. Secur. 9(6), 13–20 (2017) 17. Aliyu, M.B., Ahmad, I.S.: Anomaly detection in wearable location trackers for child safety. Microprocess. Microsyst. 91, 104545 (2022)

298

A. M. Rasim et al.

18. Akyol, K., Karacı, A., Tiftikçi, M.E.: Care4Student: an embedded warning system for preventing abuse of primary school students. Int. J. Intell. Syst. Appl. 14(4), 1–10 (2022) 19. Abdullayeva, F.C.: Development of collaborative risk assessment method for dynamic federation of clouds. Problems Inf. Technol. 2(6), 46–58 (2014) 20. Abdullayeva, F.J., Ojagverdiyeva, S.S.: Multicriteria decision making using analytic hierarchy process for child protection from malicious content on the Internet. Int. J. Comput. Network Inf. Secur. 13(3), 52–61 (2021)

Identifying the Application of Process Mining Technique to Visualise and Manage in the Healthcare Systems Arezoo Atighehchian1 , Tahmineh Alidadi2 , Reyhaneh Rasekh Mohammadi3 Farhad Lotfi4 , and Sima Ajami5(B)

,

1 Department of Industrial Engineering and Futures Studies, Faculty of Engineering,

University of Isfahan, Hezarjerib Avenue, Isfahan, Iran 2 Department of Health Information Technology and Management, School of Medical

Management and Information Sciences, Isfahan University of Medical Sciences, Hezarjerib Avenue, Isfahan, Iran 3 Faculty of Medicine, University of Belgrade, Dr. Suboti´ca Street 8, 11000 Belgrade, Serbia 4 Faculty of Organizational Sciences, University of Belgrade, Jove Ili´ca 154, 11000 Belgrade, Serbia 5 Department of Health Information Technology and Management, School of Medical Management and Information Sciences, Isfahan University of Medical Sciences, Hezarjerib Avenue, Isfahan, Iran [email protected]

Abstract. This study aims to identify the application of process mining techniques in health centres for the visualisation of healthcare activities. As a scoping review, this research was used and divided into three phases: literature collection, assessment, and selection. A literature search had done on Google Scholar, Web of Science, PubMed, Elsevier, and ProQuest, along with the impact of inclusion and exclusion criteria. Keywords have been addressed as follows: process mining, visualising, mapping, workflow mining, automated business process, discovery, process discovery, performance mining, healthcare, hospital, emergency department, emergency medical service, and apply. The findings showed that process mining can be used to analyse different activities in the field of healthcare, including workflow in healthcare, clinical and administrative processes, data analysis in information systems, events data in patients’ infectious, creation of dashboards, the discovery of unexpected, and hidden relationships. Finally, as the significance of this research, it has been argued that the use of process mining in healthcare allows health professionals to understand the actual implementation of processes. Keywords: Process Mining · Healthcare · Visualisation · Bottleneck

1 Introduction The world is faced with resource constraints such as increasing healthcare costs, complexity, and expensive therapies. The use of modern technologies to evaluate and analyse these processes is essential [1, 2]. Healthcare processes are very dynamic, complex, temporary, and increasingly multi-disciplined, so need to analyse and improve continuous © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 299–308, 2023. https://doi.org/10.1007/978-3-031-36118-0_26

300

A. Atighehchian et al.

activities of health processes [3]. Process mining is a part of techniques related to the process management and data science that helps to identify the arrangement of treatment activities [4]. Process mining has been able to distinguish characteristics and challenges in the field of healthcare [5]. In addition to these reasons, process mining has been chosen as a research area in this study. Process mining technology has also been able to improve healthcare and reduce the cost of this system by discovering the correlation in data [5–10]. Process management has been considered in many organisations, but healthcare systems had more complex processes than other industries, and it can play as an important role in different fields [11]. Therefore, the use of process mining techniques in healthcare in the form of identifying real processes and methods is essential [1]. In addition, process mining is rapidly becoming an important success for modelling the transmission of the virus, controlling infection, and emergency response analysis of pandemics and epidemic diseases. The existing solutions for process mining include discovering the actual process and presenting it in the form of a model, specifying the type and place of the required change in the process and extracting process information, and calculating KPIs for evaluation by decision-makers. This study’s focus was to identify the application of process mining techniques in health centres for visualisation and management in this regard. In this approach, researchers introduced the process mining approach which is a solution to improve productivity, reduce waiting times for receiving health services, reduce costs, improve complex, and time-consuming processes in the health and treatment system and eliminate process bottlenecks.

2 Methods This study was a scoping review the literature was on the title of visualising and managing the healthcare by process mining techniques. A sub-systematic method was used, which was divided into three phases: literature collection, assessment, and selection. A literature search had done on libraries, databases, and motor engines including Google Scholar, Web of Science, PubMed, Elsevier, and ProQuest, from 2007 to 2021 along with the impact of inclusion and exclusion criteria (Table 1 and Table 2). Table 1. Inclusion and exclusion criteria. Criterion Time period

Primary Keywords

Scope

Inclusion

14 recent years (2007–2021)

Process mining OR Workflow mining OR Process Discovery AND health OR Hospital OR Prehospital emergency

Information Technology AND Healthcare

Exclusion

Any study outside this date

Any study without this Title/Abstract/ key word

Any study outside this Scope

Identifying the Application of Process Mining Technique

301

The keywords, their synonyms, and their combinations were employed in the searching areas. Combinations in search terms, including MeSH, and Nama thesaurus (science and technology policy information exchange system) approved were used in the searching areas of title, abstract, and keywords with “AND” and “OR” (Table 2). Table 2. Search terms and strategies. SPIDER Tool

Search term

S

(Hospital OR Healthcare centre OR Pre Hospital Emergency) AND Process mining

PI

(Barrier OR Challenge) OR (opportunity OR Benefit OR Application) OR (Process mining OR Workflow mining OR Process Discovery) AND (“healthcare centre*” OR “hospital” OR “emergency department”)

D/E/R

“Qualitative” OR “Quantitative” OR “Mixed method” OR “Review”

* SPIDER = Sample, Phenomenon of Interest, Design, Evaluation, Research type

By derivation of correlation coefficient and curve alignment of WAMS/SCADA data, When 110 kinds of literature are read, it was found that very limited systematic

Fig. 1. Flow diagram: The process of PRISMA for data collection and analysis.

302

A. Atighehchian et al.

information was regarding all of the domains this study wanted to investigate. So, after de-duplication, titles and abstracts of articles and reports were screened, out of which 29 studies were selected for the review based on their relevance and access to full text (Fig. 1).

3 Results The mortality rate of patients referring to the emergency rose at the time of the COVID19, and the amount of use ambulances with COVID-19 has roughly was 10% more than other patients. Also, the acceptance rate for COVID-19 disease in hospital was about 49%, while in other groups, 22%. In addition, the acceptance rate for COVID-19 in the ICU was nearly 5% higher than the other patients [12] (Fig. 2).

Fig. 2. Process Mining Capabilities.

The global error can be got by solving the deviation by process mining offers countless benefits in the various processes, which is an example of them in Fig. 1 [13–16]. Based on the findings of Table 3, it can be stated that process mining can be used in various fields of the healthcare system, including improving the care service process to patients, reducing the cost and waiting time of patients, and improving the efficiency and effectiveness. According to Table 3 and Table 4, process mining in the analysis of care flow in different healthcare sectors provides a better understanding of processes and management in the emergency department, operating room, and patient and inpatient wards.

Identifying the Application of Process Mining Technique

303

Table 3. Examples of Practical Process Mining Applications in Healthcare. Outcome

Country

Information, analysis, and production of the model can be obtained by tools and techniques of process mining for using less structured, flexible, and multidisciplinary processes [17]

The Netherlands ProM

Tool

# 1

Three cluster models were proposed Greece according to the triage level in the emergency department. In the first cluster, green, yellow, and red triage were 34%, 57%, and 8% respectively, which included approximately 50% of patients. In the second cluster, (10% of patients), red triage was much normal, while in the third cluster (almost 40% of cases), yellow triage is exceeded [18]

ProM Disco Fuzzy, Genetic & Heuristic 2 Mining

Better insight has been obtained by analysing the pathway via a process mining approach [19]

Australia

ProM

3

The process of mining showed that the bottleneck emergency process is the "disease diagnosis stage". It took more time in comparison to examining the vital signs and sending sample tests which were 29.3, 11, and 4.5 min respectively [20]

The USA

Disco

4

Disco

5

Combining process mining methods Cuba with hospital information systems enabled unskilled managers to support clinical and administrative decisions to fix existing system problems. The first activity period was more than the second activity. The average time of activity (14.38 h) was also lower than the second activity (3.31 days) [21]

Using the process mining of different The Netherlands Questionnaire types of data in hospital information system in a spectrum of administrative systems, clinical support systems, logical health systems, and medical equipment information are fully described. There was also a lot of waiting time as follows: the average time before hospitalisation in hospital (12.87 days), outpatient intestinal clinics (average of 4.25 days), and surgery (Average of 6.63 days) [22]

6

(continued)

304

A. Atighehchian et al. Table 3. (continued)

Outcome

Country

Tool

#

Results showed that the use of analysis The Netherlands ProM of events, mapping, and process analysis was able to display the paths of the patient and employees to reduce the waiting time, patient re-acceptance, and simplification of clinical processes for better understanding of care-providers [23]

7

The combination of the mining and simulation process indicates that the use of prevention techniques significantly reduced the cost of medical care by approximately 45% [24]

Various tools in process mining

8

According to the research project, the Spain dashboard allowed discovery and increase the flow of patients based on the spatial data of patients undergoing intervention in the operating room. Therefore, 67% of users wanted to use daily dashboards to organise and monitor operation room. Between 33 and 44% of respondents agreed with quite difficult questions related to simplicity, compatibility, and intuition of the dashboard. Only two questions had 10% severe and 22% strong opposition, and these were related to the ease of use [25]

Various tools in process mining

9

Process mining was able to redesign and Italy simplify their processes which could be converted into higher-quality processes. It was probably due to underestimating stroke symptoms by patients, general relatives, and physicians. After the onset of the stroke, 27% of patients decided instead of contacting relatives (18%) prefer to contact a general practitioner (22%) [26]

Various tools in process mining

10

In this research, performance analysis was used by considering the process of cycle time and costs that helped the pre-hospital emergency, improving their processes by identifying bottlenecks. In addition, the ratio of the delay time to the total process time, 0.49 was high [5]

Disco

13

France

Iran

(continued)

Identifying the Application of Process Mining Technique

305

Table 3. (continued) Outcome

Country

Waiting times in each of the emergency Iran activities of the treatment were identified. The results showed that the emergency department activities (29% of total waiting time) it was the main emergency bottlenecks. Also, CT scan, triage, and laboratory activities were 21%, 24%, and 25% over standard time respectively [27]

Tool

#

ProM Disco

14

Table 4. The result of the application of process mining in health. Country

Application

The Netherlands Greece Australia The USA Cuba The Netherlands The Netherlands France Spain Italy Iran

A better understanding of the healthcare system [17] Studying and analysing processes in the emergency department in-depth [18] Patient path analysis from start until the end of care [19] Workflow analysis in healthcare [20] Clinical and administrational process analysis of the hospital [21] Analysis of various types of data in hospitals [22] Analysis of trend care in sepsis patients [23] Analysis of clinical pathways in patients with a hernia [24] Create dashboards in the operating room [25] Care in strokes [26] Improving pre-hospital emergency processes [5]

4 Discussion Using process mining as an appropriate and flexible tool for improving the performance of treatment processes has been chosen. To perform corrective actions, it can examine the therapeutic processes and identify existing bottlenecks in treatment [12, 21]. Discovering the actual process and identifying possible subjects before the occurrence has been introduced as some of the process mining capabilities so which in previous studies, Zhou et al. the stage of “diagnosis of any diseases”, was more than 29.3 min in the emergency department [20]. Also, the results of Mans et al., showed that there was a lot of waiting time before hospitalisation in hospital [22]. The countless capabilities related to the process mining have led to be an important success in modelling for transmitting, controlling, analysing, and responding to the virus. In brief, research findings confirm that using Information and Communications Technology (ICT) can have quite benefits in the field of healthcare. Using traditional methods and not using ICT, as well as Artificial Intelligence (AI) in the medical industry can cause several problems and challenges [28–31]. In particular, process mining techniques could be used to analyse complex data, manage them, and so on. These techniques can also focus on treatment, and the analysis

306

A. Atighehchian et al.

of diagnostics to improve the health systems in real-time responsiveness [5]. Finally, and as mentioned in the result of this study, process mining has improved conditions in hospitals, as well as operating rooms, and emergency departments. Although the previous research has argued that the process redesign stage (5.3%) and a domain expert in each stage (3.8%) were some key findings of process mining of their research [16].

5 Conclusion The goal of this research was to improve healthcare systems by using process mining. Specifying the type and place of the required change in the process, discovering the actual process and presenting it in the form of a model, extracting process information and calculating KPIs for evaluation by decision-makers, and achieving complete transparency and identifying possible subjects before occurrence were new findings of process mining capabilities in this study. Some main limitations of the process mining are as follows: access to appropriate, correct, and complete process data, also, improved usability for non-experts. This study found it easy to identify the application of process mining for the visualisation of healthcare activities. In the conclusion of this study, available research related to the use of process mining in healthcare was investigated. The results of these studies showed that the use of process mining in healthcare and treatment can have several benefits. In general, the process of mining by extracting knowledge from current reports, hospital information systems, and existing records can show a clear view of what has happened and reduce costs and the treatment length of stay.

6 Ethical Approval This research is part of the M.Sc. Thesis that is under the ethics code [IR.MUI.RESEARCH.REC. 1399.497]. Funding. This article resulted from the Master of Sciences thesis in “Health Information Technology” and research project No. 399503 and ethic code IR.MUI.RESEARCH.REC.1399.497that funded by Isfahan University of Medical Sciences, Isfahan, Iran.

Conflict of Interest. None declared.

References 1. Homayounfar, P. (eds.).: Process mining challenges in hospital information systems. 2012 federated conference on computer science and information systems (FedCSIS). IEEE (2012) 2. Wang, L.L., Lo, K.: Text mining approaches for dealing with the rapidly expanding literature on COVID-19. JBiB 22(2), 781–99 (2021) 3. Alibabaei, A., Badakhshan, P., Alibabaei, H.: Studying BPM success factors differences in various industries. JIJoM6(1), 68–74 (2017)

Identifying the Application of Process Mining Technique

307

4. Badakhshan, P., Alibabaei A.: Using process mining for process analysis improvement in prehospital emergency. ICT for an Inclusive World, 567–580 (2020). https://doi.org/10.1007/9783-030-34269-2_39 5. Munoz-Gama, J., Martin, N., Fernandez-Llatas, C., Johnson, O.A., Sepúlveda, M., Helm, E., et al.: Process mining for healthcare: characteristics and challenges. J. Biomed. Inform. 127, 103994 (2022) 6. Qiu, H.-J., Yuan, L.-X., Wu, Q.-W., Zhou, Y.-Q., Zheng, R., Huang, X.-K., et al.: Using the internet search data to investigate symptom characteristics of COVID-19: a big data study 6(S1), S40–S48 (2020) 7. Guraya, S.Y.: Transforming laparoendoscopic surgical protocols during the COVID-19 pandemic; big data analytics, resource allocation and operational considerations. JIJoS 80, 21–25 (2020) 8. Ayyoubzadeh, S.M., Ayyoubzadeh, S.M., Zahedi, H., Ahmadi, M., Kalhori, S.R.N.: Predicting COVID-19 incidence through analysis of google trends data in Iran: data mining and deep learning pilot study. JJPH, Surveillance, 6(2), e18828 (2020) 9. Haleem, A., Javaid, M., Khan, I.H., Vaishya, R.: Significant applications of big data in COVID19 pandemic. JIJOO 54(4), 526–528 (2020) 10. Bragazzi, N.L., Dai, H., Damiani, G., Behzadifar, M., Martini, M., Wu, J., et al.: How big data and artificial intelligence can help better manage the COVID-19 pandemic. JIJOER 17(9), 3176 (2020) 11. Buttigieg, S.C., Prasanta, D., Gauci, D.: Business process management in health care: current challenges and future prospects (2016) 12. Chang, H., Yu, J.Y., Yoon, S.Y., Hwang, S.Y., Yoon, H., Cha, W.C., et al.: Impact of CoViD19 pandemic on the overall diagnostic and therapeutic process for patients of emergency department and those with acute cerebrovascular disease 9(12), 3842 (2020) 13. van der Aalst, W.M., Netjes, M., Reijers, H.A.: Supporting the full BPM life-cycle using process mining and intelligent redesign. Contemporary issues in database design and information systems development: Igi Global, 100–132 (2007) 14. van der Aalst, W.: Process mining: overview and opportunities. JATOMIS 3(2), 1–17 (2012) 15. Van der Aalst, W.: Using process mining to bridge the gap between BI and BPM. MJC 44(12), 77–80 (2011) 16. de Roock, E., Martin, N.: I. Process mining in healthcare–an updated perspective on the state of the art. JJOBI 103995 (2022) 17. Gupta, S.: Technische Universiteit Eindhoven. Workflow and process mining in healthcare. JMST (2007) 18. Delias, P., Doumpos, M., Grigoroudis, E., Manolitzas, P., Matsatsinis, N.: Supporting healthcare management decisions via robust clustering of event logs. JK-BS 84, 203–213 (2015) 19. Perimal-Lewis, L., de Vries, D., Thompson, C.H. (eds.).: Health intelligence: discovering the process model using process mining by constructing Start-to-End patient journeys. In: Proceedings of the Seventh Australasian Workshop on Health Informatics and Knowledge Management-Volume 153 (2014) 20. Zhou, Z., Wang, Y., Li, L. (eds.).: Process mining based modeling and analysis of workflows in clinical care-a case study in a Chicago outpatient clinic. In: Proceedings of the 11th IEEE International Conference on Networking, Sensing and Control. IEEE (2014) 21. Orellana García, A., Pérez Alfonso, D., Larrea Armenteros, O.U.: Analysis of hospital processes with process mining techniques. MEDINFO 2015: eHealth-enabled Health. IOS Press, pp. 310–314 (2015)

308

A. Atighehchian et al.

22. Mans, R.S., van der Aalst, W.M.P., Vanwersch, R.J.B., Moleman, A.J.: Process mining in healthcare: data challenges when answering frequently posed questions. In: Lenz, R., Miksch, S., Peleg, M., Reichert, M., Riaño, D., ten Teije, A. (eds.) KR4HC/ProHealth -2012. LNCS (LNAI), vol. 7738, pp. 140–153. Springer, Heidelberg (2013). https://doi.org/10.1007/9783-642-36438-9_10 23. Hendricks, R.M.: Process mining of incoming patients with sepsis. JOJOPHI 11(2) (2019) 24. Phan, R., Augusto, V., Martin, D., Sarazin, M. (eds.).: Clinical pathway analysis using process mining and discrete-event simulation: an application to incisional hernia. In: 2019 Winter Simulation Conference (WSC). IEEE (2019) 25. Martinez-Millana, A., Lizondo, A., Gatta, R., Vera, S., Salcedo, V.T., Fernandez-Llatas, C., et al.: Process mining dashboard in operating rooms: analysis of staff expectations with analytic hierarchy process. JIJOER 16(2), 199 (2019) 26. Mans, R., Schonenberg, H., Leonardi, G., Panzarasa, S., Cavallini, A., Quaglini, S., et al. (eds.).: Process mining techniques: an application to stroke care. MIE (2008) 27. Taei, M.: Case study: emergency department of Alzahra hospital in Detection and analysis of processes in health systems using mining process techniques Isfahan University of Isfahan, Isfahan, Iran (2017). (Thesis) 28. Pramanik, M.I., Lau, R.Y.K., Demirkan, H., Azad, M.A.K.: Smart health: big data enabled health paradigm within smart cities. Expert Syst. Appl. 87, 370–383 (2017) 29. Lotfi, F., Fatehi, K., Badie, N.: An analysis of key factors to mobile health adoption using fuzzy AHP. Int. J. Inf. Technol. Comput. Sci. 12(2), 1–17 (2020) 30. Nayim, A.M.: Comparative analysis of data mining techniques to predict cardiovascular disease, vol. 14, no. 6, pp. 23–32 (2022). https://doi.org/10.5815/ijitcs.2022.06.03 31. Maphosa, V.: E-health implementation by private dental service providers in Bulawayo, Zimbabwe, vol. 15, no. 1, pp. 20–28 (2023). https://doi.org/10.5815/ijieeb.2023.01.02

New About the Principle “Like Begets Like” in Genetics Sergey V. Petoukhov(B) Mechanical Engineering Research Institute of the RAN, Moscow 101990, Russia [email protected]

Abstract. The article is devoted to the connection of the ancient principle “like begets like” with the emergent algebraic properties of a complex of a lot of mutually related binary oppositions in the genetic code system (in addition to the known particular connection of this principle with the replication phenomenon in double-stranded DNA). This holistic complex is analyzed in its algebra-matrix representation. Emergence algebraic properties of this complex are firstly studied and revealed on the basis of a matrix-algebraic approach. Appropriate families of algebra-genetic matrices have features for mutual replications of their matrices. It gives a new approach for understanding the important role of the above-named principle in the genetics and genetically inherited physiological systems and also for further development of algebraic biology. Keywords: DNA alphabets · complementary replication · binary numberings · binary-complementary replication · matrices

1 Introduction Many sciences turn to biology in connection with the anthropomorphic slogans “man is the measure of all things” and “man is the cosmos” in the hope of finding support for solving their own problems. This point of view can be seen, for example, in the book about the human phenomenon [1], whose author believed the following. To find out how the world was formed and what its further fate is, one should “decipher” the person; because of this, future synthetic science will take man as a basis; this will be a new era in science, in which there will be a complete understanding that man as a “subject of knowledge” is the key to the whole science of nature. The book [2] also supports that the theory of living systems – but not physics – is the most appropriate center for the vision of the world. Accordingly, the knowledge about the genetic code system and inherited properties of physiological ensembles is undoubtedly useful for computer science, engineering and education applications. The DNA double helix model created by J.D. Watson and F. Crick in 1953 gave a powerful impetus to the development of genetic research. It showed the world a recursive algorithm for the complementary replication of DNA strands, which ensures the replication of the genetic information recorded on these strands. Before the complementary replication, DNA is separated in two complementary strands. Each strand of the original © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 309–317, 2023. https://doi.org/10.1007/978-3-031-36118-0_27

310

S. V. Petoukhov

DNA molecule serves as a template for the production of its new complementary counterpart. This seminal work by Watson and Crick was perceived as the discovery of a key secret of life, corresponding to the ancient notions that “like begets like”. Scientists were struck by how simple and beautiful this explanation of the replication and preservation of genetic information based on the mechanism of complementarity turned out to be. It was emphasized that it is this complementarity that provides the most important properties of DNA as a carrier of hereditary information [3]. DNA complementary replication occurs in all living organisms and is required for cell division during growth and tissue repair, as well as ensuring that each new cell receives its own copy of the DNA. Because the cell has the distinct property of division, complementary replication of DNA is required. Complementary replication of DNA strands occurs at breakneck speed. The well-known bacteria E. coli, for example, replicates at a rate of over 1,000 bases per second [4]. The purpose of this article is to present results giving pieces of evidence that the principle “like begets like” is related to the emergent properties of complex of a lot of mutually related binary oppositions in the genetic code system (in addition to the known particular connection of this principle with the replication phenomenon in doublestranded DNA). This relation is revealed on the basis of analysis of an algebra-matrix representation of the complex. Data are described about algebraic binary-complementary replications in the multicomponent binary structures of the genetic coding system, which can be used to model structural features of the system on the language of algebraic matrices and dyadic groups of binary numbers. This result was received by belowdescribed methods of algebra-matrix analysis of emergent properties of the holistic genetic system of binary oppositions. Various algebraic approaches were proposed by different authors for study of genetic structures as it was noted in [5–7] but a limited volume of this article does not allow their review here. The presented algebra-matrix approach in this article is an original and testifies in favor of usefulness of its application in future genetic studies.

2 Binary Oppositions in the Genetic System and Genetic Matrices The complementary replication of DNA strands is based on the oppositions of complementary nucleobases A-T and C-G, where symbols A, T, C, and G denote correspondingly adenine, thymine, cytosine, and guanine. But these molecular oppositions are just particular cases of a wide set of mutually related binary oppositions in the molecular integral system of genetic coding. Other types of such molecular oppositions are: amino (A and C) vs. keto (G and T); three hydrogen bonds vs. two hydrogen bonds in complementary pairs of nucleotides; purines (A and G) vs. pyrimidines (C and T); n-plets with strong roots vs. n-plets with weak roots. In genetics, each of the types of binary oppositions is usually considered on its own out of their compete ensemble. But the organism is one, and this holistic system of binary oppositions may have important emergent properties that its individual parts do not have and that need to be studied. Below such emergent algebraic properties of the genetic system of binary oppositions are described. As it was shown in our previous publications [5–8], DNA alphabets of 4 letters, 16 duplets, 64 triplets, 256 tetraplets, etc. can be presented in the form of square tables

New About the Principle “LIke Begets Like” in Genetics

311

(Fig. 1), whose columns are numbered with binary indicators “pyrimidine or purine” (C = T = 1, A = G = 0), and rows are numbered with binary indicators “amino or keto” (C = A = 1, T = G = 0). These tables of DNA alphabets are members of the general tensor family of alphabetic matrices [C, A; T, G](n) , where (n) is the tensor integer power. In such matrices, each of 4 letters, 16 doublets, 64 triplets, … takes automatically its own individual place.

Fig. 1. Matrices of DNA-alphabets of 4 nucleotides, 16 duplets, and 64 triplets. Black cells in the matrices of duplets and triplets contains so-called strong duplets and triplets with strong roots. In each of the rows of the matrices of duplets and triplets, its black-and-white mosaics has a meanderlike character and corresponds to meander-like Rademacher functions shown graphically at the right.

Each of the n-plets in such matrices (Fig. 1) can be binary enumerated by the concatenation of the binary numbers of the row and column at the intersection of which it is located. For example, the triplet CAT becomes numbering 001010 because it belongs to the row 001 and the column 010. In these matrices, all complementary n-plets are located inverse-symmetrically with respect to the center of the appropriate matrix. Correspondingly, binary numbering of each n-plet is transformed into numbering of its complementary n-plet (that is, n-plet of the opposite strand of DNA) by the mutual interchanging of digits 0 ↔ 1 in it. For example, by this complementary operation, the numbering 001010 of the triplet CAT becomes numbering 110101 of its complementary triplet GTA. This interchanging 0 ↔ 1 is called the binary-complementary operation and actively used below regarding to an algebraic realization of the ancient principle “like begets like” in matrix genetics. In the matrices in Fig. 1, which are built on the basis of the noted binary oppositions, the sets of binary numberings of rows and columns correspond to dyadic groups of n-bit binary numbers in which the logic operation of modulo-2 addition serves as the group operation [5–7, 9]. For example, in the matrix of 64 triplets, the dyadic group of binary numberings of its rows and rows contains eight 3-bit binary numbers (1): 000, 001, 010, 011, 100, 101, 110, 111

(1)

Two binary numbers that are converted into each other under interchange 0 ↔ 1 are called complementary each to another. For example, in the dyadic group (1), the pairs of complementary numbers are the following: 000–111, 001–110, 010–101, 011–100 (in the decimal system, they correspond to pairs of numbers 0–7, 1–6, 2–5, 3–4). In a pair of complementary numbers, one of them is always even and the other is odd, that

312

S. V. Petoukhov

is any pair of complementary numbers is the pair of even and odd numbers. Accordingly, those columns (rows) that are enumerated by complementary binary numbers are called complementary each to another. In the genetic matrices in Fig. 1, complementary columns are located mirror-symmetrical in the left and right halves of the matrices, and complementary rows are located mirror-symmetrical in the upper and lower halves. Another type of genetic binary opposition is known. This is the binary-oppositional division of the DNA alphabet of 64 triplets into two equal sub-alphabets based on their code properties: 32 triplets with strong roots (i.e., triplets starting with the other 8 duplets CC, CT, CG, AC, TC, GC, GT, GG) and 32 triplets with weak roots (i.e., triplets starting with the other 8 duplets) [10, 11]. The coding value of triplets with strong roots is unaffected by the letter in the third position. For example, the four triplets with the same strong root CGC, CGA, CGT, CGC encode the same amino acid Arg, despite the fact that their third position is different. The coding value of triplets with weak roots, on the other hand, is determined by a letter in the third position. For example, two triplets (CAC, CAT) encode the amino acid His, while the other two (CAA, CAG) encode another amino acid Gln in the grouping of four triplets with the same weak root CAC, CAT, CAA, and CAG. In Fig. 1, all triplets with strong roots are highlighted in black, while triplets with weak roots are highlighted in white. Let us concentrate on analysis of the mosaic matrix [C, A; T, G](3) of 64 triplets in Fig. 1. A black-and-white mosaic of each of its rows has a meander-like character: black fragments and white fragments have identical length. Such meander mosaics of rows correspond to meander-like forms of Rademacher functions r n (t) = sign(sin2n πt), n = 1, 2, 3,…, which are well known in the theories of discrete signals, orthogonal series, and probabilities (here sgn is the sign function of the argument). Rademacher functions take only two values “ + 1” and “ − 1” as it is shown for the considered case in Fig. 1 at right. Black and white cells of the symbolic matrix in Fig. 1 reflect the opposition of two sub-alphabets of n-plets with strong and weak roots and therefore can be represented by elements + 1 and -1 in them. In this representation, the numeric matrix appears, which is conditionally called the genetic Rademacher matrix of the 64 triplets because each of its rows correspond to one of the Rademacher functions (Fig. 2 at the top). This matrix W is a sum of two sparse matrices W0 and W1 , one of which contains non-zero columns enumerated by even numbers 000, 010, 100, 110 and another contains non-zero columns enumerated by odd numbers 001, 011, 101, 111 (Fig. 2, at the bottom). In this even-odd decomposition of the Rademacher matrix W of 64 triplets, each of the non-zero columns of the first sparse matrix W0 is a complementary to one of the non-zero columns of the second sparse matrix W1 . By this reason, these two sparse matrices W0 and W1 are called a complementary each to another. The sparse matrix W0 with even-numerated columns is called the even-columns matrix; all its non-zero columns correspond to triplets, which contain pyrimidines C or T at their ends (by this reason, this sparse matrix can be also called the pyrimidine-columns matrix). The sparse matrix W1 with odd-numerated columns is called the odd-columns matrix; all its nonzero columns correspond to triplets, which contain purines A or G at their ends (by this reason, this matrix can be called the purine-columns matrix).

New About the Principle “LIke Begets Like” in Genetics

313

Fig. 2. The even-odd decomposition of the Rademacher genetic matrix W of 64 triplets as the sum of two sparse complementary matrices W0 and W1 : at left, the even-columns matrix W0 containing only non-zero columns having even numberings; at right, the odd-columns matrix W1 containing only non-zero columns having odd numberings. Empty cells contain zeroes.

3 Algebraic Properties of the Matrices Based on the Genetic Complex of Binary Oppositions The even-columns (8 * 8)-matrix W0 in Fig. 2 is the sum of 4 sparse (8 * 8)-matrices s0 + s1 + s2 + s3 shown in Fig. 3. The set of these 4 matrices s0 , s1 , s2 , s3 is closed relative to multiplication and corresponds to a certain multiplication table in Fig. 3 at right. This table matches to the multiplication table of the Cockle split-quaternions algebra used in the Poincare conformal disk model of hyperbolic geometry; it is also connected with the theme of the algebraic holography in genetics [12, 13]. Analogically, the odd-columns matrix W1 (Fig. 2) is the sum of 4 sparse matrices p0 + p1 + p2 + p3 shown in Fig. 4. The set of these 4 matrices p0 , p1 , p2 , p3 is closed regarding multiplication and defines the multiplication table in Fig. 4 at the right. This multiplication table coincides with the multiplication table of the 4-dimensional algebra, which was received for the even-columns matrix W0 (Fig. 3). Both the even-columns matrix and the odd-columns matrix present Cockle’s split-quaternions with unit coordinates (these split-quaternions have different forms of their matrix representations, with which these even-columns and odd-columns genetic matrices turn out to be associated). Correspondingly, both these genetic matrices are connected with the Poincare conformal disk model of hyperbolic geometry. The summation of the even-columns matrix W0 and the odd-columns matrix W1 , which are binary-complementary each other and connected to the same 4-dimensional algebra, gives the summary matrix W (Fig. 2, at the top), which is a new algebraic entity connected already with the 8-dimensional algebra. This summary Rademacher matrix W can be decomposed into the sum of 8 sparse matrices v0 + v1 + v2 + v3 + v4 + v5 + v6 + v7 shown in Fig. 5. The set of these matrices v0 , v1 , v2 , v3 , v4 , v5 , v6 , v7 is

314

S. V. Petoukhov

Fig. 3. The decomposition of the even-columns matrix W0 (from Fig. 2) into 4 sparse matrices s0 , s1 , s2 , s3 , whose set is closed relative to multiplication; s0 plays a role of the identity matrix in this set. The multiplication table for this set is shown at the bottom, which matches with the multiplication table of the 4-dimensional algebra of Cockle split-quaternions used in the Poincare conformal disk model of hyperbolic geometry. A symbol of this model is presented (from https:// commons.wikimedia.org/wiki/Category:Poincar%C3%A9_disk_models).

Fig. 4. The decomposition of the odd-columns matrix W1 (from Fig. 2) into 4 sparse matrices p0 , p1 , p2 , p3 , whose set is closed relative to multiplication; p0 plays a role of the identity matrix inside this set. The shown multiplication table for this set matches the multiplication table of the Cockle split-quaternions algebra used in the Poincare conformal disk model of hyperbolic geometry. The symbol of this model is presented.

closed relative to multiplication and matches already to the multiplication table (Fig. 5, at the right) of a certain 8-dimensional algebra. This summary matrix W generates algorithmically its complementary-replicated analogue WR by means of the interchange of numbers 0 ↔ 1 in the binary numerating of its columns with the corresponding rearrangement of the columns (that is, rearrangements of columns located mirror symmetrically in the left and right halves

New About the Principle “LIke Begets Like” in Genetics

315

Fig. 5. The decomposition of Rademacher matrix W (from Fig. 2, at the top), which is the sum of the even-columns matrix W0 and the odd-columns matrix W1 , into 8 sparse matrices v0 , v1 , v2 , v3 , v4 , v5 , v6 , v7 , whose set is closed relative to multiplication. The multiplication table for this set is shown at the bottom.

of the matrix). This interchanging algorithm 0 ↔ 1 in binary numbers provides interchanges in any pair of complementary columns that differ from each other in the content of triplets with purine and pyrimidine endings, in some analogy with the complementarity of purines and pyrimidines in DNA double strands. For example, column with number 110 (which corresponds to the nucleotide order “purine-purine-pyrimidine” in all its triplets) takes the place of column with number 001 (which corresponds to the order “pyrimidine-pyrimidine-purine” in all its triplets). Briefly speaking, molecular complementary-replicated properties of DNA strings exist jointly with algebraic binarycomplementary replication properties of the considered alphabetical matrix of the genetic code. Both of these properties are parts of genetics of the whole organisms and so interrelated. These algebraic complementary-replicated properties of genetic matrices allow applying effective algebraic methods for further study of genetics to include it in the field of modern mathematical natural sciences in connection with multi-dimensional algebras, hyperbolic geometry, theory of resonances, etc. The action of complementaryreplicated (8 * 8)-matrices W and WR on an arbitrary 8-dimensional vector X generates two new vectors that are complementary to each other: the corresponding coordinates of both generated vectors are the same in their absolute values, but have opposite signs. In addition, each of the resulting vectors X *W and X *WR is always a complementary

316

S. V. Petoukhov

palindrome: the sequence of its coordinates, which is read in forward order, coincides with the sequence, which is read in reverse order and having coordinates with the opposite sign. It is interesting because in molecular genetics the problem of complementary palindromes has long been known. The limited volume of this article allows demonstrating only a fragment of the wide theme of the important role of the principle “like begets like” connected with binarycomplemented replications. One can see more detail on this theme in the preprint [14]. The received results can be used in different scientific directions, for example, presented in [15–19].

4 Some Concluding Remarks The briefly presented algebraic complementary-replicated properties of genetic matrices allow applying effective algebraic methods for further study of genetics to include it in the field of modern mathematical natural sciences. At various levels of genetically inherited biological organization, different forms of implementation of the fundamental biological principle “like begets like” (or a complementary replication in a broad sense) can be seen. Mirror neurons, for example, are known in the human and animal brains, which have mirror complementary hemispheres. A mirror neuron is a neuron that fires both when an animal acts and when the animal observes another animal performing the same action. As a result, the neuron “mirrors” the behavior of the other, as if the observer were acting. The theme of mirror neurons, whose function is based on one of the forms of the complementary replication principle (in a broad sense), concerns cognitive functions, the origin of language, learning facilitation, automatic imitation, motor mimicry, autism, human emotional capacity such as empathy, and a variety of other issues. [20]. One of the questions that has arisen is: where do mirror neurons come from? [21]. It may be recalled here that mathematical formalisms have a creative power. As Heisenberg noted, “the Pythagoreans seem to have been the first to realize the creative force inherent in mathematical formulations” [22]. The proposed matrix formalisms have a creative power in algebraic biology and should be used in the future in theoretical and applied researches. The above-described results of our studies give pieces of evidence that the system of mirror neurons and the system of DNAs complementary replications are not isolated parts of the organism, but they are particular parts of a bio-algebraic complex realizing phenomena “like begets like”. It’s not that the molecules of two strands of DNA randomly docked, formed a complementary pair and began to repeat the process of complementary replication at breakneck speed. Another point of view is proposed: the DNA filaments replication phenomenon is part of a holistic bio-algebraic genetic complex of complementary replication, parts of which manifest themselves at different levels of organization of the living, up to the functioning of the brain with its mirror neurons and the ability to empathize and imitate external events. This bioalgebraic complex is responsible for the implementation of the ancient principle “like begets like” at different levels of biological organization in the course of biological evolution.

New About the Principle “LIke Begets Like” in Genetics

317

References 1. Teilhard de Chardin, P.: The Human Phenomenon. Brighton: Sussex Academic (1999) 2. Capra, F.: The Tao of Physics: An Exploration of the Parallels Between Modern Physics and Eastern Mysticism. Shambhala Publications Inc, New Jersey (2000) 3. Chapeville, F., Haenni, A.-L.: Biosynthese des proteins. Hermann Collection, Paris Methodes (1974). (in French) 4. Bank, E.: How much time does it take for a DNA molecule to replicate? - sciencing.com.https:// sciencing.com/much-time-dna-molecule-replicate-21660.html. Accessed 8 Nov 2022 5. Petoukhov, S.V.: Matrix genetics, algebras of genetic code, noise immunity. Moscow, RCD, p. 316 (2008). (in Russian). ISBN 978-5-93972-643-6 6. Hu, Z.B., Petoukhov, S.V., Petukhova, E.S.: I-Ching, dyadic groups of binary numbers and the geno-logic coding in living bodies. Prog. Biophys. Mol. Biol. 131, 354–368 (2017) 7. Petoukhov, S.V., He, M. Symmetrical analysis techniques for genetic systems and bioinformatics: advanced patterns and applications. Hershey, USA: IGI Global (2010) 8. Petoukhov, S.V., Svirin, V.I.: Stochastic rules in nucleotide sequences in genomes of higher and lower organisms. IJMSC 7(2), 1–13 (2021). https://doi.org/10.5815/ijmsc.2021.02.01 9. Harmut, H.F.: Information Theory Applied to Space-Time Physics. The Catholic University of America, DC, Washington (1989) 10. Rumer, Y.: Codon systematization in the genetic code. Dokl. Akad. Nauk SSSR 183(1), 225–226 (1968) 11. Fimmel, E., Strüngmann, L.: Yury Borisovich Rumer and his ‘biological papers’ on the genetic code. Phil. Trans. R. Soc. A 374, 20150228 (2016). https://doi.org/10.1098/rsta.2015.0228 12. Petoukhov, S.V.: Binary oppositions, algebraic holography and stochastic rules in genetic informatics. Biosystems 221, 104760 (2022). https://doi.org/10.1016/j.biosystems.2022. 104760 13. Petoukhov, S.V.: Hyperbolic numbers in modeling genetic phenomena. Preprints 2019, 2019080284 (2020). https://doi.org/10.20944/preprints201908.0284.v4 14. Petoukhov, S.V.: The principle “like begets like” in molecular and algebraic-matrix genetics. Preprints 2022, 2022110528 (2022). https://doi.org/10.20944/preprints202211.0528.v2 15. Ganguly, K.K., Asad, M., Sakib, K.: Decentralized self-adaptation in the presence of partial knowledge with reduced coordination overhead. IJITCS 14(1), 9–19 (2022) 16. Lone, S.A., Mir, A.H.: Smartphone-based biometric authentication scheme for access control management in client-server environment. IJITCS 14(4), 34–47 (2022) 17. Hamd, M.H., Ahmed, S.K.: Biometric system design for Iris recognition using intelligent algorithms. IJMECS 10(3), 9–16 (2018) 18. Hissain, A., Muhammad, Y.S., Sajid, M.N.: An efficient genetic algorithm for numerical function optimization with two new crossover operators. IJMSC 4(4), 42–55 (2018) 19. Almutiri, T., Nadeem, F.: Markov model applications in natural language processing: a survey. IJITCS 14(2), 1–16 (2022) 20. Ferrari, P.F., Rizzolatti, G.: Mirror neuron research: the past and the future. Philos. Trans. R Soc. Lond. B Biol. Sci. 369(1644), 20130169 (2014) 21. Heyes, C.M.: Where do mirror neurons come from? Neurosci. Biobehav. Rev. 34(4), 575–583 (2010) 22. Heisenberg, W.: Physics and Philosophy: The Revolution in Modern Science, Penguin Classics (2000)

A Study of Implementing a Blockchain-Based Forensic Model Integration (BBFMI) for IoT Devices in Digital Forensics Chintan Singh, Himanshu Khajuria(B) , and Biswa Prakash Nayak Amity Institute of Forensic Sciences Amity University, Uttar Pradesh, Sector 125, Noida 201301, India [email protected]

Abstract. The rapid advancement of Internet of Things (IoT) technology has brought about new challenges in the field of digital forensics. Traditional forensic methods are often inadequate in dealing with the decentralized and distributed nature of IoT devices. IoT has several advantages that have made it appealing to consumers and attackers alike. The technology and resources available to today’s cybercriminals allow them to launch millions of sophisticated attacks. In this paper, we propose a blockchain-based forensic model (BBFMI) for investigating IoT devices. The proposed BBFMI model utilizes the immutability and tamperproof features of blockchain technology to provide a secure and reliable way of collecting, storing, and analyzing forensic evidence from IoT devices. One of the most important benefits of BBFMI is that it provides digital forensics investigators with an immutable chain of evidence that can be used to trace the source of data and its subsequent changes. For instance, when using BBFMI, forensic investigators can easily trace the chain of events that led to the data breach of an IoT device. Our results show that the proposed blockchain-based forensic approach can provide a secure, efficient, and tamper-proof solution for investigating IoT devices in digital crime scenes. Keywords: Blockchain · Digital forensic · Internet of things · framework · decentralized storage

1 Introduction The Internet of Things (IoT) has brought about a new era of connectivity, enabling devices to collect and share data in ways that were previously impossible. From smart homes to connected vehicles and medical devices, IoT has the potential to improve our lives in countless ways. However, with this increased connectivity comes new challenges in the field of digital forensics. Traditional forensic methods are often inadequate in dealing with the decentralized and distributed nature of IoT devices [1]. The data generated and stored by these devices is often spread across multiple locations, making it difficult to establish a chain of custody and preserve the integrity of the evidence. One of the main challenges in IoT forensics is the sheer volume of data that is generated by these © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 318–327, 2023. https://doi.org/10.1007/978-3-031-36118-0_28

A Study of Implementing a Blockchain-Based (BBFMI)

319

devices. IoT devices are constantly collecting and transmitting data, and this data can be used to track the actions and movements of individuals, making it a valuable tool for forensic investigations. However, due to the decentralized and distributed nature of IoT devices, this data is often spread across multiple locations, making it difficult to collect and forensically analyze. Furthermore, the data stored on IoT devices in incidents are often unstructured and in various formats, making it difficult to extract meaningful information. Another challenge in IoT forensics is the lack of standardization in the way that data is stored and transmitted [2]. Different manufacturers use different protocols and standards for data storage and transmission, making it difficult to extract and analyze data from different devices. This lack of standardization also makes it difficult to establish a chain of custody, as it is difficult to determine who had access to the device and when. To address these challenges, researchers have proposed the use of blockchain technology for forensic investigations, which provide data security and avoid the contamination of chain of custody. Li et al. (2019) proposed a unique ID method for IoT devices and record them in blockchain to avoid single point failure attack [3]. Zhao et al. (2020) to tackle problems associated with smart homes by developing federated learning systems for IoT devices [4]. In 2021 Gong et al. proposed a framework for IoT devices to make up with computational intensive requirements by introducing blockchain of things gateways [5]. So, using blockchain technology, it is possible to create a tamper-proof and immutable record of the data collected from IoT devices, making it possible to establish a chain of custody and preserve the integrity of the evidence. The paper begins by outlining the need for better digital forensics solutions for IoT devices, definitions, guidelines processes and difficulties associated with traditional forensic techniques. It then provides an overview of blockchain technology, discussing its capabilities and potential applications [6]. We aim to demonstrate the effectiveness of our proposed model through a case study and its results. The proposed model will include the use of smart contracts to automate the process of evidence collection and chain of custody, as well as a decentralized storage solution for preserving the integrity of the collected evidence. We also aim to address the lack of standardization in IoT data storage and transmission by proposing a standard for data storage and transmission that can be used by IoT device manufacturers. In conclusion, the proposed blockchain-based forensic model for IoT is a promising solution for addressing the challenges of IoT forensics. It provides a secure, efficient, and tamper-proof way of collecting, storing, and analyzing forensic evidence from IoT devices. By utilizing the immutability and tamper-proof features of blockchain technology, it is possible to establish a chain of custody and preserve the integrity of the evidence. Furthermore, by proposing a standard for data storage and transmission, we aim to address the lack of standardization in IoT data storage and transmission and make it easier to extract and analyze data from different devices.

320

C. Singh et al.

2 Background 2.1 Digital Forensic Overview Digital forensics is the process of identifying, preserving, analyzing, and presenting digital evidence in a manner that is legally admissible. The field of digital forensics has evolved rapidly in recent years, driven by the increasing use of digital technology in all aspects of modern life. Digital forensics is used in a wide range of contexts, including criminal investigations, civil litigation, and incident response [7]. There are several ways to describe digital forensics. Horsman et al., (2022) described computer forensics as “the preservation, identification, extraction, interpretation, and documentation of computer evidence to include the rules of evidence, legal procedure, the integrity of evidence, factual reporting of the information, and providing expert opinion in a court of law or other legal and/or administrative proceedings as to what was found” [8]. The definition of digital forensics that is most often utilized, according to James et al. (2015), is “The use of scientifically derived and proven methods toward the preservation, collection, validation, identification, analysis, interpretation, documentation and presentation of digital evidence derived from digital sources to facilitate or further the reconstruction of events found to be digital in nature” [9]. 2.2 IoT in Forensics IoT forensics, according to Atlam et al., (2017), consists of device, network, and cloud level forensics [10]. Additionally, IoT forensics was defined as a part of digital forensics by Stoyanova et al., 2020 [11]. Forensics, which deals with the identification, collection, organizing, and presentation of evidence, takes place in the context of the IoT ecosystem. IoT devices typically have a range of sensors and other components that can be used to collect data, such as GPS coordinates, audio and video recordings, and even biometric data. This data can be used to reconstruct events, such as a crime or accident, and to identify suspects or victims. One of the main challenges in IoT digital forensics is the sheer volume of data that can be generated by these devices. For example, a smart home system may generate hundreds of thousands of log entries, each of which must be analyzed to extract relevant information. Additionally, IoT devices can be configured to store data in a variety of formats, which can make it difficult to extract relevant information. Another challenge is the lack of standardization in IoT devices. Many devices use proprietary protocols and data formats, which can make it difficult to extract data. Furthermore, many IoT devices are designed to be disposable, which can make it difficult to recover data from them. Overall, IoT devices are becoming an increasingly important source of digital evidence, and digital forensics experts must be prepared to deal with the challenges that they present. As IoT continues to evolve, it will be important for digital forensics experts to stay up to date with the latest technologies and techniques in order to effectively analyze the data that these devices generate.

A Study of Implementing a Blockchain-Based (BBFMI)

321

2.3 Blockchain and Digital Forensic Blockchain technology is becoming increasingly relevant in the field of digital forensics as more and more transactions are conducted on blockchain networks. Blockchain is a distributed ledger technology that records transactions across a network of computers. Each block in the chain contains a number of transactions, and each block is linked to the previous block through a cryptographic hash. One of the key features of blockchain is its immutability, meaning that once a block is added to the chain, it cannot be altered. This makes blockchain an attractive option for digital forensics as it provides a tamper-proof record of transactions that can be used as evidence in investigations. In a dispersed network of untrusted peers, the blockchain serves as “an immutable ledger for recording transactions” (Gaur et al., 2018). Transactions made by a node are confirmed by other participating nodes in the network. Following validation, the group of transactions is added to the block by a certain class of nodes known as miners, as is the case with bitcoin [11]. Forensic examiners can use blockchain analysis tools to trace the flow of digital assets, track the movement of funds, and identify the parties involved in a transaction. They can also use blockchain data to reconstruct the sequence of events that led to a crime or incident, providing valuable insights for law enforcement and other authorities. In addition, blockchain can be used to verify the authenticity of digital evidence. For example, hash values can be stored on the blockchain, providing a tamper-proof record of the integrity of the evidence.

3 Previously Related Work We examine several works that are closely similar to our own in this part. IoT security has made extensive use of blockchain technology. When it comes to IoT forensics, the idea is still in the exploring stage. Hossain et al., 2018 proposed a forensic investigation framework utilizing a public digitalized leadger to gather facts in digital criminal investigations in IoT systems [12]. The framework is set up in three nodes, namely controller to IoT device, controller to cloud, and controller to controller, using a centralised manner. However, it was demonstrated in this work’s proof of concept that the suggested system can extract forensic data from IoT devices. However, because of the centralised structure of their strategy, it is challenging to confirm the veracity of the data the framework has gathered. A digital witness strategy was also employed by Li et al., (2019) to enable sharing of IoT device records with guaranteed anonymity [13]. The authors deployed their digital witness model using the Privacy-aware IoT Forensics (PRoFIT) methodology that they had previously developed. The approach suggested in their study aims to support the collection of digital evidence in IoT contexts while preserving the privacy of the data amassed. The suggested approach supports the PRoFIT methodology’s 11 privacy criteria as documented by Nieto et al. (2018) [14]. A distributed logging approach for IoT forensics was suggested by Nieto et al. in their work from 2017 [15]. A modified information dispersal algorithm (MIDA) was employed by the authors of this study to guarantee the accessibility of logs produced in IoT contexts. The logs are distributed, authenticated, encrypted, and consolidated. The distributed methodology utilised in this study solely addresses the storage of logs, not their verification. An IoT

322

C. Singh et al.

forensics framework using a permissioned-based blockchain was presented by Noura et al., (2020) by using the immutability virtue of blockchain technology. The integrity, legitimacy, and non-repudiation of evidence are improved by this paradigm [16]. Kumar et al., (2021) provided a forensic framework for distributed computing, transparency and decentralization in digital forensic investigations [17]. The provided Ethereum virtual machine framework results in clear view of process performed under digital forensic investigations that include all the stakeholders such as cloud service providers, internet service providers and heterogeneous devices. Thus, this framework could be utilized in security operations centers also. Furthermore, in after that Jacob et al., (2022) also designed framework model for forensic IoT using blockchain to help in preventing duplication of data in investigation and also developed secure level of confidence between evidence entities. More recently Gao et al., (2022) provided a botnet framework which is enhanced by blockchain in order to provide strong resilience against Distributed Denial of Service Attack (DDoS), Sybils, and digital forensic investigations [18]. Although the proposed method of botnets lacks in terms of censorship, resilience against single point of failure, high latency, and cost.

4 Results and Discussions 4.1 Proposed Model for IoT Incidence in Digital Forensics In this section, we demonstrate our blockchain-based forensic model for the Internet of Things (BBFMI). IoT setups are divided into three layers. These are the network, device, and cloud layers. Our design allows leverage of the decentralized nature of blockchain to make sure that the logs created in IoT settings are preserved on the network and are accessible for verification by any of the participating nodes in the network. Numerous artefacts should be considered during a forensic investigation. However, our technique just considers system and event logs. The entities used in BBFMI are tabulated in Table 1, with their assigned abbreviations. Also Fig. 1. Shows the suggested BBFMI model that can enhance the IoT evidence in digital forensic investigations. Table 1. Entities used in BBFMI framework model S No

Entities Used

Abbreviations Used

1

Cloud Service Providers

(CPs)

2

Blockchain Centre

(BCs)

3

Log Processing Centre

(LPs)

4

User Centre

(UCs)

4.1.1 Blockchain Centre Entity (BCs) A blockchain is a distributed ledger technology that is used to record transactions across a network of computers. In a blockchain model, each block on the chain contains a group

A Study of Implementing a Blockchain-Based (BBFMI)

323

of transactions, and once a block is added to the chain, the information in that block cannot be altered [19]. This allows for a secure and transparent way to store and share information. The blockchain is maintained by a network of nodes, each of which has a copy of the entire ledger. The nodes work together to validate new transactions and add them to the chain, ensuring that the ledger is accurate and up to date. The distributed nature of the blockchain means that no single entity controls the data, making it more resistant to tampering and fraud. Before being written onto a block on the distributed ledger that makes up the BCs, each log is processed. The distributed ledger consists of all the blocks that have been committed to the network. Each block contains a transactional value of hashed values derived from logs. The logs are collected from the various entities, hashed, and added to the blockchain network as transactions. The nodes in the BCs are made up of forensic investigators, CPs, and IoT devices. The blocks are proposed once the nodes have reached a consensus.

Fig. 1. Proposed BBFMI Model for IoT using blockchain

4.1.2 Log Processing Centre Entity (LPs) A log processing center in a blockchain model refers to the network of nodes that are responsible for processing and validating transactions before they are added to the blockchain. In this model, each node has a copy of the entire blockchain ledger and works together to validate new transactions and add them to the chain. This is done by following a set of consensus rules, which are designed to ensure the integrity and security of the blockchain. When a new transaction is submitted to the network, it is broadcast to all the nodes in the network. Each node then verifies the transaction, checking that it is valid and that the sender has sufficient funds to complete the transaction [20]. Once a transaction is validated, it is grouped with other transactions into a block and added to the blockchain. The block is then broadcast to all the nodes in the network, who update their copies of the ledger to reflect the new block.

324

C. Singh et al.

LPs handled the processing of logs in this BBFMI model. An Application Programming Interface (API) called the LCs serves as a bridge between different entities and the blockchain network. The IoT, network, and cloud layers all produce logs that are extracted by the LCs. The hashed data from the retrieved logs are written onto the block as a transaction after being hashed with the SHA-256 hash algorithm demonstrated by formula (1). Since the logs may include sensitive information, we decided to hash the logs. So it’s not a good idea to preserve the logs in plaintext. Second, by reducing the size of the logs by hashing, processing the logs takes less time overall. Transaction = Hash × (log)

(1)

4.1.3 User Centre Entity (UCs) The user centre entity (UCs) is composed of the courts and forensic investigators. Due to the existence of this component in the model, forensic investigators may validate the accuracy of logs that service providers send to them. Furthermore, forensic investigators may still verify the veracity of the logs even when they are being passed along from one investigator to another as part of the chain of custody. The court also has the power to evaluate whether a particular piece of evidence (log) needs to be accepted and to check the validity of the logs that the prosecution has provided. As a result, investigators are prevented from altering the records to either accuse or exonerate perpetrators. 4.1.4 Endorsement of Logs

Fig. 2. Endorsement process of proposed BBFMI framework

Investigators completely rely on the CPs, networks, and IoT devices for evidence when it comes to IoT forensics, in contrast to traditional digital forensics. This dependence on CPs might result in unreliable evidence. Our suggested approach uses a decentralised ledger to ensure log integrity while also allowing several forensic process parties to examine logs. In our proposed approach, the forensic investigator still depends on

A Study of Implementing a Blockchain-Based (BBFMI)

325

CPs and IoT devices for the evidence. However, the evidence—in this case, the various logs generated by the cloud instances, network, and IoT devices—is encrypted when a forensic investigator acquires such data from a CPs. The hashed result Fig. 2. Endorsement process of proposed BBFMI framework is then checked against hashes that are used as transaction values on the public blockchain. The prosecutor then searches for the hashed value on the BCs system. If the hash value is present on the blockchain, the prosecutor validates the log and delivers it to the court as trustworthy evidence. On the other hand, if the hash value does not exist on the blockchain network, the log is rejected. When a court acquires a log from a forensics expert, it may evaluate the credibility of the log by hashing it precisely and comparing the result to the hash values on the blockchain network. The court further than decide the acceptance and rejection on the basis credibility of the evidences. Figure 2. Demonstrate attestation of the proposed BBFMI framework. Our proposed approach is based on blockchain, making it totally decentralised, in contrast to the centralised manner described in previous works [13, 15]. Because our architecture is decentralised, it is feasible to audit logs to see if they are authentic or not. Additionally, it prevents forensic analysts and service providers from surreptitiously altering logs. Our approach makes use of the blockchain’s immutability property to ensure the accuracy of logs collected in an IoT environment during digital forensic investigations. It also has the benefit of being verifiable. This benefit offers the forensic organisations the opportunity to confirm the veracity of the logs generated in IoT devices, a benefit that is either not present or is not critically examined in the previous related work [14, 15].

5 Conclusion The Internet of Things (IoT) has brought about a proliferation of connected devices in our homes, workplaces, and cities. These devices, which include everything from smartphones and smart home devices to industrial control systems and medical equipment, generate vast amounts of data that can be used for forensic investigations. IoT devices are often vulnerable to tampering, both physically and remotely, which can make it difficult to trust the authenticity of the data they store while investigation. This makes it difficult for forensic investigators to use the data as evidence in criminal and civil proceedings. Blockchain technology, with its decentralized and tamper-proof nature, can be used to ensure the integrity and immutability of digital evidence from IoT devices. Implementing and endorsement of log processes entities modification, user centre entity workflow of a BBFMI for IoT devices in digital forensic investigation has the potential to revolutionize the way of incident response and better reliability of IoT devices. By leveraging the tamper-proof and decentralized nature of blockchain technology, BBFMI can ensure the integrity and immutability of digital evidence, making it a valuable tool for both forensic investigators and organizations looking to improve their incident response capabilities. However, it is important to note that further research and development is needed to fully realize the potential of BBFMI and overcome any technical challenges that may arise due to BBFMI.

326

C. Singh et al.

References 1. Okoh, S.A., Dibal, P.Y.: Performance analysis of IoT cloud-based platforms using quality of service metrics. Int. J. Wirel. Microw. Technol. 13, 1–4 (2020). https://doi.org/10.5815/ IJWMT.2023.01.05 2. Jain, R., Tata, S.: Cloud to edge: distributed deployment of process-aware IoT applications. In: Proceedings - 2017 IEEE 1st International Conference on Edge Computing, EDGE 2017, pp. 182–189 (2017). https://doi.org/10.1109/IEEE.EDGE.2017.32 3. Li, D., Peng, W., Deng, W., Gai, F.: A blockchain-based authentication and security mechanism for IoT. In: Proceedings - International Conference on Computer Communications and Networks, ICCCN. Institute of Electrical and Electronics Engineers Inc (2018). https://doi. org/10.1109/ICCCN.2018.8487449 4. Zhao, Y., et al.: Privacy-preserving blockchain-based federated learning for IoT Devices. IEEE Internet Things J. 8, 1817–1829 (2021). https://doi.org/10.1109/JIOT.2020.3017377 5. Gong, L., Alghazzawi, D.M., Cheng, L.: Bcot sentry: a blockchain-based identity authentication framework for IoT devices. Information 12, 203 (2021). https://doi.org/10.3390/info12 050203 6. Agrawal, S., Kumar, S.: MLSMBQS: design of a machine learning based split & merge blockchain model for QoS-aware secure IoT deployments. Int. J. Image Graph. Signal Process. 14, 58–71 (2022). https://doi.org/10.5815/IJIGSP.2022.05.05 7. Mishra, A., Singh, C., Dwivedi, A., Singh, D., Biswal, A.K.: Network forensics: an approach towards detecting cyber crime. In: 2021 International Conference in Advances in Power, Signal, and Information Technology, APSIT 2021. Institute of Electrical and Electronics Engineers Inc (2021). https://doi.org/10.1109/APSIT52773.2021.9641399 8. Motha, J., Maduranga, M., Jayatilaka, N.: Design of an IoT-enabled solar tracking system for smart farms. Int. J. Wirel. Microw. Technol. 12, 1–13 (2022). https://doi.org/10.5815/IJWMT. 2022.06.01 9. James, J.I., Shosha, A.F., Gladyshev, P.: Digital forensic investigation and cloud computing. In: Cloud Technology: Concepts, Methodologies, Tools, and Applications, pp. 1231–1271. IGI Global (2014). https://doi.org/10.4018/978-1-4666-6539-2.ch057 10. Atlam, H.F., Alenezi, A., Walters, R.J., Wills, G.B., Daniel, J.: Developing an adaptive riskbased access control model for the internet of things. In: Proceedings - 2017 IEEE International Conference on Internet of Things, IEEE Green Computing and Communications, IEEE Cyber, Physical and Social Computing, IEEE Smart Data, iThings-GreenCom-CPSCom-SmartData 2017, pp. 655–661 (2018). https://doi.org/10.1109/iThings-GreenCom-CPSCom-SmartData. 2017.103 11. Stoyanova, M., Nikoloudakis, Y., Panagiotakis, S., Pallis, E., Markakis, E.K.: A survey on the internet of things (IoT) forensics: challenges, approaches, and open issues (2020) https:// ieeexplore.ieee.org/abstract/document/8950109/. https://doi.org/10.1109/COMST.2019.296 2586 12. Rauf, A., Shaikh, R.A., Shah, A.: Trust modelling and management for IoT healthcare. Int. J. Wirel. Microw. Technol. 12, 21–35 (2022). https://doi.org/10.5815/IJWMT.2022.05.03 13. Li, S., Choo, K.K.R., Sun, Q., Buchanan, W.J., Cao, J.: IoT forensics: amazon echo as a use case. IEEE Internet Things J. 6, 6487–6497 (2019). https://doi.org/10.1109/JIOT.2019.290 6946 14. Nieto, A., Rios, R., Lopez, J.: IoT-forensics meets privacy: towards cooperative digital investigations. Sens. (Switz.) 18, 492 (2018). https://doi.org/10.3390/s18020492 15. Nieto, A., Rios, R., Lopez, J.: A methodology for privacy-aware IoT-forensics. In: Proceedings - 16th IEEE International Conference on Trust, Security and Privacy in Computing and Communications, 11th IEEE International Conference on Big Data Science and Engineering

A Study of Implementing a Blockchain-Based (BBFMI)

16.

17.

18. 19. 20.

327

and 14th IEEE International Conference on Embedded Software and Systems, pp. 626–633 (2017). https://doi.org/10.1109/Trustcom/BigDataSE/ICESS.2017.293 Noura, H.N., Salman, O., Chehab, A., Couturier, R.: DistLog: a distributed logging scheme for IoT forensics. Ad Hoc Netw. 98, 102061 (2020). https://doi.org/10.1016/j.adhoc.2019. 102061 Kumar, G., Saha, R., Lal, C., Conti, M.: Internet-of-Forensic (IoF): a blockchain based digital forensics framework for IoT applications. Futur. Gener. Comput. Syst. 120, 13–25 (2021). https://doi.org/10.1016/j.future.2021.02.016 Gao, H., et al.: BlockchainBot: a novel botnet infrastructure enhanced by blockchain technology and IoT. Electron. 11, 1065 (2022). https://doi.org/10.3390/electronics11071065 Novo, O.: Blockchain meets IoT: an architecture for scalable access management in IoT. IEEE Internet Things J. 5, 1184–1195 (2018). https://doi.org/10.1109/JIOT.2018.2812239 Atlam, H.F., Alenezi, A., Alassafi, M.O., Wills, G.B.: Blockchain with Internet of Things: benefits, challenges, and future directions. Int. J. Intell. Syst. Appl. 10, 40–48 (2018). https:// doi.org/10.5815/ijisa.2018.06.05

The Method of Determining the Adequacy of Virtual Community’s Themes Synko Anna(B) Lviv Polytechnic National University, Stepana Bandery Street 12, Lviv 79013, Ukraine [email protected]

Abstract. Due to the large increase in information content, there is a need for selection, verification data and sources which contain this data. The article presents the method for determining the adequacy of a source that may contain useful information. It is known that not all open sources of information on the Internet contain the necessary data for the user searching. Themes posted in virtual communities were chosen as the data source. The proposed method consistently checks the selected theme according to the established requirements (criteria): the measure of relevance of the information content and the measures of actuality of the data. The relevance score reflects the degree of compliance of the theme of the source, which may contain valuable information, to the user’s request. To calculate the relevance, statistical measure the TF-IDF and stemming methods (lemmatization and a lookup table) were applied. The actuality score is represented through an activity of creation or placement of posts in the theme. If over certain periods of time there is a significant increase in posts (information content) in the theme, then this theme and the data in it are urgent and needed for users. Indicators that determine the threshold value for each requirement are also provided. If each requirement is successfully met, the theme as a data source can be considered adequate. This method quickly finds relevant and “active” data sources that may contain the necessary information based on the experience of community users. To display the operation of the method, the industry of software is chosen. Among the selected 29 themes of different communities, only five are adequate to the user’s request, which is 17%. This means that, with a high probability, 83% of the information posted in other themes is not necessary for the user. Keywords: Virtual communities · Theme · Adequate source · Data relevance · social networks · Actuality · TF-IDF

1 Instruction Social networks are the most popular and most visited resources of the Internet. In 2021, over 4.26 billion people were using social media worldwide, a number projected to increase to almost six billion in 2027 (Fig. 1) [1]. Social networks are one of the platforms where users can share knowledge. On average, internet users spend 144 min per day on social media and messaging apps, an increase of more than half an hour since 2015 [1]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 328–340, 2023. https://doi.org/10.1007/978-3-031-36118-0_29

Adequacy of Virtual Community’s Themes

329

Fig. 1. A chart to represent number of social media users worldwide

It is known that social networks contain virtual communities (VC) on various themes. There are thematic communities and communities containing many themes. Virtual communities become a good source of data because it contains data not only published by product manufacturers, but also consumer feedback [2, 3]. Which gives an opportunity to learn about a certain product from different points of view [4]. The amount of open information is growing rapidly every day. Data on one specific question can be on thousands of different sites. Searching for information is one of the most demanded tasks in practice, which any Internet user has to solve [5]. Therefore, there is a need to develop a method for checking the relevance and actuality of data sources that may contain useful information.

2 Related Work VC research is a long-term experience among various sciences (including communication and social sciences, mathematical analysis and graph theory, marketing, artificial intelligence and software). These industries reflect the behavior of different segments of the population in a certain period of time on the Internet. The development of the source adequacy method is based on the research of the following aspects: virtual communities and data analysis techniques. 2.1 Virtual Communities Research In previous studies, the authors [6] conducted an analysis of types of communities and their features. As a result, a community model was built, which shows the general

330

S. Anna

structure characteristic of all types of virtual communities. The main components of the VC structure are participants and information content. Therefore, many studies are devoted to this. For example, the author [7] made a classification based on the behavioral characteristics of community members to understand their content creation activities and predict their subsequent actions. In the work [8], the assessment of posts is given, which depends on the rating and the way of communication with the author who published it. 2.2 Research of Data Analysis Techniques The author [9] defined the aspects of checking the quality and authenticity of the content in the VC where the quality of the content is determined by the correspondence of the text of the post to the theme of the community; authenticity is checked on the basis of text verification services (PeopleBrowsr, Snopes.com, Geofeedia, HuriSearch and etc.) and images (Foto Forensics, Findexif.com, Google Search by Image and etc.) or with the involvement of an expert. Works [10, 11] highlight the process of searching and analyzing the content of platforms by parsing their pages. The parsing method allows you to collect, systematize, and update information. The advantages of using this technique include: 1) 2) 3) 4) 5)

Fast data collection in any mode; Prevention of errors; Carrying out a regular inspection at a given time interval; Presentation of data in any format; Ensuring an even load on the website where parsing takes place (so as not to create the effect of a DDoS attack).

This parsing technique works only for one VC. To analyze and collect data from other websites, it is necessary to perform similar actions with their HTML pages. Linguistic methods are used to analyze the content of the text. One of them is stemming, which has several variants of algorithms that differ in accuracy and performance. The study [12] proposes to compare document retrieval precision performances based on language modeling techniques – particularly stemming and lemmatization. As a result, it was established that lemmatization gives a better result in terms of accuracy in contrast to the baseline algorithm. But lemmatization has an important drawback – dependence on the correct recognition of parts of speech. As opposed to previous studies, the author proposed to consider the theme itself as a source of useful information, because communities can have many themes and, accordingly, many thematic situations [13]. This will reduce the volume of data and provide the user with the data corresponding to his request. The quality of the selected content is suggested to be determined by the end user. The innovativeness of the work lies in the fact that in previous studies there was no verification of communities – their themes as a source of data. In works [13–15], the correspondence of the content is determined by one of the stemming methods – lemmatization. In this work, two approaches are used (lemmatization and a lookup table), the sequential application of which eliminates each other’s shortcomings and

Adequacy of Virtual Community’s Themes

331

increases the accuracy of the algorithm. Also was used a dictionary containing auxiliary terms that expands and supplements the user’s information request and theme names [16]. Fast gains content in VC effect on data volume and time consumption for processing them as well. Viewing the theme as a source of data will allow: 1) To ignore themes, which don’t semantically meet the user’s query but are in communities. 2) To reduce the risk of block due to excessive activity to download data. 3) To speed up the search for information by reducing the amount of data. The purpose of the works is to develop the method for determining the adequacy of the theme for downloading discussions from them. The software industry was selected to display the operation of this method.

3 The Method for Determining of Adequacy of a Theme as Source of Valuable Data Before uploading discussions, it is necessary to check themes as sources of data for their adequacy. The adequacy is a broad term covering all science (philosophy, psychology, jurisprudence, mathematics, etc.). In probability theory the adequacy is to comply with the actual results (for example, results of the research), with the results obtained by the calculation. In the technique the adequacy is compliance of the basic characteristics of the model to specifications of the object, which simulated [17]. The adequacy source of information is the themes of virtual communities that meets requirements: 1) To be relevant. 2) To be actual – the growth of the count of posts over a certain period of time. The relevance score (μrelevance ) specifies if themes of VC have information about a particular subject area. This estimate we’ll be considering using the relevance score of the user’s query to set off all keywords characterizing the theme. Restrictions on score μrelevance are range [0, 1]. The source can be considered relevant if performed the next the following criterion: μrelevance ≥ α

(1)

α – the indicator, which threshold for estimate of relevance score of sources. The actuality score (μactual ) determines the rate of content data growth for a certain period of time. Restrictions on score μactual are range [0,1]. The source can be considered relevant if performed the next the following criterion: μactual ≥ β

(2)

β – the indicator, which threshold for estimate of the actuality score of sources. These requirements define the adequacy of the theme are performed sequentially. If the condition of relevance doesn’t fulfill, then the next rate won’t be calculating and

332

S. Anna

the source will not be able to fulfill criterion the adequacy estimate. If the condition of relevance had been fulfilled but the actuality score doesn’t satisfy the given criterion the theme is not adequate. The UML activity diagram displays the method for determining the adequacy of a theme (Fig. 2). Activity diagrams describe the dynamic aspects and behavior of the system. Activity diagrams represent the flow from one activity to another activity.

Fig. 2. The UML activity diagram to determine of adequacy of a theme

4 Determination of Relevance of the Theme The relevance is measure of correspondence of obtained result to desired or expected. Search relevance is the measure of how closely and accurately the generated search results relate to the user’s search query. There are many search engines that will offer his own search priority in terms of indices. Therefore, the same query in different systems displays different results. After analyzing the criteria of increase of relevance of the search content [18] from VC was allocated the following: the title of VC, theme and post (message) should contain keywords and logical structure about certain object or subject (for example title or function the software); absence of errors in the text (for correct indexing of the text parts of the site); presence of completed meta tags that may contain useful data (for example, Title, Description, Keywords, Heading ect.), keywords in the URL-address and in internal links to the theme page, etc.; the content of post should be unique; loading time of community pages; visual representation of the content of the VC (text size).

Adequacy of Virtual Community’s Themes

333

Usually, the relevance is determined by statistical research methods. The main functions are to collect, construction, group, synthesis, analysis and evaluation of data. So, the relevance theme is a theme which is in VC and with a certain degree of relevance contains content in the form of a set of discussions to the user’s request. For selection and further evaluation of themes of virtual communities it is necessary to perform and fill the database according to the following attributes: URL address, title of the VC, titles of themes. All this data needed for the automated data collection (parsing). After that, it is necessary to select those themes that can contain information about a software. Communities that contain software data often list this information in the community title (if this source is entirely focused on the specified software) or in the theme(s). To compare the titles, the author introduced the concept of the measure relevance of title software to themes titles. If the title of the software is not mentioned in the community themes, we will assume that this source does not contain information about it. And therefore, based on that, the author developed the method of selecting VC that may contain valuable content for a specific software (Fig. 3).

Fig. 3. A flowchart of the selection of VC themes

The method of selection VC themes is to perform the next steps: 1) For each VC themes calculated the relevance measure according to the user’s request – μrelevance (SW (Keyword ), VC(Themei )). 2) If the relevance value of the title of the i-th theme to the keywords of the software is greater than or equal to the specified value, then we will assume that the source (theme) is relevant. 3) Otherwise, we will consider that the source (theme) is not relevant.

334

S. Anna

4) The method is made until all themes are mapped to the keywords of the software. The result of the method has selected communities and rated themes that contain information about a certain software. Therefore, the relevance indicator is fulfilled. Later, the selected themes should be investigated according to the following assessment – an actuality of the content. After all, it is this evaluation that influences the subsequent decision regarding acceptance or rejection of the information content of a certain theme of the VC. 4.1 Calculation of Relevance of the Theme To determine the estimate of relevance of the theme to the user’s request, it is necessary to perform text normalization [19]. Normalization is a method in which a set of words in a sentence is turned into a sequence to reduce search time. Words that have the same meaning but have some differences depending on the context or sentence are normalized. A stemming operation is usually used to normalize the text. The result of stemming is finding the base of the word (stem) by removing auxiliary parts such as endings and suffixes. The results of stemming are sometimes very similar to determining the root of a word, but its algorithms are based on different principles [20]. Therefore, the word after processing by the stemming algorithm may differ from the morphological root of the word. To eliminate this issue, it is necessary to use a combination of stemming algorithms (hybrid approach): lemmatization and a lookup table [21]. The algorithmic process of lemmatization consists in finding the lemma of a word depending on its meaning. Lemmatization usually refers to the morphological analysis of words, the purpose of which is to remove inflectional endings. Therefore, this approach is considered better than other stemming algorithms because the result is a lemma, which is the original, basic form of its inflectional forms. The disadvantage of using lemmatization is the dependence on the correct recognition of parts of speech. The stemming algorithm of a lookup table is implemented on the basis of a table that contains all possible variants of words and their forms after stemming. The advantages of this method are simplicity, speed and convenience of processing exceptions to language rules. The disadvantages include the fact that the search table must contain all forms of words: that is, the algorithm will not work with new words. But this drawback is eliminated when using the lemmatization algorithm. Based on stemming algorithms, a text normalization algorithm was developed, which takes place in the following steps: 1) Divide the text into lexemes (words). 2) Clean words from apostrophes, punctuation marks. 3) Remove words that are not taken into account during the linguistic analysis because they do not carry important information for the text (pronouns, articles, prepositions, conjunctions) [22]. 4) Apply the lemmatization algorithm, which will allow you to obtain lemmas (the basic form of words). And if a lemma for a word is not found, it is handed over to an expert for consideration. 5) Apply the stemming algorithm – lookup table.

Adequacy of Virtual Community’s Themes

335

6) The words that were not processed as a result of passing the search algorithm according to a lookup table are transferred to the expert, and the algorithm for cutting off endings and suffixes is performed. It should be taken into account that the keywords specified by the user may not be sufficient for the selection of themes. Therefore, in order to expand the query of the selection of sources, it is necessary to develop a dictionary of terms containing important, useful words for the selected subject area. As a result, it is possible to combine the user query and the dictionary as follows into an expression (Vyraz) containing a set of meaningful (unique, non-repeated) words: Vyraz = norm(Query) ∪ Dicrionary

(3)

where norm(Query) – set of normalized words that are in the user’s request; Dicrionary – a dictionary of terms for the selected subject area. We will calculate the degree of correspondence of the theme of the VC of the chosen subject area based on the word frequency formula (TF-IDF measures): μrelevance =

||Vyraz|| ∩ ||norm(Theme)|| ||norm(Theme)|

(4)

where ||Vyraz|| – set of normalized words that are in the user’s request; ||norm(Theme)| – a dictionary of terms for the selected subject area. The text normalization process according to a user request is depicted using the UML sequence diagram in Fig. 4. The result of passing this algorithm is the normalization of the words contained in the text using stemming methods. The consistent application of several methods allowed to eliminate the shortcomings of each and make the normalization method universal for any words. 4.2 Determining the Actuality Requirement of Virtual Communities’ Themes The actuality is the subjective value that people attribute to information if it meets their needs and interests. The actuality of data is a property that reflects their significance in today’s reality [23]. Accordingly, an actuality post within the community discourse is a post containing relevant data for user queries. The actuality of the discourse as an element of the theme is determined by the presence of relevant posts in it. So, the actuality of the theme is characterized by the number of published relevant discourse. To measure the value of relevance, the author introduced the concept of the level of actuality, which is determined by the number of relevant posts in the theme, as a source of data. It allows the following statement: the more posts that have been created in a certain period of time (T) in a certain theme, the greater the importance of the actuality level. The natural limitations of the actuality level are [0, 1]. Therefore, to calculate the level of actuality, it is necessary to apply the concept of activity (A), which determines the rate of increase in information content in a certain theme.

336

S. Anna

Fig. 4. UML sequence diagram of the text normalization process and determination of the degree of conformity

To calculate the actuality of the theme, author applied a logarithmic function, which has the following advantages: 1) The set of definitions of the logarithmic function D = (0, ∞), and the range of obtained results for the theme will be within the natural limits [0,1]. 2) The logarithmic function is monotonic and increasing (to calculate the activity of the theme). When b > 1 (the increase of messages in the theme must be more than one message): a) if b = 1 – an increase in the theme by one post, which is not a good indicator for confirming activity. Accordingly, we will consider the theme not actual; b) if b = 0 – there was no information growth during the comparison period or the theme is new and has no post yet. Accordingly, the logarithm will be undefined (which does not meet the condition for calculating the logarithm). Such a theme will not be considered; 3) Logarithmic functions are proportional to different bases. 4) Simplicity when working with a large amount of data (posts), since the logarithm function grows more slowly with increasing numbers [24]. Taking into account the above conditions for the use of the logarithm, the author developed a formula for calculating the activity of the theme using the binary logarithm: A=

log2 (yi ) log2 (yn )

(5)

Adequacy of Virtual Community’s Themes

337

where yi – logarithmic value of the number of posts of the current level (compared period); yn – the logarithmic value of the total number of posts. Therefore, using this method will allow you to eliminate insignificant themes that have a small increase in the number of posts over a certain period of time, as well as get rid of sources that have no posts (new theme) or have only one post, which cannot be considered a discussion.

5 Results An experimental study was conducted to determine the adequacy of themes as sources of valuable information. A parsing process was applied to automatically collect data from the VC. User request: «Development of web applications in Java». Accordingly, it was necessary to find exactly those themes that will contain actual and relevant information for this request. Five communities were selected for information selection – Dou.ua, replace.org, senior.ua, proger.com, cyberforum.ru (Fig. 5). These communities contain themes on various subjects. Each theme has a score calculated according to the following estimate relevance. For indicatorα, which threshold for these estimated the following value is applied α ≥ 0, 75. Therefore, the theme is adequate if all indicators are higher than 0, 75.

Fig. 5. A diagram to represent a selection of communities and themes by relevance of data

338

S. Anna

As can be seen from the research, among 29 themes, only 6 turned out to be relevant sources of data (Fig. 5). The next step is to check relevant themes about the activity of creating posts for certain periods – the last six months (Table 1). Table 1. Results regarding the growing of count of post in adequate themes over the past six months Community title

Theme title

Post growth rate (monthly)

Total score

Aug

Sept

Oct

Nov

Dec

Jan

0.61

0.69

0.79

0.9

0.82

0.81

dou.ua/forums

Developing

0.77

Gamedev

0.93

0.94

0.92

0.96

0.99

0.96

0.95

proger.com.ua

Programming

0.72

0.8

0.84

0.79

0.85

0.88

0.81

cyberforum.ru

Webmaster

0.56

0.59

0.52

0.66

0.69

0.59

0.6

Software

0.72

0.68

0.71

0.75

0.81

0.8

0.75

Programmers

0.88

0.89

0.86

0.94

0.92

0.91

0.9

Total score in Table 1 represents the arithmetic mean of the activity of filling messages in the theme for the last six months. For indicator β, which threshold for these estimates the following value is applied β ≥ 0, 75. As can be seen from the research, among six adequate themes, only 5 satisfy the condition of relevance.

6 Summary and Conclusion Due to the large amount of available data on the Internet, there is a need to check sources that may contain valuable and useful information. Today, virtual communities are the most visited sites where users share their experiences. Since the communities contain different themes, the author proposed to research the themes as a source of data. Therefore, it was suggested to check each theme for adequacy according to requirements: the measure of relevance and actuality. The research was conducted using data parsing – necessary attributes from virtual communities (title of the community, rating and date of publication of the post and ect). The obtained data was determined as a measure of relevance according to the user’s request using a TF-IDF and linguistics methods. After that, the assessment of the actuality of the adequate theme, which determines the growth of count of posts over a certain period of time, is calculated. Each estimate of the given measures is in the range [0, 1]. To determine the adequacy is introduced indicators α and β, which threshold for estimates. To conduct the research, it was established that α ≥ 0, 75 (the relevance assessment) and β ≥ 0,65 (the actual assessment). The developed method made it possible to speed up the search for data from the VC and to focus the end user’s attention on those themes that are relevant and actual. And therefore, with a high probability, adequate themes contain the necessary data for the

Adequacy of Virtual Community’s Themes

339

end user. Selection and verification of data posted in adequate themes should be carried out by the user searching for information. As a result of the research, it was established that only 17% of themes are adequate for the user. These results are compatible with the Pareto principle, which states that roughly 80% of consequences come from 20% of causes. Based on this, it can be argued that 20% of themes contain data that provide 80% of the necessary data to the user.

7 Future Research Today, it is important not only to research data sources which can have useful information, but also the content of this source. Posted information can be analyzed according to the following components: content and rating of post. In VC, users rate a post, thereby forming its rating. The rating reflects how relevant the post was for other users. Published data can be analyzed for unwanted content, links, malicious files, etc., as well as establish authenticity and uniqueness based on text verification services. In the next study, the author plans to develop a software complex that collects authority posts from adequate themes.

References 1. Statista. Number of social media users worldwide from 2017 to 2027. Social Media & User-Generated Content. https://www.statista.com/statistics/278414/number-of-worldwidesocial-network-users/ 2. Korzh, R., Peleschyshyn, A., Syerov, Y., Fedushko, S.: The cataloging of virtual communities of educational thematic. Webology 11(1), 117 (2014) 3. Trach, O., Peleshchyshyn, A., Korzh, R.: Methods for creating a team for managing a virtual community. CEUR Workshop Proceedings. In: Proceedings of the 1st International Workshop IT Project Management (ITPM 2020), vol 2565, pp. 83–92 (2020) 4. Winanti, F., Goestjahjanti, S.: Component learning community for informal education to support culinary community at era new normal covid-19: A systematic literature review. Int. J. Inf. Technol. Comput. Sci. 14(1), 20–31 (2022). https://doi.org/10.5815/ijitcs.2022.01.03 5. Faryad, S., Batool, H., Asif, M., Yasin, A.: Impact of internet of things (IoT) as persuasive technology. Int. J. Inf. Technol. Comput. Sci. (IJITCS) 13(6), 16–28 (2021). https://doi.org/ 10.5815/ijitcs.2021.06.02 6. Peleshchyshyn, A., Kravets, R., Sierov, Y.: Analysis of existing types of virtual communities on the internet and building a virtual community model based on a web forum. Inf. Syst. Netw. J. Lviv Polytech. Nat. Univ. 699, 212–221 (2011) 7. Fedushko, S.: Development of a software for computer-linguistic verification of sociodemographic profile of web-community member. Webology, 11(2), 126 (2014). http://www. webology.org/2014/v11n2/a126.pdf 8. Fedushko, S., Mastykash, O., Syerov, Y., Shilinh, A.: Model of search and analysis of heterogeneous user data to improve the web projects functioning. In: Zhengbing, H., Petoukhov, S., Dychka, I., He, M. (eds.) Advances in Computer Science for Engineering and Education IV, pp. 56–74. Springer International Publishing, Cham (2021). https://doi.org/10.1007/9783-030-80472-5_6

340

S. Anna

9. Korzh, R., Peleshchyshyn, A., Trach, O., Tsiutsiura, M.: Increasing the efficiency of the processes of formation of the informational image of the HEI. In: Shakhovska, N., Medykovskyy, M.O. (eds.) CSIT 2019. AISC, vol. 1080, pp. 661–679. Springer, Cham (2020). https://doi. org/10.1007/978-3-030-33695-0_45 10. Zhou, B., Hang, Z., Xavier, P., Sanja, F., Adela, B., Antonio, T.: Scene parsing through ADE20K dataset. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 633–641 (2017). https://doi.org/10.1109/CVPR.2017.544 11. Zhao, H., Hengshuang, Z., Jianping, S., Xiaojuan, Q., Xiaogang, W., Jiaya, J.: Pyramid scene parsing network. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2881–2890 (2017) 12. Balakrishnan, V., Lloyd-Yemoh, E.: Stemming and lemmatization: A comparison of retrieval performances. In: Proceedings of SCEI Seoul Conferences, Seoul, Korea, pp.174–179 (2014) 13. Synko, A.: The method of trust level of publications hosted in virtual communities. Sci. J. TNTU (Tern.) 105(1), 68–79 (2022). https://doi.org/10.33108/visnyk_tntu2022.01.068 14. Lamberth, C.: The power of cross-indication testing: agrochemicals originally stemming from a different indication. Pest Manage. Sci. 78(11), 4438–4445 (2022). https://doi.org/10.1002/ ps.7100 15. Alhaj, Y.A., Xiang, J., Zhao, D., Al-Qaness, M.A., Dahou, M.A.: A study of the effects of stemming strategies on Arabic document classification. IEEE Access 7, 32664–32671 (2019). https://doi.org/10.1109/ACCESS.2019.2903331 16. Babich, O., Vyshnyvskiy, V., Mukhin, V., Zamaruyeva, I., Sheleg, M., Kornaga, Y.: The technique of key text characteristics analysis for mass media text nature assessment. Int. J. Modern Educ. Comput. Sci. (IJMECS) 14(1), 1–16 (2022). https://doi.org/10.5815/ijmecs. 2022.01.01 17. Tedeschi, L.O.: Assessment of the adequacy of mathematical models. Agric. Syst. 89(2–3), 225–247 (2006) 18. Dogan, A., Birant, D.: Machine learning and data mining in manufacturing. Expert Syst. Appl. 166, 114060 (2021) 19. Singh, D., Singh, B.: Investigating the impact of data normalization on classification performance. Appl. Soft Comput. 97, 105524 (2020) 20. Ballabio, D., Grisoni, F., Todeschini, R.: Multivariate comparison of classification performance measures. Chemom. Intell. Lab. Syst. 174, 33–44 (2018) 21. Jivani, A.G.: A comparative study of stemming algorithms. J. Comput. Sci. Technol. 2(6), 1930–1938 (2011) 22. Manalu, S.R.: Stop words in review summarization using TextRank. In: 14th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON) (2017) 23. Ngiam, K.Y., Khor, I.W.: Big data and machine learning algorithms for health-care delivery. Oncology 20(5), 262–273 (2019) 24. Reynolds, R., Stauffer, A.: A method for evaluating definite integrals in terms of special functions with examples. Int. Math. Forum 15(5), 235–244 (2020)

A Feature Selection and Regression Refinement Network for Power Facility Detection in Remote Sensing Images Yu Zhou1 , Wenhao Mo2 , Changyu Cai2 , Yuntao Yao2(B) , and Yanli Zhi1 1 State Grid Jiangxi Electric Power Supply Co. Ltd., Nanchang 330096, China 2 China Electric Power Research Institute, Beijing 100192, China

[email protected]

Abstract. Intelligent identification and detection of power facility is crucial to power grid security, and remote sensing equipment is usually used to collect images in power routing inspection. However, the characteristics of remote sensing images, such as complex background, large scale change and arbitrary orientation, make it challenging to detect the objects in them. To improve the detection accuracy of power facility in remote sensing images, a feature selection and regression refinement network is proposed in this paper. Specifically, a bidirectional feature fusion module (BFFM) is devised which fuses deep and shallow features to obtain a multi-scale feature map with stronger representation ability. A feature selection module (FSM) with hybrid attention is devised which utilizes the fusion of spatial and channel attention to obtain attention response maps and generates the features specific to classification task and regression task through different polarization functions. Furthermore, a regression refinement module (RRM) is designed to adjust the anchors based on regression features to improve the regression accuracy. Experiments on DOTA and a power facility dataset achieve excellent detection results, the mAP on DOTA reaches 73.1 and the mAP of high voltage tower reaches 71.2. Keywords: Object detection · Remote sensing · Power facility

1 Instruction Detection of power facility plays an important role in maintaining the security of the grid system. Large power facilities such as high-voltage towers and thermal power chimneys have high spatial complexity, and require remote sensing acquisition equipment with wide coverage to obtain their source images, so it is of great significance to research the object detection of power facility in remote sensing images. However, the background of remote sensing images is complex, which cause irrelevant background features to seriously disturb the feature information of objects. Moreover, some objects are densely arranged and have small sizes, making it impossible to effectively capture the feature distribution of objects. In addition, the orientation of objects is arbitrary, which makes the object detection of remote sensing image more difficult. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 341–348, 2023. https://doi.org/10.1007/978-3-031-36118-0_30

342

Y. Zhou et al.

With the development of deep learning, many object detection algorithms [1–3] have achieved relatively accurate detection results. RetinaNet [4] is a typical one-stage algorithms, it uses Focal Loss to balance positive and negative samples, while using feature pyramids to make predictions on multiple layers of feature maps, which achieves high accuracy and speed. CAD-Net [5] and method [6] alleviate the problem that the object scale varies greatly in remote sensing images by constructing more powerful feature representations. R3Det [7] develops a feature refinement module to increase detection performance by obtaining more accurate features. However, there are few researches on object detection for power facilities in remote sensing images. DeepWind [8] is a weakly supervised localization model which automatically localize wind turbines in satellite imagery. The model takes a satellite image as input and output a predicted response map which indicates the location of the turbines in the image. In the response map, a peak detection module finds local maxima and combines them based on according to proximity. Using deep convolutional neural networks, Sal-MFN [9] is an end-to-end detection framework for Thermal Power Plants. It employs a multi-scale feature network for the adaptation of the thermal power plants of various sizes and a saliency enhanced module to reduce background distractions for enhanced feature maps. In this paper, a feature selection and regression refinement network (FSRR-Net) for power facility detection in remote sensing images is proposed. Specifically, to enhance the network’s feature representation ability, a bidirectional feature fusion module (BFFM) is designed. To generate task-specific features, a feature selection module (FSM) with hybrid attention is designed. To improve the regression accuracy, a regression refinement module (RRF) is designed to adjust the anchors for better spatial alignment with the object feature.

2 Proposed Method Figure 1 displays the overall framework of the proposed FSRR-Net. ResNet [10] is used as the backbone. Firstly, the bidirectional feature fusion module (BFFM) fuses shallow features and deep features to generate a multi-scale feature pyramid. Then, the features specific to classification task and regression task are extracted separately through the feature selection module (FSM) with hybrid attention. Finally, the regression refinement module (RRM) adjusts anchors in multiple levels based on regression features. The BFFM can improve the ability of network to represent features, the FDM can generates the features specific to different tasks, the RRM can improve spatial misalignment between anchor and object feature. In this way, the detection performance can be significantly improved. 2.1 Bidirectional Feature Fusion Module In order to detect objects of various scales, we add a bottom-up fusion path on the basis of the top-down fusion path of FPN [11]. Shallow features no longer need to pass through as many convolution layers to reach the top layer due to the architecture, so as to reduce the feature loss during the transfer process. We also introduce P6 and P7 layers in the feature pyramid to predict larger scale objects. The new feature extraction module is

A Feature Selection and Regression Refinement Network

343

Fig. 1. The overall framework of FSRR-Net

called Bidirectional Feature Fusion Module (BFFM). Details of the module are shown in Fig. 2, where 1 × 1Conv denotes the convolution process with the 1 × 1 convolution kernel, which serves to change the feature map’s channel number; 3×3Conv denotes the convolution process with the step size of 1 and the 3 × 3 convolution kernel; 2 × Down denotes the two-fold down-sampling operation, which is implemented with the step size of 2 and the 3 × 3 convolution kernel; 2 × Up represents the two-fold up-sampling operation, which is implemented with the bilinear interpolation.

Fig. 2. Bidirectional Feature Fusion Module

2.2 Feature Selection Module with Hybrid Attention The majority of object detection frameworks use the same feature maps for classification and regression tasks. However, due to the inconsistency between the two tasks, the shared features reduce the detection performance [12]. We design Feature Selection Module (FSM) with hybrid attention to avoid this problem, which can extract the features specific to the classification and the regression separately. Figure 3 shows the structure of FSM. Firstly, a hybrid attention mechanism with channel attention [13] and spatial attention

344

Y. Zhou et al.

[14] generate the attention response maps of feature maps. Then, the features specific to different tasks are generated by different polarization functions.

Fig. 3. Feature Selection Module with hybrid attention

The channel attention model achieves enhancement or suppression of certain channels of the feature map by learning the importance of different channels. The spatial attention model selects feature regions that contribute significantly to the task for processing. After attention response maps Wreg ∈ RC×W ×H and Wcls ∈ RC×W ×H are generated by the hybrid attention model, the polarization functions for different tasks are used to obtain the task-specific features. In the regression task branch, network focuses more on the boundary features of the object. Therefore, the polarization function needs to enhance the boundary information of the feature map. In the classification task branch, network focuses more on semantic information. Therefore, the polarization function needs to enhance the useful information of the feature map and ignore the background noise information. The polarization functions for regression and classification tasks respectively are as follows: ⎧ ⎨ freg (x) = 4x(1 − x) (1) 1 ⎩ fcls (x) = 1 + e−(x−0.5) The extracted features for the regression task are spread uniformly over the object, which helps to recognize the boundaries of the object and locate it accurately. The extracted features for the classification task are mainly clustered in the most easily recognized object’s parts to prevent disruptions from its other parts, which leads to more accurate classification results.

A Feature Selection and Regression Refinement Network

345

3 Regression Refinement Module Anchor-based object detection algorithms need to pre-set anchor of a certain scale and a certain aspect ratio, and during the training, the class which the anchor belongs to and the offset from the ground-truth are learned. In remote sensing images, because of the large scale-variation and arbitrary orientation of objects, it is difficult to align the fixed anchor with the object. Based on the above analysis, we design Regression Refinement Module (RRM), Based on the above analysis, this paper designs an anchor frame refinement network based on image features, which apply multi-level refinement to pre-set anchors to alleviate spatial unalignment between anchors and objects, enabling anchors to capture more object features. Figure 4 displays the structure of RRM.

Fig. 4. Feature Selection Module with hybrid attention (r in spatial attention – dilation rate)

The proposed algorithm uses the five-parameter approach to describe the bounding box. The anchor after the nth refinement is noted as (xn , yn , wn , hn , θ n ), which represents the pre-set anchor when n = 0. The offset learned in the n + 1 th refinement is (txn+1 , tyn+1 , twn+1 , thn+1 , tθn+1 ), defined as follows: ⎧ xn+1 − xn n+1 yn+1 − yn ⎪ txn+1 = , ty = , ⎪ ⎪ wn hn ⎪ ⎨ wn+1 hn+1 twn+1 = log( ), thn+1 = log( ), ⎪ ⎪ wn hn ⎪ ⎪ ⎩ n+1 tθ = tan(θn+1 − θn )

(2)

where (xi , yi )wi , hi , θi represent the center point coordinates, width, height and angle of the anchor in the i th refinement, respectively. Because of the serious boundary problems with the five-parameter approach, we use IoU-Smooth L1 loss to address this problem [15]. IoU-Smooth L1 is defined as follows:    ,v Lreg vnj  nj   (3) IoU − Smooth L1 = − log(IoU )  j∈{x,y,w,h,θ} Lreg vnj , vnj  represents the prediction’s offset, v where Lreg (·) represents the Smooth L1 loss [1], vnj nj represents the ground-truth’s offset, and IoU represents the intersection ratio between

346

Y. Zhou et al.

the predicted and true bounding boxs. The loss function is approximately equal to zero in the boundary situation, which eliminates the abrupt increase in loss. The proposed detector contains two tasks. The loss of the regression task consists of refinement regression loss and detection regression loss, both calculated by IoU-Smooth L1. The loss of the classification task is the anchor category prediction loss, calculated by Focal Loss [16]. The weighted sum of the classification and regression losses constitutes the overall loss function of the proposed detector.

4 Experiments Experiments are carried out on a public remote sensing dataset DOTA [16] and a power facility dataset collected from the Internet. DOTA contains 2806 aerial images, fully annotated with 188, 282 instances. There are 15 categories in DOTA, We select 7 categories related to facilities for experiments, which are storage tank (ST), basketball court (BC), ground track field (GTF), harbor (HA), bridge (BR), soccer ball field (SBF) and swimming pool (SP). The power facility dataset contains 1534 remote sensing images of 4 categories, including high-voltage tower, oil tank, photovoltaic power panel and thermal power chimney. 4.1 Results on DOTA To evaluate the performance of the proposed FSRR-Net, we compare it to RetinaNet [4], CAD-Net [5], R3Det [7] on the DOTA dataset. In the experiments, FSRR-Net uses ResNet152 as the backbone and uses rotation and random flip for image enhancement. Table 1 displays the experimental results. Table 1. Detection results on DOTA (%) (The best results of each category are bolded.) Methods

BR

GTF

BC

ST

SBF

HA

SP

mAP

RetinaNet

43.7

66.7

80.2

75.5

53.9

63.9

65.9

68.7

CAD-Net

49.4

66.5

79.2

73.3

48.4

62.0

67.0

69.9

R3 Det

48.5

62.5

81.4

83.5

62.0

65.4

67.5

71.7

FSRR-Net

52.1

68.6

86.7

86.3

64.8

66.9

68.7

73.1

From Table 1, the mAP of the proposed FSRR-Net reaches 73.1% for fifteen categories of objects, which is the best among the four algorithms, and achieves best detection results in all seven categories objects related to facilities. Results from the experiments demonstrate the proposed detector’s superior performance.

A Feature Selection and Regression Refinement Network

347

Table 2. Detection results on the power facility dataset (%) Category

mAP

High-voltage tower

71.2

Oil tank

87.7

Thermal power chimney

76.1

Photovoltaic power panel

21.5

4.2 Results on the Power Facility Dataset Table 2 displays the FSRR-Net’s detection mAPs of the high-voltage tower, oil tank, photovoltaic power panel and thermal power chimney, and Fig. 5 shows the corresponding visualization From Table results. 2, it can be seen that the proposed FSRR-Net obtains high mAPs on the high-voltage tower, oil tank, and thermal power chimney, indicating that FSRR-Net is able to accurately detect objects of different scale and densely arranged within the images. However, the mAP is low on the photovoltaic power panel, which is probably due to the fact that the size of a photovoltaic panel unit in remote sensing images is too small and the arrangement of photovoltaic panels is too dense, making it difficult for the network to extract useful features.

Fig. 5. The visualization detection results on the power facility dataset

5 Summary and Conclusion To increase the detection accuracy of power objects in remote sensing images, this paper proposes a feature selection and regression refinement network (FSRR-Net). Specifically, the bidirectional feature fusion module (BFFM) is designed to detect objects of various scales. The feature selection module (FSM) is designed to generate the features required for classification task and regression task correspondingly, which can avoid the interaction between the two tasks. Furthermore, the regression refinement module

348

Y. Zhou et al.

(RRM) is designed to enable the anchor to capture more effective object features. Experiments on DOTA dataset and a power facility dataset achieves excellent detection results, indicating that our method has superior detection performance for typical power facilities and can be well used for intelligent monitoring of power facilities. Funding Information. This work is funded by the Science and Technology Project of State Grid Corporation(Grant No. 5700-202225178A-1-1-ZN).

References 1. Girshick, R.: Fast R-CNN. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 1440–1448 (2015) 2. Li, L., Zhou, Z., Wang, B., et al.: A novel CNN-based method for accurate ship detection in HR optical remote sensing images via rotated bounding box. IEEE Trans. Geosci. Remote Sens. 59(1), 686–699 (2020) 3. Xiao, J., Guo, H., Zhou, J., et al.: Tiny object detection with context enhancement and feature purification Expert Syst. Appl. 118665 (2023) 4. Lin, T.Y., Goyal, P., Girshick, R., et al.: Focal loss for dense object detection. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 2980–2988 (2017) 5. Zhang, G., Lu, S., Zhang, W.: CAD-Net: A context-aware detection network for objects in remote sensing imagery. IEEE Trans. Geosci. Remote Sens. 57(12), 10015–10024 (2019) 6. Xiao, J., Guo, H., Yao, Y., et al.: Multi-scale object detection with the pixel attention mechanism in a complex background. Remote Sens. 14(16), 3969 (2022) 7. Yang, X., Yan, J., Feng, Z., et al.: R3det: Refined single-stage detector with feature refinement for rotating object. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, no. 4, pp. 3163–3171 (2021) 8. Zhou, S., Irvin, J., Wang, Z., et al.: Deepwind: weakly supervised localization of wind turbines in satellite imagery. In: 33rd Conference on Neural Information Processing Systems (NeurIPS 2019) (2019) 9. Yin, W., Sun, X., Diao, W., et al.: Thermal power plant detection in remote sensing images with saliency enhanced feature representation. IEEE Access 9, 8249–8260 (2021) 10. He, K., Zhang, X., Ren, S., et al.: Deep residual learning for image recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770–778 (2016) 11. Lin, T.Y., Dollár, P., Girshick, R., et al.: Feature pyramid networks for object detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2117– 2125 (2017) 12. Wu Y, Chen Y, Yuan L, et al. Rethinking classification and localization for object detection[C]. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 10186–10195 13. Hu, J., Shen, L., Sun, G.: Squeeze-and-excitation networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 7132–7141 (2018) 14. Woo, S., Park, J., Lee, J.-Y., Kweon, I.S.: CBAM: convolutional block attention module. In: Ferrari, V., Hebert, M., Sminchisescu, C., Weiss, Y. (eds.) ECCV 2018. LNCS, vol. 11211, pp. 3–19. Springer, Cham (2018). https://doi.org/10.1007/978-3-030-01234-2_1 15. Yang, X., Yang, J., Yan, J., et al.: Scrdet: towards more robust detection for small, cluttered and rotated objects. In: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 8232–8241 (2019) 16. Xia, G.S., Bai, X., Ding, J., et al.: DOTA: a large-scale dataset for object detection in aerial images. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 3974–3983 (2018)

Development of Arithmetic/Logical Computational Procedures: Computer Aided Digital Filter Design with Bilinear Transformation Abdul Rasak Zubair1(B) and Adeolu Johnson Olawale1,2 1 Electrical and Electronic Engineering Department, University of Ibadan, Ibadan, Nigeria

[email protected] 2 Electrical & Electronics Engineering Department, Lead City University, Ibadan, Nigeria

Abstract. The development of a software system for computer aided digital filter design using bilinear transformation is presented. Starting from first principles, patterns in the manual determination of filter coefficients are studied and converted to arithmetic/logical computational procedures or operations which are coded into computer subprograms. These subprograms are packaged as a software system which is user friendly, and which can be regarded as a filter coefficients’ calculator. The software system is subjected to rigorous tests and is found to be accurate. The software system also plots pole-zero diagram and the frequency response. Mathematics coupled with the Arithmetic/Logical computational capabilities of the computer have been used to assist digital filter design in Electrical and Electronic Engineering. Keywords: Filter · Approximation functions · Filter coefficients · Bilinear transformation · Patterns · Arithmetic and logical operations · Frequency response

1 Introduction A filter is said to be a frequency selective electrical network. It passes some frequencies and blocks some frequencies. Filtering is used to extract a desired signal from the available mixture of signals. Analog filter processes analog signals while digital filter processes discrete time signals. Analog signal or continuous-time signal can be converted to discrete-time signal before being processed by a digital filter and later converted back to analog signal [1–3]. Designing a digital filter basically entails determination of a set of coefficients of the filter. Infinite Impulse Response (IIR) Filter design begins with normalized analog low pass filter specifications of Fig. 1 [3, 4]. ε is the passband ripple parameter, δp is the passband deviation, δs is the stopband deviation, wpass is the passband edge frequency, and wstop is the stopband edge frequency. Apass and Astop are the passband Gain and the stopband Gain respectively as described in Eqs. (1) and (2) [3, 4]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 349–360, 2023. https://doi.org/10.1007/978-3-031-36118-0_31

350

A. R. Zubair and A. J. Olawale

An approximation function provides a realizable frequency response which is approximate to the ideal frequency response. Five approximation functions are studied in details in [4]. These are the Butterworth, Chebyshev, Inverse Chebyshev, Elliptic, and Bessel approximation functions [4–10]. A narrow transition band and a high negative stopband Gain are desirable for effective frequency discrimination between the passband and stopband frequencies. The narrower the transition band (lower wstop /wpass ratio), the higher the order number n. Higher negative stopband Gain Astop also requires higher order number n. The higher the order number n, the more complex are the design computations and the filter network [4].

Fig. 1. Low pass filter specifications [3, 4]

Apass = 10log 10 (1 + ε2 ) = −20log 10 (1 − δp )

(1)

Astop = −20log10 (δs )

(2)

Digital Filter design is an active area of research in Electrical and Electronic Engineering. A number of complex approaches are used to arrive at the filter coefficients using optimization techniques or tedious approaches using extensive tables coupled with manual expansion and computations [11–15]. A simple approach is developed in this work. There are basic equations in the literature for designing normalized analog low pass filter with passband edge frequency of 1 rad/sec for the different approximation functions. There are substitutions in the form of additional equations to be made to convert a normalized analog low pass filter to analog low pass or high pass or band pass or band stop filter with given edge frequencies. There is another equation which is known as bilinear transformation to convert analog filter to digital filter. The collection of these equations are referred to as first principles or theory of filter design [1–10]. Filter design requires complex manual mathematical expansions of expressions and complex computations of filter’s coefficients. The manual expansions and computations from first principles are

Development of Arithmetic/Logical Computational Procedures

351

studied, observed patterns in these tasks are converted to arithmetic/logical computational procedures or operations which are coded into computer programs. Mathematics coupled with the Arithmetic/Logical computational capabilities of the computer are employed to assist digital filter design in Electrical and Electronic Engineering.

2 Methodology 2.1 Normalized Analog Low Pass Filter Design from First Principles wpass of a normalized analog low pass filter is 1 rad/sec. wstop , Apass , and Astop must be specified. There are equations for finding the filter order n for four of the five approximation functions except the Bessel approximation function [4–10]. The filter function is the ratio of two polynomials as given by Eqs. (3), (4), and (5). s is equal to jw, where w is the angular frequency. There are equations for finding the filter function for all the five approximation functions. The polynomials PolyNF and PolyDF are matrices with filter coefficients as elements. The Gain of the filter is given by Eq. (6). K is a constant. H(s) =

K(a1 sn + a2 sn−1 + a3 sn−2 + ....+an−1 s2 + an s + an+1 ) KPolyNF = PolyDF (b1 sn + b2 sn−1 + b3 sn−2 + ....+bn−1 s2 + bn s + bn+1 )   PolyNF = a1 , a2 , a3 , a4 , ..., an−2 , an−1 , an , an+1

(3) (4)

  PolyDF = b1 , b2 , b3 , b4 , ..., bn−2 , bn−1 , bn , bn+1

(5)

Gain = 20log10 |H(jw)|

(6)

2.2 Analog Filter Design Analog low pass filter, Analog high pass filter, Analog band pass filter, and Analog band stop filter are obtainable from the normalized analog low pass filter function. New polynomials PolyNNF and PolyNDF are formed for the proposed analog filter from the polynomials PolyNF and PolyDF of the normalized analog low pass filter function as shown in Eq. (7). H(s) =

KPolyNNF K(A1 sn + A2 sn−1 + A3 sn−2 + ....+An−1 s2 + An s + An+1 ) = PolyNDF (B1 sn + B2 sn−1 + B3 sn−2 + ....+Bn−1 s2 + Bn s + Bn+1 ) (7)

352

A. R. Zubair and A. J. Olawale

2.2.1 Analog Low Pass Filter Design To obtain an Analog low pass filter with bandpass edge frequency of wc from the normalized analog low pass filter with bandpass edge frequency of 1 rad/sec, Eq. (8) is substituted into Eq. (3). Steps 1, 2 and 3 are carried out. s (8) s= wc Step 1: To obtain PolyNNF, substitution of Eq. (8) in the polynomial PolyNF a1 sn + a2 sn−1 + a3 sn−2 + .... + an−1 s2 + an s + an+1 leads to n n−1 n−2 2 1 0 a1 ws n + a2 s n−1 + a3 s n−2 + ... + an−1 ws 2 + an ws 1 + an+1 ws 0 wc wc c c c c Step 2: Multiplication of all terms by wcn gives a1 sn + a2 wc1 sn−1 + a3 wc2 sn−2 + .... + an−1 wcn−2 s2 + an wcn−1 s1 + an+1 wcn s0 Compare with A1 sn + A2 sn−1 + A3 sn−2 + .... + An−1 s2 + An s1 + An+1 s0 ∴ Am = am wcm−1 Step 3: Repeat steps 1 and 2 for the polynomial PolyDF to obtain PolyNDF. Similarly, Bm = bm wcm−1 The observed patterns in steps 2 and 3 are converted to arithmetic/logical computational procedures or operations which are coded into a computer program. The filter order remains as n. 2.2.2 Analog High Pass Filter Design To obtain an Analog high pass filter with bandpass edge frequency of wc from the normalized analog low pass filter with bandpass edge frequency of 1 rad/sec, Eq. (9) is substituted into Eq. (3). Steps 1, 2 and 3 are carried out. wc (9) s= s Step 1: To obtain PolyNNF, substitution of Eq. (9) in the polynomial PolyNF. a1 sn + a2 sn−1 + a3 sn−2 + ....+an−1 s2 + an s + an+1 leads to wcn wcn−1 wcn−2 w2 w1 w0 + a2 n−1 + a3 n−2 + ... + an−1 2c + an 1c + an+1 0c n s s s s s s n Step 2: Multiplication of all terms by s gives a1

1

a1 wcn s0 + a2 wcn−1 s + a3 wcn−2 s2 + ....+an−1 wc2 s

n−2

+ an wc1 s

n−1

+ an+1 wc0 s

n

Flip from left to right n

an+1 wc0 s + an wc1 s

n−1

+an−1 wc2 s

n−2

1

+ .... + a3 wcn−2 s2 + a2 wcn−1 s + a1 wcn s0

Compare with A1 sn + A2 sn−1 + A3 sn−2 + ....+An−1 s2 + An s1 + An+1 s0 ∴ Am =an+2−m wcm−1 Step 3: Repeat steps 1 and 2 for the polynomial PolyDF to obtain PolyNDF. Similarly, Bm = bn+2−m wcm−1 The observed patterns in steps 2 and 3 are converted to arithmetic/logical computational procedures or operations which are coded into a computer program. The filter order remains as n.

Development of Arithmetic/Logical Computational Procedures

353

2.2.3 Analog Band Pass Filter Design To obtain an Analog band pass filter with bandpass edge frequencies of wc1 and wc2 from the normalized analog low pass filter with bandpass edge frequency of 1 rad/sec, Eq. (10) is substituted into Eq. (3). Steps 1, 2, 3, and 4 are carried out. s=

s2 + wc1 wc2 s2 + p s p = = + s(wc2 − wc1 ) sg g sg

(10)

where p is the product (wc2 wc1 ) and g is difference (wc2 − wc1 ). Step 1: To obtain PolyNNF, substitution of Eq. (10) in the polynomial PolyNF. a1 sn + a2 sn−1 + a3 sn−2 + ....+an−1 s2 + an s + an+1 leads to 

     s s s p n−1 p 1 p 0 + + + + ... + an + an+1 g sg g sg g sg   p m Step 2: The expansion of gs + sg in Table 1 is studied for obvious patterns. The matrices in Table 1 can be built logically and sequentially from m = 0, 1, 2 up to a desired value of m by simply starting with 1, adding two consecutive elements in the matrix for (m-1) to get new element in the matrix for m and ending with a 1. Negative powers of s are introduced in addition to the positive powers of s. The number of coefficients will therefore increase from (n + 1) to (2n + 1). The order of the filter is doubled. a1

s p + g sg



n

+ a2

  p m Table 1. Expansion of gs + sg .



p s g + sg

0

Expansion

Matrix

m

1

[1]

0



 p 1 s g + sg

sp0 p1 g + sg

[1, 1]

1



 p 2 s g + sg

s2 p0 p1 p2 +2 2 + 2 2 g2 g s g

[1, 2, 1]

2



 p 3 s g + sg

s3 p0 s1 p1 p2 p3 +3 3 +3 3 + 3 3 g3 g sg s g

[1, 3, 3, 1]

3



 p 4 s g + sg

s4 p0 s2 p1 p2 p3 p4 +4 4 +6 4 +4 2 4 + 4 4 g4 g g s g s g

[1, 4, 6, 4, 1]

4

A1 sn + A2 sn−1 + ... + An s1 + An+1 s0 + An+2 s−1 + ... + A2n s−(n−1) + A2n+1 s−n Each coefficient an−m+1 in PolyNF will contribute to the terms sm , sm−2 , sm−4 , … up to s−m in PolyNNF.

354

A. R. Zubair and A. J. Olawale

Step 3: Multiplication of all terms by sn gives A1 s2n + A2 s2n−1 + ... + An sn+1 + An+1 sn + An+2 sn−1 + .... + A2n s1 + A2n+1 s0 Step 4: Repeat steps 1, 2 and 3 for the polynomial PolyDF to obtain PolyNDF. The observed patterns in steps 2, 3, and 4 are converted to arithmetic/logical computational procedures or operations which are coded into a computer program. The filter order becomes 2n. 2.2.4 Analog Band Stop Filter Design To obtain an Analog band stop filter with bandstop edge frequencies of wc1 and wc2 from the normalized analog low pass filter with bandpass edge frequency of 1 rad/sec, Eq. (11) is substituted into Eq. (3). Steps 1, 2, 3, and 4 are carried out. s=

s(wc2 − wc1 ) = s2 + wc1 wc2



s2 + p sg

−1

 =

p s + g sg

−1 (11)

where p is the product (wc2 wc1 ) and g is difference (wc2 − wc1 ). Step 1: To obtain PolyNNF, substitution of Eq. (11) in the polynomial PolyNF. a1 sn + a2 sn−1 + a3 sn−2 + ....+an−1 s2 + an s + an+1 leads to  a1

s p + g sg

−n

 + a2

s p + g sg

−(n−1) 

Multiplication of all terms by  a1

s p + g sg



0 + a2

s p + g sg

s g

+

 + ... + an

 p n sg

s p + g sg

−1

 + an+1

s p + g sg

0

leads to 

1 + ... + an

s p + g sg



n−1 + an+1

s p + g sg

n

It is noted that the difference between band pass filter and band stop filter is that m decreases from n to 0 in the step 1 for band pass filter (Sect. 2.2.3) but m increases from   m

p 0 to n in the step 1 for band stop filter (Sect. 2.2.4). The coefficient in front of gs + sg is an−m+1 for band pass filter but am+1 for band stop filter. Steps 2 to 4 are the same for band pass filter in Sect. 2.2.3. The computer program for the band pass filter is copied and adjusted slightly to form the program for band stop filter. The filter order is also 2n.

2.3 Digital Filter Design with Bilinear Transformation An analog filter can be converted to a digital filter of the same “type” with the substitution of Eq. (12) in Eq. (7). New polynomials PolyZNF and PolyZDF are formed for the

Development of Arithmetic/Logical Computational Procedures

355

proposed digital filter from the polynomials PolyNNF and PolyNDF of the corresponding analog filter function as shown in Eq. (13). s= H(z) =

2 [z − 1] Ts [z + 1]

(12)

K(C1 zn + C2 zn−1 + C3 zn−2 + ....+Cn−1 z2 + Cn z + Cn+1 ) KPolyZNF = PolyZDF (D1 zn + D2 zn−1 + D3 zn−2 + ....+Dn−1 z2 + Dn z + Dn+1 ) (13)

where z = ejw . “type” refers to low pass, high pass, band pass and band stop. Ts is the sampling interval which is the reciprocal of the sampling frequency fs . The order n of the Digital filter is the same as the order n of the corresponding analog filter. Equation (12) is known as Bilinear Transformation. Analog frequency (wc ) and digital frequency (wdc ) are related by Eqs. (14) and (15). The frequency response of a digital filter is plotted over the range of 0 to π radian. Beyond this range, the frequency response repeats itself. Steps 1, 2, 3, and 4 are carried out. w  2 dc tan Ts 2   Ts wc = 2tan−1 2

wc = wdc

(14) (15)

Step 1: To obtain PolyZNF, substitution of Eq. (12) in the polynomial PolyNNF A1 sn + A2 sn−1 + A3 sn−2 + ....+An−1 s2 + An s + An+1 leads to  A1

2 [z − 1] Ts [z + 1]



n + A2

2 [z − 1] Ts [z + 1]



n−1 .. + An

2 [z − 1] Ts [z + 1]



1 + An+1

2 [z − 1] Ts [z + 1]

0

Multiplication of all terms by [z + 1]n leads to 

  n−1 2 n 2 n 0 A1 [z − 1] [z + 1] + A2 [z − 1]n−1 [z + 1]1 + . Ts Ts  1  0 2 2 1 n−1 .. + An + An+1 [z − 1] [z + 1] [z − 1]0 [z + 1]n Ts Ts  m1 Each term is like Am T2s [z − 1]m1 [z + 1]m2 where m1 = n + 1 − m, m2 = m − 1, and n = m1 + m2. m varies from 1 to n + 1, m1 varies from n to 0, and m2 varies from 0 to n. Step 2: The expansions of [z − 1]m1 and [z + 1]m2 in Table 2 are studied for obvious patterns. The matrices similar to that in Table 1 are involved in the expansions in Table 2 and can be handled in the same way. The product [z − 1]m1 [z + 1]m2 is obtainable by multiplying the expansions of [z − 1]m1 and [z + 1]m2 . Each of the coefficients Cm in

356

A. R. Zubair and A. J. Olawale

PolyZNF will receive contributions from the coefficients A1 , A2 , A3 , . . ., and An+1 in PolyNNF. C1 z n + C2 z n−1 + C3 z n−2 + .... + Cn−1 z 2 + Cn z + Cn+1 Step 3: Multiplication of all terms by z−n gives C1 z 0 + C2 z −1 + C3 z −2 + .... + Cn−1 z −(n−2) + Cn z −(n−1) + Cn+1 z −n

Table 2. Expansions of [z − 1]m1 and [z + 1]m2 . m1

Expansion of [z − 1]m1 (m1 + 1) terms

m2

Expansion of [z + 1]m2 (m2 + 1) terms

0

1

0

1

1

z−1

1

z+1

2

z2 − 2z + 1

2

z2 + 2z + 1

3

z3 −3z2 + 3z − 1

3

z3 +3z2 + 3z + 1

4

z4 −4z3 +6z2 − 4z + 1

4

z4 +4z3 +6z2 + 4z + 1

5

z5 −5z4 + 10z −10z2 + 5z − 1

5

z5 +5z4 + 10z +10z2 + 5z + 1

3

3

Step 4: Repeat steps 1, 2, and 3 for the polynomial PolyNDF to obtain PolyZDF. The observed patterns in steps 2, 3, and 4 are converted to arithmetic/logical computational procedures or operations which are coded into a computer program. The filter order of the digital filter is the same as the order of the corresponding analog filter. 2.4 Digital Filter Design Software The computer programs developed for the normalized analog low pass filter, the four types of analog filter and the digital filter are packaged as a software system for digital filter design. The software system developed in MATLAB working environment can plot the pole-zero diagram for the normalized analog low pass filter and frequency response for the various filters. User friendliness is incorporated in the software system which can be regarded as a filter coefficients’ calculator. The software and its user manual are available on the research gate pages of the authors.

3 Tests and Results The software system is subjected to rigorous tests. Sample results are presented in this section. The user is expected to supply some inputs based on the desired filter type and specifications.

Development of Arithmetic/Logical Computational Procedures

357

3.1 Digital Band Pass Filter 3.1.1 Inputs wpass = 1 rad/sec, wstop = 2 rad/sec, Apass = −3 dB, and Astop = − 100 dB for the normalized analog low pass filter. Elliptic approximation function was selected. Bandpass edge frequencies are wc1 = 1.8850 × 104 rad/sec (3 kHz) and wc2 = 6.2832x104 rad/sec (10 kHz) for the analog band pass filter. Sampling period Ts was selected as 1/50000 s (Sampling Frequency is 50 kHz which is greater than twice the maximum frequency of 10 kHz). 3.1.2 Outputs n was found to be 7 for the normalized analog low pass filter and 14 for the analog band pass filter and the digital band pass filter. Equivalent bandpass edge digital frequencies were found to be wdc1 = 0.3726 rad/sec and wdc2 = 1.122 rad/sec. Filter functions of Eqs. (16), (17), and (18) are obtained for the normalized analog low pass filter, analog band pass filter and digital band pass filter respectively. The frequency response curves for the normalized analog low pass filter, the analog band pass filter, and the digital band pass filter are presented in Fig. 2 (a), 2(b), and 3(a) respectively. 3.533 x 10−5 (s6 + 29.3437s4 + 222.78s2 + 491.5639) H(s) = 7 6 s + 0.5053s + 1.9405s5 + 0.7659s4 + 1.1025s3 + 0.2939s2 + 0.1640s + 0.174

(16)

11

H(s) =

H(z) =

3.533x10−5 (1.3814x10−28 s13 + 8.8232x10−18 s + 1.5522x10−7 s9 + 834.9448s7 +2.1773x1011 s5 + 1.7360x1019 s3 + 3.8126x1026 s) 14

13

12

3.1409x10−33 s + 6.9800x10−29 s + 3.7830x10−23 s + 7.0068x10−19 s11 +1.7530x10−13 s10 + 2.5901x10−9 s9 + 3.9778x10−4 s8 + 4.4189s7 + 4.7111x105 s6 +3.6331x109 s5 + 2.9122x1014 s4 + 1.3786x1018 s3 + 8.8153x1022 s2 + 1.9264x1026 s1 +1.0266x1031

(17)

3.533x10−5 (0.3430z0 − 1.4837z−1 + 2.6771z−2 − 3.2476z−3 + 3.7185z−4 − 3.1998z−5 +1.3503z−6 + 1.1912x10−15 z−7 − 1.3503z−8 + 3.1998z−9 − 3.7185z−10 + 3.2476z−11 −2.6771z−12 + 1.4837z−13 − 0.3430z−14 ) 0.108z0 − 1.08z−1 + 5.284z−2 − 16.739z−3 + 38.250z−4 − 66.580z−5 +90.935z−6 − 98.929z−7 + 86.128z−8 − 59.726z−9 + 32.496z−10 − 13.469z−11 +4.027z−12 − 0.780z−13 + 0.074z−14

(18)

3.2 Digital High Pass Filter 3.2.1 Inputs wpass = 1 rad/sec, wstop = 2 rad/sec, Apass = −3 dB, and Astop = −50 dB for the normalized low pass filter. Chebyshev approximation function was selected. Bandpass edge frequency wdc = 1. 2 rad/sec for the digital high pass filter.

358

A. R. Zubair and A. J. Olawale

Fig. 2. Frequency response curves for the normalized low pass filter and analog pass filter

3.2.2 Outputs n was found to be 5 for the normalized analog low pass filter, the analog high pass filter and the digital high pass filter. Equivalent bandpass edge analog frequency was found to be wc = 5.654x104 rad/sec (9 kHz). Sampling period Ts was selected as 1/50000 s (Sampling Frequency is 50 kHz which is greater than twice the maximum frequency of 9 kHz). Filter functions of Eqs. (19), (20), and (21) are obtained for the normalized analog low pass filter, the analog high pass filter and the digital high pass filter respectively. The frequency response of the digital high pass filter is presented in Fig. 3(b). H(s) = H(s) =

H(z) =

s5

+ 0.5745s4

+ 1.415s3

0.0626 + 0.5489s2 + 0.4080s1 + 0.0626

0.0626(5.6549x104 s5 ) 3.5427x103 s5 + 1.3046x109 s4 + 9.9264x1013 s3 + 1.447x1019 s2 + 3.322x1023 s +3.2699x1028

0.626(0.5655z0 − 2.8274z−1 + 5.6549z−2 − 5.6549z−3 + 2.8274z−4 − 0.5655z−5 ) 4.7576z0 − 2.5992z−1 + 5.207z−2 + 0.7633z−3 + 0.9356z−4 + 1.3994z−5

(19)

(20)

(21)

3.3 Discussion Manual calculation is easily handled for order 1 and 2. Manual calculation for order 3 and above is not convenient and is prone to human error. With this software, calculation for higher order is made easy and error free.

Development of Arithmetic/Logical Computational Procedures

359

Fig. 3. Frequency response curves for the digital band pass filter and digital high pass filter

4 Conclusion The development of a software system for the computation of filter coefficients has been presented and tested. Analog and Digital filter design has been made simple. Test results and frequency response plots confirm the accuracy of the arithmetic/logical computational procedures deduced from patterns in manual computations. The software is recommended for use in Electrical and Electronic Engineering practice.

References 1. Oppenheim, V.A.: Discrete Time Signal Processing. Prentice-Hall, Englewood Cliffs (1999) 2. Antoniou, A.: Digital Signal Processing: Signals, Systems and Filters. McGraw Hill, New York (2006) 3. Orfanidis, S.J.: Introduction to Signal Processing. Prentice Hall, Rutgers (2010) 4. Zubair, A.R., Olawale, A.J.: Active learning strategy: computer aided numerical class project on pole-zero plot and transfer function of five low pass filter approximation functions. Global J. Eng. Technol. Adv. 12(1), 038–063 (2022) 5. Thede, L.: Practical Analog and Digital Filter Design. Artech House, Inc., Ohio (2004) 6. Rice, J.R., Usow, K.H.: The Lawson algorithm and extensions. Math. Comput. 22, 118–127 (1968) 7. Smith, S.W.: The Scientist and Engineer’s guide to Digital Signal Processing. California Technical Publishing, California (1999) 8. Ingle, V.K., Proakis, J.G.: Digital Signal Processing Using MATLAB, 3rd edn. Global Engineering, Stamford (2010) 9. Rader, B.G.: Digital Processing of Signals. McGrawHill, New York (1969) 10. Abramowitz, M., Stegun, I.A. (eds.): Handbook of Mathematical Functions with Formulas. Dover Publications, New York (1965) 11. Ferdous, H., Jahan, S., Tabassum, F., Islam, M.: The performance analysis of digital filters and ANN in de-noising of speech and biomedical signal. Int. J. Image Graph. Signal Process. 15(1), 63–78 (2023) 12. Adamu, Z.M., Dada, E.G., Joseph, S.B.: Moth flame optimization algorithm for optimal FIR filter design. Int. J. Intell. Syst. Appl. 13(5), 24–34 (2021)

360

A. R. Zubair and A. J. Olawale

13. Kaur, R., Patterh, M.S., Dhillon, J.S.: Digital IIR filter design using real coded genetic algorithm. Int. J. Inf. Technol. Comput. Sci. 07(7), 27–35 (2013) 14. Jain, M., Gupta, M.: Design of fractional order recursive digital differintegrators using different approximation techniques. Int. J. Intell. Syst. Appl. 12(1), 33–42 (2020) 15. Hea, L., Zhonga, X., Qua, D., Fana, Z., Shib, K.: Design on open-loop bandpass filter with multiple harmonics suppression. Int. J. Wirel. Microw. Technol. 8(5), 27–36 (2018)

Multidimensional Ranking and Taxonomic Analysis of the Regional Socio-Economic Development in Ukraine Mariana Vdovyn and Larysa Zomchak(B) Ivan Franko National University of Lviv, Lviv 79000, Ukraine [email protected]

Abstract. The research paper deals with the problems of evaluation, analysis and modeling of socio-economic development of the regions of Ukraine. The main focus is on the main indicators that affect the maintenance of the physical, moral, economic and intellectual state, such as the income of the population and the level of wages. The population’s expenses and the unemployment rate were also considered. Taxonomic analysis and multidimensional ranking approach was used for modeling of socio-economic development. Reasons for using multidimensional statistical methods for assessing regional socio-economic development are substantiated in the paper. A brief description of the method of taxonomic analysis and multidimensional ranking was made. A comparative analysis of the obtained results was also carried out. The correlation coefficient was calculated for the evaluation of ranking results using different methods, such as the method of taxonomic analysis and multidimensional ranking. The place of each region among other regions of Ukraine was determined and recommendations were made for outsider regions. The value and importance of the research is to develop these recommendations for improving the socio-economic indicators for the outsider regions, according to the determination of the place of the region in the aggregate of the rest of the regions. It is the methods of taxonomic analysis and multidimensional ranking that make it possible to identify leaders and outsiders. Keywords: Socio-economic development · Socio-economic indicators · Taxonomic analysis · Multidimensional ranking

1 Introduction Socio-economic development reflects the dependence between the level of economic development and the solution of social problems of the country as a whole, its regions, business entities, etc. Government institutions need the development of anti-crisis programs and strategies of socio-economic development at the state and regional levels. So, the study of socio-economic development is permanently relevant. Clarification of regional socio-economic development in Ukraine is important for their further recovery and development stimulation. There are many indicators of socio-economic development, but it is important to focus on those that are closely related to the standard © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 361–370, 2023. https://doi.org/10.1007/978-3-031-36118-0_32

362

M. Vdovyn and L. Zomchak

of living of the population. Such indicators are: unemployment rate, average nominal wage, income and expenditure of the population. Obviously, there are regional socioeconomic development programs that offer mechanisms for improving socio-economic development indicators. However, it is important to understand what positions and places a specific region has in the rating. Multidimensional classification methods provide an opportunity to evaluate and analyze indicators of socio-economic development. Taxonomic analysis and multidimensional ranking are precisely those methods that help to identify the place of a specific element among the rest, in our case, the region.

2 Literature Review The study of socio-economic development of the country is one of the topical issues. In particular, socio-economic development in EU countries is considered in [1], the impact of socio-economic factors on population poverty in Pakistan is investigated in [2] and the correlation between socio-economic development and the level of tourism development in EU countries is offered in [3]. Rodríguez Martin and others in [4] offer an indicator of social and economic development for southern European countries based on the DP2 distance method. The socio-economic development in the EU countries, especially those that became members of the integration association after 2004, is described in detail in [5]. Unemployment as one of the indicators of socio-economic development is described in many researches. In particular, unemployment during COVID 19 D.L. Blustein, P.A. and others research in [6] and [7]. Chi-Wei Su, Ke Dai, Sana Ullah & Zubaria Andlib also investigate the impact of COVID 19 on the unemployment rate in European countries in [8]. Some scholars focus on the costs of health insurance as a form of social protection [9], others consider COVID 19 modeling [10] and regions development modeling [11]. A quantitative approach for analysis of the socio-economic state of the country is offered in many papers. Pareto, A. in [12] suggests the correlation and regression analysis of socio-economic indicators, Barrington-Leigh, C. and Escande, A. in [13] offer comparative analysis of the indicators of welfare. The evaluation and modeling of socio-economic processes are considered in [14]. Cluster analysis of data is suggested in [15] ang forecasting of economic processes in [16]. Logistic regression analysis of economic development was made in [17] and modeling the impact of the instability of the external environment on the enterprise in [18]. Methods of taxonomic analysis and multidimensional ranking were used in a lot of researches too. Mlodak, A. in [19] offers the indices of feature variability in a multidimensional space-time model and individual feature variability. Malina, A. in [20] considers the taxonomic analysis of economic structures in Poland. Bogliacino, F., & Pianta, M. in [21] investigate manufacturing and services using taxonomy. A statistical study of economic regional development was carried out in [22]. The authors use the taxonomic method for the analysis of economic processes. Hristov Metodi uses the methods of assessing the socio-economic indicators of EU countries in [23] and rating analysis in [24]. Different methods of data standardization are considered in [25–27]. So, the literature review showed that, despite the large number of scientific works on the problems of studying socio-economic development and various methods of modeling

Multidimensional Ranking and Taxonomic Analysis of the Regional

363

economic processes, there is still a need to study regional socio-economic development and find out the place of each Ukrainian region in the ranking.

3 Methodology. Methods of Multidimensional Statistics 3.1 Taxonomic Analysis in Modeling of the Regional Socio-Economic Development Various methods are used to estimate the state of socio-economic development, in particular, methods of descriptive statistics, game-theoretic modeling, optimization methods, heuristics, etc. The research considers the methods of multidimensional statistics, such as taxonomic analysis and multidimensional ranking. The paper deals with the problems of determining the place of each region of Ukraine among the rest, in order to understand the socio-economic development of which regions should be paid attention to by state administration bodies. That is why the methods chosen for the study are rating methods. Taxonomic analysis involves the formation of a matrix of observations, standardization of data, formation of a reference vector and determination of the distance between observations and this vector, as well as the calculation of a taxonomic indicator. The distance between each element and the reference vector is often defined as follows  m 1/2  2 (zij − z0 ) (1) Cj0 = i=1

where zij − standardized indicators and z0 − the reference value. This is the Euclidean distance. The taxonomic index is defined as the ratio of the Euclidean distance of the distance to the distance between the upper and lower points of the range. Taxonomic indicators are often ranked to clearly visualize the place of each element. 3.2 Multidimensional Ranking in Modeling of the Regional Socio-Economic Development Multidimensional ranking is the construct of an appropriate sequence of placement of units of a statistical population according to priorities, which makes it possible to determine the superiority of one unit over another at once based on several characteristics. The method involves the calculation of an integral indicator. The integral estimate (Gj) is geometrically interpreted as a point in multidimensional space whose coordinates indicate the scale or position of the j-th unit. The integral estimate Gj is often defined as the mean of the standardized values of features Zij. All indicators are standardized and analyzed, which indicators are stimulators, and which indicators are dissimulators. The results of the integrated assessment are also ranked. The following stages of calculating the integral indicator can be distinguished: 1. formation of a feature set (matrix of input information);

364

M. Vdovyn and L. Zomchak

2. implementation of data standardization, since, as a rule, indicators have different units of measurement; 3. choice of methods for calculating weighting factors, or understanding that all indicators are equally important; 4. determination of the procedure for aggregation of indicators; 5. ranking of the obtained results. The methods of multidimensional ranking and taxonomic analysis are precisely those methods that will help to identify outsiders, so that it is possible to propose specific steps for improving certain processes for them. In our case, to identify outsider regions for which it is necessary to develop a mechanism for improving socio-economic indicators.

4 Results. Model the Socio-Economic Development of Ukraine 4.1 General Overview of the Socio-Economic Development of Ukraine In 1991, with the restoration of independence, Ukraine was seen as a country with great prospects for potential high socio-economic development. However, the great integration into the unified national economic complex of the USSR did not give the opportunity to fully use this potential. Obstacles to high socio-economic development, in our opinion, were also a distorted system of political power influence, a high level of corruption and, as a consequence, shadow processes in the economy (black economy). In 2000–2020, there was an increase in indicators of socio-economic development, except, of course, the global crisis in 2008. The impact of COVID-19 cannot be ignored either. However, we believe that the full-scale invasion of the aggressor country into Ukraine in 2022 will have the greatest impact. Given the possibility of using official statistical data, we will conduct our analysis for 2021. Indicators of socio-economic development may change rapidly in 2022 due to the war, but our main task is to examine the situation in peacetime. It is clear that the need to develop a mechanism for improving indicators of socio-economic development is a priority for the leadership in the regions of Ukraine. It is also important to forecast the state of socio-economic development. Forecasting methods are carried out in [28, 29]. 4.2 Indicators of Socio-Economic Development in Different Regions of Ukraine The following main indicators of socio-economic development were chosen for the study: income of the population (million UAH), unemployment rate in the age group of 15–70 years (%), average nominal wage (thousand UAH), population expenses (million UAH). All indicators are considered in terms of 24 regions of Ukraine. Crimea is not taken into account, since it is annexed by Russia and the State Statistics Service does not publish official statistical data. Luhansk and Donetsk regions were also considered without taking into account the temporarily occupied territories. It is worth noting that the highest level of unemployment in 2021 was in Luhansk region (16%), and the lowest in Kharkiv region (6.7%). Despite the Covid-19 pandemic, Kharkiv region managed to create 9.5 thousand new jobs only for the first quarter of 2021 at the expense of all sources of funding.

Multidimensional Ranking and Taxonomic Analysis of the Regional

365

Regarding the average nominal salary in 2021: the highest in the Kyiv region (17,409 UAH), and the lowest - in the Chernihiv region (13,537 UAH). The income of the population was the highest in the Dnipropetrovsk region (434,791 million UAH), and the lowest in the Chernivtsi region (72,792 million UAH). Household expenses are the highest in Kyiv region (328,874 million UAH), and the lowest in Luhansk region (55,870 million UAH). Obviously, according to different indicators, different regions are leaders and outsiders. Therefore, it is advisable to use the methods of multivariate statistics. In our opinion, it is appropriate to consider ranking methods, since it is important for us to determine the place of each region. We chose two methods: taxonomic analysis and multidimensional ranking. 4.3 Taxonomic and Integral Indicators of Regional Socio-Economic Development In the taxonomic analysis, Euclidean distance was used and in the multidimensional ranking, standardization was performed on the basis of root mean square deviation taking into account that personal income and average nominal wages are stimulators, while spending and unemployment are dissimulators. The results are presented in Table 1 and visualized in Fig. 1. Table 1. Assessment of the socio-economic development in Ukrainian regions Region

Taxonomic indicators

Rank

Integral indicators

Rank

Vinnytsia

0,795

6

−0,001

11

Volyn

0,744

21

−0,426

21

Dnipropetrovsk

0,729

22

0,302

4

Donetsk

0,755

18

0,007

10

Zhytomyr

0,769

17

−0,232

18

Zakarpattia

0,788

8

−0,024

12

Zaporizhzhia

0,807

4

0,218

8

Ivano-Frankivsk

0,776

12

−0,078

13

Kyiv

0,803

5

0,522

1

Kirovohrad

0,709

24

−0,719

24

Luhansk

0,716

23

−0,478

22

Lviv

0,797

7

0,293

5

Mykolayiv

0,808

3

0,482

2

Odesa

0,786

9

0,272

6

Poltava

0,783

10

−0,072

14 (continued)

366

M. Vdovyn and L. Zomchak Table 1. (continued)

Region

Taxonomic indicators

Rank

Integral indicators

Rank

Rivne

0,825

1

0,436

3

Sumy

0,781

11

−0,092

15

Ternopil

0,751

19

−0,348

19

Kharkiv

0,769

16

0,266

7

Kherson

0,749

20

−0,393

20

Khmelnytsk

0,810

2

0,175

9

Cherkassy

0,775

13

−0,158

17

Chernivtsi

0,770

14

−0,115

16

Chernihiv

0,725

15

−0,588

23

Source: authors’ construct based on the data of the State Statistics Service[30, 31]

The table shows that Rivne and Mykolaiv regions are the clear leaders in both methods, and Luhansk and Kirovohrad regions are outsiders. The current standard of living of the population of the Kirovohrad region creates significant limitations for regional development due to low wages. In general, in the Kirovohrad region there are conditions for regional development, in particular large areas of land resources, significant potential in engineering and food industry, a developed agro-industrial complex. There are also deposits of unique minerals in the region, sufficient human resources. There are opportunities for expansion in the region’s transit potential, the growth of tourism. All these characteristics, of course, affect regional development, but they are not decisive. Thus, it decreases labor productivity, the outflow of qualified engineers and technicians increases personnel, employees of labor professions outside the region and Ukraine. The regional leadership should take into account all factors and make decisions on improving social and economic development. Certain recommendations for decision-making are offered in [32]. As for the Luhansk region, the main reason for the low socio-economic development is the armed aggression of Russia in the east of Ukraine and the temporary occupation of a part territories of Luhansk region. In 2022, the situation became worse, as the majority of the territory of Luhansk region remains occupied as a result of a full-scale invasion. Figure 1. Clearly shows that according to the results of the taxonomic analysis, the difference in the distribution of different regions is insignificant, but according to the integrated assessment, the picture is slightly different. It is also noticeable (Table 1) that certain regions are leaders according to the first method, and outsiders according to the other. First of all, this concerns the Dnipropetrovsk region. It is worth illustrating the results also by the ranks according to both methods (Fig. 2). Figure 2 shows that only a few regions received different rating results. The correlation coefficient between the results of multidimensional ranking and taxonomic analysis CORREL = 0.683, which indicates a noticeable correlation according to the Chaddock scale.

Multidimensional Ranking and Taxonomic Analysis of the Regional

367

Fig. 1. Taxonomic and integral indicators of socio-economic development in the regions of Ukraine, Source: authors’ construct based on the data of the State Statistics Service [30, 31]

Fig. 2. Chart Comparison of rating results by taxonomic analysis and integral assessment, Source: authors’ construct based on the data of the State Statistics Service [30, 31]

5 Discussion So, by examining the place of each region, we were able to understand who is a clear outsider and what conclusions should be made by the healers of these regions regarding the future improvement of socio-economic development. Taxonomic analysis and multidimensional ranking show that the outsider regions are the Luhansk and Kirovohrad

368

M. Vdovyn and L. Zomchak

regions. In our opinion, the reason for the fall of the Luhansk region is the russian occupation of its part in 2014. Regarding to the Kirovohrad region as an outsider, everything is also clear due to the low level of wages. Obviously, it is necessary to help reduce the level of unemployment, including at the expense of small businesses, to raise the level of wages. Salaries largely depend on the type of economic activity as a result of uneven technical and technological development. It is worth developing innovative industries, because there the level of wages is higher. Of course, it is not only the outsider regions that need improvement. It is important to ensure favorable conditions for all regions of Ukraine without exception.

6 Summary and Conclusion The indicators of regional socio-economic development of the regions of Ukraine are quite different. It is clear that there are leaders and there are outsiders. Actually, it is important for outsiders to understand the reasons for their positions in the rating. The methods of economic analysis and multidimensional ranking make it possible to identify outsiders, and in dynamics to assess the change in positions. In general, the following main proposals for improving the socio-economic development of the region in terms of well-being and the standard of living of the population can be identified: 1. Contribute to increasing labor productivity and reducing the gender pay gap. 2. To ensure an increase in payments of social funds for vulnerable sections of the population. 3. The fight against illegal business, corruption and assistance in the detection and legal regulation of non-labor income of the population. 4. Allocation of funds from the regions for retraining of persons who have lost their jobs. 5. Cooperation with employers and financing of programs involving young professionals, graduates of higher educational institutions. However, all these conclusions based on the indicators of 2021 would be relevant if there was no war. We believe that first, after the liberation of the occupied territories, it will be necessary to rebuild the country. We believe that it is necessary to monitor the place of each region of Ukraine among the rest of the regions every year. This is necessary for the further development of mechanisms for improving indicators of socio-economic development. The methods of multidimensional statistics will help in this case. It is important, in the future, to take into account various indicators of socio-economic development for rating.

References 1. Fura, B., Wang, Q.: The level of socioeconomic development of EU countries and state of ISO 14001 certification. Qual. Quant. 51(1), 103–119 (2017). https://doi.org/10.1007/s11135015-0297-7

Multidimensional Ranking and Taxonomic Analysis of the Regional

369

2. Ayyoub, M., Gillani, D.Q.: Role of socioeconomic factors in poverty alleviation: an assessment of urban informal sector. Kashmir Econ. Rev. 26(1), 28–40 (2017) 3. Stec, M., Grzebyk, M.: Socio-economic development and the level of tourism function development in European Union countries – a comparative approach. European Rev. 30(2), 172–193 (2022). https://doi.org/10.1017/S106279872000099X 4. Rodríguez Martin, J.A., Holgado Molina, M.D., Salinas Fernández, J.A.: An index of social and economic development in the community’s objective-1 regions of countries in southern Europe. Eur. Plan. Stud. 20(6), 1059–1074 (2012) 5. Stec, M., Filip, P., Grzebyk, M., Pier´scieniak, A.: Socio-economic development in the EU Member States – concept and classification. Eng. Econ. 25(5), 504–512 (2014) 6. Blustein, D.L., Duffy, R., Ferreira, J.A., Cohen-Scali, V., Cinamon, R.G., Allan, B.A.: Unemployment in the time of COVID-19: a research agenda. J. Vocat. Behav. 119, 103436 (2020) 7. Blustein, D.L., Guarino, P.A.: Work and unemployment in the time of COVID-19: the existential experience of loss and fear. J. Humanist Psychol. 60(5), 702–709 (2020) 8. Chi-Wei, S., Dai, K., Ullah, S., Andlib, Z.: COVID-19 pandemic and unemployment dynamics in European economies. Econ. Res.-Ekonomska Istraživanja 35(1), 1752–1764 (2022) 9. Tkachenko, R., et al.: Piecewise-linear approach for medical insurance costs prediction using SGTM neural-like structure. CEUR Workshop Proc. 2255, 170–179 (2018) 10. Abriham, A., Dejene, D., Abera, T., Elias, A.: Mathematical modeling for COVID-19 transmission dynamics and the impact of prevention strategies: a case of Ethiopia. Int. J. Math. Sci. Comput. (IJMSC) 7(4), 43–59 (2021). https://doi.org/10.5815/ijmsc.2021.04.05 11. Koroliuk, Y., Hryhorenko, V.: ANN model of border regions development: approach of closed systems. Int. J. Intell. Syst. Appl. (IJISA) 11(9), 1–8 (2019). https://doi.org/10.5815/ijisa. 2019.09.01 12. Pareto, A.: A robust method for regression and correlation analysis of socio-economic indicators. Qual. Quant.2022). https://doi.org/10.1007/s11135-022-01599-z 13. Barrington-Leigh, C., Escande, A.: Measuring progress and well-being: a comparative review of indicators. Soc. Indic. Res. 135, 893–925 (2018). https://doi.org/10.1007/s11205-0161505-0 14. Podgorna, I., Babenko, V., Honcharenko, N., Sáez-Fernández, F.J., Fernández, J.A.S., Yakubovskiy, S.: Modelling and analysis of socio-economic development of the European Union countries through DP2 method. Wseas Trans. Bus. Econ. 17, 454–466 (2020). https:// doi.org/10.37394/23207.2020.17.44 15. Lytvynenko, V., Kryvoruchko, O., Lurie, I., Savina, N., Naumov, O., Voronenko, M.: Comparative studies of self-organizing algorithms for forecasting economic parameters. Int. J. Modern Educ. Comput. Sci. (IJMECS) 12(6), 1–15 (2020). https://doi.org/10.5815/ijmecs. 2020.06.01 16. Chittineni, S., Bhogapathi, R.B.: Determining contribution of features in clustering multidimensional data using neural network. Int. J. Inf. Technol. Comput. Sci. (IJITCS) 4(10), 29–36 (2012). https://doi.org/10.5815/ijitcs.2012.10.03 17. Zomchak, L., Starchevska, I.: Macroeconomic determinants of economic development and growth in Ukraine: logistic regression analysis. In: Hu, Z., Wang, Y., He, M., eds Advances in Intelligent Systems, Computer Science and Digital Economics IV. CSDEIS 2022. Lecture Notes on Data Engineering and Communications Technologies. Springer Cham, p. 158 (2023). https://doi.org/10.1007/978-3-031-24475-9_31 18. Poplavska, Z., Komarynets, S.: Modelling the External Economic Environment Instability Impact on the Organizational Flexibility of the Enterprise. In: Kryvinska, N., Greguš, M. (eds.) Developments in Information & Knowledge Management for Business Applications. SSDC, vol. 330, pp. 1–20. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-621 51-3_1

370

M. Vdovyn and L. Zomchak

19. Mlodak, A.: Evaluation of the variability of statistical features in a taxonomic model. Wiadomo´sci Statystyczne Polish Stat. 9, 5–18 (2005). (in Polish) 20. Malina, A.: Multi-criteria taxonomy in comparative analysis of economic structures in Poland [M], [w:] A. Zelia´s (red.) Przestrzenno-czasowe modelowanie i prognozowanie zjawisk gospodarczych, Wydawnictwo Akademii Ekonomicznej w Krakowie, pp. 305–312 (2002) (in Polish) 21. Bogliacino, F., Pianta, M.: The Pavitt taxonomy, revisited. Patterns of innovation in manufacturing and services. Documento Escuela de Economía, p. 57 (2015) 22. Andrusiv, U., et al.: Analysis of economic development of Ukraine regions based on taxonomy method. Manag. Sci. Lett. 10(3), 515–522 (2020) 23. Metodi, H.: Opportunities for assessing the socio-economic Condition of the countries of the European Union, March 30 (2022). https://doi.org/10.2139/ssrn.4070368.SSRN: https://ssrn. com/abstract=4070368 24. Metodi, H.: Rating analysis of cohesion policy of the European Union and the member countries 29 June 2020. SSRN https://ssrn.com/abstract=3637950 25. Zelias, A.: Some notes on the selection of normalization of diagnostic variables. Stat. Transit. 5(5), 787–802 (2002) 26. Singh, D., Singh, D.: Feature wise normalization: an effective way of normalizing data. Pattern Recogn. 122, 108307 (2022). https://doi.org/10.1016/j.patcog.2021.108307. ISSN 0031–3203 27. Jain, S., Shukla, S., Wadhvani, R.: Dynamic selection of normalization techniques using data complexity measures. Expert Syst. Appl. 106, 252-262 (2018). https://doi.org/10.1016/ j.eswa.2018.04.008. ISSN 0957-4174 28. Izonin, I., et al.: Stacking-based GRNN-SGTM ensemble model for prediction tasks. In: Proceedings of the DASA, pp. 60–66 (2020) 29. Hassan, M.M., Mirza, T.: Using time series forecasting for analysis of GDP growth in India. Int. J. Educ. Manag. Eng. 11(3), 40–49 (2021). https://doi.org/10.5815/ijeme.2021.03.05 30. Official site: State Statistics Service of Ukraine. http://www.ukrstat.gov.ua. Accessed 20 Jan 2023 31. Official site: State Employment Center. https://www.dcz.gov.ua. Accessed 20 Jan 2023 32. Sahai, A., Sankat, C.K., Khan, K.: Decision-making using efficient confidence-intervals with Meta-analysis of spatial panel data for socioeconomic development project-managers. Int. J. Intell. Syst. Appl. (IJISA) 4(9), 92–103 (2012). https://doi.org/10.5815/ijisa.2012.09.12

Advanced Method for Compressing Digital Images as a Part of Video Stream to Pre-processing of UAV Data Before Encryption Zhengbing Hu1 , Myroslav Ryabyy2 , Pylyp Prystavka2 , Sergiy Gnatyuk2(B) , Karina Janisz3 , and Dmytro Proskurin2 1 Hubei University of Technology, Wuhan 430068, Hongshan, China 2 National Aviation University, 1 Liubomyra Huzara Avenue, Kyiv 03058, Ukraine

{m.riabyi,s.gnatyuk}@nau.edu.ua 3 University of Applied Sciences in Nowy S˛acz, 1 Staszica Street, 33-300 Nowy S˛acz, Poland

Abstract. Reducing the volume of video and photo information while maintaining high-quality visual perception remains an urgent task that specialists in specific subject areas are constantly working on. However, without questioning the success of the developers of compression methods (both lossy and lossless), reducing the size during streaming of video information carries a deterioration in the quality of information that is quite noticeable to the human eye or distorts the information in such a way that its further processing becomes impossible. In the work, a basic method to increase the level (percentage) of compression of a digital image (as a part of the video stream) with losses was proposed, and it is based on both transformations that ensure the invariance of that digital image and background extraction methods. The main value of result is that a high level of compression was demonstrated. Future study will be related to the universal dataset of crypto-algorithms development as well as AI/ML subsystem formation. Keywords: Data compression · digital images · video stream · image processing · digital image compression · UAV

1 Introduction The comprehensive development of information technologies, aimed at improving life, also brings some problems that need to be solved. One of the global problems is related to data storage, including video information, the volume of which is constantly and relentlessly growing. The other side of this problem is the communication channels required to transmit this information. In the absence of broadband access to the network, compression of source information without loss or with loss, that is not visible to the human eye, remains an urgent task. Reducing the volume of video and photo information, while maintaining high-quality visual perception, is a problem on which specialists from specific subject areas are © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 371–381, 2023. https://doi.org/10.1007/978-3-031-36118-0_33

372

Z. Hu et al.

constantly working. However, without questioning the successes of the developers of compression methods (both with loss and without), it is worth noting that usually, today, reducing the size of streaming video information brings a deterioration in the quality of information that is quite noticeable to the human eye or distort information so that its further processing becomes impossible.

2 Literature Review Up-to-date image processing methods [1–3] analysis allowed defining compression approaches and technologies [4–6], including AI/ML-based techniques use to improve efficiency of the image processing [7–10]. Also there are some cases of image processing in the various spheres of UAV implementation from civil needs to special military tasks [11–15]. All mentioned approaches have some advantages as well as disadvantages that can be object for research and improvement.

3 Problem Statement Mentioned approaches and existing solutions have many disadvantages but can be realized in practical cases after improvements. The main objective of the work is to propose an improved method of digital image compression for data pre-processing before encryption in the absence of broadband network access.

4 Methodology Compression methods are divided into two groups, lossy compression and lossless compression. Lossless compression methods give a lower compression ratio, but retain the exact value of the original pixel. Lossy compression methods provide high compression ratios, but do not allow reproduction of the original data with accuracy to the intensity of the color components of a pixel. However, when viewing photos and video information, the human eye does not perceive all shades of colors, therefore, some details can be ignored without significantly distorting the image. Therefore, when compressing photo and video data with losses, a part of the information that is not visible to the observer, or that which does not have a significant impact on the perception of the information, will be discarded. Consider a digital image raster (a video data frame can act as a digital image (DI)) (Fig. 1 a) 729 * 601 pixels with a color depth of 24 bits/pixel, which is a two-dimensional array of 729 elements in width and 601 elements in length, and each element carries 24 bits of information about the color of a given image point. The size of this image is 1.25 MB, the corresponding image size in PNG format is 585 KB and in JPEG format is 286 KB (in this work, compression using known compression methods will be used to retain the best quality). The image shown in the figure (Fig. 1b) is also 729 * 601 pixels with a color depth of 24 bits/pixel, its size in BMP format is also 1.25 MB, but in PNG format it is reduced to 2.59 Kb and in JPEG format - to 7.44 Kb. This difference is explained by the different detailing of the data of

Advanced Method for Compressing Digital Images as a Part of Video Stream

373

Fig. 1. Color DI

the DI, and more precisely, in the second case, a straight line will be presented on the graph. After considering the three-dimensional graph (Fig. 2) of a part of the image (Fig. 1a) and the two-dimensional graphs 1st and 25th lines of the DI for one color component (Fig. 3), it can be seen that the image function is oscillating, which affects the compression level of the DI [18–20], since smoother functions obviously compress better.

Fig. 2. A three-dimensional plot of a part of the image (Fig. 1)

To reduce the oscillation of the function, we can apply a “smoothing” (low frequency) filter to the image (Fig. 1a) [17, 18], which will make the function smoother, which in turn will reduce oscillation and possibly increase the compression level. Applying the filter to the image (Fig. 1a) immediately shows that the distortions are not visually identifiable (Fig. 4). After looking at the graphs shown in Figs. 5 and 6, we can say that the function has become smoother. New image size in PNG format is 443 KB and in JPEG format - 258 KB. More precisely, for the PNG format the image size decreased by 11.36%, and for the JPEG format by almost 4% [21, 22].

374

Z. Hu et al.

Fig. 3. Two-dimensional graph of the 1st and 25th lines of the image by one color component: a) 1st line; b) 25th line)

Fig. 4. Filtered DI (Fig. 1a)

However, if a video frame is a raster image in RGB format, and a stream of video frames is a sequence of video frames received from a digital video recording device (digital video camera), then the movement in the frame (frame change) is the translational movement of a certain group of pixels, which is observed in the adjacent frames of the video stream. A necessary condition for finding motion in a frame is the possibility of locating the same group of pixels in adjacent frames.

Advanced Method for Compressing Digital Images as a Part of Video Stream

375

Fig. 5. Two-dimensional plot of 1st and 25th lines of the filtered image using one color component: a) 1st line; b) 25th line)

A moving object is a group of pixels that keeps its shape on adjacent frames, and the rectangle surrounding it, which keeps its dimensions, and with its center point slightly displaced between those frames (within a few pixels).

Fig. 6. A 3D plot of a portion of the filtered image (Fig. 5)

376

Z. Hu et al.

A group of pixels with a rectangle surrounding it is shown in Fig. 7. The center of the group is located at the point Ci(xi, yi) and Ci + 1(xi + 1, yi + 1) on the i-th and i + 1-th frames, respectively.

Fig. 7. A group of pixels on adjacent frames

Taking into account all the above, these result can be improved by combining methods of invariant transformations and image background separation, namely, a general method of increasing the level (percentage) of lossy DI compression based on transformations that ensure DI invariance and methods for background identification. This method is based on the following actions: • • • • •

preliminary analysis of the first frame of the flow; removing the background parts of the image by the selected method; image filtering using a low-pass filter of the required power; DI compression by any method or algorithm with or without losses; restoration of the image using a CF filter, which is the pseudo-inverse of the LF filter, that was applied to smooth the DI.

Fig. 8. Color DI

Figures 8, 9 present the frames of the video stream from the video surveillance and the UAV cameras with the results of their processing. After putting a filter on the image

Advanced Method for Compressing Digital Images as a Part of Video Stream

377

where the moving part was separated from the background (Fig. 8), it is immediately clear that the distortions are not visually identifiable (Fig. 9, 10). The size of the original and processed images is given in the table (Table 1) for images a and b under numbers 1 and 2, respectively.

Fig. 9. Color DI

Fig. 10. Color DI

5 Results and Discussion Based on the data in the table (Table 1), we can talk about the effectiveness of the proposed method, which poses the following tasks: • • • •

research and analysis of existing and promising background extraction methods; research of possible options for improving the use of invariant transformations; verification of the proposed method in “real-time” operation mode; statistical substantiation of the results of the study of the PSNR indicator and the relative error.

After the compression (pre-processing stage) UAV data will be encrypted in AI-based security system (Fig. 11). To achieve this target, it is necessary to create the universal set (dataset) of crypto algorithms that could solve various types of problems in different conditions. The encryption algorithm will be selected by AI component (Fig. 12).

378

Z. Hu et al. Table 1. Picture’s size

№

Output size, MB

The size of the processed image, MB

Compression percentage, %

1

2,64

0,34

87,12

2

2,63

0,24

90,87

Fig. 11. Concept of the AI/ML-based encryption

There are many cryptographic algorithms that can be used in UAVs, it will be necessary to conduct analysis based on the selected criteria [23–26], for example: size of the encrypted data block; encryption key sizes; resistance against crypto-analytical attacks known at this time; structure of the algorithm; no weak and equivalent keys; data encryption speed; requirements for operational and non-volatile memory; restrictions on the use of the algorithm.

Fig. 12. Concept of the AI/ML-based encryption

Also other criterion and requirements should be added for more detail analysis based on USA, EU and domestic crypto contests.

6 Conclusions and Future Work In the work, a new method is proposed to increase the level (percentage) of compression of the DI (as a part of the video stream) with losses. It is based on transformations that ensure the invariance of the DI and methods for extracting the background. The main value of result is that a high level of compression was demonstrated and it requires further refinement and experimental research.

Advanced Method for Compressing Digital Images as a Part of Video Stream

379

It is promising to focus further research in the direction of choosing or refining an existing background identification method or creating an own one, which can be used on different types of devices for receiving a video stream. Also the universal dataset of crypto-algorithms will be developed as well as AI/ML subsystem [27] will be proposed in the next research studies. Acknowledgement. This work is carried out within the framework of research project № 0122U002361 “Intellectualized system of secure packet data transmission based on intelligence and search UAV” (2022–2024), funded by the Ministry of Education and Science of Ukraine.

References 1. Pang, S., et al.: SpineParseNet: spine parsing for volumetric MR image by a two-stage segmentation framework with semantic image representation. IEEE Trans. Med. Imaging 40(1), 262–273 (2021). https://doi.org/10.1109/TMI.2020.3025087 2. Perdios, D., Vonlanthen, M., Martinez, F., Arditi, M., Thiran, J.-P.: CNN-based image reconstruction method for ultrafast ultrasound imaging. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 69(4), 1154–1168 (2022). https://doi.org/10.1109/TUFFC.2021.3131383 3. Küstner, T., et al.: LAPNet: non-rigid registration derived in k-space for magnetic resonance imaging. IEEE Trans. Med. Imaging 40(12), 3686–3697 (2021). https://doi.org/10.1109/TMI. 2021.3096131 4. Tellez, D., Litjens, G., van der Laak, J., Ciompi, F.: Neural image compression for gigapixel histopathology image analysis. IEEE Trans. Pattern Anal. Mach. Intell. 43(2), 567–578 (2021). https://doi.org/10.1109/TPAMI.2019.2936841 5. Xu, Y., Zhang, J.: Invertible resampling-based layered image compression. In: 2021 Data Compression Conference (DCC), Snowbird, UT, USA, pp. 380–380 (2021). https://doi.org/ 10.1109/DCC50243.2021.00064 6. Mali, A., Ororbia, A.G., Kifer, D., Giles, C.L.: Neural JPEG: end-to-end image compression leveraging a standard JPEG encoder-decoder. In: 2022 Data Compression Conference (DCC), Snowbird, UT, USA, pp. 471–471 (2022). https://doi.org/10.1109/DCC52660.2022.00082 7. Baba Fakruddin Ali, B.H.,Prakash, R.: Overview on machine learning in image compression techniques. In: 2021 Innovations in Power and Advanced Computing Technologies (i-PACT), Kuala Lumpur, Malaysia, pp. 1–8 (2021). https://doi.org/10.1109/i-PACT52855.2021.969 6987 8. Pilikos, G., Horchens, L., Batenburg, K.J., van Leeuwen, T., Lucka, F.: Deep data compression for approximate ultrasonic image formation. In: 2020 IEEE International Ultrasonics Symposium (IUS), Las Vegas, NV, USA, pp. 1–4 (2020). https://doi.org/10.1109/IUS46767. 2020.9251753 9. Khan Gul, M.S., Suleman, H., Bätz, M., Keinert, J.: RNNSC: recurrent neural network-based stereo compression using image and state warping. In: 2022 Data Compression Conference (DCC), Snowbird, UT, USA, pp. 455–455 (2022). https://doi.org/10.1109/DCC52660.2022. 00066 10. Patwa, N., Ahuja, N., Somayazulu, S., Tickoo, O., Varadarajan, S., Koolagudi, S.: Semanticpreserving image compression. In: 2020 IEEE International Conference on Image Processing (ICIP), Abu Dhabi, United Arab Emirates, pp. 1281–1285 (2020). https://doi.org/10.1109/ ICIP40778.2020.9191247 11. Zhou, J., Xi, W., Li, D., Huang, H.: Image enhancement and image matching of UAV based on histogram equalization. In: 2021 28th International Conference on Geoinformatics, Nanchang, China, pp. 1–5 (2021). https://doi.org/10.1109/IEEECONF54055.2021.9687650

380

Z. Hu et al.

12. Liu, B., Zhang, J., Li, Z.: An improved APAP algorithm via line segment correction for UAV multispectral image stitching. In: IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium, Kuala Lumpur, Malaysia, pp. 6057–6060 (2022). https://doi. org/10.1109/IGARSS46834.2022.9884605 13. Gademer, A., et al.: Faucon noir UAV project development of a set of tools for managing, visualizing and mosaicing centimetric UAV images. In: 2009 IEEE International Geoscience and Remote Sensing Symposium, Cape Town, South Africa, pp. III-228–III-231 (2009). https://doi.org/10.1109/IGARSS.2009.5417833 14. Yang, X.H., Xin, S.X., Jie, Z.Q., Jing, X., Dan, Z.D.: The UAV image mosaic method based on phase correlation. In: 11th IEEE International Conference on Control & Automation (ICCA), Taichung, Taiwan, pp. 1387–1392 (2014). https://doi.org/10.1109/ICCA.2014.6871126 15. Zhang, R., Xu, F., Yu, H., Yang, W., Li, H.-C.: Edge-driven object matching for UAV images and satellite SAR images. In: IGARSS 2020 - 2020 IEEE International Geoscience and Remote Sensing Symposium, Waikoloa, HI, USA, pp. 1663–1666 (2020). https://doi.org/10. 1109/IGARSS39084.2020.9324021 16. Naveen Kumar, R., Jagadale, B.N., Bhat, J.S.: An improved image compression algorithm using wavelet and fractional cosine transforms. Int. J. Image Graph. Signal Process. (IJIGSP) 10(11), 19–27 (2018). https://doi.org/10.5815/ijigsp.2018.11.03 17. Prystavka, P., Cholyshkina, O.: Pyramid image and resize based on spline model. Int. J. Image Graph. Signal Process. (IJIGSP) 14(1), 1–14 (2022). https://doi.org/10.5815/ijigsp. 2022.01.01 18. Prystavka, P.O., Matsuga, O.M.: Data Analysis. Education. Manual, Publishing House of Dnipro (2008). 92 p. 19. Kumar, M., Vaish, A.: Lossy compression of color images using lifting scheme and prediction errors. Int. J. Mod. Educ. Comput. Sci. (IJMECS) 8(4), 1–8 (2016). https://doi.org/10.5815/ ijmecs.2016.04.01 20. Sharath Kumar, P.N., Devanand, P., Alexander, G., Saritha, Sujathan, K., Deepak, R.U.: Automated matching of pixel of interest between two digital images from two different microscope imaging devices. In: 2020 IEEE Recent Advances in Intelligent Computational Systems (RAICS), Thiruvananthapuram, India, pp. 96–100 (2020). https://doi.org/10.1109/RAICS5 1191.2020.9332522 21. Wang, Z., Liu, Z.: Image quality improvement algorithm in the process of film shooting under digital imaging technology. In: 2022 International Conference on Applied Artificial Intelligence and Computing (ICAAIC), Salem, India, pp. 957–960 (2022). https://doi.org/10. 1109/ICAAIC53929.2022.9792797 22. Tang, M.: Image segmentation technology and its application in digital image processing. In: 2020 International Conference on Advance in Ambient Computing and Intelligence (ICAACI), Ottawa, ON, Canada, pp. 158–160 (2020) 23. Gnatyuk, S., Akhmetov, B., Kozlovskyi, V., Kinzeryavyy, V., Aleksander, M., Prysiazhnyi, D.: New secure block cipher for critical applications: design, implementation, speed and security analysis. In: Hu, Z., Petoukhov, S., He, M. (eds.) AIMEE 2019. AISC, vol. 1126, pp. 93–104. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-39162-1_9 24. Kalimoldayev, M., Tynymbayev, S., Gnatyuk, S., Ibraimov, M., Magzom, M.: The device for multiplying polynomials modulo an irreducible polynomial. News Natl. Acad. Sci. Repub. Kazakhstan Ser. Geol. Techn. Sci. 2(434), 199–205 (2019) 25. Hu, Z., Gnatyuk, S., Okhrimenko, T., Tynymbayev, S., Iavich, M.: High-speed and secure PRNG for cryptographic applications. Int. J. Comput. Netw. Inf. Secur. 12(3), 1–10 (2020) 26. Saxena, A.K., Sinha, S., Shukla, P.: Design and development of image security technique by using cryptography and steganography: a combine approach. Int. J. Image Graph. Signal Process. (IJIGSP) 10(4), 13–21 (2018). https://doi.org/10.5815/ijigsp.2018.04.02

Advanced Method for Compressing Digital Images as a Part of Video Stream

381

27. Lanubile, F., Calefato, F., Quaranta, L., Amoruso, M., Fumarola, F., Filannino, M.: Towards productizing AI/ML models: an industry perspective from data scientists. In: 2021 IEEE/ACM 1st Workshop on AI Engineering - Software Engineering for AI (WAIN), Madrid, Spain, pp. 129–132 (2021). https://doi.org/10.1109/WAIN52551.2021.00027

Ternary Description Language and Categorical Systems Theory G. K. Tolokonnikov(B) VIM RAS, Moscow, Russia [email protected]

Abstract. Categorical systems theory is an integral part of algebraic biology that predicts the properties of organisms based on the genome, including intellectual properties, which leads to artificial intelligence models. A wide range of system approaches is formalized in the language of ternary description, so it is very important to find relationships between the systems of the ternary description language (TDL) and category systems theory. This paper proposes a version of the TDL, in which it is possible to present the definition of the system in the TDL in the required form, in which it is possible to establish the relationship of this type of systems with the system block of algebraic biology. The irreducibility of categorical systems to systems based on TDL is shown. The proposed new version of the TDL is also useful for solving existing problems in the TDL formalism. Keywords: Categorical systems · Algebraic biology · Logic · Ternary description language · Artificial intelligence

1 Introduction There are dozens of definitions of the concept of a system, the most profound of which is the functional system of P.K. Anokhin. These approaches largely fall under the language of ternary description (TDL) developed for systems [5–9], where the definition of functional systems was also considered. However, this consideration turns out to be incomplete, since the role of the system-forming factor, which also underlies categorical systems, remains unclear. This paper proposes a version of the TDL, in which it is possible to present the definition of the system in the TDL in the required form, in which it is possible to establish the relationship of this type of systems with the system block of algebraic biology. The irreducibility of categorical systems to systems based on TDL is shown.

2 Comparison of the Concept of a System in TDL and Categorical Systems Theory The development of various logics is usually based on the analysis of natural language. The modern predicate calculus was built in the works of G. Frege, the well-known contradiction discovered in them by B. Russell led to the creation of type theory, axiomatics © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 382–390, 2023. https://doi.org/10.1007/978-3-031-36118-0_34

Ternary Description Language and Categorical Systems Theory

383

of set theory, where this contradiction was eliminated, and no other contradictions were found. The justification of G. Frege’s logic is based on his fundamental works “On the meaning and meaning” and “On the concept and subject” and his deep analysis of natural language, adopted in modern classical logic [10]. A rigorous construction of constructive logic in the form of a Markov tower, which has many applications (for example, [11]), on which the theory of Markov algorithms, the theory of computation and programming languages is based, was carried out in [12] and in the works of A.A. Markov cited in [12]. At the same time, the construction of the Markov tower also relies on the analysis of natural language. The TDL logic of A.I. Uyemov, which interests us, is also based on the analysis of natural language, however, as we will see, it has a number of flaws and is actually not completed in a number of essential sections (well-formed formulas, inference rules, interpretation of symbols, etc.), which is easy is seen when compared with the construction of the Markov tower [12]. Nevertheless, A.I. Uyemov’s approach contains original ideas that, apparently, can be saved. First of all, it is desirable to do this in relation to the definition of systems that interests us, for which we have proposed in this paper a suitable interpretation of TDL. The fundamental pillar of logic according to G. Frege is the assertion that for science the key logical concept is the concept of truth (falsity) of natural language statements, which are narrative sentences. The goal of science is to find true statements about the objects under study. G. Frege cleared the content of natural language statements from everything that is not related to truth (falsity) and received classical logic in a modern way. The Markov tower was built on the same support, its construction was made possible thanks to a very narrow class of studied real-life objects in the form of words in alphabets, in contrast to which G. Frege considers arbitrary objectively existing things and actual infinity. As can be seen from the texts of A.I. Uyemov, it is not based on the specified requirement to discard everything in scientific statements that does not relate to questions of the truth of statements. As a result, the logic he develops is not strictly scientific; here, in order to compare his systems approach with the scientific approach of the categorical systems theory, we propose an adjustment to the TDL. We will demonstrate a couple of places where A.I. Uyemov deviates from scientific justification. TDL is built on the basis of two triads of philosophical categories “thing - property relation” and “definite-indefinite-arbitrary”. In contrast to the position accepted in mathematical logic (in the formulation of G. Frege: “What is simple cannot be decomposed, and what is logically simple cannot be determined … if we find something that is simple, … then it must be given a clear name … A definition intended to give a name to something that is logically simple is impossible. Therefore, there is nothing left but to let the reader or listener know by hints how to understand what is meant by this word” ([13], p. 253)) the following is accepted in the TDL, which, as is well known, leads to contradictions due to logical circles, the position: “Between the definitions of these categories (things, properties and relations) there is the following dependence (Fig. 1): All these categories are defined through each other, and the central main category among them is the category of a thing. Through it, the category of property and relation is directly determined, while the category of property is determined with the help of

384

G. K. Tolokonnikov

Fig. 1. Things, properties, and relations

the category of relation, and vice versa indirectly, through the category of thing” [7]. In mathematical logic, it is unacceptable to lay such circles as the basis of the theory. The second of the unacceptable features of the TDL is the violation of A.I. Uyemov’s requirement in mathematical logic, also discovered by G. Frege, of the semantic triangle “sign - meaning - meaning” in relation to the sign: “It must be demanded from a logically perfect language that … it is not provided with a value” ([13] p. 240). Note that just as G. Frege does not explicitly separate syntax from semantics (unlike the formal theories of mathematical logic), so there is no such separation in TDL. But in TDL there is no specified requirement that signs have an interpretation in the form of a meaning. A.I. Uyemov introduces three symbols in TDL t, a, A. At the same time, he associates the symbols with the words of the natural language: t - “definite” (“this”), a - “indefinite” (“something”), A - “arbitrary” (“any”) ([8] p. 70). So, t stands for a certain thing, a stand for an indefinite thing. But “… a certain object t, … an indefinite object a” contradicts the main further statement of A.I. Uyomov that “thing (object)-property-relationship” makes sense only in the context. Let’s see what t, a, A are from their description in the second book [9]: “… if we replace some words or groups of words with symbols, it will be much easier for us to understand what is written. … denote the defined by t. Indefinite, let’s denote some as a. Arbitrary, we will denote as A. Thus, the symbols t, a, A Uyemov A.I. compared some words of natural language. A sign (a name, a proper name, a word, and so on, something that points to an object) has a single meaning (an object, including an imaginary one) and a single meaning. The utterance is written as a declarative sentence in natural language. The value of a proposition is the truth value “false” or “true”. The sign differs from the concept, which is predicative, for example, the word “definite” defines the concept (“___ is certain”), since there are certain objects, that is, just the meanings of the signs from the Frege triangle, and there are many certain objects. We emphasize that we do not give definitions of the listed: sign, meaning, meaning (subject), concept. We refer to Frege’s thought cited above, only “hint” at these entities, which we will further treat as the basis for rigorous definitions. Let’s return again to attempts to understand what connects with the signs t, a, A in TDL. Of the two contradictory interpretations, the second, published 20 years after the first, is more understandable, as well as what A.I. Uyemov does. He tries to write out some part of the natural language in the form of signs for groups of words in the natural language, using the two triads “thing-property-relation” and “definite-indefinitearbitrary”. So, having written out a set of characters, he tries to match it with a set of words of a natural language. It has two kinds of sets of words, declarative sentences (judgments) and, apparently, one can say concepts (“a phrase denoting a concept” [14]).

Ternary Description Language and Categorical Systems Theory

385

Further, he determines well-formed formulas from symbols, providing such formulas with an interpretation in the natural language, builds axioms and inference rules, again relying on the interpretation of symbols and formulas from them in the form of natural language words. Unfortunately, he did not give a complete construction of the ternary description language, he refers to the fact that this is not bad, since it is possible to introduce, in addition to the existing ones, new operations for constructing formulas and new inference rules. Of course, here one cannot bypass the questions of consistency and other questions common to logical calculi. A.I. Uyemov provided some fragment of his language in [9]. So, A.I. Uyemov does not explain what the strict unambiguous interpretation of the symbols he introduced consists in, that is, he does not give a clear definition based on his understanding of the words “definite”, “indefinite”, “arbitrary”. In his opinion, A.I. Uyemov believes that one of the motivations for constructing TDL is the existing difficulties in mathematical logic, in particular, in relation to manyplace predicates. However, the examples he cites do not stand up to scrutiny. The main example, also cited by his followers [15], is as follows. Properties and relations in classical logic are specified by one-place and many-place predicates, respectively (see p.81 in [13]), while the number of places and the order of arguments are fixed. “… Mathematical logic requires fixing both the number of objects entering into a relationship and their order. But from the point of view of the ordinary understanding of the relationship, this is not required. For example, let’s take the statement “the merchants quarreled”. It is clear to everyone that a certain relationship between merchants is expressed here, and for understanding this relationship, both the number of these merchants and their order are completely indifferent. Therefore, it is quite possible to speak of relationships with an indefinite number of related objects. In mathematical logic, such relations are considered incorrect” (p. 82 in [8]). The example is, in fact, incomplete. Obviously, “it is clear to everyone” not only that there is a ratio of quarreling merchants among the merchants, but also how many merchants quarreled in that clear situation. The transition to an indefinite number of merchants is not an easy abstraction, it is clear to A.I. Uyemov, but far from “everyone”, to whom he referred. But, indeed, it is clear to everyone when any person directly observes the fight and sees how many of these merchants are there. The second point of incomprehensibility is that the predicate P(x1 , . . . , xn ) that n merchants fought contains variables that run through the set of K merchants. But the candidate for the relation “merchants fought” must have an interpretation, otherwise it will remain indefinite: how is the term “merchants” interpreted in this candidate? A.I. Uyemov does not give an answer to the question on what (on what set and how it is connected with K) the candidate for the relationship is defined. In fact, there is no particular difficulty in defining the predicate “the merchants fought” in the classical predicate calculus. This is a simple task, there is, for example, the following solution. Let M denote the collection of people. Denote by Q(x) = “__x__” is a merchant”, - the predicate “Pn (x1 , . . . , xn )n people x1 , . . . , xn fought”. Then the formula Pn (x1 , . . . , xn ) & Q(x1 ) & … & Q(xn ) has the interpretation “the merchants fought.” This is a second-order weak logic formula.

386

G. K. Tolokonnikov

As you can see, there is no such difficulty with multi-place predicates in the traditional predicate calculus, and the example cited by A.I. Uyemov is a simple problem of mathematical logic that has the above solution. In the section on systems, we will discuss other deviations from the scientific method present in A.I. Uyemov’s systems theory based on TDL.

3 Ternary Description Language Fragment with Interpretation Unlike A.I. Uyemov, we will consider the roots of words and from them (from the roots) we will consider the formation of things, properties and relationships: “brother” - the root coincides with the thing (Uyemov’s triads “thing-property-relationship”), further, “fraternal” - property, “fraternal” - attitude. We will also present in our interpretation the indicated fragment of the ternary description language. We propose to consider the set of word roots as the basis of explicit interpretation for the language of ternary description. On this set, we introduce further some operations that will generate new elements in such a way that the roots of words will be generators for the complete set of interpretation of the language. The analogue here is the free algebra with generators and n-ary operations. Let us introduce variables x, y, z, . . . The symbols t, a, A will be defined as partial functional symbols applicable to all word roots, that is, we have expressions tx, ay, Az interpreted as follows, but possibly having limitations in application both to these expressions and to some other words of the language. A.I. Uyemov, in addition to the symbols t, a, A, introduces into the language brackets (), [], an asterisk *, the symbol i (more precisely, the Greek letter “iota” ι, but it is more convenient for us to use the Latin i). These symbols and their use, he interprets as follows. For the time being, let B denote a thing, O a relation, C a property. Then: (B)C is interpreted as “thing B has property C”. O(B) - “thing B has the property O”. (B∗)C - “property C is attributed to thing B”. O(∗B) - “relation O is established in thing B”. Brackets [] translate the listed judgments into things, properties and relations: [(B)C] - “a thing with property C”. [(B∗)C] - “property C attributed to thing B” and so on. A.I. Uyemov postulates that t, a, A become things, properties and relations (respectively, defined, indefinite and arbitrary) depending on the “context”, as in the language - what is in the formula to the right of the parentheses - it’s a property, what’s on the left is a relation, what’s in parentheses is a thing. Taking into account the interpretation of the symbols t, a, A, which should be substituted for B, C, O, we have, for example, the expression for the last formula [(t∗)a], interpreted by A.I. Uyemov as “an indefinite property attributed to a certain thing”. With such a formalization, the problem arises of the occurrence of the same symbol in the formula twice.

Ternary Description Language and Categorical Systems Theory

387

Let us immediately give the definition of the system (denoted S), which is given by A.I. Uyemov (iA)s = ([a(∗iA)])t or t([iA ∗ a]). The letter i plays the role of indicating that the symbol A on the right and on the left means the same thing. A.I. Uyemov interprets this formula as follows. “Any object is a system by definition, if some relation that has a predetermined property is realized in this object, or if some properties that are in a predetermined relation are realized in this object” (p. 35 and p. 40 in [nine]). We interpret this definition using the proposed set of word roots. We see two variants of the formulas of the language of ternary description, the proposition (for example, (a)t) and the proposition placed in square brackets. Note that a “thing” (property or relation) makes sense only in a judgment corresponding to the place it occupies, therefore it is impossible to call some judgments in square brackets a thing. In our interpretation, this roughness is removed: we postulate that the judgment placed in square brackets is a root (perhaps a new root, in addition to the roots of individual words). In our version of the language of ternary description LTO - the above formulas have the form:     (B)C ∼ (ξ )η ∼ Pξ xηy, ξ, η = t, a, A P ∗ for ξ ∗ η . For example, when choosing ξ = a, η = t, we have the judgment Paxty with the interpretation “some thing ax has a particular property ty”.    O(B) ∼ η(ξ ) ∼ Qξ xηy, ξ, η = t, a, A Q∗ forη ξ ∗ . When choosing ξ = a, η = t, we have the judgment Qaxty with the interpretation “some thing ax has a particular relation ty”, and so on. Let us introduce the symbol μ into the language corresponding to the comparison of the judgments of the roots, so, for example, μPaxty is the root. This is a new root because it corresponds to the two existing x and y word roots. To this root are applicable, as well as to other roots t, a, A. Since, for different x and y, the expressions Ax and Ay are graphically different, and for x = y they coincide and, thus, have the same meaning, there is no need to introduce the operator i in our version of TDL. Let us introduce a binary operation δ on the judgments U , V , which corresponds in interpretation to the union of the disjunction “or” (“either one or the other”): δUV . Now we can write out the definition of the system according to A.I. Uyemov in our version of the language of ternary description. A system is any root of the following proposition. δPμQ∗ AUaVtWQμP ∗ AUaVtW . U , V , W - arbitrary roots.

388

G. K. Tolokonnikov

4 Comparison of the Concept of a System in TDL and Categorical Systems Theory In this section, we will begin a comparison of systems theory according to A.I. Uyemov and categorical systems theory. In a separate paper, we will continue the analysis of TDL and its logic, in particular, the section on implications and rules of inference. Here, as indicated in [15], there are unfinished and contradictory aspects of A.I. Uyemov’s approach. Having explained these points, we will be able to advance in understanding the attributive theory of systems developed by A.I. Uyemov on the basis of TDL. Above, we developed the concept of a system according to A.I. Uyemov, correcting the fragment of the TDL to a variant that can be understood, while obtaining a definition of the system, the rationale for which A.I. Uyemov is discussed below. In the categorical theory of systems, which generalized the theory of functional systems by P.K. Anokhin, a system is understood as a convolution of categorical splices corresponding to subsystems of a given system [16]. The concept of a system is functional in nature, the system is formed on the basis of a system-forming factor, which is formalized by convolution. In one of the main interpretations, a polyarrow as a system translates a set of objects that have a certain property, expressed by the corresponding predicate, into a new property of these (or transformed) objects, which is compared toresult, according to the terminology of P.K. Anokhin. The categorical splices themselves can be objects for which it is possible to construct categorical splices of the next order, similarly to how higher-order categories arise in category theory. In our opinion, one of the main differences between the definitions of systems is the static definition of the system according to A.I. Uyemov, it states the state of a set of objects with some properties that implements a certain order on a set of objects, described by the corresponding predicate. The second difference, which shows the impossibility of covering the concept of a categorical system by the definition of a system according to A.I. Uyemov (relation, and a relation having a predetermined property) are capable of realizing only the first two levels of the hierarchy. Systems according to A.I. Uyemov do not satisfy the important requirement for systems put forward in [22], which is fulfilled for functional systems and their generalizing categorical systems and consists in the need to include sources for its occurrence and functioning in the concept of a system. We proceed from the fact that the system is an objective phenomenon, it can be revealed by other researchers, while the system properties are as objective as the properties of the electron that physicists study. It is generally accepted in science that the real object under study can have various mathematical models, and in the process of studying these models, although they more and more accurately describe the properties of the object, it is impossible to achieve the “final” model. When one model replaces another, the previous model is said to be partially wrong, and the next model is said to be more true. A classic example here is a set of models of the atom: the Thomson atom, the Bohr atom, the atom in nonrelativistic quantum mechanics and in quantum field theory. The method applied by A.I. Uyemov, in this case, is reduced to the analysis of various definitions of the atom in these theories and the identification of what is common in these definitions. Obviously, this method has nothing to do with science. With a scientific approach to such a phenomenon as a system, its definition in one theory should also turn out to be more accurate, closer to the essence of the phenomenon. Thus, the substantiation of the definition of the system

Ternary Description Language and Categorical Systems Theory

389

given by A.I. Uyemov is not scientific. Therefore, the definition of a system given by A.I. Uyemov can only be viewed as another one separate from other definitions. It is necessary, which is part of our task, to find out the relationship between the concept of a system in TDL and the concept of a system in categorical systems theory.

5 Conclusions There are numerous systems theories. In particular, the general theory of systems of A.I. Uyemov (GTS), based on TDL and covering several dozens of system approaches. Categorical systems theory is an integral part of algebraic biology that predicts the properties of organisms based on the genome, including intellectual properties, which leads to models of artificial intelligence (AI) related to strong AI, in contrast to approaches based on neural networks [17–21]. In this paper, we propose a version of the TDO fragment, in which it was possible to present the definition of the system according to A.I. Uyemov in the form necessary for comparison with the system block of algebraic biology. It is shown that categorical systems are not reducible to GTS at TDO. The proposed new version of the ternary description language is also useful for solving existing problems in the formalism of the ternary description language.

References 1. Anokhin, P.K.: Fundamental questions of the general theory of functional systems. In: Principles of Systemic Organization of Functions, pp. 5–61. Nauka (1973) 2. Tolokonnikov, G.K.: Informal category systems theory. Biomachsystems 2(4), 7–58 (2018) 3. Tolokonnikov, G.K., Petoukhov, S.V.: From algebraic biology to artificial intelligence. In: Hu, Z., Petoukhov, S., He, M. (eds.) CSDEIS 2019. AISC, vol. 1127, pp. 86–95. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-39216-1_9 4. Tolokonnikov, G.K.: Convolution polycategories and categorical splices for modeling neural networks. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2019. AISC, vol. 938, pp. 259–267. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-16621-2_24 5. Uyemov, A. The language of ternary description as a deviant logic: Part 1, 2, 3. Boletim da Sociedade Paranaense de Matematica 15(1–2), 25–35 (1995), 17(1–2), 71–81 (1997), 18(1–2), 173–190 (1998) 6. Uyemov A.I. Some issues of the development of modern logic. In: Scientific Notes of the Tauride National University and Vernadsky, vol. 21, no. 1 (2008) 7. Uyemov, A.I.: Things, properties and relations. Publishing House of the Academy of Sciences of the USSR (1963). 184 p. 8. Uyemov A.I.: System approach and general theory of systems. Thought (1978). 272 p. 9. Uyemov, A.I., Saraeva, I.N., Tsofnas, A.Yu.: General systems theory for the humanities. W., Wydawnictwo “Uniwersitas Rediviva” (2001). 276 pp. 10. Church, A.: Introduction to mathematical logic, Moscow, IL (1960). 486 pp. 11. Rajan, E.G.: Theory and Application of Symbolic Computing with Artificial Intelligence Perspective. London Journals Press (2023). 1074 p. 12. Markov, A.A., Nagorny, N.M.: Theory of Algorithms. Nauka (1984). 432 p. 13. Frege, G.: Logic and logical semantics. Aspkt Press (2000). 512 p. 14. Uyemov, A.I. Some questions of the development of modern logic. In: Scientific notes of the Taurida National University and Vernadsky. Philosophy, Sociology, vol. 21(60), no. 1 (2008)

390

G. K. Tolokonnikov

15. Leonenko, L.L.: Definitions in the language of ternary description. In: Scientific Notes of the Taurida National University and Vernadsky. Philosophy, Culturology, Political Science, Sociology, vol. 24(63), no. 3–4. pp. 397–404 (2011) 16. Tolokonnikov, G.K.: Categorical gluings, categorical systems and their applications in algebraic biology. Biomachsystems 5(1), 148–235 (2021) 17. Karande, A.M., Kalbande, D.R.: Weight assignment algorithms for designing fully connected neural network. IJISA (6), 68–76 (2018) 18. Dharmajee Rao, D.T.V., Ramana, K.V.: Winograd’s inequality: effectiveness for efficient training of deep neural networks. IJISA (6), 49–58 (2018) 19. Hu, Z., Tereykovskiy, I.A., Tereykovska, L.O., Pogorelov, V.V.: Determination of structural parameters of multilayer perceptron designed to estimate parameters of technical systems. IJISA (10), 57–62 (2017) 20. Awadalla, H.A.: Spiking neural network and bull genetic algorithm for active vibration control. IJISA 10(2), 17–26 (2018) 21. Ojo, J.S., Ijomah, C.K., Akinpelu, S.B.: Artificial neural networks for earth-space link applications: a prediction approach and inter-comparison of rain-influenced attenuation models. IJISA 14(5), 47–58 (2022) 22. Lektorsky, V.A., Sadovsky, V.N.: On the principles of systems research. Quest. Philos. (8), 67–79 (1960)

Radial Mean LBP (RM-LBP) for Face Recognition Shekhar Karanwal(B) CSE Department, Graphic Era University (Deemed), Dehradun, UK, India [email protected]

Abstract. In literature various local descriptors has been invented for Face Recognition (FR) in unconstrained conditions. Most of these local descriptors are inspired from the LBP concept. With the result numerous LBP variants are introduced. Some achieved good results, and some are not. The reason behind unsatisfactory performance is the lack of building the discriminant concept in forming the descriptor. The major significance of any research is to develop the novel concept/method in form of descriptor, and it should be discriminant and robust, which should justify it by defeating the various benchmark methods. In proposed work such novel concept is introduced in the form of novel local descriptor called as Radial Mean LBP (RM-LBP), in different unconstrained conditions. In RM-LBP, initially mean value is obtained from 4 radial pixels located at 4 different radii, in eight directions of 9 × 9 patch. This forms the 3 × 3 patch, in which neighbors are filled with mean values and center place is filled with median of those. Ultimately for making the RM-LBP code, the neighbors are compared with the center pixel. After obtaining RM-LBP map image (by computing codes in each location) the size derived is 256. To compress and match reduced feature, PCA and SVMs are used. For experiments evaluation ORL face dataset is used. Results suggests that RM-LBP is better and efficient descriptor than many others. The 6 other evaluated descriptors and many literature approaches are easily outclassed by RM-LBP. RM-LBP achieves the ACC of [72.22% 84.37% 92.85%], which proves higher than many others. The significance of RM-LBP is justified when it defeats the numerous benchmark methods. The matlab version used for evaluation is MATLAB R2021a. Keywords: Face recognition · local feature · global feature · compression · matching · gray images

1 Introduction In Computer Vision and Pattern Recognition fields, the one class of descriptors which has significantly marks its influence is the local descriptors. The local descriptors achieve astonishing outcomes in such fields as a result invention of local descriptors has been progressing day by day. The local descriptors build its size by working on different image portions by using the small and larger size patches. Some descriptor used small © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 391–400, 2023. https://doi.org/10.1007/978-3-031-36118-0_35

392

S. Karanwal

scale patches and some uses the large scale patches for feature extraction. The descriptors which build the feature size by using integration of small and large scale patches produces finer results than either of them. In local there are two ways of feature extraction, (i) Either local features are extracted globally or (ii) the regional manner feature extraction (with integration) is done to achieve the entire feature size. Each phase has its advantages. In first phase, the time complexity is on the lower side and in second phase the time complexity is on the higher side. The accuracy in second phase is much superior in contrast to the first phase. The various image transformations which lower the performance are occlusion, corruption, blur, light, age, scale, emotion and pose. One of the major patches based local descriptor developed in literature is LBP [1]. LBP is the first descriptor in which extraction of features are done by using 3 × 3 patch. In LBP, the neighbors are replaced with binary value 1 or 0 as per the thresholding function. For higher or equal value to center, 1 is given as the label else 0 is given. The 8 bit pattern is then transformed to LBP code by weights allocation. By continuation of calculating codes (in each position) constructs LBP image, from which the histogram size produced is 256. LBP has the two basic advantages, (i) invariance monotonic property of gray scale and (ii) less complex concept. There are numerous disadvantages are also noticed in LBP and these are, (i) noisy function, which is used for thresholding, (ii) limited spatial patch ability, (iii) large size of the feature and (iv) un-affective in harsh lightning variations. This makes the researchers to propose the various variants of LBP. Some achieves good results, and some are not. In recent times, other class of methods which attracted significant attention is deep learning methods. The success achieved by deep learning methods is relied on the complex structure built up as per the unconstrained conditions. As a result, their performances are much superior to local and global methods. CNN, LetNet, AlexNet and VGG are some of the versions of deep methods. In CNN methods, the completely connected final layers of pre-trained CNN models are used for making the size. The other activation functions (like softmax and classification layers) can also be used for developing the feature size. Most of these deep methods possesses complex structure, which results in various demerits, and these are, (i) high computational cost, (ii) training data requirement in huge amount and, (iii) adaption as per the parameter settings. These demerits put a question in the mind for developing such method. Literature portrays that there are some works based on the local descriptors, which achieves better performance than these deep methods. So, to develop such descriptor is more beneficial in contrast to methods which are complex and whose performances are not adequate. In literature various local descriptors has been invented for Face Recognition (FR) in unconstrained conditions. Most of these local descriptors are inspired from the LBP concept. With the result numerous LBP variants are introduced. Some achieved good results and the performance of others are not satisfactory. The main limitation behind unsatisfactory performance is the lack of building the discriminant concept in forming the descriptor size. The developed concept in form of descriptor should be discriminant and robust which should prove it by defeating the various benchmark methods. This should be prime objective for any research. The proposed descriptor fulfills this objective (the prime) comprehensively. In proposed work such novel concept is introduced in form of novel local descriptor called as Radial Mean LBP (RM-LBP), in different unconstrained

Radial Mean LBP (RM-LBP) for Face Recognition

393

conditions. In RM-LBP, initially mean value is obtained from 4 radial pixels located at 4 different radii, in eight directions of 9 × 9 patch. This forms the 3 × 3 patch, in which neighbors are filled with mean values and center place is filled with median of those. Ultimately for making RM-LBP code, the neighbors are compared with center pixel. After obtaining RM-LBP map image (by computing codes in each location) the size derived is 256. To compress and match the reduced feature PCA [2] and SVMs [3] are used. For experiments evaluation ORL [4] face dataset is used. Results suggests that RM-LBP is better and efficient descriptor than many others. The 6 other evaluated descriptors i.e. LBP [1], HELBP [5], NI-LBP [6], LPQ [7], RD-LBP [6] and 6x6 MBLBP [8], and many literature approaches are easily outclassed by RM-LBP. RM-LBP achieves the ACC of [72.22% 84.37% 92.85%], which proves higher than many others. The matlab version used for evaluation is MATLAB R2021a. Road map: In Sect. 2 the work related to local descriptors are discussed, Sect. 3 provides proposed descriptor description, results and discussions are communicated in Sect. 4 and 5, Finally conclusion and future prospect are given in Sect. 6.

2 Related Works Karanwal et al. [9] presented two LBP variants in Face Recognition (FR) called as NDLBP and NM-LBP. In ND-LBP, clockwise directioned pixels are collated to each other, for making its feature size. In NM-LBP, the feature size is formed after the neighbors are compared with mean of those. Eventually the most robust feature is formed by merging ND-LBP and NM-LBP features, called as ND-LBP + NM-LBP. PCA and SVMs are consumed for the compaction and classification. For evaluating results ORL and GT face datasets are used. Result shows that ND-LBP + NM-LBP performance is far better than individual descriptors and also from various literature methods. The performance of ND-LBP, NM-LBP and ND-LBP + NM-LBP is not as effective as it should be in the unconstrained conditions. The methodology used for making their feature size is not discriminant and robust, which reflects in their results. If the developed methodology is not effective, then the results are not encouraging. In [10] Khanna et al. discovered the Emotion Recognition (FR) method, by utilizing the two best performing local methods i.e. LBP and STFT. LBP is used as local feature for extracting features and STFT is used as the frequency feature for extracting the features. The feature amalgamation of both is further shortened by FDR, variance threshold and chi-square test. SVMs is the classification method used for matching. Results on different datasets confirms the capability of invented feature. In this, those two methods are used which are already exists in literature. No new methodology is evaluated in this research, as a result the outcome is not as encouraging as it should be in the unconstrained conditions. Karanwal et al. [11] discovered three LBP versions in FR i.e. MLBP, MnLBP and CLP. MLBP is mean based method in which desired code is developed by neighbor’s comparison to whole patch mean. MnLBP is the median based method in which desired code is developed by neighbor’s comparison to whole patch median. In contrast to the center pixel comparison (as LBP does), the mean and median comparison proves much better and discriminant. To build more informative and influential face descriptor the LBP, MLBP and MnLBP features are merged totally. This feature is termed as CLP. For compression and matching

394

S. Karanwal

PCA and SVMs are used. On ORL and GT, the CLP conquers performance of alone and various literature techniques. Due to the usage of 3 × 3 patch, the existing methodology is not as impressive as it should be in the unconstrained conditions. This sacrifices the robustness to huge extent. Under noisy conditions, the higher patch must be used for attaining the better results. In proposed work the higher patch is used for feature extraction and it achieves better results than the other patches. Bedi et al. [12] presented MLBP for classification of Liver images, ultrasound in nature (LUI). MLBP transforms mutual relationship among neighboring pixels, to the pattern (binary) based on the Standard deviation and Euclidean distance templates from center place. The GLCM and the color features are also used for making the size. On two distinct applications i.e. natural and facial images, MLBP proves its ability, by defeating various methods. The one thing which is lacking in this work is that the feature extraction is performed from small scale patch. It must be performed from the higher scale patch for achieving better results. Karanwal et al. [13] invented novel FR descriptor MBZZBP. By taking 6x6 patch, initially mean is generated in different regions (of size 2x2). Then the zigzag ordered pixels are compared to each other. The higher ordered pixel is differentiated from the lower ordered pixel to form MB-ZZLBP code. On several datasets MB-ZZLBP defeats the accuracy of numerous methods. In MB-ZZLBP, the feature extraction is performed through the 6x6 patch. Although 6x6 patch achieve better accuracy but not effective as proposed descriptor achieve in the unconstrained conditions. The RM-LBP extracts features from higher patch and achieves encouraging results. Kola et al. [14] imposed noise discriminant descriptor for ER. LBP feature is generated first, by computing the diagonal and four neighbors individually. For effective description of feature, adaptive window process and radial orientation mean are also launched. For matching SVMs is used. On distinct datasets, result confirms invented method ability. Despite using various things the discriminativity achieved by this method is not as impressive as required in the unconstrained conditions. Some better methodology could be created in unconstrained conditions to achieve the better outcome.

3 The Invented Descriptor RM-LBP Literature presented various local descriptors in unconstrained conditions. Most of these local descriptors are inspired from the methodology of LBP. LBP encodes the neighbors based on gray pixel value and value of center pixel. After the development of LBP, various LBP variants were introduced. Most of these descriptors (compared descriptors, implemented in proposed work) lacks somewhere in their methodology as a result, the desired results are not generated when it is required. The main objective of this research is to develop novel methodology, and which beats accuracy of several existing methods. In proposed RM-LBP, the research objective i.e. discriminativity and robustness is achieved by introducing the novel descriptor Radial Mean LBP (RM-LBP) for FR. The RM-LBP beats several descriptors in terms of accuracy. The RM-LBP is implemented by using 9 × 9 patch for the FR application. Results suggests that RM-LBP is far better descriptor than the compared descriptors. The advancing features of RM-LBP is the extraction of features from 9 × 9 patch. The detailed RM-LBP illustration is given as.

Radial Mean LBP (RM-LBP) for Face Recognition

395

In RM-LBP, 9 × 9 patch is used for extracting the feature size. Initially, the mean value is derived from 4 radial pixels (located at different radii) in eight directions of the 9 × 9 patch. After mean value generation there is the evolution of 3 × 3 patch. In 3 × 3 patch, the eight neighbor locations are filled with the mean values (derived from the respective directional radial pixels (located at different radii). The center location is replaced with the median of those eight neighborhoods. Further, the neighbors are thresholded to 1 for being larger or same mean value to median else 0 is given. This construct the pattern of size 8 bit, which is transfigured to RM-LBP code by deployment of the weights (binomial) and values summation. The RM-LBP code computation process is done for every pixel location and it results in the map image of RM-LBP. The RM-LBP map forms 256 size. Equation 1 and Eq. 2 shows RM-LBP code computation procedure for single location. In Eq. 1, the mean is obtained from each radial directions. In Eq. 1, P, (R1-R4) and U(R1−R4),p signifies the size of the neighborhood, radial pixels (at R1-R4 i.e. 4) and individual pixels located at different radii. In Eq. 2, P, R, µ(R1−R4),p and Uc(median) signifies the size of the neighborhood, radius, individual mean values and median value. Figure 1 shows RM-LBP illustration. The proposed FR framework is delivered in Fig. 2. 1 P/2−1 U(R1−R4),p (for each direction) (1) p=0 P/2   P−1   p 1x≥0 RM − LBPP,R (xc ) = f µ(R1−R4),p − Uc(median) 2 , f(x) = (2) p=0 0x forAll(v | v.oclIsKindOf(File))

effective text processing approaches, for example [4], into analysis of requirement specification. In addition, proposed approach [2] supports four types of OCL expressions. The same ideas may be expressed using different types of OCL expressions. For example, OCL 1.1.1 from the Table 1, may be written from the context of class “User”. Context User (computers c):boolean Pre: c.policy = true Result: return true This OCL expression describes the same condition as the OCL 1.1.1, namely from which computers user can get access to some resource. From the point of view of network audit procedures semantic of this operation is the same. It is noticed, that procedures of monitoring or controlling access to resources require procedure of selecting objects from some collections. It is recommended, to add new types of OCL for processing collection of objects for model of analytical representation of OCL expressions. Proposed model is flexible and allow doing it. In addition, the recommendation for domain expert is to express the same condition using all possible types of OCL expressions. This recommendation expects that domain

518

O. Chebanyuk

expert has high experience of OCL expressions designing and increases the time of domain analysis. It is recommended to add to practices of company using plug-in or environment for automated verification of designed OCL expressions as well as using of methods of extracting facts from text [9]. Selected plug-ins must be convenient for usage, and be as close as possible to family of OCL standards. 4.3 Software Specification Analysis 4.3.1 Requirement Analysis. Designing of Software Specification for Considering Problem Domain with Semantic Attributes Small part of designed requirement specification, prepared in authors’ investigation is represented in the Table 2. Designing of requirement specification touches to the same problem as designing of repository table. When requirement engineer designs OCL expressions different types of OCL expressions may be used to express the same semantic. Recommendation for requirement engineer – is to read existing OCL expressions and try to use the same type of OCL that were used by domain experts for verifying the same condition expressed in natural language. Additional requirement for requirement engineer also high level of knowing of OCL and experience of work with the corresponding plug-in or environments. 4.4 Comparison of Semantic Attributes 4.4.1 Defining Common Keywords in Repository Tree and Requirement Specification The list of the common keywords: “access matrix, log journal, and statistical characteristics of traffic” (Table 1 and Table 2). This point fully available for the chosen application domain. 4.4.2 OCL Expressions Comparison Results of OCL expressions comparison for problem domain repository tree and requirement specification are represented in the Table 3. Comparison operation is performed realizing ideas explained in model of approximate OCL comparison. Words that are differ in project and repository (reference) OCL expressions are marked by yellow color (Table 3). OCL 1.2.2 and OCL F2.1, OCL 1.2.1 and OCL F2.1.1, OCL 1.2.2 and OCL F2.1.1 and OCL 1.2.2 and OCL F3.1 have different types – they are not equal. Estimation of comparison model – for the chosen problem domain it needs precision of conditions for comparison of collections used in OCL expressions. Other words – it needs to design comparative model for the new type of OCL expressions working with collections of objects. Analysis of the Table 3 shows that coefficients for OCL comparison in paper [2] were chosen correctly because those OCL expressions that looks like the same from human point of view are appeared equal after calculations.

Investigation of Drawbacks of the Software Development

519

Table 2. Requirement specification of security model for monitoring access to disk in local network RQ code RQ description and limitation in natural language

Keywords and OCL OCL code and expressions [7] for comparison

F1

Plug-in for monitoring access to disk resources

1.1 access matrix

OCL F1.1.1 Context disk_access (User u):boolean

F1.1

Access for disk memory must be given only to verified user

OCL 1.1.1

Pre: u.checked = true Result: return true;

F 1.2

If assess to 1.1 access matrix specific folder for OCL 1.1.1 domain user and your security group is allowed in domain computer then user can modify files in this folder

OCL F1.1.2 Context GroupPolicy inv: Comp implies User.ID-> forAll(v | v.oclIsKindOf(User))

F2

Calculate a network traffic statistics

1.2 log journal OCL 1.2.1 OCL 1.2.2

OCL F2.1 Set{time1,time2,…,timen} -> iterate(i: Integer, sum: Integer = 0 | sum + time(i))

F2.1

If computer has more than 10000 open connections than access to it is blocked

1.2 statistical Characte-ristics of traffic OCL 1.2.1 OCL 1.2.2

OCL F2.1.1 Context NetworkTraficc::block(stat acess, User u): EBoolean pre: stat.con> 10000 post: result = true

F3

Every 10 s save 1.2 log journal collection of OCL 1.2.1 users that get OCL 1.2.2 access to database( their logins, and access matrixes)

OCL F3.1 Context buffer Set{ Tuple{login:’Jordi Cabot’, 11111}, Tuple{login:’Marco Brambilla’,00000,Tuple(login:’Massimo Tisi’,11001)

520

O. Chebanyuk Table 3. Comparison of project and reference OCL expressions.

Reference OCL expressions (from repository)

Project OCL expressions

Comparison results, explanations, and recommendations for reuse

Context comp_access (User u):boolean Pre: u.checked = true Result: return true;

Context disk_access (User u):boolean Pre: u.checked = true Result: return true;

OCL 1.1.1 and OCL F1.1.1 are considered approximately equal Solution 1 - Software modules of network audit can be reviewed for reuse to realize RQ F1.1

Context comp_access (User u):boolean Pre: u.checked = true Result: return true;

Context GroupPolicy OCL 1.1.1 and OCL F1.1.2 have inv: Comp implies User.ID-> forAll(v different types and considered not | v.oclIsKindOf(User)) equal There is no software asset for reuse

Context Router Set{time1,time2,…,timen} -> iterate(i: Integer, sum: float = 0 | sum + time(i))

Set{time1,time2,…,timen} -> iterate(i: Integer, sum: Integer = 0 | sum + time(i))

OCL 1.2.1 and OCL F2.1 are considered approximately equal Solution 1 is used

Context Router Set{time1,time2,…,timen}-> iterate(i: Integer, sum: float = 0 | sum + time(i))

Context buffer Set{Tuple{login:’Jordi Cabot’,11111},Tuple{login:’Marco Brambilla’,00000}}

OCL 1.2.1 and OCL F3.1 are not considered approximately equal

5 Comparative Analysis of the Proposed and Other Approaches This chapter represents comparative analysis of different approaches allowing analyzing semantic of software development artifacts according to defined. Table 4 summarizes advantages and disadvantages of different software engineering approaches. The first row represents drawbacks defined during the analysis represented in the previous point. The seconds’ row – drawbacks of the proposed approach, the third and the fourth rows describe drawbacks of other approaches. Paper about MoC tools represents EmS formal design tools, i.e. the formal description of the semantics of models (including the execution semantics). Different scenarios of semantic comparison of specification and design documents are described [14]. Paper [15] proposes an ontology-based method to reason about the correctness of SysML models that include OCL constraints. In order to do this Transformation language from is proposed SysML models with OCL constraints to OWL id proposed. Some operation need to be implemented in mind of stakeholder, this factor influences to different representation of the same semantic. Analysis of the table shows that defined drawbacks that were recognized in authors’ approach are typical for development of approaches that process semantic of software development artifacts.

Investigation of Drawbacks of the Software Development

521

Table 4. Defined drawbacks of approaches using semantic comparison of software development artifacts, Proposed approach MoC Tools [14] Sys ML OWL [15] +

+

+

Different varians to consider the same + semantics aspects when domain model is designed ( p3.2)

+

+

Different varians to represent the same + semantics when domain model is used ( p3.3)

+−

+

−

+–

-

Representation different results of systematization information about domain (p 3.1)

Not correct semantic model for comparison

6 Conclusions and Recommendations 6.1 Domain Analysis Area In order to design domain model reusable in application engineering stage the next problems need to be solved: 1) Different variants of domain models may represent the same semantic. (Examples for domain trees: different domain experts may assign different names of the processes or different structure of domain trees may represent the same semantic (Fig. 1), In texts or OCL the same semantics may be described by different expressions. The solution is the next: follow naming rules - for naming of classes, attributes, and defining keywords using definitions from international (regional or company) standards. In addition, it is necessary to think about possible synonyms of used terms. Alternative way to solve this problem is to involve approaches for natural languages text processing [12]. They may be realized in different ways – reusing of cloud services, taking software complexes for processing of texts, etc. From the other hand such factors as grammar errors or different styles for text representation, different encodings, quality of texts in images etc. Solution is the next: it is recommended to use definitions from enterprise standards where it possible; (ii) keep a style of the text allowing formal extraction of answers to questions (avoid dialects, non-formal style of facts representing, grammar errors). 2) Follow strict rules for representing of semantic for software development artifacts. In order to simplify the process of OCL expressions recognition in application engineering stage it is recommended for domain expert to write OCL expressions using all possible types of them.

522

O. Chebanyuk

6.2 Evaluation of Drawbacks of the Proposed Approach 1) It is recommended to add the new type of OCL expression allowing to process collections. Represented approach proposes the flexible model for analytical representation of OCL expression (based on graph representation) with possibility to extend supported types of OCL expressions. In addition, the comparison model allows setting up different weights for comparison coefficients. 2) Estimation of comparison model. Chosen coefficients for comparison of OCL expressions for considering problem domain are set correctly. It is proved by the fact that the from human point of view and results of calculation, preformed using proposed model. Are matched. 6.3 Application Engineering Area It is recommended for requirement engineer to obtain high qualification with using of plug-ins or environments for verification of OCL expressions in automated mode. In addition, composition of such type a specifications needs more time and qualification in comparison with approaches allowing composing simple requirement specification. Advantage of the proposed approach – pre and post conditions allow providing a good background for designing of UML sequence (or activity) diagrams. OCL expressions may be used as initial information for automated generation of metamodel (class diagrams + OCL expressions) [8]. It allows to save time of performing processes of software designing and testing. Additional activity recommended for domain expert and for stakeholders is to involve approaches and software systems for extracting important facts from requirement specification and performing other operations of text processing [1], for example documentation about software modules.

References 1. Andonov, F., Slavova, V., Petrov, G.: On the open text summarizer. Inform. Content Process. 3, 278–287 (2016). http://www.foibg.com/ijicp/vol03/ijicp03-03-p05.pdf 2. Chebanyuk, O.: An approach to software assets reusing. In: Zlateva, T., Goleva, R. (eds.) Computer Science and Education in Computer Science. CSECS 2022. Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, vol. 450. Springer, Cham (2022). https://doi.org/10.1007/978-3-031-17292-2_6 3. Gnatyuk, S.: Critical Aviation Information Systems Cybersecurity, Meeting Security Challenges Through Data Analytics and Decision Support, NATO Science for Peace and Security Series, D: Information and Communication Security. IOS Press Ebooks, vol. 47, no. 3, pp. 308–316 (2016) 4. Ivanova, K.: NLA-Bit: a basic structure for storing big data with complexity O(1). Big Data Cogn. Comput. 5(1), 8 (2021). https://doi.org/10.3390/bdcc5010008 5. Lisboa, L., Garcia, V.D., Lucrédio, D., et al.: A systematic review of domain analysis tools. Inform. Softw. Technol. 52(1), 1–13 (2016) 6. Odarchenko, R.R, Abakumova, A., Polihenko, O., Gnatyuk, S.: Traffic offload improved method for 4G/5G mobile network operator. In: Proceedings of 14th International Conference on Advanced Trends in Radioelectronics, Telecommunications and Computer Engineering (TCSET-2018), pp. 1051–1054 (2018)

Investigation of Drawbacks of the Software Development

523

7. OMG standard Object Constraint Language 2.3.1 (2020). https://www.omg.org/spec/OCL/2. 3.1/About-OCL/ 8. Pérez, B., Porres, I.: Reasoning about UML/OCL class diagrams using constraint logic programming and formula. Inf. Syst. 81, 152–177 (2018) 9. Slavova, V.: Emotional valence coded in the phonemic content–statistical evidence based on corpus analysis. Cybern. Inform. Technol. 2, 3–21 (2020) 10. TZI. https://tzi.ua/ua/nd_tz_1.1-003-99.html 11. Akuma, S., Anendah, P.: A new query expansion approach for improving web search ranking. Inform. Technol. Comput. Sci. 2, 1–16 (2022). https://doi.org/10.5815/ijitcs.2023.01.05 12. Almutiri, T.F.: Nadeem markov models applications in natural language processing: a survey I.J. Inform. Technol. Comput. Sci. 2, 1–16 (2022). https://doi.org/10.5815/ijitcs.2022.02.01 13. Goel, V., Kumar, V., Jaggi, A.S., Nagrath, P.: Text extraction from natural scene images using OpenCV and CNN I.J. Inform. Technol. Comput. Sci. 9, 48–54 (2019). https://doi.org/10. 5815/ijitcs.2019.09.06 14. Diallo, P.I.: A framework for the definition of a system model MoC-based semantics in the context of tool integration. Embedded Systems. Université de Bretagne occidentale - Brest (2014). English. ffNNT : 2014BRES0067ff. fftel-03258224f 15. Lu, S., Tazin, A., Chen, Y., et al.: Detection of Inconsistencies in SysML/OCL models using owl reasoning. SN Comput. Sci. 4, 175 (2023). https://doi.org/10.1007/s42979-022-01577-0

Perfection of Computer Algorithms and Methods

Improving Piezoceramic Artificial Muscles for Flying Insect-Sized Mini Robots Sergey Filimonov, Constantine Bazilo(B) , Olena Filimonova, and Nadiia Filimonova Cherkasy State Technological University, Shevchenko Blvd, Cherkasy 46018006, Ukraine {s.filimonov,n.filimonova}@chdtu.edu.ua, [email protected]

Abstract. The relevance of the work is associated with the development of artificial muscles of flying insect-sized mini robots. The development of such robots is due to a significant decrease in the population of insects, namely bees. The purpose of the article is to improve the design of the piezoceramic artificial muscle of flying insect-sized mini robot by determining the rational dimensions of the actuator and developing its design. 3D-dimensional numerical simulation of the process was carried out using the COMSOL Multiphysics software package to study and determine the maximum mechanical vibrations of the piezoactuator. On the basis of the obtained data the graphical dependencies were constructed to determine the rational parameters of the piezoelectric actuator. In addition, a new design of an artificial muscle based on a cruciform piezoelectric actuator has been created and studied. The research results showed an increase in the oscillation amplitude by 7 times compared with the basic design. These results will reduce the power consumption by about 7 times by reducing the amplitude of the control voltage of artificial muscles based on piezoactuators. In addition, this will reduce the weight of the entire structure and make it more autonomous and lighter. The obtained results can be used in the design of piezoelectric actuators for insect-sized mini robots. Keywords: Air mini robot · Piezoelectric element · Actuator · Artificial muscles · Robobee

1 Introduction Today, the loss of bees, to which we owe 1/3 of the entire crop, cannot pass without consequences. Having lost the population of bees, humanity will lose not only honey [1]. Since most of the crop depends on bees. Simultaneously with the search for a solution to stop the increase in bee mortality, scientists are looking for a replacement for them. In the modern world, many processes are automated [2, 3], and it is possible to automate the process of pollination of plants as well. One of the analogues to replace the bee is a flying insect-sized mini robot RoboBee. In addition, robots of this type can be used for numerous applications, from search and rescue operations to environmental monitoring [4] and biological research. Bee-sized robots (−100 mg) impose greater restrictions on the use of the components. One of the most important components in flying insect-sized mini robots is the element © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 527–538, 2023. https://doi.org/10.1007/978-3-031-36118-0_47

528

S. Filimonov et al.

that sets the wings in motion. As is known, common drives used in larger robots drive the propellers in most drones. Such electric motor is not advantageous to scale down to an insect in terms of efficiency and power density [5]. This is due to the fact that depending on the surface area, losses, such as Coulomb friction and electrical resistance, take on a greater value, the smaller is scale [6]. Flying insect-sized mini robots mostly use piezoceramic actuators (artificial muscles) to create wing movement. However, the main disadvantage of these artificial muscles is the high control voltage to generate enough wing movement to get airborne. In addition, obtaining a high voltage in autonomous mode leads to a heavier circuitry. All these factors greatly limit the carrying capacity. To solve the problem, it is necessary to sequentially perform a number of tasks: – to analyse the design features of modern flying insect-sized mini robots and the main results of research on working processes in them; – to simulate a piezoelectric actuator (artificial muscle) and determine its rational dimensions; – to develop a new and improved design of an actuator (artificial muscle) for flying insect-sized mini robot and conduct an experimental study. Thus, the development of a new type and improvement of known types of piezoceramic actuators is an important task. The object of research is the processes of bending mechanical oscillations of bimorph piezoelectric element of piezoactuator. The subject of the study is a piezoelectric actuator based on a bimorph piezoelectric element (artificial muscle). The aim of the work is to improve the design of the piezoceramic artificial muscle of flying insect-sized mini robot by determining the rational dimensions of the actuator and developing its design.

2 Literature Review Figure 1 shows the design of one of the analogues of flying mini robots (small-sized flying systems – quadcopters), which uses electromagnetic motors [7–9]. Further reduction of such structures requires a reduction in the dimensions of electromagnetic motors, which in turn leads to a decrease in the efficiency [10], and this is not acceptable. One of the most relevant ways to reduce the size of flying mini robots is the use of piezoelectric actuators [11, 12]. Researchers from the Wyss Institute at Harvard University (USA) are developing an insect-sized mini robot RoboBee (Fig. 2) [13, 14]. First of all, in the framework of this work it is important for us to note the design features of RoboBee piezoelectric actuators which is the main analogue of almost all designs of flying insect-sized mini robots. The main element for lifting into the air in RoboBee are “unimorph” piezoelectric actuators [15]. The use of actuators of this type allows to reduce the dimensions and weight, and also increases the reliability and durability of the structure [16]. Additional modifications allow some models of RoboBee to switch from swimming underwater to flying, as well as to “hook” on surfaces using static electricity [17, 18]. However, piezoelectric actuators remain the main elements for lifting into the air.

Improving Piezoceramic Artificial Muscles for Flying Insect-Sized Mini Robots

529

It should be noted that currently RoboBee is autonomous only under certain conditions (bright light with a certain wavelength). The next example of an insect-sized mini robot is the development of engineers from the University of Washington (USA). They presented their version of the 74-mg miniature flying robot RoboFly (Fig. 3) [19]. As in the previous design, piezoelectric actuators are used for lifting into the air.

Fig. 1. Mini robots based on electric motors “flying monkey”

Fig. 2. Mini robot RoboBee: 1 – piezoelectric actuator, 2 – reducer mechanism, 3 – wings

Another type of mini robot design based on bimorph elements is the Four Wings design (Fig. 4) [20]. It differs from the previous ones in the number and location of piezoelectric actuators. Having considered the designs of mini robots RoboBee, RoboFly and Four Wings, which are based on a piezoelectric actuator, we can say that their construction is almost the same. Some models differ only in execution (bimorph or monomorph), location and number of piezoelectric actuators in operation. However, options for choosing the geometry, the location of the piezoelectric material on the surface of a metal plate or other geometric structure are not given. Another disadvantage of piezoelectric actuators is the high control voltage, about 200 V. All these shortcomings lead to the impossibility of further improvement of the technical characteristics of piezoelectric actuators, except for adding the number of actuators in one node. However, this leads to an increase in the mass and dimensions of the air mini robot.

530

S. Filimonov et al.

Fig. 3. Mini robot RoboFly

Fig. 4. Mini robot Four Wings

Professor YuFeng Chen from Massachusetts Institute of Technology has developed insect-sized drones with unprecedented agility. Aerial robots are equipped with a new class of soft drives, allowing them to withstand the physical stress of flight (Fig. 5). The actuators of this flying mini robot can make up to 500 strokes per second, making the mini robot resistant to external influences. The robot weighs only 0.6 g, which is approximately the same as the mass of a large bumblebee [21]. Despite all the advantages of this design, it has certain disadvantages. The disadvantages of this mini robot are the high control voltage, which is 2000 V, and the great complexity of manufacturing, because elements of nanotechnology are used in the process of creation. High control voltage leads to certain PCB requirements, increased weight and power consumption. Thus, having considered the designs of flying insect-sized mini robots, we can draw next conclusions. The main element of the lifting force of the mini robot is the actuator. Actuators are divided into piezoelectric and based on dielectric elastomer materials. The main disadvantages of actuators based on elastomer dielectric materials are mentioned above. Having considered the designs of mini robots Robobee, RoboFly and Four Wings, which are based on a piezoelectric actuator, we can say that their construction is almost the same. Some models differ only in execution (bimorph or unimorph), location and number of piezoelectric actuators in the robot. However, options for choosing the geometry, the location of the piezoelectric material on the surface of a metal plate or other geometric structure are not given. Another disadvantage of piezoelectric actuators is the high control voltage, of about 200 V. All these shortcomings lead to the impossibility

Improving Piezoceramic Artificial Muscles for Flying Insect-Sized Mini Robots

531

of further improving the technical characteristics of piezoelectric actuators, except for adding the number of actuators in one node. In turn, this leads to an increase in the mass and dimensions of the aerial mini robot.

Fig. 5. Mini robot based on dielectric elastomer actuators

So, improving the design of the artificial muscle of a flying insect-sized mini robot by determining the rational dimensions of the actuator and developing its design, as well as creating a new actuator geometry is an important and urgent task.

3 Materials and Methods Considering the technical features of piezoelectric actuators, which complicate the experimental determination and selection of the correct form of their oscillations, it is optimal to use for this purpose numerous calculation methods implemented by specialized CAD systems [22, 23]. To study the influence of the design parameters of the piezoelectric actuator, numerical simulation of the piezoelectric element operation process was carried out using the COMSOL Multiphysics software package [9], which makes it possible to carry out numerical simulation of three-dimensional models of piezoelectric actuators with the necessary parameters and limiting conditions. In the process of modelling Lagrange finite elements with elementary basis functions of the second order (Lagrange-Quadratic) are used. Firstly, we determine the rational location of the piezoelectric element on a metal plate. Secondly, we determine the rational length of the piezoelectric plate, taking into account the first study. Certain results will make it possible to obtain the maximum oscillation amplitude of the bimorph piezoceramic actuator, which in turn can reduce the amplitude of the control voltage. Initial boundary conditions of bimorph piezoelectric element. The metal plate is a console that is rigidly fixed on one side and the other side is free. The piezoelectric element is symmetrically located relative to the axis of symmetry of the metal plate, and its bottom side is located at a distance of 1 mm from the place where the metal plate is fixed. An input voltage of 100 V is applied to the top and bottom surfaces of the piezoceramic plate. The dimensions of the piezoelectric element are 16 × 4 × 0.1 mm, and the dimensions of the metal plate are 38 × 6 × 0.1 mm. During the experiment, the dimensions of the metal plate remain stable, while the dimensions of the piezoelectric plate change (Fig. 6).

532

S. Filimonov et al.

Fig. 6. Basic model of a bimorph piezoelectric actuator: 1 – metal plate; 2 – piezoelectric element; 3 – the place of the cantilever fixing of the metal plate

The principle of operation of a bimorph piezoceramic actuator is as follows. When an alternating electrical voltage is applied, namely, the signal output is connected to the free side of the piezoceramic element 2 and the common output to the metal plate 1, the piezoceramic actuator begins to deform. Since the actuator is fixed on one side, the maximum vibrations will be on the free side. In the general design of a flying mini robot, the free side is connected through a gear mechanism to the wings, which in turn begin to perform movements.

4 Numerical Experiments and Results Using the functionality of the COMSOL Multiphysics software package, we obtain the amplitude-frequency dependence of bimorph actuator (Fig. 7). The largest oscillation amplitude of the bimorph piezoelectric actuator corresponds to the frequency of 70 Hz and 280 Hz. A visual display of the oscillation form of a bimorph piezoactuator at these frequencies is shown in Fig. 8. Figure 8 shows that the most favourable form of oscillations corresponds to a frequency of 70 Hz. It has only the first mode of bending vibrations, but under certain conditions, vibrations at a frequency of 280 Hz can also be used. Thus, our oscillation frequency search range can be narrowed down to 300 Hz.

Fig. 7. Amplitude-frequency dependence of bimorph piezoelectric actuator

Improving Piezoceramic Artificial Muscles for Flying Insect-Sized Mini Robots

533

Fig. 8. Visual display of oscillations of bimorph piezoelectric actuator: a) at a frequency of 70 Hz; b) at a frequency of 280 Hz

To obtain maximum oscillations of the actuator, it is necessary to determine the rational location of the piezoceramic element on the metal plate. To do this, we will change the distance between the edge of the piezoelectric element and the place of fixing with a step of 5 mm. The results of the study (simulation) are shown in Fig. 9. Figure 9 shows that the largest oscillation amplitude is obtained when the piezoelectric element is located at a distance of 5 mm from the fixed edge of the metal plate. The next step is to change the length of the piezoelectric plate upward in increments of 3 mm, but taking into account the preliminary results (the initial state of the location from the edge of the fixing is 5 mm). The obtained simulation results with increasing the length of the piezoelectric plate are shown in Fig. 10. Figure 10 shows that the largest oscillation amplitude is obtained with a piezoelectric element length of 25 mm at a distance of 5 mm from the edge of the fixed metal plate. Thus, after conducting research with the place of fixation and the geometric dimensions of the piezoelectric element, it is possible to design a piezoelectric actuator for insect-sized mini robot that is lighter and less energy consuming. Therefore, it is possible to reduce the size of the control board, which will also lead to a reduction in the weight of the entire structure. But to get a better result, it is necessary to change the design of the piezoelectric actuator. After carefully analyzing the design of piezo actuators used in insect-sized mini robots, the authors proposed an improved design based on a bimorph cruciform actuator, the model of which is shown in Fig. 11.

534

S. Filimonov et al.

Fig. 9. Dependence of the oscillation amplitude on the distance of the piezoelectric element relative to the fixed edge of the metal plate

Fig. 10. Dependence of the oscillation amplitude on the change in the length of the piezoelectric element relative to the fixed edge of the metal plate

A feature of the proposed design is an increase in the amplitude of oscillations due to the displacement of the fixing point. Thanks to this, we got a kind of lever, in which there are two arms, the ratio of which leads to an increase in the amplitude of oscillations and an increase in mobility. Initial boundary conditions for a bimorph cruciform piezoelectric actuator. The metal plate is connected by a metal jumper, and is free on the other sides. The metal jumper has a rigid fixing on the edge sides. The piezoelectric element is symmetrically located relative to the axis of symmetry of the metal plate, and its bottom side is located at a distance of 5 mm from the lower edge of the metal plate fixing. An input voltage of 100 V is applied to the top and bottom surfaces of the piezoceramic plate. The dimensions of the piezoelectric element are 25 × 4 × 0.1 mm, and the dimensions of the metal plate are 38 × 6 × 0.1 mm. Using the COMSOL Multiphysics software package, the amplitude-frequency dependence of this actuator is obtained (Fig. 12). Figure 12 shows that the largest oscillation amplitude of the bimorph cruciform piezoelectric actuator corresponds to a frequency of 410 Hz, which significantly increases the resistance to external factors, for example, air.

Improving Piezoceramic Artificial Muscles for Flying Insect-Sized Mini Robots

535

Fig. 11. Bimorph cruciform piezoelectric actuator: 1 – metal plate; 2 – metal jumper; 3 – piezoelectric element; 4 and 5 – places of rigid fixing of the metal jumper

Fig. 12. Amplitude-frequency dependence of a bimorph cruciform piezoelectric actuator

A visual display of the oscillation form of a bimorph cruciform piezo actuator is shown in Fig. 13. The obtained results are approximately 7 times higher than the results of the basic bimorph piezoelectric actuator design presented above. To test the proposed design, an experimental sample was developed (Fig. 14). The test results of the experimental sample confirmed the operability of the developed cruciform design.

536

S. Filimonov et al.

Fig. 13. Visual display of the oscillations of bimorph cruciform piezoelectric actuator at a frequency of 410 Hz

Fig. 14. Experimental sample of a bimorph cruciform piezoelectric actuator

5 Conclusions The main disadvantages of artificial muscles (piezoelectric actuators), which are the main elements for lifting into the air, have been established. 3D-dimensional modelling was carried out to determine the most rational location of the piezoelectric element on the metal plate, as well as to determine the rational length of the piezoelectric element at its specific location on the metal plate. A new design of a bimorph cruciform piezoelectric actuator was proposed. The research results showed an increase in the amplitude of vibrations by 7 times compared to the basic design. Thus, the results obtained will reduce energy consumption by about 7 times by reducing the amplitude of the control voltage of artificial muscles based on piezoelectric actuators. In addition, this will reduce the weight of the entire construction and make it more autonomous and lighter. These results accelerate development in this direction. The obtained results can be used not only in designing flying insect-sized mini robots, but also in designing crawling and floating mini robots. Further research is planned to

Improving Piezoceramic Artificial Muscles for Flying Insect-Sized Mini Robots

537

be directed to the creation of a new type of piezoceramic actuator assembly with a new type of force transfer to the wing of flying insect-sized mini robot.

References 1. Sánchez-Bayo, F., Wyckhuys, K.A.G.: Worldwide decline of the entomofauna: a review of its drivers. Biol. Cons. 232, 8–27 (2019). https://doi.org/10.1016/j.biocon.2019.01.020 2. Ahemed, R., Amjad, M.: Automated water managemenrt system (WMS). Int. J. Educ. Manage. Eng. (IJEME) 3, 27–36 (2019). https://doi.org/10.5815/ijeme.2019.03.03 3. Li, J.: Design of active vibration control system for piezoelectric intelligent structures. Int. J. Educ. Manage. Eng. (IJEME) 2(7), 22–28 (2012). https://doi.org/10.5815/ijeme.2012.07.04 4. Abutu, I.M., Imeh, U.J., Abdoulie, T.M.S., et al.: Real time universal scalable wireless sensor network for environmental monitoring application. Int. J. Comput. Netw. Inf. Secur. (IJCNIS) 6, 68–75 (2018). https://doi.org/10.5815/ijcnis.2018.06.07 5. Wood, R.J., Finio, B., Karpelson, M., et al.: Progress on ‘pico’ air vehicles. Int. J. Robot. Res. 31(11), 1292–1302 (2012). https://doi.org/10.1177/0278364912455073 6. Trimmer, W.S.: Microrobots and micromechanical systems. Sens. Actuat. 19(3), 267–287 (1989) 7. Mulgaonkar, Y., Araki, B., Koh J., et al.: The flying monkey: a mesoscale robot that can run, fly, and grasp. In: 2016 IEEE International Conference on Robotics and Automation (ICRA) (2016). https://doi.org/10.1109/ICRA.2016.7487667 8. Kushleyev, A., Mellinger, D., Kumar, V.: Towards a swarm of agile micro quadrotors. In: Proceedings of Robotics: Science and Systems Conference (RSS), Sydney, Australia (2012). https://doi.org/10.15607/RSS.2012.VIII.028 9. Bazilo, C., Filimonov, S., Filimonova, N., Bacherikov, D.: Determination of geometric parameters of piezoceramic plates of bimorph screw linear piezo motor for liquid fertilizer dispenser. In: Hu, Z., Petoukhov, S., Yanovsky, F., He, M. (eds.) ISEM 2021. LNNS, vol. 463, pp. 84–94. Springer, Cham (2022). https://doi.org/10.1007/978-3-031-03877-8_8 10. Spanner, K., Vyshnevskyy, O., Wischnewskiy, W.: New linear ultrasonic micro motors for precision mechatronic systems. In: Proceedings of the 10th International Conference on New Actuators, pp. 439–443 (2006) 11. Sharapov, V.: Piezoceramic sensors. Springer (2011).https://doi.org/10.1007/978-3-642-153 11-2 12. Aladwan, I.M., Bazilo, C., Faure, E.: Modelling and development of multisectional disk piezoelectric transducers for critical application systems. Jordan J. Mech. Indust. Eng. 16(2), 275–282 (2022) 13. Chen, Y., Zhao, H., Mao, J., et al.: Controlled flight of a microrobot powered by soft artificial muscles. Nature 575, 324–329 (2019). https://doi.org/10.1038/s41586-019-1737-7 14. Mam, K.Y., Chirarattananonm, P., Fullerm, S.B., Woodm, R.J.: Controlled flight of a biologically inspired, insect-scale robot. Science 340(6132), 603–607 (2013). https://doi.org/10. 1126/science.1231806 15. Yang, X., Chen, Y., Chang, L., et al. Bee+: A 95-mg four-winged insect-scale flying robot driven by twinned unimorph actuators. IEEE Robot. Autom. Lett. 4270–4277 (2019). https:// doi.org/10.1109/LRA.2019.2931177 16. Jafferis, N.T., Graule, M.A., Wood, R.J.: Non-linear resonance modeling and system design improvements for underactuated flapping-wing vehicles. In: IEEE International Conference on Robotics and Automation (ICRA), pp. 3234−3241. Stockholm, Sweden (2016). https:// doi.org/10.1109/ICRA.2016.7487493

538

S. Filimonov et al.

17. Graule, M.A., Chirarattananon, P., et al.: Perching and takeoff of a robotic insect on overhangs using switchable electrostatic adhesion. Science 352(6288), 978–982 (2016). https://doi.org/ 10.1126/science.aaf1092 18. Chen, Y., Wang, H., Helbling, F., et al.: A biologically inspired, flapping-wing, hybrid aerialaquatic microrobot. Sci. Robot. 2(11), eaao5619 (2017). https://doi.org/10.1126/scirobotics. aao5619 19. Chukewad, Y.M., Singh, A.T., James, J.M., Fuller, S.B.: A new robot fly design that is easier to fabricate and capable of flight and ground locomotion. In: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 4875−4882. Madrid, Spain (2018). https:// doi.org/10.1109/IROS.2018.8593972 20. Fuller, S.B.: Four wings: an insect-sized aerial robot with steering ability and payload capacity for autonomy. IEEE Robot. Autom. Lett. 4(2), 570−577 (2019). https://doi.org/10.1109/LRA. 2019.2891086 21. Chen, Y., Xu, S., Ren, Z., Chirarattananon, P.: Collision resilient insect-scale soft-actuated aerial robots with high agility. IEEE Trans. Rob. 37(5), 1752–1764 (2021). https://doi.org/ 10.1109/TRO.2021.3053647 22. Halchenko, V.Y., Filimonov, S.A., Batrachenko, A.V., Filimonova, N.V.: Increase the efficiency of the linear piezoelectric motor. J. Nano- Electron. Phys. 10(4), 04025 (5pp) (2018). https://doi.org/10.21272/jnep.10(4).04025 23. Sharapov, V.M., Filimonov, S.A., Sotula, Z.V., Bazilo, K.V., Kunitskaya, L.G., Zaika, V.M.: Improvement of piezoceramic scanners. In: 2013 IEEE XXXIII International Scientific Conference Electronics and Nanotechnology (ELNANO), Kiev, pp. 144–146 (2013). https://doi. org/10.1109/ELNANO.2013.6552063

Platform Construction and Structural Design of Passionfruit Picking and Sorting Robot Under the Guidance of TRIZ Caigui Huang1,2(B) 1 School of Intelligent Manufacturing, Nanning University, Guangxi 530200, China

[email protected] 2 Intelligent Manufacturing and Virtual Simulation Research Center of Nanning University,

Nanning University, Guangxi 530200, China

Abstract. The picking and sorting of passionfruit is mainly manual currently, which has problems such as high labor intensity and low efficiency. Aiming at this problem, this paper, guided by TRIZ theory, discusses the principles and methods of solving the invention problem, probes the solution matrix of contradictions and conflicts, analyzes the improvement parameters and deterioration factors, and puts forward the invention principle of solving the design conflict of robot platform. Based on the “material - field” theory of TRIZ, the description model of the passionfruit picking and sorting robot platform is established to identify and resolve the technical conflict, and the passionfruit sorting robot platform is designed, which includes CCD visual recognition system, sorting manipulator body, grab manipulator end actuator, translation module, automatic control system and collection module. According to the principle of the invention, the structure of the robot platform is map out and the end actuator structure is designed to meet the functional requirements of the picking and sorting of passionfruit. Keywords: Passionfruit · Pick · Sorting · Robot · TRIZ

1 Introduction Fruit picking and picking is the most common, basic and necessary operation in the harvest season. Using advanced automation technology and equipment can greatly improve the efficiency of fruit harvest [1–3]. The robot designed to complete these operations highlights its unique advantages [4, 5]. Passion fruit originates from the United States and Brazil, has a strong fragrance and high medicinal value, and is rich in vitamins. China’s main producing areas are Guangxi, Guangdong, Yunnan and other temperate regions [6–8]. Passion fruit is gradually being used as a fruit crop to get rid of poverty and become rich in southern China, but the current planting, grabbing and sorting are basically realized by manual means. At present, there is little research on the mechanical device of Passion fruit at home and abroad, so the mechanical harvesting and sorting of Passion fruit is particularly important at present. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 539–549, 2023. https://doi.org/10.1007/978-3-031-36118-0_48

540

C. Huang

The Theory of Inventive Problem Solving, abbreviated as TRIZ, is derived from the invention problem solving theory summarized by Archishuler and his team from the Soviet Union in 1946 in order to solve millions of patents in the invention problem analysis patent database, and has been recognized as a widely popular topic of innovation in the technical field [9, 10]. TRIZ provides a structured problem-solving method, which overcomes unreliable decision-making methods and finds real direct solutions for complex problems. It is a well-structured and innovative problem-solving method in technical and non-technical fields [11, 12]. The combination application analysis of TRIZ and other tools is also common, such as the integration of TRIZ and axiomatic design [13]. TRIZ theory has been widely used in aerospace science and technology, military science and technology and other fields in the former Soviet Union, which has achieved significant development, solved some scientific research problems, and improved the efficiency of solving problems. After more than half a century of development, through continuous research and practice, it is found that TRIZ theory has obvious effect in solving scientific research problems in many countries, such as Europe and the United States, and is known as an internationally recognized innovation method [14, 15]. Due to the universality of TRIZ theory and the objective existence of scientific principles and rules in the process of solving invention problems, in recent years, TRIZ theory has also been constantly tried to apply to other fields of innovation, such as education, management, medicine, agriculture, social relations, etc. to solve related problems [16, 17]. Compared with the traditional innovation method of brainstorming, TRIZ theory is easier to operate, and it is systematic and process-based in solving problems. However, it depends on the personal experience, knowledge, inspiration, etc. of designers, which is suitable for most people to make bold innovation [18]. As an advanced design method, TRIZ aims to define and overcome some key problems that may affect product development through potential innovative solutions [19]. This paper analyzes the problems of low efficiency, low safety, high labor intensity, and the integrity of the fruits in the sorting process of passionfruit, puts forward the conflict resolution matrix of the platform design of passionfruit sorting robot based on TRIZ theory, and establishes the problem solution of structural design through the proposed invention principle.

2 TRIZ Theory Content and Solution to Invention Problems 2.1 TRIZ Theory Content The main contents of TRIZ theory include 40 invention principles, 39 engineering technical characteristic parameters, and the conflict resolution matrix established by using 39 general parameters and 40 invention principles, as well as the material field analysis theory and the invention problem solving algorithm (ARIZ) [20]. Among them, 40 principles of invention, also known as the conflict resolution principle, were discovered and proposed by Archishuler on the basis of the analysis and research of patents around the world, including 40 principles such as segmentation, refining, local change and asymmetry. In addition, based on this, TRIZ proposed a solution to the separation principle, including the separation of space, time, condition-based, whole and part, and the corresponding relationship with the invention principle is shown in Table 1.

Platform Construction and Structural Design of Passionfruit Picking

541

Table 1. Correspondence between separation principle and invention principle Separation principle

Principle of invention

Spatial separation

1, 2, 3, 4, 7, 13, 17, 24, 26, 30

Time separation

9, 10, 11, 15, 16, 18, 19, 20, 21, 29, 34, 37

Separate parts from whole

12, 28, 31, 32, 35, 36, 38, 39, 40

Conditional separation

1, 7, 25, 27, 5, 22, 23, 33, 6, 8, 14, 25, 35, 13

In TRIZ theory, the conflicting characteristics in different fields are highly summarized and abstracted into 39 technical characteristic parameters, which can uniformly and clearly describe the various conflicts contained in different problems, mainly including 39 engineering technical parameters such as the mass of moving objects, the mass of stationary objects, speed, productivity, etc. In the process of solving system problems, to improve a certain technical feature will often lead to the deterioration of other technical parameters, that is, technical conflict [21]. The conflict resolution matrix establishes the contradiction relationship between the improved technical parameters and the deteriorated technical parameters, and lists the invention principles corresponding to the conflict resolution, as shown in Table 2. Table 2. Conflict Resolution Matrix (partial)

Improved technical parameters

Principle of invention

Deteriorated technical parameters 1

2

3

…

39

1

/

－

15,8,29,34

…

35,3,24,37

2

－

/

－

…

1,28,15,35

3

8, 15 29,34

－

/

…

14,4,28,29

…

….

….

….

…

…

…

…

…

…

…

39

35,26,24,37

28,27,15,3

18,4,28,38

…

/

2.2 Methods to Solve the Invention Problem TRIZ theory can solve the problem of invention. Its core is the principle of technological system evolution and the principle of conflict resolution. It also establishes the method of conflict resolution, and uses the method of general solution to solve the technical conflict and then solve the problem of invention. The method flow of TRIZ to solve the

542

C. Huang

invention problem is shown in Fig. 1, which can be carried out in four steps [22]; In the process of using TRIZ theory to solve the invention problem, the invention problem to be solved should be clarified first, and then the problem to be solved by the object to be designed should be converted into a general problem expression by means of matter-field analysis and other methods, that is, the general model of the problem should be established, and then the general solution of the general problem should be found by using the 40 invention principles and tools provided by TRIZ theory. According to the tips of the general problem-solving methods obtained, and referring to the existing design experience and knowledge, the innovative solutions to the specific problems of the structural design are determined.

Fig. 1. TRIZ’s solution to the invention problem

3 Scheme Design of Robot based on TRIZ Fruit grabbing and sorting plays an important role in agricultural production, with high labor intensity and seasonality, and high cost of grabbing and sorting. The fruit grabbing and sorting robot has been applied earlier in the United States, Japan, the Netherlands and other countries, and its technology maturity is high. At present, it has realized the grabbing and sorting of grapes, oranges, strawberries, cucumbers, tomatoes and other common fruits and vegetables, with great development potential. There are various robot platforms for grasping and sorting, which are mainly based on machine vision for positioning and recognition, and are used for grasping and sorting by motion control manipulators and end actuators [23, 24]. Based on TRIZ theory, this paper designs the overall structure and system of the thymus fruit grabbing and sorting robot platform to meet the various requirements of grabbing and sorting.

Platform Construction and Structural Design of Passionfruit Picking

543

3.1 Analysis of Problems to Be Solved and Establishment of General Problem Model In order to make the design of the grab and sort robot platform more creative and adaptive, this paper uses TRIZ theory to guide the design, but the key is to do a good job in the pre-and post-processing of specific problem analysis. The process is as follows: first, analyze the specific engineering problems to be solved in R&D and design, establish a general problem model according to TRIZ, analyze the improved and deteriorated engineering technical parameters, determine the corresponding invention principle in the technology conflict matrix, and obtain the general solution, and finally determine the best specific solution by combining engineering knowledge and experience. In view of the design of the Passion Fruit Picking and Sorting Robot, it is first necessary to determine the problems and technical conflicts to be solved in the process of picking and sorting mechanization. Passion fruit grabbing and sorting robot should first be able to recognize the fruit maturity and quality, the size and color difference of Passion fruit, etc., which requires the robot to have more extensive adaptability and recognition accuracy. In addition, it is necessary to ensure that the manipulator has sufficient working range during grasping, otherwise it will cause missing picking, and it is required that the manipulator has enough flexibility to work so as to be closer to the fruit target. This requires that the robot platform for grabbing and sorting has enough degrees of freedom. The more degrees of freedom, the higher the flexibility of the working range, and the higher the grabbing efficiency. But this means that the structural design of the grabbing robot platform will be more complex, and the more complex the structural reliability will be reduced, and the design difficulty will increase. In addition, the damage of the end effector in the process of picking and sorting should also be considered, otherwise the appearance quality will be affected. In the process of grasping, the damage of fruit skin should be reduced, which requires that the holding force of the end effector of the grasping robot should not be too large in the process of grasping, and the gripped mechanical finger should not be rigid, or the flexible clamping structure should be added, which increases the overall design difficulty and processing manufacturing difficulty, and the cost will also increase.

Fig. 2. “Material-field” analysis model

The “material field” analysis of TRIZ theory is used to establish the description model of the passionfruit grabbing and sorting robot platform to determine the technical

544

C. Huang

conflict. As shown in Fig. 2, the material-field analysis model shows that the object of grasping and sorting is Passion Fruit. It is assumed that the grasping and sorting robot platform can achieve efficient and rapid grasping and sorting functions, but it also has harmful effects that make the structure of the grasping robot platform (material) more complex, manufacturing difficulties and high costs, which also leads to technical conflicts. 39 general technical parameters are used to analyze and describe the above technical conflicts. The parameters for improvement and deterioration are shown in Table 3. Table 3. Parameter analysis of improvement and deterioration Content

Improved technical parameters Deteriorated technical parameters

1. The structure of the robot platform is simple and stable

Structural stability (13): the Productivity (39): Useful integrity of the system and the value created per unit time relationship between its components

2 The grabbing robot can accurately identify and avoid obstacles

Reliability (27): the ability of the system to complete the required functions in the expected manner and state

Productivity (39): Useful value created per unit time

3. The grasping process does not damage the fruit and is flexible

Harmful factors generated by objects (31): Harmful factors will reduce the efficiency of objects or systems or the quality of completing functions

Device complexity (36): number and diversity of components in the system

4. It has a wide working range Applicability and versatility and good adaptability (35): the ability of an object or system to respond to external changes or to apply under different conditions

Manufacturability (32): The ability to process, manufacture and assemble easily

3.2 Ways Reaching Solutions According to the above technical conflicts, the technical parameters that need to be improved and the deterioration parameters that will occur in the design of the thymus fruit grabbing and sorting robot platform are analyzed, and all the corresponding TRIZ invention principles are listed in the technical conflict matrix in Table 2. Since the method to solve the problem is not unique, the invention principle obtained by referring to the technical conflict matrix table in Table 2 is not single, and the TRIZ invention principle obtained represents the possible direction to solve the problem, not the specific solution, so not all invention principles can solve the problem, which requires us to make specific

Platform Construction and Structural Design of Passionfruit Picking

545

analysis according to the actual situation, Select the most suitable scheme to apply to the design of the grab and sort robot platform structure. According to Table 2, the serial numbers of technical parameters that need to be improved are 13, 27, 31 and 35, and the serial numbers of technical parameters that cause deterioration are 32, 33, 36 and 39. Combined with Table 3, the corresponding invention principle serial numbers in TRIZ’s 40 invention principles can be obtained, and the invention principle serial numbers to solve the problem are shown in Table 4. Table 4. Invention principles for solving problems Improve technical parameters

Deteriorating technical parameters

Solved invention principle

Screening appropriate invention principles

13

39

23, 35, 40, 3

27

33

27, 17, 40

31

36

19, 1, 31

35

32

1， 13, 31

1 (segmentation); 13 (reverse); 17 (dimension change), 23 (feedback), 35 (parameter change), 40 (composite material)

4 Solution for the Design of Thymus Fruit Grabbing and Sorting Robot For the design of the thymus fruit grabbing and sorting robot, this paper mainly describes the design of the overall structure of the grabbing and sorting robot platform, the design of the terminal actuator structure responsible for completing the grabbing action, and the machine vision grabbing and grading control scheme responsible for feedback and adjustment. 4.1 Design of Grab and Sort Robot Platform Structure The principle of “segmentation” is adopted to increase the independence of the object and facilitate disassembly and assembly. Specifically, the robot platform for passionfruit grabbing can be modular designed, and the robot platform structure is simple and stable. The “reverse” principle is adopted to make the moving part of the object stationary and the stationary part moving. Drive and control the robot platform base, manipulator, and end effector separately. If you want to grab and sort, you need to keep one or two of the modules stationary, and then the rest of the modules are controlled by motion to grab and sort. The principle of “dimension change” is adopted to convert one-dimensional space motion into two-dimensional or three-dimensional space motion. The robot platform is designed as a multi-degree of freedom operation platform. In this paper, the robot operation platform is set as a four-degree of freedom to make the robot operation more

546

C. Huang

flexible. The grasping robot can accurately identify and avoid obstacles, and the multidegree of freedom can improve the working range of the manipulator. According to the guidance of the above invention principles, the overall design of the thymus fruit grabbing and sorting robot platform can obtain the overall structure shown in Fig. 3, including the CCD visual recognition system, the main body of the sorting manipulator, the end effector of the grabbing manipulator, the translation module, the automatic control system and the collection module. The main component of CCD visual recognition system is CCD industrial camera, the main body of sorting manipulator includes adjusting arm, motor, gear and bracket, etc., and the translation module includes base, pulley, synchronous belt, motor, connecting frame, guide mechanism, etc. The end grabbing and sorting actuator is installed at the front end of the telescopic arm. At the same time, an industrial camera is installed at this position for fixation. The main body of the sorting manipulator is installed on the mobile platform module.

Fig. 3. Overall design structure of robot platform

4.2 Structure Design of End Actuator Use the principle of “parameter change” to change the flexibility of the object. The joint and contact plate of the gripper finger of the end actuator are made of soft rubber, and the spring is set as a buffer to increase flexibility and solve the problem of fruit damage during grasping. In addition, based on the principle of “composite materials” and different functional requirements, materials with single material are replaced with composite materials. Here, the key stressed finger joints are made of lightweight highstrength aviation materials or aluminum alloy materials. The structure of the end actuator is shown in Fig. 4, including joint, contact plate, transmission mechanism and support part.

Platform Construction and Structural Design of Passionfruit Picking

547

Fig. 4. Structure diagram of end actuator

4.3 Hierarchical Control Scheme of Machine Vision Grabbing During the operation of robots and other automatic equipment, it is necessary to use feedback information as the basis for further adjustment. Use the “feedback” principle to adapt to different needs by adjusting the size and sensitivity of the feedback signal. A CCD visual recognition system is added to the end effector of the manipulator, which can effectively identify the position and forward path of the passionfruit, and feed back to the control center to accurately and effectively carry out the grabbing and sorting action, so as to ensure that the grabbing robot can accurately identify and avoid obstacles. The grab grading system based on machine vision recognition is composed of image acquisition and recognition part, controller, grab actuator, transmission grading and collection part as shown in Fig. 5.

Fig. 5. Workflow of visual grabbing grading scheme

548

C. Huang

First, turn on the light source to adjust the contrast to assist the camera’s recognition, and process the acquired image for contrast recognition, and then control the operation of grabbing and grading.

5 Conclusion In view of the vacancy of thymus fruit grabbing and sorting mechanization, this paper designs the overall structure of thymus fruit grabbing and sorting robot platform based on TRIZ theory. The main work of this paper is as follows: 1) Analyze the specific engineering problems to be solved in the development and design of the thymus fruit grabbing and sorting robot platform, and establish a general problem model according to TRIZ. 2) Analyze 4 improved engineering technical parameters and 4 deteriorated engineering technical parameters in the design process. 3) According to the conflict matrix, the invention principle suitable for the design is obtained. Finally, the invention principle of “segmentation, reverse, dimension change, feedback, parameter change, composite material” is selected to design the overall structure of the passionfruit grabbing and sorting robot. 4) According to the design results, the structure of the passionfruit grabbing and sorting robot platform is complete, which can meet the set functional requirements and meet the design requirements. Acknowledgment. This thesis is supported by (1) the project of the Yongning District Scientific Research and Technological Development Plan of Nanning City “Research on Passion Fruit Harvesting Device Based on Computer Vision Recognition Technology” (20180206A) and (2) Nanning University “Machinery Design and Manufacturing and Automation Professional Certification Construction” (ZYRZ03).

References 1. Wang, Z., Xun, Y., Wang, Y., Yang, Q.: Review of smart robots for fruit and vegetable picking in agriculture. Int. J. Agric. Biol. Eng. 15(1), 33–54 (2022) 2. Bogue, R.: Fruit picking robots: has their time come?. Ind. Robot. 47(2), 141–145 (2020) 3. Dewi, T., Mulya, Z., Risma, P., Oktarina, Y.: BLOB analysis of an automatic vision guided system for a fruit picking and placing robot. Int. J. Comput. Vis. Robot. 11(3), P315–P327 (2021) 4. Polishchuk, M., Opashnianskyi, M., Suyazov, N.: Walking mobile robot of arbitrary orientation. Int. J. Eng. Manufac. 8(3), 1–11 (2018) 5. Marie, M.J., Mahdi, S.S., Tarkan, E.Y.: Intelligent control for a swarm of two wheel mobile robot with presence of external disturbance. Int. J. Mod. Educ. Comput. Sci. 11(11), 7–12 (2019) 6. Fang, Y.: Key points of high-yield cultivation technology of passionfruit. Agric. Develop. Equip. 09, 162–163 (2021). (in Chinese)

Platform Construction and Structural Design of Passionfruit Picking

549

7. Liang, Q., Li, Y., Long, M., et al.: Research progress in chemical constituents and pharmacological activities of passionfruit. Food Indust Sci. Technol. 39(20), 343–347 (2018). (in Chinese) 8. Yao, L., Chen, Y., Zhao, Y., et al.: Cultivation technology of mountain passionfruit hedge. Mod. Agric. Sci. Technol.(18), 86–87+90 (2021). (in Chinese) 9. Mann, D.: An introduction to TRIZ: the theory of inventive problem solving. Creat. Innov. Manage. 10(2), 123–125 (2001) 10. Chaoton, S., Lin, C., Chiang, T.: Systematic improvement in service quality through TRIZ methodology: an exploratory study. Total Qual. Manag. Bus. Excell. 19(3), 223–243 (2008) 11. Sauli, S.A., Ishak, M.R., Mustapha, F., et al.: Hybridization of TRIZ and CAD-analysis at the conceptual design stage. Int. J. Comput. Integrat. Manufac. 32(9), 890–899 (2019) 12. Lin, C., Chaoton, S.: An innovative way to create new services: applying the triz methodology. J. Chin. Inst. Indust. Eng. 24(2), 142–152 (2007) 13. Duflou, J.R., Dewulf, W.: On the complementarity of TRIZ and axiomatic design: from decoupling objective to contradiction identification. Procedia Eng. 9, 633–639 (2011) 14. Ye, C., Wang, M., Zhu, Y.: Innovative design of feeding equipment based on TRIZ theory. Manufac. Technol. Mach. Tool 11, 47–52 (2018). (in Chinese) 15. Akay, D., Demıray, A., Kurt, M.: Collaborative tool for solving human factors problems in the manufacturing environment: the Theory of Inventive Problem Solving Technique (TRIZ) method. Int. J. Prod. Res. 46(11), 2913–2925 (2008) 16. Lau, D.K.: The role of TRIZ as an inventive tool in technology development and integration in China. In: International Conference on the Business of Electronic Product Reliability and Liability, pp. 157–161 (2004) 17. Kim, T.-Y., Kim, J.-H., Park, Y.-T.: Improving the inventive thinking tools using core inventive principles of TRIZ. J. Korean Soc. Qual. Manage. 46(2), 259–268 (2018) 18. Zhang, C., Zhao, Z.: Mechanical Innovation Design (Fourth Edition). Machinery Industry Press, Beijing (2021). (in Chinese) 19. Frizziero, L., Francia, D., Donnici, G., et al.: Sustainable design of open molds with QFD and TRIZ combination. J. Ind. Prod. Eng. 35(1), 21–31 (2018) 20. Wang, L., Chen, G., Zhao, H.: Blueberry grabbing robot design based on TRIZ theory. J. Northeast Forest. Univ. 37(06), 45–47 (2009). (in Chinese) 21. Wang, J., Ge, Z., Zhu, H.: Research on fruit grabbing robot based on TRIZ theory. Agric. Mechan. Res. (06), 34–36+59 (2007). (in Chinese) 22. Donnici, G., Frizziero, L., Francia, D., et al.: TRIZ method for innovation applied to an hoverboard. Cogent Eng. 5(1), 1–24 (2018) 23. Mahfouz, A.A., Aly, A.A., Salem, F.A.: Mechatronics design of a mobile robot system. Int. J. Intell. Syst. Appl. 5(3), 23–36 (2013) 24. Masrour, T., Rhazzaf, M.: A new approach for dynamic parametrization of ant system algorithms. Int. J. Intell. Syst. Appl. 10(6), 1–12 (2018)

Development and Application of Ship Lock System Simulation Modules Library Yuhong Liu(B) and Qiang Zhou College of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China [email protected]

Abstract. System simulation is a crucial technology for complex engineering system, while current simulation systems lack of library for ship lock systems. In this paper, a new library of simulation modules for ship lock systems is proposed. Firstly, based on the object-oriented modeling technology and system simulation method, the composition, association relationship between the subsystems and system configuration of the ship lock logistics system are analyzed, and then the object model and dynamic model of the lock system is established. After that, the universality of the development of ship lock system simulation module library is expounded, and the ship lock system simulation module library is constructed based on the object model and dynamic model. Meanwhile, the structure and usage instructions of the module library are introduced in detail. Finally, the simulation model of Three Gorges Lock is quickly built by using the ship lock system simulation module library and the simulation test is carried out to verifies the engineering application value of the module library. Keywords: Ship lock system · Simulation module library · Quickly built

1 Introduction System simulation technology is one of the most effective tool to analyze complex engineering systems, and attracting much attention of many researchers in recent decade [1, 2]. Many works have used system simulation technology to solve the real problems of discrete event dynamic systems in transportation, mathematics, engineering application [3–5]. Some researchers focus on taking advantage of system simulation on ship lock. As a typical discrete event dynamic system, the lock system has complex characteristics such as nonlinearity and uncertainty, so many scholars emphasis on analyzing the passing ability and reliability of the lock system through simulation, so as to conduct more in-depth research on the ship lock system [6–10]. For instance, Li simulated the hydraulic system of the lock opening and closing machine in Weicun Hub based on AMESim, and puts forward ideas for the detailed design of the subsequent hydraulic system [11]. Chen et al. used Simulink to build a simulation model and found that the bilateral herringbone gate follow-up control system can greatly improve the rapidity of system response without affecting the stability and accuracy of the system [12]. Liu Ying © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 550–559, 2023. https://doi.org/10.1007/978-3-031-36118-0_49

Development and Application of Ship Lock System Simulation

551

et al. established a simulation model of the Three Gorges Lock based on the simulation software SIVAK, analyzed the waiting time of the ship, and studied the lock scheduling strategy [13]. However, when building the simulation model of the lock system, the system boundary, system construction, and logical flow of the lock should be analyzed first, and then the model is built, which requires not only familiarity with the logistic activities and operational operations of the lock, but also proficiency in using different simulation software. Therefore, the modeling period is often long. In view of this, this paper deeply analyzes the complex relationship of the lock system, expounds the development ideas of the system simulation module library with wide applicability, and develops a set of object-oriented lock system simulation module library, which users can use to quickly build the lock model through simple steps such as “dragging in the module “→” establishing logical connection “→” entering basic data” on the simulation platform, and in-depth study of several problems in the lock system.

2 Module Library Development Ideas The object-oriented modeling method is an effective method to study complex problems in complex systems by constructing software systems, in which the object-oriented object modeling technique is a programming-independent graphical representation method. The development of the lock system simulation module library is carried out. 2.1 Object Model of the Ship Lock System The object model of the ship lock system is mainly based on the object modeling. It represents the static structure of the system by describing the objects in the system, the relationships between objects, attributes and describing the attributes and operations of each object class [14]. The main objects of the extracted lock system are: anchorage, dispatch center, approach channel, waiting area, lock door, lock chamber, and filling and emptying system. The specific descriptions are as follows. 1) Anchorage: for ships to wait at anchor, with two states of idle and occupied. 2) Scheduling Center: Scheduling the lock chamber according to the scheduling plan, generating the lock chamber scheduling chart and issuing scheduling instructions to the vessels at berth. 3) Approach channel: Has dimensions and can be occupied or vacant. 4) Waiting area: When the water level adjustment inside the gate is not completed, vessels can wait in the waiting area. 5) Lock door: Open or close. 6) Lock chamber: With dimensions for ships to berth waiting for water level adjustment, vacant or occupied. 7) Filling and emptying system: Regulate the water level elevation inside the gate by watering and sluicing water. The object model of the ship lock system is shown in Fig. 1.

552

Y. Liu and Q. Zhou

Fig. 1. Object model of the ship lock system

2.2 Dynamic Model of the Lock System As a typical complex discrete-event dynamic system, the dynamic model of the ship lock system is mainly used to describe the instantaneous and behavioral control characteristics of the system. The state diagram is used as a descriptive tool to establish the dynamic model of the ship lock system, which represents the state of the object and the change of state between objects (event operation). The main events in the lock system are: ships declaration, ships entering to anchorage, dispatch center generating the lock chamber scheduling chart, ship entering to approach channel, filling and emptying system regulating the water level, lock door opening, lock door closing, ships entering to

Fig. 2. Dynamic model of the ship lock system

Development and Application of Ship Lock System Simulation

553

the lock chamber, and ship leaving the lock. The dynamic model of the ship lock system based on the above events is shown in Fig. 2.

3 Systematic Assumptions for Building Module Libraries Before developing the library of simulation modules for the ship lock system, following assumptions must be made: 1) The boundary of the lock system simulation model is the anchorage of the lock where the ship enters and the end of the channel after the ship leaves the lock chamber. 2) It is assumed that each passing ship is an independent individual, and the ship can run as soon as it enters the simulation without breaking anchor in the middle. 3) Vessels pass through the lock door according to the principle of “priority arrangement for key ships and first-come-first-served for general ships”. 4) The ship is the temporary entity of the simulation system, and it is assumed that the time interval of ship arrival obeys the negative exponential distribution, and the ship type of crossing the lock door is classified according to the size and the quantity obeys a certain probability distribution, and it is assumed that the distribution of ship type is consistent with the distribution of ship deadweight tonnage. 5) When the ship is moored in the lock chamber, there are safety distances between the ships in the longitudinal and transverse directions, which are reflected in the mooring length and berthing width of the lock chamber, that is, the two are calculated by subtracting the safety distance from the original size of the lock chamber.

4 Systematic Construction of the Module Library and Description of the Module Structure 4.1 Systematic Construction of Module Libraries The object model and dynamic model based on the object-oriented modeling method are the important basis for the division of the lock system simulation module library, and the main systems of the lock system simulation module library are: ship plan generation and scheduling module, approach channel module, lock chamber module, data statistics module and public variable module. Among them, the approach channel module is subdivided into the upstream approach channel module and the downstream approach channel module because of the diversity of the lock system; the lock chamber module is subdivided into the upstream lock chamber module, the downstream lock chamber module and the upstream and downstream staggered lock chamber module. The specific components and functions of each module are as follows. 1) Ship plan generation and scheduling module. Components: Vessel, dispatch plan, vessel security check area, anchorage, input and output interfaces. Function: Generate ships, complete the declaration and security check activities respectively, and enter the berth system for berthing; After the scheduling plan is completed according to the registered ship attributes, the dispatching system will issue a lock door crossing command to the ship to realize the scheduling function.

554

Y. Liu and Q. Zhou

2) Approach channel module. Components: approach channel, waiting area, input and output interfaces. Function: The navigation route of the ship from the anchorage to the lock door is specified, and the ship can queue up at the waiting area near the lock door to prepare for entering the lock door. 3) Lock chamber module. Components: locks, chambers, filling and emptying systems, input and output interfaces. Function: The water level inside the lock chamber is adjusted by the filling and emptying system, so that the water level in the two adjacent levels of the lock chamber is equal and the ship completes its passage. At the same time, the ship can moor and wait inside the lock chamber during the water level adjustment. 4) Data statistics module. Components: histograms, pie charts, variables and other tools used to count each indicator. Function: In the simulation process, the data such as the passing capacity of the lock system, ship waiting time, ship crossing time, and the occupancy rate of the lock chamber area are counted, and quantitative data are obtained to analyze the problems studied and find ways to solve them. 5) Public variables module. Components: public variables and attribute variables for each module entity element. Function: As an auxiliary module, it mainly consists of common variables and attribute variables of each module. 4.2 Module Structure and Instructions for Use The main purpose of establishing a library of ship lock system simulation modules is to enable the rapid construction of a ship lock model and to obtain the required statistical parameters for in-depth study of the system. The main structural components of each module in the lock system module library are: module input interface, module logic body and module output interface. Among them, the module input interface is represented in the system module library as the input interface sub-module of the module, and the main component element is the variable, which can be used to receive the basic parameters input by external users, and the user only needs to import the data parameters to the input interface as required after completing the rapid construction of the model with the module library. The module output interface is represented in the system module library as the output interface sub-module of the module. On the one hand, it is used to output statistical parameters to the data statistics module for statistics; on the other hand, it is connected with the module input interface, and the output ship is received by the module input interface and then input to the logical body part of the module for the transfer of entity flow and information flow between models.

Development and Application of Ship Lock System Simulation

555

5 Application Validation of the Module Library In the following, the Three Gorges double-line continuous V-stage locks are used as the application object of the simulation module library, and this module library is used to build the simulation model of Three Gorges locks on the WITENSS simulation platform to verify the usability and convenience of this module library. 5.1 Module Selection and Model Building The Three Gorges locks are double-line continuous five-stage locks, and the basic components of each line of locks are: anchorage, approach channel, lock chamber, lock door, and filling and emptying system. The layout of the Three Gorges Locks system model is built based on the Fig. 3 [15].

Fig. 3. Layout of Three Gorges lock system

5.1.1 Module Selection Each line of locks consists of anchorage, approach channel and 5 levels of lock chambers. The selected basic modules and the required quantities are shown in Table 1. 5.1.2 Model Building On WITNESS simulation platform, drag in modules in the order of public variable module, data statistics module, ship plan generation and scheduling module, upstream approach channel module, upstream first stage lock chamber module, upstream intermediate stage lock chamber module, upstream last stage lock chamber module, then establish the logical relationship between modules, import data according to the data parameters of Three Gorges ship lock system, and complete the simulation model of north line ship lock. The process of building the south line locks is similar. In addition to the infrastructure data of the Three Gorges Locks, the ship crossing scheduling follows the principle of “priority arrangement for VIP ships and first-come-first-served for general ships”, and the ship type distribution is set according to the “Standard Ship Type

556

Y. Liu and Q. Zhou Table 1. Number of base modules

M odule Name

Quantity

Ship plan generation and scheduling module

2

Upstream approach channel module

1

Upstream first stage lock chamber module

1

Upstream intermediate stage lock chamber module

3

Upstream last stage lock chamber module

1

Downstream approach channel module

1

Downstream first stage lock chamber module

1

Downstream Intermediate stage lock chamber M odule

3

Downstream last stage lock chamber module

1

Data Statistics M odule

1

Public Variables M odule

1

Main Scale Series for Inland River Transportation Ships - Yangtze River System, and the interval of the scheduling plan is set to 85 min according to the scheduling plan issued by the Yangtze River Three Gorges Navigation Administration [16]. The screenshot of the established Three Gorges Lock simulation model in WITNESS is shown in Fig. 4.

Fig. 4. Simulation model of Three Gorges lock system

5.2 Model Simulation Verification According to the navigation data statistics released by the Yangtze River Three Gorges Navigation Administration, taking October 2022 as an example, excluding 3.12 h of

Development and Application of Ship Lock System Simulation

557

suspension of the Three Gorges South Line due to high winds, 4.2 h of suspension of the Three Gorges North Line locks due to high winds, and 5 h of suspension of maintenance, the design model simulation time is 739 h, 44340 min. From the following statistical parameters to carry out the validation of the model built with the module library. Table 2. Simulation data statistics of Three Gorges Locks Statistical parameters

Model Simulation Data

Realistic statistics

Number of ships(ships)

3631

3606

Passing capacity (million tons)

1669

1360

Lock door opening times (times)

938

936

From Table 2, we can analyze. 1) Number of ships In terms of the number of ships, the difference between the model simulation data and the real statistics is 0.7%, which indicates that the model simulation is very similar to the actual operation. 2) Passing capacity It can be seen from the table that the cargo volume obtained by the model simulation increased by 22.72% compared with the actual statistics, but this does not explain the incorrectness of the model, because the distribution of ship types set in the model simulation is set according to the standard of large-scale ships, and the load factor is uniformly distributed between (0.7, 1). But in the operation of the Three Gorges Lock, because the freight volume of inland water transportation is unbalanced or there are still more small ships, the actual statistical cargo volume will be less than the model simulation data. 3) Lock door opening times The simulation data of the Three Gorges Lock model is almost the same as the actual reality statistics, and the model is in line with the real situation. The comparison of the above statistical parameters effectively verifies the correctness of the model; moreover, in the process of model building of the Three Gorges locks with the module library, it is time-consuming and efficient, and the user only needs to complete a few fixed steps to quickly model and realize the simulation of the Three Gorges locks, obtain the statistical parameters, and conduct the bottleneck problem study. Therefore, it can be determined that the library of simulation modules for ship lock system based on object-oriented development is not only usable, but also convenient for users to build the required models quickly.

558

Y. Liu and Q. Zhou

6 Conclusion The lock system simulation module library based on object-oriented development can quickly build lock models, reduce the construction time of models required for research, and more systematically and intuitively explore the logistics technology problems in the planning ship lock systems or existing ship lock systems, reveal the mechanism of various complex factors on the passing ability and reliability of lock systems. The module library also has excellent scalability and portability, which can facilitate users to modify the parameters and module structure according to special needs. The simulation application case of Three Gorges Lock shows that this module library has important practical value and will greatly improve the research efficiency of locks. The development of this lock system simulation module library is based on the principle of “priority for key ships and first-come-first-served for general ships” for ship crossing, and in the subsequent research process, different principles of ship crossing can be set to give priority to the utilization rate of the lock chamber area, and the scheduling plan can be made optional, which will make the simulation more complete.

References 1. Ma, G., Xinmei A., He, L.E.I.: The numerical manifold method: a review. Int. J. Comput. Meth. 7(01), 1–32 (2010) 2. Machaˇcek, J., Wichtmann, T., Zachert, H., et al.: Long-term settlements of a ship lock: measurements vs. FE-prediction using a high cycle accumulation model. Comput. Geotechn. 97, 222–232 (2018) 3. Numerical study on hydrodynamic interaction between a berthed ship and a ship passing through a lock. Ocean Eng. 88, 409–425 (2014). (in Chinese) 4. Smith, L.D., Sweeney, D.C.: Campbell. Simulation of alternative approaches to relieving congestion at locks in a river transportion system. J. Oper. Res. Soc. (2009) 5. Richter, J., Verwilligen, J., Dilip Reddy, P., et al.: Analysis of full ship types in high-blockage lock configurations. Masin 2012, 1–9 (2012) 6. Shang, J., Liu, C., Tang, Y., Guo, Z.: Review of lock passing capacity. Water Transp. Eng. 07, 103–108 (2018). (in Chinese) 7. Hu, X.: Study on the passing capacity of cruise ship locks under the Minjiang River. Southwest Jiaotong University, Chengdu (2011). (in Chinese) 8. Huang, H., Zhang, W., Li, X.: Research on the passing capacity of Beijing-Hangzhou Canal locks based on queuing theory. J. Wuhan Univ. Technol. (Transp. Sci. Eng.) 33(03), 604–607 (2009). (in Chinese) 9. Zhang, W., Gu, D., Wang, Q.: Navigation analysis and simulation study of Yangzhou section of Yanshao Line. Water Transp. Eng. (05), 122–127 (2015). (in Chinese) 10. Liu, Z.: Reliability analysis of herringbone gate lock hydraulic system of Subei Canal. China Water Transp. (Second Half Month) 18(11), 85–86+89 (2018). (in Chinese) 11. Li, J.: Design and simulation of hydraulic system of lock opening and closing machine based on AMESim. China Water Transp. 08, 72–74 (2022). (in Chinese) 12. Chen, K., Gao, S., Chen, X.: Simulation study on follow-up control of double-sided herringbone gates of Three Gorges Locks. Yangtze River 52(08), 235–238 (2021). (in Chinese) 13. Liu, Y., Mou, J.: Simulation of passage capability of Three Gorges Lock based on SIVAK. J. Dalian Maritime Univ. 41(04), 37–41 (2015). (in Chinese)

Development and Application of Ship Lock System Simulation

559

14. Zhang, H., Liu, X.: Research on object-oriented object modeling technology and its application. Software 03, 66–68 (2011). (in Chinese) 15. Cheng, M.: Three gorges ship lock capacity evaluation and impact factor research. Chongqing Jiaotong university, Chongqing (2019). (in Chinese) 16. JTS 196-6-2012. Technical regulations for navigation scheduling of Three Gorges Locks. Industry standard – transportation, Wuhan (2012) 17. Yin, Y.: Study on main scale series of standard ship types of typical inland waterway transportation vessels. China Maritime 05, 33–35 (2021). (in Chinese)

Hydrodynamics of Inhomogeneous Jet-Pulsating Fluidization Bogdan Korniyenko(B) , Yaroslav Kornienko, Serhii Haidai, and Andrii Liubeka National Technical University of Ukraine «Igor Sikorsky Kyiv Polytechnic Institute», Kyiv 03056, Ukraine [email protected]

Abstract. The method of creating inhomogeneous (non-uniform) fluidization in the self-oscillating (auto-oscillating) mode by using a special gas distribution device (GDD) with a slit-type construction and a granulator chamber is presented. The impact of the design parameters of the gas distribution device on the quality of the hydrodynamics of the jet-pulsating mode of fluidization was determined, due to the reduction of the risk of formation of stagnant zones on the horizontal working surface of the gas distributing device for the case when the height of the vertical gas jet is three or more times less than the initial height of bed of solid particles (granules) in the granulator chamber. This condition follows from the determination of the heat and mass exchange surface during the granulation process of heterogeneous liquid systems in a fluidized bed. Keywords: Granulation · Fluidization · Inhomogeneous Fluidization · Jet-Pulsating Mode

1 Introduction To obtain granulated organic-mineral fertilizers of a new generation that contain nutrients of mineral and organic origin, deoxidizing and stimulating impurities it is most appropriate to use fluidization technique, which provides obtaining a granulated product with a uniform distribution of components throughout the volume of granule with a heat utilization coefficient of more than 50% [1–4]. For further intensification of transfer processes, it is proposed by the authors [5–7] to carry out the processes of dehydration and granulation with inhomogeneous fluidization. In the works [8–12] was carried out a study of the non-uniform (inhomogeneous) jet-pulsating mode of fluidization at the ratio of the gas jet breakdown height zf to the height of the initial stationary bed of solid granulated material H 0 in zf /H 0 = 0.33. The use of such a fluidization mode during the granulation of a liquid heterogeneous systems that based on an ammonium sulfate solution with sunflower ash impurities made it possible to obtain granulated organic-mineral fertilizers of the composition [Humates]:[K]:[S]:[N]:[Ca]:[Mg]:[P] = 1.5:21.5:13.8:9.1:4.6:3.2:1.8 with a layered structure [13–15]. At the same time, the granulation coefficient was ψ ≥ 90% and the © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 560–573, 2023. https://doi.org/10.1007/978-3-031-36118-0_50

Hydrodynamics of Inhomogeneous Jet-Pulsating Fluidization

561

intensity of moisture removal from a unit surface of the bed of solids was 1.5 times greater than in the bubbling mode [16]. An increase in the productivity of the device based on evaporated moisture while preserving the layer-by-layer granulation mechanism is associated with an increase in the total surface of the granular material, which is accompanied by an increase in the total height of the bed of solids. The purpose of the article is to determine the influence of the design parameters of the gas distribution device (GDD) on the intensity of movement of solid granular material along the working surface of the gas distribution device without the formation of stagnant zones.

2 Analysis of Scientific Papers In the works [17–19] it is proposed to use a pulsating supplying of the heat carrier when processing heat-resistant materials. In particular, the authors [20] carried out the pulsating mode of supplying the heat carrier to the fluidized bed is using a mechanical pulsator, that is, the reliability of the device depends on the functioning of the mechanical pulsator operating in the environment with the high-temperature heat carrier. In this case, there is a time interval in the working cycle when the supplying of the liquefying agent to the device is completely stopped. This leads to the formation of stagnant (low-moving) zones on the working surface of gas distribution device (GDD). This increases the risk of melting of the material during the granulation of heat-labile substances when the heat carrier is supplied, the temperature of which exceeds the melting point temperature of the components of solid particles. To eliminate this disadvantage by the authors [21] is proposed a method of inhomogeneous (non-uniform) jet-pulsating fluidization by using of original construction of the gas distribution device (GDD). An innovative method of interaction between the gaseous heat carrier and the solid granular material is proposed, in which the self-oscillating jet-pulsating mode of fluidization is realized without the formation of stagnant zones on the working surfaces of the gas distribution device, which contributes to the intensification of diffusion-controlled processes. For this, two jets in orthogonal planes are injected into the granulator chamber through the slit type gas distribution device. The basic jet is injected horizontally along the curved surface of the gas distributing device (GDD) and vertical jet (fountaining jet) is injected at a certain distance. The speed of the horizontal jet is selected in such a way that its merging with the vertical jet can be achieved. During the dehydration and granulation processes, the total surface of the bed of solids is determined from the conditions of mass transfer, therefore the height of the initial bed of solid granular material H 0 is three times greater than the height of the breakdown height of the vertical jet zf . Therefore, a gas bubble intensively forms on the upper part of the jet, which, after reaching a critical size, begins to rapidly move vertically upwards. The potential energy accumulated by the bubble causes the emission of solid granular material into the space above bed of solids, which, thanks to the guiding insert, is intensively moved to the left part of the chamber of the apparatus in the area affected

562

B. Korniyenko et al.

by the base slit. This leads to an instant increase in the hydraulic resistance of the gas ejecting from the horizontal slit, which is accompanied by a decrease in the speed of the dispersed system. Such design of gas distributing device (GDD) in combination with structural changes in the granulator chamber ensures the realization of inhomogeneous (non-uniform) jetpulsating fluidization, which under certain conditions goes into the self-oscillating mode of fluidization [14–20]. In works [9–13] it is proposed to evaluate the quality of inhomogeneous jet-pulsating fluidization in the absence of stagnant zones on working surfaces of gas distributing device (GDD) with continuous supplying of the fluidizing agent. The physical model of hydrodynamics in the zone of gas distributing device (GDD) is shown in Fig. 1 [7–9].

Fig. 1. Physical model of inhomogeneous fluidization [19], a) general scheme of the granulator chamber b) features of hydrodynamics in zone D, 1 – granulator chamber; 2 – slit-type GDD; 3 – guiding insert; 4 – mechanical disperser; 5 – slow moving granules

The peculiarity of the hydrodynamic mode of fluidization is that the liquefying agent is injected into the chamber of the granulator with gas distributing device (GDD) of slit type 2 through two slits, at points p and k, Fig. 1, in the horizontal (m1 ) and vertical (m2 ) directions. The distance between the slits t is determined by the horizontal range of the gas jet hhor , and the shape of the working surface of the gas distributing device (GDD) plate repeats the shape of the gas jet, which determines the need for the location of the second slit at a height  relative to the first slit. At the point k, the two jets merge, which leads to the formation of a combined jet with the breakdown height of zf . Conventional planes drawn through points p and k divide the apparatus chamber into three zones of equal width A/3. It was established that the height of the initial fixed bed of solids H 0 = 0.32 m is three times greater than the height of the breakdown of the gas jet zf /H 0 ≤ 0.33. Therefore, a gas bubble begins to form cyclically at the top of the jet, which, upon exiting the bed of

Hydrodynamics of Inhomogeneous Jet-Pulsating Fluidization

563

solids, causes the inertial removal of solid granular material into the space above bed of solids of zones II and III. After contact with the guiding insert 3, solid particles move to zone I, after which they quickly return to the initial volume of the bed of solids. In the case when the energy of the horizontal gas jet coming out of the slit in (point p) and moving horizontally over the working surface of the gas distributing device (GDD) is insufficient, stagnation of granular material is formed in zone D.    2 [δ] − δ(i) 2 , (1) LD = K1 [εD ] − εD(i) + K2 l where K1 and K2 – are the coefficients of proportionality (K1 = 0.3 and K2 = 0.7); εD(i) – experimentally determined the current value of porosity in zone D; [δ] = 0.01l, m; l – chord length of GDD plate, m. Such a hydrodynamic mode is considered qualitative when the coefficient of quality loss L D ≤ 0.1, and ideally – L D → 0. Taking into account the cyclic nature of the jet-pulsating mode of fluidization, it is proposed to evaluate the quality according to the dynamic quality index: iquality = τquality /τcycle

(2)

where τ quality is the time during which the quality loss function L D ≤ 0.1, s; τ cycle – duration of one cycle, s.

3 Experimental Setup To determine the effect of the height of the initial fixed bed of solids (granular material) on the qualitative indicators of hydrodynamics was used a specially developed method of providing the experiment [1–6].

Fig. 2. Schematic representation of the experimental setup, 1 – fluidized bed granulator chamber; 2 – slit type gas distribution device (GDD); 3 – guiding insert; 4 – elastic mesh-bumper; 5 – cyclone; 6 – container for collecting dust; 7 – gas blower; 8 – chamber diaphragm; 9 – differential manometers with pressure drop sensors; 10 – video camera; 11 – weights.

564

B. Korniyenko et al.

Studies of the hydrodynamic mode of fluidization were carried out on a pilot plant with the dimensions of the granulator chamber A × B × H = 0.3 × 0.11 × 1.5 m, Fig. 2. The slit-type gas distribution device is located in the lower part of the granulator chamber [9], and the upper one has a guiding insert and an elastic mesh-bumper. The cross-section coefficient of the gas distributing device (GDD) ϕ varies from 3.2% to 4%. For video and photo analysis, the front wall of the granulator chamber is transparent. The pressure drop in the bed of solids was measured using pressure drop sensors with an accuracy of ±0.1 Pa.

4 Materials To determine the coefficients of hydraulic resistance, was used a two-slit gas distributing device (GDD) with two values of the cross-section coefficient of GDD ϕ = 4% and ϕ = 3.2%. As granular material was used granulated ammonium sulfate with humate impurities with equivalent diameter d e = 2.07 mm, density ρ solids = 1450 kg/m3 . The mass of the bed of solid granules loaded into the apparatus chamber varied from 11.32 to 17.46 kg, which determined the height of the bed of solids H 0 (0.42, 0.52 and 0.6 m) and the nominal hydrostatic pressure Pnom. With porosity ε0 = 0.4, the height of the injection of the first jet is shifted by  = 40 mm.

5 Results and Discussion On Fig. 3 is given the comparison of hydraulic resistance coefficients of gas distributing device (GDD) when the cross-section coefficient of GDD ϕ changes from 3.2 to 4.9%, where A = wslits ρ solids /2. The obtained results indicate that the reducing of the ratio of the cross-section coefficient ϕ by 34.6% leads to an adequate increase in hydraulic resistance value of gas distributing device (GDD) by 32.4%, Fig. 3.

Fig. 3. Determination of hydraulic resistance coefficient of gas distributing device (GDD)

The fluidization curve was taken for the initial height of the bed of solids H 0(1) = 0.42 m at the GDD cross-section coefficient ϕ = 4%, Fig. 4, where waverage is the average

Hydrodynamics of Inhomogeneous Jet-Pulsating Fluidization

565

velocity of fluidizing agent in granulator chamber, m/s, wslits is velocity of gas flow in the slits of GDD, m/s, K w is the fluidization number (K w = waverage /wfluidization ), wfluidization is the average velocity of fluidizing agent at which fluidization begins, m/s, P nominal is the nominal hydraulic resistance of fluidized bed equal to the hydrostatic pressure of the fixed bed of solids, Pa. Jet breakdown height ratio is zf /H 0 = 0.1/0.42 = 0.24. The obtained fluidization curve can be conditionally divided into four zones. In zone 1, when the fluidization number is 0 ≤ K w < 0.54, gas movement in the apparatus occurs in filtration mode.

Fig. 4. The fluidization curve (H 0(1) = 0.42 m; wfluidization = 0.735 m/s; ϕ = 4%)

In zone 2 at values of the fluidization number 0.54 ≤ K w < 1.0 occurs the breakdown of gas jets that come out of two slits and merge into one with an upward motion, as a result of which a pulsation mode is realized in the apparatus. In zone 3 at values of the fluidization number is within 1.0 ≤ K w < 1.71 the system operates in the bubbling mode of fluidization, which is characterized by an insufficient volumetric mixing index and the formation of stagnant zones on the GDD surface. In zone 4 where K w ≥ 1.71 and the speed in the GDD slits is 31.4 m/s the device implements a high-quality self-oscillating jet-pulsating mode of fluidization, which ensures the absence of stagnant zones on the working surface of GDD. Photo fixation of cycle of the self-oscillating jet-pulsating fluidization in the granulator chamber is given on Fig. 5. After analyzing the obtained cyclogram was determined the dynamics of changes in the porosity of the bed of solids in the zone D, Fig. 6a, and was established that during the entire pulsation cycleεD > 0.85, and the quality loss function L D < 0.1, Fig. 6b. That is, with the value of K w = 1.71, the index of the dynamic quality of hydrodynamics is iquality → 1 while the gas velocity in the GDD slits is wslits = 31.4 m/s and the ratio of the gas jet breakdown height to the initial height of the granular material bed zf /H 0 = 0.24.

566

B. Korniyenko et al.

Fig. 5. Photo fixation of the state of the bed of solid granular material in the granulator chamber (H 0(1) = 0.42 m; K w = 1.71; wslits = 31.4 m/s; f = 2 Hz; τ cycle = 1/f = 0.5 s)

Fig. 6. Evaluation of the hydrodynamics quality in zone D (H 0(1) = 0.42 m; ϕ = 4.0%; f = 2 Hz; τ cycle = 1/f = 0.5 s)

Increasing the initial height of the bed of solids by 23% to H 0(2) = 0.52 mm led to a decrease in the index of dynamic quality of hydrodynamics since zf /H 0 = 0.1/0.52 = 0.19. In order to ensure iquality → 1 when increasing the initial height of the bed of solids there was installed a gas distributing device (GDD) with a cross-section coefficient ϕ = 3.2% in the apparatus, which ensured an increase in the breakdown height of the gas jet zf to zf = 0.12 m, then zf /H 0 = 0.12/0.52 = 0.23. For this case, the fluidization curve was taken for the initial height of the bed of solids H 0(2) = 0.52 m, Fig. 7.

Hydrodynamics of Inhomogeneous Jet-Pulsating Fluidization

567

Fig. 7. The fluidization curve (H 0(2) = 0.52 m; wfluidization = 0.735 m/s; ϕ = 3.2%)

Fig. 8. Photo fixation of the state of the bed of solid granular material in the granulator chamber (H 0(2) = 0.52 m; K w = 1.71; waverage = 1.26 m/s; wslits = 38.7 m/s; f = 2.5 Hz; τ cycle = 1/f = 0.4 s)

Photo fixations of the state of the bed of solids for the self-oscillating jet-pulsating mode of fluidization for a bed of solids with H 0(2) = 0.52 m, that was achieved at K w = 1.71 and wslits = 38.73 m/s are given on Fig. 8. The dynamics of changes in the porosity of the bed of solids in zone D and the the quality loss function of hydrodynamics are shown in Fig. 9. From the graphs it follows that at the values of K w = 1.71 and of gas velocity in the GDD slits wslits = 38.73 m/s a high-quality self-oscillating jet-pulsating mode of fluidization is ensured during the entire cycle time τ quality = τ cycle . The porosity in the

568

B. Korniyenko et al.

Fig. 9. Evaluation of the hydrodynamics quality in zone D (H 0(2) = 0.52 m; ϕ = 3.2%; f = 2.5 Hz; τ cycle = 1/f = 0.4 s)

zone D is εD > 0.85 and the quality loss function L D < 0.1, iquality → 1, which eliminates the risk of formation of stagnant zones on the GDD surface.

Fig. 10. The fluidization curve (H 0(3) = 0.6 m; wfluidization = 0.735 m/s; ϕ = 3.2%)

Similarly, the fluidization curve was determined with a further increase in the height of the fixed bed of solids by 15% to H 0(3) = 0.6 m. While maintaining the crosssection coefficient of GDD ϕ = 3.2% was taken the fluidization curve, Fig. 10, with photo fixations of the state of the bed of solid granular material in the high-quality self-oscillating mode of fluidization, Fig. 11. The dynamics of changes in the porosity of the bed of solid granular material in the zone D corresponds to the conditions of the qualitative hydrodynamic mode εD > 0.85 and the quality loss function L D < 0.1, Fig. 12, and so iquality → 1.

Hydrodynamics of Inhomogeneous Jet-Pulsating Fluidization

569

Fig. 11. Photo fixation of the state of the bed of solid granular material in the granulator chamber (H 0(3) = 0.6 m; K w = 1.71; waverage = 1.26 m/s; wslits = 38.73 m/s; f = 2 Hz; τ cycle = 1/f = 0.5 s)

Fig. 12. Evaluation of the hydrodynamics quality in zone D (H 0(3) = 0.6 m; ϕ = 3.2%; f = 2 Hz; τ cycle = 1/f = 0.5 s)

Based on the results of the researches, it was found that reducing the live crosssection coefficient of GDD allows increasing of the gas velocity in the GDD slits and increasing the kinetic energy of the gas jet, which ensures a high-quality hydrodynamic mode in the zone D (iquality = 1.0) at the ratio zf /H 0 = 0.12/0.6 = 0.2. A decrease in the cross-section coefficient of GDD contributed to an increase in the height of the breakdown of the gas jet, Fig. 13a. At the same time, the diameter of the gas bubble Dbubble increases, Fig. 13b, as a result of which the inertial removal of a bed of solid granular material increases, Fig. 14, as a result of which stable kinetics of the granulation process is not ensured.

570

B. Korniyenko et al.

Fig. 13. Determination of the influence of the cross-section coefficient of GDD on the geometric dimensions of the vertical gas jet and the diameter of gas bubble

Impulse removal of granular material into the space above bed of solids causes inertial removal of granular material from the apparatus. To eliminate this disadvantage, it was proposed to install an elastic bumper in the separation zone of the granulator. The results of the experiments are shown on Fig. 14 and a comparison of this data is given Table 1.

Fig. 14. Dependence of the mass of removed material mblowout from the average speed of the gas flow in the apparatus waverage

Thus, from the Table 1 it follows that the expediency of using an elastic mesh bumper was confirmed, as it made it possible to eliminate inertial removal of solids at the initial height of the bed of granular material of 0.37 m, and to reduce it from 94.28 kg/h to 6.24 kg/h i.e. by 15.1 times for the maximum investigated bed of solids height.

Hydrodynamics of Inhomogeneous Jet-Pulsating Fluidization

571

Table 1. Comparison of data of the mass of removed material at different values of the initial height of bed with and without mesh bumper waverage , mblowout , kg/h m/s H 0 = 0.37 m Mesh bumper without with

H 0 = 0.42 m 

Mesh bumper

H 0 = 0.52 m 

without with

Mesh bumper

H 0 = 0.6 m 

without with

Mesh bumper



without with

0.77

0.006

0.005 0.001 0.012

0.008 0.004

0.101

0.019

0.081

1.03

0.037

0.99

0.91

0.012

0.010 0.001 0.022

0.016 0.006

0.34

0.056

0.285

2.6

0.095

2.46

0.97

0.018

0.014 0.004 0.024

0.021 0.003

0.58

0.088

0.488

3.8

0.143

3.63

1.05

0.032

0.019 0.012 0.045

0.032 0.013

1.16

0.162

0.997

6.3

0.246

6.08

1.09

0.042

0.024 0.018 0.060

0.038 0.022

1.64

0.219

1.424

8.2

0.323

7.87

1.15

0.064

0.032 0.032 0.095

0.051 0.044

2.77

0.344

2.429 12.1

0.485 11.59

1.20

0.090

0.040 0.050 0.139

0.065 0.074

4.29

0.501

3.787 16.7

0.681 16.00

1.26

0.137

0.054 0.084 0.219

0.087 0.132

7.24

0.790

6.449 24.6

1.023 23.58

1.30

0.182

0.065 0.117 0.297

0.106 0.191 10.26

1.070

9.193 31.9

1.341 30.52

1.40

0.366

0.104 0.263 0.635

0.172 0.463 24.54

2.260 22.278 60.8

2.642 58.20

1.49

0.688

0.158 0.530 1.258

0.267 0.991 53.81

4.460 49.347 94.28

6.242 88.04

1.59

1.387

0.253 1.134 2.688

0.434 2.255

1.71

3.217

0.445 2.772 6.685

0.776 5.908

6 Summary and Conclusion Thus, it was experimentally confirmed that the implementation of jet-pulsating inhomogeneous fluidization in the self-oscillating mode without the formation of stagnant zones on the working surfaces of the gas distributing device (GDD) can be is provided by increasing the initial height of the stationary bed of solids H 0 by 42% by reducing the live cross-section coefficient of the gas distributing device. When ϕ decreases, the height of the breakdown of the gas jet increases but the diameter of the bubble increases and this causes an inertial drift of granular material, which can be eliminated by using an elastic mesh bumper. However, the diameter of the formed bubble occupies more than a third of the device’s width, Dbubble /A > 1/3, which makes it impossible to ensure the necessary volumetric mixing of the bed of solids in the middle of the apparatus. To eliminate this disadvantage, it is necessary to provide a corresponding increase in the length of the device.

References 1. Kornienko, Y.M., Haidai, S.S., Sachok, R.V., Liubeka, A.M., Korniyenko, B.Y.: Increasing of the heat and mass transfer processes efficiency with the application of non-uniform fluidization. ARPN J. Eng. Appl. Sci. 15(7), 890–900 (2020) 2. Korniyenko, B., Kornienko, Y., Haidai, S., Liubeka, A., Huliienko, S.: Conditions of nonuniform fluidization in an auto-oscillating mode. In: Hu, Z., Petoukhov, S., Yanovsky, F., He, M. (eds.) ISEM 2021. LNNS, vol. 463, pp. 14–27. Springer, Cham (2022). https://doi.org/10. 1007/978-3-031-03877-8_2

572

B. Korniyenko et al.

3. Korniyenko, B., Kornienko, Y., Haidai, S., Liubeka, A.: The heat exchange in the process of granulation with non-uniform fluidization. In: Hu, Z., Petoukhov, S., Yanovsky, F., He, M. (eds.) ISEM 2021. LNNS, vol. 463, pp. 28–37. Springer, Cham (2022). https://doi.org/10. 1007/978-3-031-03877-8_3 4. Tuponogov, V., Rizhkov, A., Baskakov, A., Obozhin, O.: Relaxation auto-oscillations in a fluidized bed. Thermophys. Aeromech. 15(4), 603–616 (2008) 5. Shevchenko, Y.M., Kornienko, Y.M., Haidai, S.S., Denisenko, V.R.: Gas distributing device of the apparatus of fluidized bed. Patent UA 136196 U Ukraine, IPC B01J 8/44, 18 February 2019, Bulletin № 15 (Ukr.) 6. Kornienko, Ya.N., Podmogilnyi, N.V., Silvestrov, A.N., Khotyachuk, R.F.: Current control of product granulometric composition in apparatus with fluidized layer. J. Autom. Inform. Sci. 31(12), 97–106 (1999) 7. Korniyenko, B., Ladieva, L.: Mathematical modeling dynamics of the process dehydration and granulation in the fluidized bed. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2020. AISC, vol. 1247, pp. 18–30. Springer, Cham (2021). https://doi.org/10. 1007/978-3-030-55506-1_2 8. Korniyenko, B., Ladieva, L., Galata, L. Control system for the production of mineral fertilizers in a granulator with a fluidized bed. In: ATIT 2020 - Proceedings: 2020 2nd IEEE International Conference on Advanced Trends in Information Theory, № 9349344, pp. 307–310 (2020). https://doi.org/10.1109/ATIT50783.2020.9349344 9. Korniyenko, B., Galata, L., Ladieva, L.: Research of Information Protection System of Corporate Network Based on GNS3. In: 2019 IEEE International Conference on Advanced Trends in Information Theory, ATIT 2019 - Proceedings, № 9030472, pp. 244–248 (2019). https:// doi.org/10.1109/ATIT49449.2019.9030472 10. Korniyenko, B., Galata, L., Ladieva, L.: Mathematical model of threats resistance in the critical information resources protection system. CEUR Workshop Proc. 2577, 281–291 (2019) 11. Kornienko, Y.M., Liubeka, A.M., Sachok, R.V., Korniyenko, B.Y.: Modeling of heat exchangement in fluidized bed with mechanical liquid distribution. ARPN J. Eng. Appl. Sci. 14(12), 2203–2210 (2019) 12. Korniyenko, B.Y., Ladieva, L.R., Galata, L.P.: Mathematical model of heat transfer process of production of granulated fertilizers in fluidized bed. ARPN J. Eng. Appl. Sci. 16(20), 2126–2131 (2021) 13. Korniyenko, B., Ladieva, L., Galata, L., Nesteruk, A., Matviichuk-Yudina, O.: Information control system for the production of mineral fertilizers in the granulator with a fluidized bed. In: 2021 IEEE 3rd International Conference on Advanced Trends in Information Theory, ATIT 2021 - Proceedings, pp. 250–255 (2021) 14. Korniyenko, B., Zabolotnyi, V., Galata, L.: The optimization of the critical resource protection system of a mineral fertilizers manufacturing facility. In: Proceedings of the 11th IEEE International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications, IDAACS 2021, vol. 1, pp. 172–178 (2021) 15. Kravets, P., Shymkovych, V.: Hardware Implementation Neural Network Controller on FPGA for Stability Ball on the Platform 2nd International Conference on Computer Science, Engineering and Education Applications, ICCSEEA 2019, Kiev, Ukraine, 26 January 2019 - 27 January 2019 (Conference Paper), vol. 938, pp. 247–256 (2019) 16. Malekzadeh, M., Khosravi, A., Noei, A.R., Ghaderi, R.: Application of adaptive neural network observer in chaotic systems. Int. J. Intell. Syst. Appl. 6(2), 37–43 (2014). https://doi. org/10.5815/ijisa.2014.02.05 17. Bhagawati, K., Bhagawati, R., Jini, D.: Intelligence and its application in agriculture: techniques to deal with variations and uncertainties. Int. J. Intell. Syst. Appl. 8(9), 56–61 (2016). https://doi.org/10.5815/ijisa.2016.09.07

Hydrodynamics of Inhomogeneous Jet-Pulsating Fluidization

573

18. Wang, W., Cui, L., Li, Z.: Theoretical design and computational fluid dynamic analysis of projectile intake. Int. J. Intell. Syst. Appl. 3(5), 56–63 (2011). https://doi.org/10.5815/ijisa. 2011.05.08 19. Patnaik, P., Das, D.P., Mishra, S.K.: Adaptive inverse model of nonlinear systems. Int. J. Intell. Syst. Appl. 7(5), 40–47 (2015). https://doi.org/10.5815/ijisa.2015.05.06 20. Babak, V.P., Babak, S.V., Myslovych, M.V., Zaporozhets, A.O., Zvaritch, V.M.: Methods and models for information data analysis. In: Diagnostic Systems For Energy Equipments. SSDC, vol. 281, pp. 23–70. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-44443-3_2 21. Polishchuk, M., Tkach, M., Parkhomey, I., Boiko, J., Eromenko, O.: Experimental studies on the reactive thrust of the mobile robot of arbitrary orientation. Indonesian J. Electric. Eng. Inform. 8(2), 340–352 (2020)

Approaches to Improving the Accuracy of Machine Learning Models in Requirements Elicitation Techniques Selection Denys Gobov1(B) and Olga Solovei2 1 National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Kyiv,

Ukraine [email protected] 2 Kyiv National University of Construction and Architecture, Kyiv, Ukraine [email protected]

Abstract. Selecting techniques is a crucial element of the business analysis approach planning in IT projects. Particular attention is paid to the choice of techniques for requirements elicitation. One of the promising methods for selecting techniques is using machine learning algorithms trained on the practitioners’ experience considering different projects’ contexts. The effectiveness of ML models is significantly affected by the balance of the training dataset, which is violated in the case of popular techniques. The paper aims to analyze the efficiency of the Synthetic Minority Over-sampling Technique usage in Machine Learning models for elicitation technique selection in case of the imbalanced training dataset and possible ways for positive feature importance selection. The computational experiment results confirmed the effectiveness of using the proposed approaches to improve the accuracy of machine learning models for selecting requirements elicitation techniques. Proposed approaches can be used to build Machine Learning models for business analysis activities planning in IT projects. Keywords: Requirement elicitation technique · machine learning · decision tree · over-sampling technique · binary classification problem

1 Introduction The choice of techniques for effectively identifying requirements in developing IT solutions is essential in planning business analysis work. A thorough understanding of the variety of techniques available, their advantages and disadvantages assists the business analyst in adapting to a particular project context [1]. The business analyst must create a combination of techniques to guarantee the success of software requirements identification activities, as it is impossible to fulfill all project stakeholders’ needs using just one technique [2]. One approach to solving this problem is to use machine learning models, where training samples are formed based on practitioners’ experience or recommendations. For example, in studies [3, 4], a machine learning model was built that recommends the usage of elicitation techniques depending on the combinations of © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 574–584, 2023. https://doi.org/10.1007/978-3-031-36118-0_51

Approaches to Improving the Accuracy of Machine Learning Models

575

factors. However, the use of machine learning models for commonly used techniques and the Accuracy of their work is associated with several difficulties. If we select the most frequently used techniques as a target class, the gathered dataset got imbalanced, i.e., most observations belong to a target class with a positive value equal to “1,” which indicates that the technique was used in the project. The mentioned model from [3] was constructed with Decision Jungle Tree (DJT) algorithm that was empirically selected as the most efficient. DJT learner, like a binary decision tree learner, creates a tree that is biased to the majority class when a dataset is imbalanced [5]. The reason causes are the following: while recursively partitioning the dataset so that the observations with similar target values are grouped together, it qualifies the candidate split of the node m using the parameters that minimize the impurity. Q = argmint G(Qm , t)

(1)

where    r  rm m left right H left Qm (t) + H right Qm (t) rm rm   left H left Qm (t) left

G(Qm , t) =

left

  right H right Qm (t)

(2) (3) (4)

are gini or entropy measures of split’s impurity for the classification task. The lower the value of   left H left Qm (t) (5)   right H right Qm (t)

(6)

the better the split. When the dataset is imbalanced, there is a significant probability that the majority of the class samples are included in the same nodes, which makes entropy close to zero. As a result, the created tree is biased toward the majority class [6, 7]. The mentioned problem negatively impacts the model’s performance which is visible in Table “Accuracy metrics’ values” in the study [3]: the high value of Accuracy (>0.9) for techniques: “Interview”, “Document Analysis” and relatively low value (~0.7) of the area under the ROC curve (AUC). According to studies [8, 9], AUC is an accurate indicator of the model’s prediction ability when a dataset is imbalanced because it shows the ability of the classifier to rank the positive instances relative to the negative ones. The ROC curve graph is based upon true positive rate (tpr) and false positive rate (ftp), so it does not depend on the class’s distributions’ changes. In contrast, Accuracy measures the minimization of the overall error to which the minority class contributes very little [10]. To tackle the problem and improve the performance, in the current work, we will propose to enhance the built ML model by adding a data preprocessing step, which is to balance the dataset before training a machine learner algorithm. The prediction’s quality

576

D. Gobov and O. Solovei

will be measured based on the same metrics used in work [3] to obtain the predictions’ comparable results. Another point is a “Positive feature importance” algorithm used in [3] – according to which features’ importance is identified based on the build-in capability of the decision tree learner when the feature used in the node’s split most often receives the higher score. Such an approach depends on the learner’s prediction Accuracy and can result in a wrong feature’s score when a created decision tree is biased toward the majority class. In the current work, we propose to use machine learner-independent methods to identify the feature’s importance score. The paper is structured as follows. Section 2 reviews the related works on approaches for handling imbalanced datasets in ML models. Section 3 contains the description of the input data and the experiment’s scheme: dataset characteristics, data pre-processing procedure, train and test procedures. Section 4 is devoted to the experiments’ results of using proposed improvements for the ML model. Section 5 concludes the paper with the main study findings and future work.

2 Related Work In the study [11], to improve classification models’ results for imbalance datasets are considered the following data preprocessing strategies: 1) random resampling: 1.1) resampling the small class at random until it contains the number of samples that is equal to the majority class; 1.2) focused resampling the small class only with data occurring close to the boundaries with a factor equal to 0.25 between the concept and its negation. 2) downsizing: 2.1) random down-sizing – eliminating at random the elements of the over-sized class until it matches the size of the majority class; 2.2) focused down-sizing - eliminating only elements further away from the boundaries; 3) learning by recognition - to use unsupervised machine learning algorithms and ignore target class in dataset. The study concluded that resampling and down-sizing methods are very effective compared to the recognition-based approach. However, it was left without a recommendation, which is a preferable method – random or focused resampling/down-sizing. Furthermore, it was specified that by random down-sizing of the majority class, there is a chance to exclude meaningful information; therefore, the preferences to be given to resampling or focused down-sizing. In the study [12] it was proposed an over-sampling approach – Synthetic Minority Over-sampling Technique (SMOTE) in which the minority class is over-sampled by creating “synthetic” examples that are identified as k minority class nearest neighbors. If the amount of over-sampling needed is 100%, then the five nearest neighbors are chosen. The authors specified that the SMOTE approach could improve the accuracy of classifiers because, with an over-sampled dataset, the classifier builds larger decision regions. After all, the created tree isn’t biased. Except for the Accuracy, oversampling algorithm SMOTE can improve the CV score, F1 Score, and recall of classifiers [13]. Further enhancements to SMOTE were proposed in the study [14] – the new method was called SMOTEBoost – according to which SMOTE is applied on each boosting round. With SMOTEBoost, the higher prediction Accuracy of the classifier was achieved. The study [15] proposed to combine the sampling technique and ensemble idea, i.e., to make the dataset balanced by down-sizing or over-sampling and after to grow each

Approaches to Improving the Accuracy of Machine Learning Models

577

tree. The Random Forest algorithm was used in experiments with methods: SMOTE, SMOTEBoost (called Balanced Random Forest), and the results are compared with the Weighted Random Forest algorithm. The conclusions were that both Weighted Random Forest and Balanced Random Forest outperformed other learners, and there was no winner between those two learners. The comparison was made by applying the ROC convex hull method [16]. The better the “more northwest” ROC convex hull because it corresponds to the classifier with a lower expected cost. Several filter methods are proposed in studies [17] to have the most predictive feature selected independently from machine learning algorithms. The appropriate filter’s method is selected depending on the parameters: a type of machine learning task; in case of classification – several unique values in target class (binary or multi-class); type of predictive features – continuous/discrete or categorical; type of data in the dataset – flat feature; structure feature; linked data; multi-source; multi-view; streaming feature; streaming data. However, the selection of the best-fit method for the concrete dataset was not formalized; therefore, an empirical test is required. In the current work, based on the results of studies [11–16], to balance a dataset, we will apply the over-sampling method SMOTE. We also use a learning algorithm called Random Forest Classifier (RFC) to build a model. The mentioned enhanced methods: SMOTEBoost and Weighted Random Forest, won’t be used to avoid over complication of our model. We will use the ROC convex hull method to evaluate the built model’s ability to recommend an elicitation technique. However, classification task metrics like precision and recall will be calculated to compare the effect of the proposed enhancements with the results achieved in the work [3]. Because the study [17] shows no recommendations on how systematically identify the filter technique which fits best the concrete dataset, in the current study, we will empirically find an optimal for our dataset a filter method.

3 Methodology To apply and evaluate the effects of the proposed enhancements, we will include additional blocks in the traditional supervised machine learning work cycle (“blue” rectangles in Fig. 1). A new block, “Data Balancing,” will generate synthetic samples that will diminish the class-imbalance problem with method SMOTE. The applied technique will get a new sample   (7) S = x + k · xR − x as linear combinations of two samples from the minority class (xR and x) with 0 ≤ k ≤ 1; xR is randomly chosen among the 5-minority class nearest neighbors of x. In the added new “Feature scoring” block, we use the filter methods, independent from the learning algorithm, to estimate the score of each feature. Different strategies can be used to get feature scores without knowing which filter method fits the dataset the best: 1) to average the feature’s scores received by applying different filters; 2) to choose the feature’s highest score from the scores received by applying different filters; 3) to apply learning algorithm with feature selected by different methods and select the best

578

D. Gobov and O. Solovei

Fig. 1. Supervised machine learning work cycle with added blocks.

filter method to get features’ scores based on the analysis of ROC convex hull. The last strategy will be used in the current work. In a new “Recommendations Forming” block with the model’s forecasts and feature scores as input information, recommendations regarding recommended techniques will be formed.

4 Experiment We execute experiments to assess the effectiveness of the proposed enhancements. The effectiveness of the built machine learning model was measured via Accuracy, AUC, precision, and recall metrics. The received metrics’ values are compared with the results achieved in work [3], and validations of the proposals are based on the comparison results. Our databases’ characteristics and imbalanced ratios calculated as majority-tominority samples are specified in Table 1. The features that are included in the dataset are two types: 1). Features to describe the project’s context; 2) features to list all elicitation techniques used in the project [18]. The following features belong to the first type: Country, ProjectSize, Industrial Sector, Company Type, Company Size, System/Service Class, Team Distribution, Experience, Way of Work, Project Category, BA Activities (BA Only Role), and Certified. The following features belong to the second type: Benchmarking and Market Analysis, Brainstorming, Business rules analysis, Collaborative games, Data mining, Design Thinking, Interface analysis, Interviews, Observations, Process analysis, Prototyping, Reuse database and guidelines, Stakeholders list, map or Personas, Survey or Questionnaire, Document Analysis, Workshops and focus groups, Mind Mapping. The features’ information that is included in the dataset with target class names: “Elicitation”, “Document Analysis”, “Interface Analysis”, and “Process Analysis” is the same with the following exception when the feature’s name is equal to the target class, then this feature isn’t included in features’ list. The data pre-processing procedure has included: 1) removing features whose values are not unique; 2) applying SMOTE to over-sample a minority class to match 100% of the majority class. Train and test procedures included: 1) Random Forest tree classifier is trained on the sub-dataset, which is received as a result of a random split with the proportion 80/20 for train and test correspondingly. 2) Trained learner is tested with the test subset. Identification of feature’s importance score: based on the dataset’s parameters from Table 1 to identify the feature’s score can be used methods: 1) Chi-squared stats measures dependency between non-negative feature and class, so that irrelevant for classification feature is set a low score [19]; 2) Analysis of variance (ANOVA f-test) measures the ratio of explained variance to unexplained variance, the higher values mean

Approaches to Improving the Accuracy of Machine Learning Models

579

Table 1. The characteristics of datasets

Target class name Interviews Document Analysis Interface Analysis Process Analysis

M ajority M inority Imbalance M achine Feature class class ratio learning task type 282 276 232 213

41 47 91 110

6.9 5.9 2.5 1.9

Binary Discrete classification

M issing values, Y/N? N

more feature’s importance to a target class [20]; 3) mutual information (MI)– measures the amount of shared information between independent feature and target class. It can detect non-linear dependencies, which is its main strength, and it can be applied when the dataset includes continuous, discrete, or both types of data [21, 22]. Because the types of the results of the mentioned filters are different, to understand which filter best fits datasets from Table 1, we will do the following: 1) create ROC curves calculated by Random Forest with balance data and feature selected by methods: chi-squared stats, mutual information, Anova f-value; 2) select the best-fit method based on ROC convex hull graphs analysis.

5 Results and Discussions Figure 2 illustrates how the split’s impurity measured by entropy increased after the dataset was balanced by applying SMOTE, which indicates that more samples from different target classes are included in the node; hence both classes are presented before the decision in favor of the majority, or minority class is made, and the created tree isn’t biased. The gap between mean entropies is ≈ 10% for datasets with imbalance ratios of more than 5 (Fig. 2- pictures (a) and (b)), and it is 5% and 4% for datasets with imbalance ratios of 2.5 and 1.9, correspondingly. Therefore, the graphs in Fig. 2 display the direct dependency between the decision tree’s bias and the dataset’s imbalance ratio.

Fig. 2. Split’s impurity of imbalance and balance datasets calculated by entropy

The values of the performance metrics of Random Forest learners with a balanced test subset are recorded in Table 2. The area under ROC curves for the balanced dataset increased significantly compared to the imbalanced dataset, and accuracy increased as well.

580

D. Gobov and O. Solovei

Similar improvements are recorded in Fig. 3, where SMOTE-Random Forest computed ROC convex hull colored by orange, and Random Forest computed ROC convex hull shown by blue color with the imbalanced dataset. The SMOTE-Random Forest ROC convex hull dominates the ROC convex hull without SMOTE, making SMOTE a more optimal classifier. Table 2. Accuracy and AUC for balanced by SMOTE and original dataset Techniques

Accuracy

AUC

Imbalanced Balanced Improving Imbalanced Balanced Improving Interviews

0,938

0.94

0%

0,720

0.97

35%

Document Analysis 0,901

0.91

1%

0,728

0.97

33%

Interface Analysis

0,790

0.84

6%

0,676

0.9

33%

Process Analysis

0,778

0.82

5%

0,757

0.88

16%

Fig. 3. ROC convex hull of imbalance and balance datasets calculated by RFC

The p-value of paired t-test for precision and recall received for the balanced dataset and imbalanced dataset from [3] are recorded in Table 3. p-value = 0.018 indicates a statistically significant difference in the precision of the Random Forest classifier with a balanced dataset compared to imbalance which means that the ability of the classifier not to label as positive a sample that is negative is affected when the negative’s class distribution’s changes because of over-sampling. The p-value = 0.087 indicates the statistically insignificant difference in the recall, i.e., the classifier’s ability to correctly find all positive samples when the positive class is a majority class has not been affected by applying the oversampling of the minority class.

Approaches to Improving the Accuracy of Machine Learning Models

581

Table 3. p-value of paired t-test of precision and recall Dataset

Interviews

Document Analysis

Interface Analysis

Process Analysis

p-value

Precision, balanced data

0.89

0.88

0.81

0.78

0.018

Precision, imbalanced data

0.936

0.899

0.831

0.818

Recall, balanced data

1

0.96

0.9

0.88

Recall, imbalanced data

1

1

0.922

0.9

0.087

To select the best feature selection method, we used ROC convex hull graphs presented in Fig. 4. For each dataset from Table 1, graphs show the “more northwest” ROC curve when features are selected by the mutual information method. The ten features which have the biggest importance score for each dataset calculated by the mutual information method are recorded in Tables 4, 5. To type of recommendations can be produced based on the results in Tables 4, 5: 1). Collaborating filtering - the list of elicitations techniques that can be used in the project in addition to the elicitation technique that is predicted by a machine learning model. 2). Content-based filtering – based on the similarity of the project’s context the system can recommend the elicitation technique to be used. Due to form the recommendations of the first and the second types, the system selects from Tables 4, 5 the features whose score is higher than a particular number and which correspondingly belong to the first and the second feature’s type. The value of the score, which the system uses to select the feature, is proposed to be a user’s choice.

Fig. 4. ROC convex hull calculated by Random Forest with balance data and feature selected by methods: chi-squared, mutual information, Anova f-value

582

D. Gobov and O. Solovei Table 4. Features’ importance score. Interviews and Document analysis.

Interviews

Document analysis

Feature

Score

Feature

Score

Project Size

0.3

Experience

0.32

Experience

0.28

Project Size

0.27

WoW

0.27

WoW

0.25

Prototyping

0.25

Industrial Sector

0.22

Project Category

0.23

Company Size

0.2

Company Type

0.21

Project Category

0.2

Process analysis

0.19

Process analysis

0.19

Industrial Sector

0.18

Company Type

0.19

Company Size

0.18

Interface analysis

0.18

Interface analysis

0.17

Observations

0.17

Table 5. Features’ importance score. Interface analysis and Process Analysis Interface analysis

Process Analysis

Feature

Score

Feature

Score

Experience

0.22

Business rules analysis

0.15

Project Size

0.15

Project Category

0.15

WoW

0.15

Company Type

0.14

Company Type

0.15

Industrial Sector

0.13

Project Category

0.14

WoW

0.12

Team Distribution

0.14

Experience

0.11

Observations

0.13

Workshops and focus groups

0.11

Prototyping

0.12

Stakeholders list, map or Personas

0.1

Document analysis

0.12

Benchmarking and Market Analysis

0.1

Company Size

0.12

Team Distribution

0.09

6 Conclusions The current study proposes several enhancements to the machine learning model for getting recommendations on elicitation techniques usage based on the combinations of factors. The first proposal is to balance the dataset by applying SMOTE methods before the training and testing classifier. As a result, improvement in the “Accuracy” indicator was obtained in the range from 0% to 6%, and the “AUC” indicator from 16% to 35%. The dependency between the value of the imbalance ratio and created tree biased to the majority class was visual. The results can be used in future works to formalize whether

Approaches to Improving the Accuracy of Machine Learning Models

583

applying SMOTE to balance a dataset is necessary, as an increased dataset will demand more operation time for machine learning. The second proposal is to calculate the features’ importance score independently from the machine learner classifier. A mutual information method was identified as the best fit for the considered datasets. The produced list of the most predictable features is proposed to use in recommended systems of both types: collaborating filtering and content-based filtering. The results obtained allow the construction of more precise models for the recommendation of requirements elicitation techniques contingent upon the project context. Further research can be conducted to explore additional BA tasks to determine correlations and make suggestions for selecting techniques for requirement specification and modeling, validation, and verification based on balanced datasets.

References 1. Rehman, T., Khan, M., Riaz, N.: Analysis of requirement engineering processes, tools/techniques and methodologies. Int. J. Inform. Technol. Comput. Sci. 5(3), 40–48 (2013). https://doi.org/10.5815/ijitcs.2013.03.05 2. Manzoor, M., et al.: Requirement elicitation methods for cloud providers in IT industry. Int. J. Mod. Educ. Comput. Sci. 10, 40–47 (2018). https://doi.org/10.5815/ijmecs.2018.10.05 3. Gobov, D., Huchenko, I.: Influence of the software development project context on the requirements elicitation techniques selection. Lect. Notes Data Eng. Commun. Technol. 83, 208–218 (2021) 4. Gobov, D., Huchenko, I.: Software requirements elicitation techniques selection method for the project scope management. CEUR Workshop Proc. 2851, 1–10 (2021) 5. Derya, B.: Data Mining in Banking Sector Using Weighted Decision Jungle Method. Data Mining-Methods, Applications and Systems. IntechOpen (2020) 6. Visa, S., Ralescu, A.: Issues in mining imbalanced data sets-a review paper. In: Proceedings of the Sixteen Midwest Artificial Intelligence and Cognitive Science Conference, vol. 2005, pp. 67–73 (2005) 7. Krawczyk, B.: Learning from imbalanced data: open challenges and future directions. Progress Artific. Intell. 5(4), 221–232 (2016). https://doi.org/10.1007/s13748-016-0094-0 8. Fawcett, T.: An introduction to ROC analysis. Pattern Recogn. Lett. 27(8), 861–874 (2006) 9. Flach, P.A.: ROC analysis. In: Sammut, C., Webb, G.I. (eds.) Encyclopedia of machine learning and data mining, pp. 1–8. Springer US, Boston, MA (2016). https://doi.org/10.1007/9781-4899-7502-7_739-1 10. Orallo, H., et al.: The 1st workshop on ROC analysis in artificial intelligence (ROCAI-2004). ACM SIGKDD Explor. Newsl 6(2), 159–161 (2004) 11. Japkowicz, N.: Learning from imbalanced data sets: a comparison of various strategies. AAAI workshop on learning from imbalanced data sets, vol. 68, pp. 10–15 (2000) 12. Chawla, N., et al.: SMOTE: synthetic minority over-sampling technique. J. Artific. Intell. Res. 16, 321–357 (2002) 13. Chakraborty, V., Sundaram, M.: An efficient smote-based model for dyslexia prediction. an efficient smote-based model for dyslexia prediction. Int. J. Inform. Eng. Electron. Bus. 13(6), 13–21. (2021). https://doi.org/10.5815/ijieeb.2021.06.02 14. Chawla, N., et al.: SMOTEBoost: improving prediction of the minority class in boosting. Lect. Notes Comput. Sci. 2838, 107–119 (2003) 15. Díez-Pastor, J., et al.: Random balance: ensembles of variable priors classifiers for imbalanced data. Knowl. Based Syst. 85, 96–111 (2015)

584

D. Gobov and O. Solovei

16. Bettinger, R.: Cost-sensitive classifier selection using the ROC convex hull method. SAS Institute, pp. 1–12 (2003) 17. Li, J., et al.: Feature selection: a data perspective. ACM Comput. Surv. 50(6), 1–45 (2017) 18. Gobov, D., Huchenko, I.: Requirement elicitation techniques for software projects in Ukrainian IT: an exploratory study. In: Proceedings of the Federated Conference on Computer Science and Information Systems, pp. 673–681 (2020). https://doi.org/10.15439/2020f16 19. Su, C.T., Hsu, J.H.: An extended chi2 algorithm for discretization of real value attributes. IEEE Trans. Knowl. Data Eng. 17(3), 437–441 (2005) 20. Lindman, H.R.: Analysis of Variance in Experimental Design. Springer Science & Business Media (2012) 21. Gao, W., Kannan, S., Oh, S., Viswanath, P.: Estimating mutual information for discretecontinuous mixtures. Adv. Neural Inform. Process. Syst. 30 (2017) 22. Ross, B.C.: Mutual information between discrete and continuous data sets. PLoS ONE 9(2), e87357 (2014)

Machine Learning Based Function for 5G Working with CPU Threads Maksim Iavich(B) Caucasus University, 1 Paata Saakadze St., 0102 Tbilisi, Georgia [email protected]

Abstract. A huge amount of information is transmitted over wireless networks. Moreover, the volume of transmitted information is constantly increasing, one of the many factors in this is that new mobile devices are continuously communicate with each other through the network, and the number of multimedia applications, streaming, video conferencing, social networks and other things are growing behind them. To meet current and expected needs, Transition to 5G networks is taking place at great pace. As of 2021 and 2022, approximately one million minutes of video per second was transmitted over the Internet worldwide, and as data grows rapidly, the 4G system needs to be replaced with a more powerful 5G system with increased bandwidth and improved quality of service to ensure a secure and stable connection. 5G have the security problems. During our research, we were able to identify various 5G security vulnerabilities and then study them in detail. On the basis of which we have implemented a new machine learning based cybersecurity model with its Firewall and IDS/IPS, which we describe in this article. We have analyzed the use of different machine learning algorithms for our task. Finally, decision tree algorithm was chosen. We have implemented this algorithm using CPU threads to increase the efficiency. The main significance of the offered security function is that it is very efficient and works very quickly; therefore, it can be considered for real time usage. The offered security function is trained with attack patterns created in a simulation lab. We have tested this mechanism in the simulated 5G network. The experiment was carried out in our laboratory, where we used 10 Raspberry Pi modems to simulate attacks on the server. A similar approach will be useful for future versions of 5G. Keywords: 5G · machine learning · networks · CPU threads

1 Introduction A large amount of data is transmitted over wireless networks. Moreover, the size of transmitted data is increasing. New mobile devices are continuously added to the network, and the multimedia applications, streaming, video conferencing, social networks and other things are growing behind them. To meet ever-increasing demand, the telecommunications industry is rapidly deploying 5G and beyond [1–4]. At 2021–2022 more than one million minutes of video per second was transmitted over the internet, the size of data grows rapidly. Therefore, the 4G system needs to be replaced with a more powerful © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 585–594, 2023. https://doi.org/10.1007/978-3-031-36118-0_52

586

M. Iavich

5G system with increased bandwidth and improved quality. This will ensure a secure and stable connection and communication. The introduction of 5G requires the introduction of new data storage and processing technologies, which will certainly set new tasks for us in terms of the functionality and security of 5G systems. The world’s leading scientists are actively researching the problems related to the security of 5G systems. Research shows that 5G security is not yet at the desired level. This is confirmed by a number of successful attacks on these networks, such as vulnerabilities of MNmap, MiTM, and a battery drain attack that can lead to malicious code infiltrating the system. So, it is needed to look for new ways to secure 5 and future 6G networks by creating new architectures and new algorithms based on AI/ML [5–7]. 5G networks must be able to transmit very large amounts of information with very low latency. All this can be achieved by deploying high-frequency towers at short distances. In addition, there will be a need for a new network architecture, new data storage technologies with great capabilities, the introduction of which will inevitably lead to new cybersecurity problems. Infrastructure will soon be fully migrated to 5G system, so to ensure security, strong security systems must be established, Which requires comparing the long-established 4G systems with 5G and studying the differences between them. Weaknesses in 5G systems also need to be identified and ways to address them need to be found. It must be proposed the efficient security function to protect 5G and beyond networks. The major research objectives is to offer the security function, which will be very efficient, using machine learning approach. As the machine learning approach we use decision tree algorithm. There are the related works, which use decision tree algorithms, but they don’t describe the parallel approaches using CPU threads. Our approach is also unique, because it uses the manual approach, which increases the detection precision. The main limitation of the existing approaches is that they cannot be used in the real time. Our main goal is to increase the efficiency in order to make it usable in the real time.

2 About 5G Security Features The LTE systems providing communication and security of 4g and 5g systems are similar. Therefore, we can divide the 5g security features into two categories, and we can take care of the device and the radio node to ensure safety. The network access security devices, which ensure security of connection between radio node and device, and also devices ensuring security of access various services for users are included in first category (Fig. 1). WAMS data and SCADA data can be stable between two short periods. Before determining the correlation coefficient, the upper and lower bounds of the correlation coefficient are required. The Pearson correlation coefficient proposed in this paper is widely used to measure sequence correlation. In a 5G system, access authentication is at the forefront. The operation is completed at the starting stage of identification, if authentication completes successfully the session keys are determined, which are used for network – device communication.

Machine Learning Based Function for 5G Working with CPU Threads

587

Fig. 1. LTE/4G and 5G security architecture

This the operation is executed in compliance with 3GPP 5G security regulations to protect EAP – Extended Authorization Protocol, whose purpose is to use data, certificates, public key passwords, and usernames – credentials stored on SIM cards. Identification and key agreement mechanisms are used to pass authorization procedures of device and network. As a result of this process, The KSEAF keys generated by this process are sent to the secure channel – SEAF service network. Thus, KSEAF is used in situations where re-authentication is not required. In 3GPP network authentication, keys are generated using 3GP and UE functions. 5G uses SUPI, which assigns a globally unique, permanent subscriber ID to each user. Permanent caller identifier supports IMSI-international mobile subscriber identifier and NAI – network access identifier. It is important that SUPI remains secret when connected to a specific mobile device, unlike 3 and 4G systems. in which the IMSI identification of a specific item on the network is exposed. Please note that the SUCI-secret signature identifier should be used until verification the authenticity of the network and device completes. After completing identification, the home network SUPI can be opened to service network to ensure that subscriber’s identity is not stolen. To avoid this a device is connected to RBS while trying to sniff unencrypted traffic. To conceal SUPI out of SUCI the SIDF – Subscription ID disclosure feature should be used. Which one uses privacy key from a privacy-related home network public/private key pair that is securely stored on the home network of operator. Access parameters in SIDF should allow only the home network to request SIDF. Disclosure must go through so called UDM – the UniFi Dream Machine. Port settings should be set so that only the home network should be able to request the port [8–10]. To ensure the security of messages sent through the N32 interface around the perimeter of the public land mobile network (PLMN), the 5G architecture includes a proxy security boundary protection proxy. All network function (NF) service layers drop messages to SEPP, which are being kept till they are sent through the N32 interface. On the other way each message is sent to SEPP by N32 interface, at his turn SEPP checks data and transmits back it to the suitable network functional level. SEPP protects all information from application layer that moves among two NFs in two various PLMNs. In order to achieve the highest best security level, 5G uses a network slicing method. The

588

M. Iavich

network segment is responsible for an corresponding quality of service. Thus, we can divide the whole network into slices so that they do not overlap. This mechanism allows you to segment the network and securely manage the corresponding services. Thus, the system is easy to manage and moreover, it is more secure.

3 5G Security Issues Our research gives us reason to say that 5G security is still not perfect and faces challenges. The presented article analyzes the problems related to 5G security: 1. Various types of attacks are constantly being carried out on the 5G network, it has much more entry points for attackers, because the 5G network architecture allows them to exploit numerous weak points, flaws and bugs. 2. With the introduction of 5G networks we need to introduce new functionality, and with new functionality comes new vulnerabilities. Attackers can attack base points and control functions. 3. Since mobile operators are dependent on providers may cause the development of new attack routes and may greatly increases attack damage. 4. After the introduction of 5G systems, almost all vital IT applications will use it, so we may encounter security issues related to the integrity and availability. 5. There will be a lot of devices on the 5G network, which can provoke DoS and DDoS attacks. 6. Network segmentation also may cause some security clauses, as attackers can attempt to connect a certain device using a desired network segment. It should be noted that an attacker can use the flaws found in 5G with the help of which attacker can perform illegal actions. We can group these attacks as follows: • MNmap To conduct an experiment, our researchers sent a message to the network, the purpose of this action was to reveal fake network. So they fabricated base station and were able to get all the information about the devices that were connected to the networks. Finally managed to determine device type, model, manufacturer, operating system, and version. • MiTM The design of 5g systems also allows for Man-in-the-Middle (MiTM) attacks. By means of which it Bidding and battery drain attacks may be performed possible. As it’s possible to slow down 5G speeds by disabling Multiple Input–Output (MIMO). Thus, the speed may fall to the of 2G/3G/4G networks speed. • Battery drain attack During this attack, small information packets are sent to the victim device, which drains the battery. The attacker adapts PSM, while the victim device is connecting network the hacker offers it a desired network, thereby gaining control of the device. It is worth noting that for the new design of 5G and future 6G networks, it is necessary to develop new AI/ML-based algorithms that provide a appropriate level of cybersecurity and appropriate defense of mobile subscribers, industry and government.

Machine Learning Based Function for 5G Working with CPU Threads

589

4 Advantages of Our Approach Our proposal is to create a security module where IDS/IPS and firewall are integrated on the same server and used on each base station (Fig. 2).

Fig. 2. Cybersecurity module

In the article, we have presented materials proving that we can still hack 5G security with Probe and Dos or software attacks. Our IDS initial version is based on M/L. Effective protection against these attacks can be performed by using different datasets in the training process. Most often, NSL-KDD kits are used for such studies. Which is widely used for creating IDS prototypes and conducting various M/L based experiments. It should be noted that some of the data collected in this set is outdated and no longer corresponds to today’s realities. We propose to build our own template and teach IDS on these templates. The laboratory consists of a server that includes a network protocol analyzer – TSark and 20 pieces of a RASPBERRY PI device. We use a network analyzer to create four datasets, each consisting of different data patterns ranging in size from 1 to 2 gigabytes. We performed brute force and DoS/DDOS attacks (ICMP, slow Loris, UDP flood, SYN flood, volume-based and Ping of Death attacks). The first set of data includes: UDP and SYN flood attack patterns. Second: ICMP and Slow Loris attacks. Volume Based and Ping of Death attacks the third one and Brute Force attack is the fourth set of data.

5 Machine Learning Algorithms We have checked our system using different machine learning algorithms in order to choose the most suitable one. We have tried decision tree, Deep Learning and random forest classifier algorithms [11–15].

590

M. Iavich

For every experiment, we split our data into training and test datasets. The experiment results for decision tree logic are illustrated in the Table 1. And the experiment results for Deep Learning and random forest classifier algorithms are illustrated on Tables 3 and 4 respectively. The best results were achieved when using Decision tree algorithm. The Table 1. Decision tree without threads Training data size

Test data size

Accuracy

Time spent for predicting and calculating

Time spent for learning data

Full session time

0.9

0.1

0.9847

0.71(s)

35(s)

404.73(s)

0.8

0.2

0.9821

1.39(s)

34.57(s)

0.7

0.3

0.978

2.10(s)

38.59(s)

0.6

0.4

0.975

2.81(s)

34.19(s)

0.5

0.5

0.9715

3.56(s)

36.54(s)

0.4

0.6

0.9689

4.22(s)

32.47(s)

0.3

0.7

0.9664

4.96(s)

31.67(s)

0.2

0.8

0.9621

5.62(s)

30.26(s)

0.1

0.9

0.9561

6.27(s)

28.34(s)

0.05

0.95

0.9326

6.62(s)

27.44(s)

0.01

0.99

0.8337

6.89(s)

28.15(s)

Table 2. Decision tree with threads Training data size

Test data size

Accuracy

Time spent for predicting and calculating

Time spent for learning data

Full session time

0.9

0.1

0.9847

8.61(s)

269.6(s)

350.55(s)

0.8

0.2

0.9821

9.22(s)

270.03(s)

0.7

0.3

0.978

5.24(s)

272.42(s)

0.6

0.4

0.975

5.11(s)

271.32(s)

0.5

0.5

0.9715

15.61(s)

265.17(s)

0.4

0.6

0.9689

13.41(s)

266.84(s)

0.3

0.7

0.9664

11.01(s)

250.17(s)

0.2

0.8

0.9621

20.46(s)

258.02(s)

0.1

0.9

0.9561

6.77(s)

62.6(s)

0.05

0.95

0.9326

9.83(s)

56(s)

0.01

0.99

0.8337

41.18(s)

206.56(s)

Machine Learning Based Function for 5G Working with CPU Threads

591

best split of dataset was 90% for training and 10% for testing. The received accuracy was 0.9847. We have implemented the same algorithm using the threads, the results are illustrated in Table 2. Table 3. Deep learning Training data size

Test data size

Accuracy

Time spent for predicting and calculating

Time spent for learning data

Full session time

0.9

0.1

0.9847

0.37(s)

108.52(s)

1065.39(s)

0.8

0.2

0.9821

0.70(s)

103.03(s)

0.7

0.3

0.978

11.69(s)

131.42(s)

0.6

0.4

0.975

1.39(s)

125.50(s)

0.5

0.5

0.9715

14.22(s)

69.32(s)

0.4

0.6

0.9689

5.99(s)

79(s)

0.3

0.7

0.9664

1.66(s)

61.96(s)

0.2

0.8

0.9621

1.94(s)

63.04(s)

0.1

0.9

0.9561

2.79(s)

65.76(s)

0.05

0.95

0.9326

42.68(s)

42.26(s)

0.01

0.99

0.8337

17.72(s)

67.27(s)

Table 4. Random forest classifier Training data size

Test data size

Accuracy

Time spent for predicting and calculating

Time spent for learning data

Full session time

0.9

0.1

0.987640

3.25(s)

434.49

1870.54 (s)

0.8

0.2

0.986984

6.27(s)

216.34(s)

0.7

0.3

0.983707

9.59(s)

193.83(s)

0.6

0.4

0.982442

11.52(s)

168.52(s)

0.5

0.5

0.980148

14.34(s)

144.25(s)

0.4

0.6

0.978150

16.49(s)

121.83(s)

0.3

0.7

0.976001

20.23(s)

140.72(s)

0.2

0.8

0.971416

20.61(s)

96.22(s)

0.1

0.9

0.963210

23.63(s)

67.71(s)

0.05

0.95

0.947818

24.04(s)

60.52(s)

0.01

0.99

0.882751

22.25(s)

53.69(s)

592

M. Iavich

6 The Offered Function Our proposal is the following IDS model. The TShark network protocol analyzer and our new smart IDS are installed on the main server. IDS is trained using decision tree algorithm. A decision tree algorithm is a supervised non-parametric learning algorithm, it is used in both regression and classification tasks [16–18]. The approach has a hierarchical, tree structure, and it has the root node, the branches, leaf nodes and the internal nodes. In order to choose the best parameter for our decisional tree, we choose the attributes with the minimal entropy values. We calculate the entropy using the following formula:  p(C) log2 pr(C) (1) E(D) = − cl∈CL

D – is the data set. C – are the classes in the set, D pr(c) – is the proportion of the data points of cl according to the number of total points of data in D. After the IDS learning process is completed, the system waits for data input. To store all traffic patterns, we have to add new server, log server. Each successful attack is analyzed manually and supplemented to our database. Our IDS system will be permanently trained using the updated datasets as described and will be pushed to main server. We offer to use decision tree algorithm for the training purposes. To increase its efficiency, we use CPU treads. Thus, we can break down our IDS model into the following stages: 1) 2) 3) 4) 5) 6) 7)

Data training on the main server Traffic analysis on the main server Pushing data to the log server Data training on the log server The trained model updating on the main server In case an attack is detected, information is immediately sent to IPS In the event attack succeeds, the data is analyzed manually and supplemented to the log server. The process is repeated endlessly.

7 Tests We have generated the attacks vectors in the simulation lab. Using these attack patterns we have trained IDS and checked it in Google Colab. We have got the following results (Table 5):

Machine Learning Based Function for 5G Working with CPU Threads

593

Table 5. Results of the experiment Attack name

Performed attacks

Identified attacks

UDP flood

20000

19979

SYN flood

20000

19896

ICMP, Ping of Death

20000

19870

Slowloris

20000

18980

Brute force attacks

20000

19960

Volume Based Attacks

20000

18982

Based on the analysis of the obtained results, we can claim that the obtained IDS is quite effective and can be proposed as a real IDS prototype. It must be also mentioned that the whole work of the system was rather effective and the whole session time was 350.55(s), that was much quicker than in the case of other machine learning algorithms. The parallel programming using threads also made the speed up of the algorithm. The system was checked in Google Colab. The results collide with the expectations of the experiment.

8 Conclusion and Plans for the Future The model presented in this article covers most of the registered attacks that threaten 5G systems. The approach we propose is in principle different from the strategies developed before, where IDS training is mainly carried out only on the KDD99 dataset. Presented approaches are purely academic and cannot be applied in practice. Our approach is rather efficient and is uses CPU threats. Conducted experiments show that proposed mechanism is able to reveal majority of the listed attacks and gradually will become more precise. Accordingly, IDS detection results will become subtle. Besides, new attack vectors are emerging. The work advances the field from the present state of knowledge, because the proposed IDS is constantly updated and is in permanent learning regime. Respectively, proposed IDS model will be applied in 5G systems. In the future, – it is interesting to work on automating the analysis process of attacks detected by our IDS system. Acknowledgment. This work was supported by Shota Rustaveli National Science Foundation of Georgia (SRNSF) [STEM-22-1076].

References 1. Bingham, J.A.C.: Multicarrier modulation for data transmission: an idea whose time has come. IEEE Commun. Mag. 28(5), 5–14 (1990). https://doi.org/10.1109/35.54342 2. Weinstein, S., Ebert, P.: Data transmission by frequency-division multiplexing using the discrete Fourier transform. IEEE Trans. Commun. Technol. 19(5), 628–634 (1971). https://doi. org/10.1109/TCOM.1971.1090705

594

M. Iavich

3. Doelz, M.L., Heald, E.T., Martin, D.L.: Binary data transmission techniques for linear systems. Proc. IRE 45(5), 656–661 (1957). https://doi.org/10.1109/JRPROC.1957.278415 4. Chang, R., Gibby, R.: A theoretical study of performance of an orthogonal multiplexing data transmission scheme. IEEE Trans. Commun. Technol. 16(4), 529–540 (1968). https://doi.org/ 10.1109/TCOM.1968.1089889 5. Iavich, M., Gnatyuk, S., Odarchenko, R., Bocu, R., Simonov, S.: The novel system of attacks detection in 5G. In: Barolli, L., Woungang, I., Enokido, T. (eds.) AINA 2021. LNNS, vol. 226, pp. 580–591. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-75075-6_47 6. Pan, F., Wen, H., Song, H.H., Jie, T., Wang, L.Y.: 5G security architecture and light weight security authentication. In: 2015 IEEE/CIC International Conference on Communications in China – Workshops (CIC/ICCC), pp. 94–98 (2015). https://doi.org/10.1109/ICCChinaW. 2015.7961587 7. Park, S., Kim, D., Park, Y., Cho, H., Kim, D., Kwon, S.: 5G Security threat assessment in real networks. Sensors 21, 5524 (2021). https://doi.org/10.3390/s21165524 8. Iwamura, M.: NGMN view on 5G architecture. In: 2015 IEEE 81st Vehicular Technology Conference (VTC Spring), pp. 1–5 (2015). https://doi.org/10.1109/VTCSpring.2015. 7145953 9. Gupta, A., Jha, R.K.: A survey of 5G network: architecture and emerging technologies. IEEE Access 3, 1206–1232 (2015). https://doi.org/10.1109/ACCESS.2015.2461602 10. Agyapong, P.K., Iwamura, M., Staehle, D., Kiess, W., Benjebbour, A.: Design considerations for a 5G network architecture. IEEE Commun. Mag. 52(11), 65–75 (2014). https://doi.org/ 10.1109/MCOM.2014.6957145 11. Safavian, S.R., Landgrebe, D.: A survey of decision tree classifier methodology. IEEE Trans. Syst. Man Cybern. 21(3), 660–674 (1991). https://doi.org/10.1109/21.97458 12. Shrestha, A., Mahmood, A.: Review of deep learning algorithms and architectures. IEEE Access 7, 53040–53065 (2019). https://doi.org/10.1109/ACCESS.2019.2912200 13. Wang, H., Chen, S., Xu, F., Jin, Y.-Q.: Application of deep-learning algorithms to MSTAR data. In: 2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp. 3743–37452015). https://doi.org/10.1109/IGARSS.2015.7326637 14. Devetyarov, D., Nouretdinov, I.: Prediction with confidence based on a random forest classifier. In: Papadopoulos, H., Andreou, A.S., Bramer, M. (eds.) AIAI 2010. IAICT, vol. 339, pp. 37–44. Springer, Heidelberg (2010). https://doi.org/10.1007/978-3-642-16239-8_8 15. Kulkarni, V.Y., Sinha, P.K.: Pruning of random forest classifiers: a survey and future directions. In: 2012 International Conference on Data Science and Engineering (ICDSE), pp. 64–68 (2012). https://doi.org/10.1109/ICDSE.2012.6282329 16. Rifat-Ibn-Alam, M.: A comparative analysis among online and on-campus students using decision tree. Int. J. Math. Sci. Comput. 8(2), 11–27 (2022). https://doi.org/10.5815/ijmsc. 2022.02.02 17. Maharjan, M.: Comparative analysis of data mining methods to analyze personal loans using decision tree and naïve Bayes classifier. Int. J. Educ. Manage. Eng. 12(4), 33–42 (2022). https://doi.org/10.5815/ijeme.2022.04.04 18. Nutheti, P.S.D., Hasyagar, N., Shettar, R., Guggari, S., Umadevi, V.: Ferrer diagram based partitioning technique to decision tree using genetic algorithm. Int. J. Math. Sci. Comput. 6(1), 25–32 (2020). https://doi.org/10.5815/ijmsc.2020.01.03

Analysis of Threat Models for Unmanned Aerial Vehicles from Different Spheres of Life Hanna Martyniuk1(B) , Bagdat Yagaliyeva2 , Berik Akhmetov2 , Kayirbek Makulov2 , and Bakhytzhan Akhmetov3 1 Mariupol State University, Kyiv 03037, Ukraine

[email protected]

2 Yessenov University, Aktau 130000, Kazakhstan 3 Abay University, Almaty 050010, Kazakhstan

Abstract. Unmanned aerial vehicles are currently used in various spheres of life: they are used for the agricultural industry, for photo and video filming, for rescue operations, in military operations, etc. Depending on the areas of application of UAVs, they can be used both for attacks on various objects and for obtaining some information from objects. The purpose of this work is to classify UAVs and consider the cybersecurity problems that arise when using them. Based on the above classification, a threat model for UAVs in various fields of activity is considered and analyzed. There is a large number of works related to the creation of a threat model. However, such works do not consider the scope of drones. It should be noted that the threat model for military UAVs should be different from the threat model for UAVs in the agricultural sector. In this regard, the authors set themselves the goal of constructing and analyzing a UAV threat model based on their classification depending on the scope of application. Keywords: Cybersecurity · UAV · Threat model

1 Instruction Unmanned aerial vehicles (UAVs) are becoming more and more popular in the world. The scope of their application is expanding: from military measures to the operations of rescuers, doctors, firefighters, agriculture and simply the supply of small-sized cargo. UAVs are capable of conducting aerial reconnaissance and surveillance, transmitting photo and video information in real time, being carriers and targets, operating in extreme conditions, in particular in areas subjected to radiation, chemical or biological contamination, in areas of disasters and intense fire countermeasures. In modern conditions, the use of UAVs and robotics has become the usual norm. With their help, the tasks of ensuring military security, as well as issues in research, security and other areas are solved [1–8]. In the fight against terrorism, drones and other types of robotics are becoming an increasingly effective and regularly used tool [9–11]. However, it should be noted that today there are problems with the protection of such UAVs from unauthorized hacking and use. Thus, a number of publications [12–17] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 595–604, 2023. https://doi.org/10.1007/978-3-031-36118-0_53

596

H. Martyniuk et al.

describe the main problems of UAV cybersecurity. Based on this analysis, the authors decided to compile a threats model, depending on the possible methods of attacking aircraft. At the same time, terrorist organizations are trying to keep up with progress and are actively using unmanned vehicles in their destructive activities [19–22]. Therefore, consideration of issues on the use of UAVs and countering them must be carried out in parallel. Only the general application of UAVs is considered in the literature for compiling a threat model. At the same time, it is impossible to equally assess the threats and risks that affect the confidentiality of these drones in the civil and military spheres. So, for example, the loss of data confidentiality during aerial reconnaissance of the state of crops will not cause such damage as the loss of data confidentiality during aerial reconnaissance of enemy sites in war. In this regard, the work will consider a model of UAV security threats depending on their areas of application.

2 Classification of Unmanned Aerial Vehicles by Features An unmanned aerial vehicle is a device that is controlled remotely without the direct participation of the pilot. The UAV consists of the following components: an air platform with a special landing system, a power plant, a power supply system for all components, and avionics equipment. The general classification of UAVs is given in Table 1. This classification is not complete, but it is quite enough to determine the cybersecurity problems that exist today with UAVs. For creating and analyzing threat models authors decided to use only usage method like classification feature,

3 Methodology of Building a Threats Model for a UAV As already mentioned, many authors have paid attention to UAV safety issues. Summing up, we can say that attackers can attack software, a communication network, or directly hardware. An illustration of such threats is shown in Fig. 1. All such attacks are used to violate the confidentiality or integrity of data, or in the worst case, take control of the UAV in their hands. In order to assess the damage that can be caused by UAVs in various areas of human activity, the authors proposed to use threat model. In general, a threat model is a formalized description of methods and means of implementing threats to information. According to [23], there are four main criteria for assessing the security of an information and telecommunication system (ITS): – confidentiality criterion - unauthorized access to information; – availability criterion - violation of the possibility of using the UAV or the information being processed; – integrity criterion - unauthorized modification of information transmitted by the UAV (forgery, distortion of information); – observation criterion - refusal to identify, authenticate and log actions. The following threats to UAVs have been identified in [19]: sniffing, spoofing, DoSattack. Using [16], threats can be classified like active interfering, backdoor malware, collisions, de-authentication etc.

Analysis of Threat Models for Unmanned Aerial Vehicles

597

Table 1. Classification of UAVs by characteristic features Classification feature

UAVs

Usage method

military civil antiterrorist

Way of solving the problem

tactical (flight range up to 80 km) operational-tactical (flight range up to 300 km) operational-strategical (flight range up to 700 km)

Weight

small-sized (up to 200 kg) medium-sized (up to 2000 kg) oversized (up to 5000 kg) heavy (over 5000 kg)

Flight duration

short duration (up to 6 h) medium duration (up to 12 h) large duration (over 12 h)

Classification feature

UAVs

Maximum altitude

low-altitude (up to 1 km) medium-altitude (up to 4 km) high-altitude (up to 12 km) stratospheric (over 12 km)

The authors also analyzed and divided the main threats in [12, 16, 17, 19] depending on the criteria listed above (Table 2). For a better understanding of the types of threats, their description is given below. Eavesdropping – attackers listen to unencrypted messages over a communication channel. Sniffing – intercepting data that is delivered within the observed network in the form of packets. Malware Infection – most UAV controllers are based on the use of mobile phones, laptops or wireless systems. Their use makes it possible to illegally install malicious software on UAV controllers, which leads to violations of the ability to use aircraft. Manipulation – allows you to change the specified route of the UAV, which can lead to the loss of the drone or the cargo it is carrying. Jamming – the introduction of interference signals, which disrupts the connection of the UAV with the controller and distorts the transmitted information. Spoofing – by intercepting an encrypted message, the attacker subsequently disguises himself as the legitimate sender of messages. De-authentication attack – this type of threat prevents the user from gaining access to their UAV.

598

H. Martyniuk et al.

Fig. 1. Classification of attacks used on UAVs [1]

Table 2. Classification of UAV threats depending on the criteria of the threats model Criterion name

Type of threats

confidentiality criterion

Eavesdropping Sniffing

availability criterion

Malware Infection Manipulation

integrity criterion

Jamming

observation criterion

De-authentication attack

Spoofing

3.1 Threat Model for Civilian UAVs To date, the scope of UAVs for civilian purposes is very extensive [1–3, 8] For example, amateur drone technology is being used to acquire high-resolution imagery data of remote areas, islands, mountain peaks, and coastlines. UAVs have made performance mapping, construction monitoring and site inspection efficient, simple and fast. UAVs can be used in precision agriculture to collect data from ground sensors (water quality, soil properties, moisture, etc.), pesticide spraying, disease detection, irrigation planning, weed detection, and crop monitoring and management. The traffic monitoring system is an area where the integration of UAVs has also generated great interest. Some possibilities of using UAVs for civilian purposes are shown in Fig. 2.

Analysis of Threat Models for Unmanned Aerial Vehicles

599

Fig. 2. UAV Applications

Taking into account the areas of application of UAVs for civilian purposes, the authors proposed an assessment of the risks and damages from attacks on UAVs (Table 3). As estimates, it is proposed to use such a scale: 1. low damage/risk; 2. medium damage/risk; 3. high damage/risk. The authors believe that threats such as “Manipulation” bring the greatest harm to civilian UAVs. This is due to the fact that with this type of threat, it is possible to lose valuable cargo or disrupt the growth of plants in agriculture, etc. The least harm will be caused by threats related to the loss of confidentiality, since in most cases the information transmitted by UAV data can be obtained by any other legal means. 3.2 Threat Model for Military UAVs Unmanned aerial vehicles play an important role in military surveillance missions. Several countries have added UAVs to their defense strategic plans [1, 9, 10, 18, 22]. Countries use these flying robotic vehicles for enemy detection, anti-poaching, border control, and maritime surveillance of critical shipping lanes. Inexpensive, reliable and versatile UAVs are now making a significant contribution to aerial surveillance, monitoring and

600

H. Martyniuk et al. Table 3. Calculation of risks and damage for civilian UAVs

Type of threats

Risk assessment

Damage assessment

Overall Threat

Eavesdropping

1

1

2

Sniffing

1

1

2

Malware Infection

2

2

4

Manipulation

3

3

6

Jamming

2

3

5

Spoofing

2

3

5

De-authentication attack

1

3

4

inspection of any specific area to prevent any illegal activity. For example, surveillance of any threat can be detected by drones and they can be used to track any movement in any restricted area. The UAV can provide these services by receiving automatic alerts with minimal manual effort. In addition, using the example of Ukraine, in the conditions of a full-scale war in Ukraine, the issue of conducting aerial reconnaissance by means of unmanned aircraft is an important and relevant task [9, 22]. The collection and transmission of operational information requires the solution of a number of tasks, among which a separate place is occupied by the processing of data from the payload cameras directly on board the UAV, the preparation of such data for transmission and direct transmission to a ground workstation. Accordingly, the transmitted information must be protected from unauthorized access, depending on the degree of access restriction. Taking into account the areas of application of UAVs for military purposes, the authors proposed an assessment of the risks and damages from attacks on UAVs (Table 4). Table 4. Calculation of risks and damage for military UAVs Type of threats

Risk assessment

Damage assessment

Overall Threat

Eavesdropping

3

3

6

Sniffing

3

3

6

Malware Infection

1

2

3

Manipulation

1

2

3

Jamming

2

3

5

Spoofing

3

3

6

De-authentication attack

1

1

2

Analysis of Threat Models for Unmanned Aerial Vehicles

601

When conducting military operations, including aerial reconnaissance, the loss of confidential data poses the greatest threat, since not only the military, but also the civilian population can suffer from this. In this regard, the authors decided that the highest threat to military UAVs is the methods of “Eavesdropping”, “Sniffing” and “Spoofing”. 3.3 Threat Model for UAVs Used as a Means of Destruction of Targets In this section, attention will be paid to UAVs that can directly harm civilian or military targets. These aircraft include [19] reconnaissance UAVs, which are used to collect information about a certain telecommunications network with a subsequent attack on this network. This also includes UAVs for electronic warfare. This type of UAV is capable of affecting the performance of the telecommunications network, introducing distortions and creating false base and subscriber stations. It is also necessary to note strike UAVs (Fig. 3), which are capable of causing material damage and incapacitating the entire object of information activity. Taking into account the considered UAVs, the authors proposed an assessment of the risks and damages from attacks on UAVs (Table 5). Table 5. Calculation of risks and damage for UAVs as means of destruction targets Type of threats

Risk assessment

Damage assessment

Overall Threat

Eavesdropping

3

1

4

Sniffing

3

1

4

Malware Infection

1

1

2

Manipulation

1

1

2

Jamming

3

3

3

Spoofing

3

3

6

De-authentication attack

1

3

4

For such UAVs, the loss of confidentiality will not lead to a violation of the integrity or functional features of information activity objects. The accessibility criterion for such aircraft is also not in the first place. At the same time, breaching the integrity of information has the highest priority for attackers, as it will compromise the integrity of the information and result in the most damage. In this regard, the authors decided that for UAVs that are used as a means of destruction targets, the greatest total threat is posed by attacks such as “Jamming” and “Spoofing”.

602

H. Martyniuk et al.

Fig. 3. Types of strike UAVs [17]

4 Summary and Conclusion Cybersecurity issues and methods of attacks on UAVs have received much attention in the literature. There are a number of publications devoted to threats models. However, the threats models are of a general nature and do not depend on the scope of the UAV in practice. In this paper, the authors carried out work to distinguish between attacks and the damage that these attacks can cause, depending on the scope of the UAV. The authors proposed to consider the threats model for aircraft used for civil and military purposes. In addition, a threats model for UAVs is given, which are used to cause damage at the objects of information activity. These types of attacks on UAVs pose the greatest threat: 1) “Manipulation” threats do the most harm to civilian UAVs; 2) the highest threat to military UAVs are the methods of “Eavesdropping”, “Sniffing” and “Spoofing”; 3) the greatest total threat is posed by attacks such as “Jamming” and “Spoofing” when using UAVs as a means of destroying objects. The authors believe that such detailing of threat models, depending on the areas of application of UAVs, will make it possible to build the best protection for them.

References 1. Mohsan, S.A.H., Khan, M.A., Noor, F., Ullah, I., Alsharif, M.H.: Towards the Unmanned Aerial Vehicles (UAVs): A Comprehensive Review. Drones 6, 147 (2022). https://doi.org/10. 3390/drones6060147 2. Bauk, S., Kapidani, N., Sousa, L., Lukši´c, Ž., Spuža, A.: Advantages and disadvantages of some unmanned aerial vehicles deployed in maritime surveillance. The 8th Maritime Conference - MT2020 At: Barcelona, pp. 1–12 (2020) 3. Chan, K., Nirmal, U., Cheaw, W.: Progress on drone technology and their applications: a comprehensive review. AIP Conf. Proc. 2030, 020308 (2018). https://doi.org/10.1063/1.506 6949

Analysis of Threat Models for Unmanned Aerial Vehicles

603

4. Hryb, O.G., Karpaliuk, I.T., Shvets, S.V., Zakharenko, N.S.: Increasing the reliability of the power supply system through unmanned aerial vehicles. Zapysky VNTU 1, 1–6 (2020). (in Ukrainian) 5. Macrina, G., Pugliese, L.D.P., Guerriero, F., Laporte, G.: Drone-aided routing: a literature review. Transp. Res. Part C Emerg. Technol. 120, 102762 (2020) 6. Khan, M.A., Alvi, B.A., Safi, A., Khan, I.U.: Drones for good in smart cities: a review. In: Proceedings of the International Conference on Electrical, Electronics, Computers, Communication, Mechanical and Computing (EECCMC). Tamil Nadu, India (28–29 January 2018) 7. Alshbatat, A.I.N.: Fire extinguishing system for high-rise buildings and rugged mountainous terrains utilizing quadrotor unmanned aerial vehicle. Int. J. Ima. Graph. Sig. Proce. (IJIGSP) 10(1), 23–29 (2018). https://doi.org/10.5815/ijigsp.2018.01.03 8. Ramesh, K.N., Chandrika, N., Omkar, S.N., Meenavathi, M.B., Rekha, V.: Detection of rows in agricultural crop images acquired by remote sensing from a UAV. Int. J. Ima. Graph. Sig. Proce. (IJIGSP) 8(11), 25–31 (2016). https://doi.org/10.5815/ijigsp.2016.11.04 9. Prystavka, P., Sorokopud, V., Chyrkov, A., Kovtun, V.: Automated complex for aerial reconnaissance tasks in modern armed conflicts. CEUR Workshop, vol. 2588, 1–10 10. Unmanned Aerial System DEVIRO “Leleka-100”, http://deviro.ua/leleka-100, last accessed 16 October 2019. (in Ukrainian) 11. Chyrkov, A., Prystavka, P.: Suspicious Object Search in Airborne Camera Video Stream. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2018. AISC, vol. 754, pp. 340–348. Springer, Cham (2019). https://doi.org/10.1007/978-3-319-91008-6_34 12. SavasIlgi, G., Ever, Y.K.: Chapter Eleven - Critical analysis of security and privacy challenges for the Internet of drones: a survey Drones in Smart-Cities. Security and Performance 207–214 (2020) 13. Bakkiam, D.D., Fadi, A.-T.: Chapter One - Aerial and underwater drone communication: potentials and vulnerabilities. Drones in Smart-Cities. Security and Performance 1–26 (2020) 14. Ly, B., Ly, R.: Cybersecurity in unmanned aerial vehicles (UAVs). J. Cyb. Secu. Technol. 5(2), 120-137 (2021). https://doi.org/10.1080/23742917.2020.1846307 15. Gudla, C., Rana, S., Sung, A.H.: Defense techniques against cyber attacks on unmanned aerial vehicles. Int’l Conf. Embedded Systems, Cyber-physical Systems, & Applications, ESCS’18, pp. 110–116 16. Tsao, K.-Y., Girdler, T., Vassilakis, V.G.: A survey of cyber security threats and solutions for UAV communications and flying ad-hoc networks. Ad Hoc Networks, 1–39 (2022) 17. Siddiqi, M.A., Iwendi, C., Jaroslava, K., Anumbe, N.: Analysis on security-related concerns of unmanned aerial vehicle: attacks, limitations, and recommendations. MBE 19(3), 2641–2670 (2021) 18. Hartmann, K., Giles, K.: UAV Exploitation: A New Domain for Cyber Power. In: Pissanidis, N., Rõigas, H., Veenendaal, M. (eds.) 2016 8th International Conference on Cyber Conflict Cyber Power, pp. 205–241. 2016 © NATO CCD COE Publications, Tallinn 19. Voitenko, S., Druzhynin, V., Martyniuk, H., Meleshko, T.: Unmanned aerial vehicles as a source of information security threats of wireless network. International Journal of Computing 21(3), 377–382 (2022) 20. The Drone Cyberattack That Breached a Corporate Network. Accessed December 2022. https://blogs.blackberry.com/en/2022/10/the-drone-cyberattack-that-breached-acorporate-network 21. Yeo, M.: Electronic warfare in the land domain - The threats continue to increase. Asia-Pacific Defence Reporter 45(7), 20–22 (2019)

604

H. Martyniuk et al.

22. Kjellèn, J.: Drone-based jamming (REB-N). Russian Electronic Warfare. The role of Electronic Warfare in the Russian Armed, Swedish Defence Research Agency, p. 105 23. RD TPI 2.5-004-99. Criteria for assessing the security of information in computer systems from unauthorized access (in Ukrainian)

Identification of Soil Water Potential Sensors Readings in an Irrigation Control System Using Internet-of-Things (IoT): Automatic Tensiometer and Watermark 200SS Volodymyr Kovalchuk1(B) , Oleksandr Voitovich1 , Pavlo Kovalchuk1 , and Olena Demchuk2 1 Institute of Water Problems and Land Reclamation NAAS, Kyiv 03022, Ukraine

[email protected]

2 National University of Water and Environmental Engineering, Rivne 33028, Ukraine

[email protected]

Abstract. The analysis of world practice showed that various authors identified the relationship between soil water potential, Watermark 200SS gypsum block resistance and soil temperature P = P(R, T) for light soils. However, as our research shows, such dependencies cannot be applied to other types of soils. The method of identifying the parameters of the presented nonlinear model is also not given in the works of other authors. In the development of the Internet of Things (IoT), a method for identifying the dependences of soil water potential on the resistance of the Watermark 200SS gypsum block and soil temperature has been developed. In a laboratory experiment, the soil water potential (using a tensiometer) and the resistance value of a Watermark 200SS gypsum block are measured in parallel at certain temperatures using monoliths of an undisturbed structure of heavy loamy chernozem soil. Non-linear dependence is reduced to a linear relationship. The parameters of the linear model, which are used to identify the nonlinear model, are found by the method of least squares. Based on the non-linear dependence, the values of soil water potential are calculated based on Watermark 200SS indicators and soil temperature. Based on the soil water potential, watering timings are determined in irrigation control systems using the Internet of Things sensor system. Keywords: Sensor calibration · irrigation control · parameter identification · soil water potential · least squares method

1 Introduction 1.1 Formulation of the Problem In the context of climate change, the rational use of irrigation water is the basis for the sustainable development of agriculture. It is impossible without the use of the Information System for Operational Irrigation Control, which is based on field networks of various © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 605–614, 2023. https://doi.org/10.1007/978-3-031-36118-0_54

606

V. Kovalchuk et al.

sensors, that is, on the use of the Internet of Things (IoT) [1–7]. Many researchers use Watermark 200SS sensors to control irrigation in their field moisture sensor networks. Irrometer Inc. Developed and patented a whole series of such sensors [8]. However, the given formula for the relationship between soil water potential and Watermark 200SS resistance and soil temperature P = P(R,T) is recommended for soils of light mechanical composition. The problem is to develop a method for conducting experimental studies and identifying dependencies based on experimental data for an arbitrary type of soil. 1.2 Literature Review Modern systems for measuring soil moisture content use the physical properties of the soil, such as volumetric moisture and soil water potential. The fundamental mathematical model of the relationship between the volumetric soil moisture and the soil water potential Θ = f (P) was proposed by Van Genuchten, M.T. [9]. Functional features of the relationships model, depending on the soil mechanical composition, are calculated by the Rosetta program [10]. In this way, the measured soil moisture or soil water potential can easily be converted from to each other and used in the irrigation control system to determine watering. In a previous publication [5] we recommended to use an automatic tensiometer and Watermark 200SS for building field networks of sensors that measure soil water potential in irrigation control systems. Our recommendations are supported by a significant body of literature, where Watermark 200SS sensors [11–15] or automatic tensiometers [1, 14, 16–18] are also used for these purposes. The Watermark 200SS measures moisture indirectly, namely by measuring the variable resistance of a wet gypsum block [8], as water exchange between the soil and the sensor changes the electrical resistance between the electrodes in the sensor. Sensor readings are identified in units of soil water potential, and electrical resistance depends on soil moisture and temperature [11]. The authors [11, 12] identified the Watermark 200SS readings for soils of light mechanical composition based on research. Studies [19] revealed a significantly lower correlation coefficient for soils of heavy mechanical composition when obtaining the dependence between the soil moisture and soil water potential. The work [14] indicates an overestimated value of the dryness of the Watermark 200SS readings for heavy soils compared to other measurement methods. Therefore, it is necessary to calibrate such sensors of soil water potential for soils of heavy mechanical composition for their use in irrigation control systems. It is known that a mechanical [20] or modern automatic tensiometer [16–18, 21] directly measures the vacuum pressure, which is equal in modulus to the soil water potential. The tensiometer consists of a sealed system: a sensitive ceramic porous sensor element buried in the soil, a tube that goes to the surface and a hydraulically connected mechanical indicator (vacuum meter) or an automatic sensor for vacuum measuring. This equipment (tensiometer) is also often used in irrigation control systems [1, 5, 16–18] to monitor soil water potential and make decisions about watering. The accuracy and reliability of the tensiometer in determining the soil water potential is high [1, 14, 15, 20, 21] and largely depends on the accuracy of vacuum pressure

Identification of Soil Water Potential Sensors Readings

607

measuring instruments. To a lesser extent, the accuracy depends on the quality of the ceramic probe immersed in the soil layer [20]. Since the measurement data using this sensor does not require any additional interpretation, it is used to identify other sensors readings with indirect measurement methods [1, 12, 14, 15], such as the Watermark 200SS [12, 14, 15]. 1.3 The Purpose of the Study The purpose of this study is to develop a method for identifying the relationship between soil water potential and Watermark 200SS readings and soil temperature values based on data from laboratory studies of soil samples for various soils types. The identified dependencies are used to calculate the soil water potential and assign watering in irrigation control systems using the IoT sensor system.

2 Research Methods The following there is the identification method based on experimental studies and mathematical modeling of the relationship between soil water potential and electrical resistance measured by Watermark 200SS at different temperatures. 2.1 Experimental Equipment for Data Obtaining Experimental equipment (Fig. 1), which is able to maintain the soil temperature in the soil monoliths of an undisturbed structure within the specified limits using the controller (5) Arduino Nano V3.0 AVR ATmega328P with a relay that controls the aquarium heater (4) was designed and manufactured. The transfer of commands by the user to the server and from the server to the relay through the Arduino controller is carried out remotely using the GPRS module. The water temperature and the temperature inside the soil monoliths are measured by digital sensors (3) Dallas DS18B20, located inside a stainless flask using thermally conductive paste. The data of vacuum pressure and electrical resistance of the Watermark are received and collected using the Arduino Nano V3.0 AVR ATmega328P controller (6) (Fig. 1) and transmitted to the server using Wi-Fi. Watermark 200SS sensors (2) were calibrated using automatic tensiometers (Fig. 1), which were developed at the Institute of Water Problems and Land Reclamation (IWPLR) of the National Academy of Agrarian Sciences of Ukraine (UA utility model patent No. 132271 [21]) based on a BMP280 digital vacuum gauge (1). And although the BMP280 has a significant drawback, it is afraid of water getting on the sensitive element; its advantage is high accuracy in measuring vacuum pressure. It can be argued that both sensors measure the water potential of soil moisture: the automatic tensiometer directly measures the vacuum pressure equal to the modulus of the soil water potential, and the Watermark 200SS is calibrated in potential units.

608

V. Kovalchuk et al.

Fig. 1. Experimental equipment for measurement of readings of soil water potential sensors and Watermark 200SS (the designations are given in the text of the publication)

2.2 Model Identification Method At present, the relationship between the parameters P = P(R,T) of Watermark 200SS sensors is determined by the formula [11]: P=

−a − bR . 1 − cR − dT

(1)

where P is the soil water potential (module of vacuum pressure), kPa; R is the resistance, k; T is the temperature, °C; a, b, c, d are desired coefficients. This formula was proposed by American scientists [11], verified [12] for soils of light mechanical composition. They obtained the following calibration coefficients: a = 3.213; b = 4.093; c = 0.009733; d = 0.01205. However, for soils of heavy mechanical composition, the identification of dependence (1) is required on the basis of experimental studies and mathematical modeling. To identify the coefficients of non-linear Eq. (1), we rewrite it in the form: S(1 − cR − dT ) = −a − bR,

(2)

S = −a − bR + cSR + dST .

(3)

or.

Identification of Soil Water Potential Sensors Readings

609

Dependence (3), in contrast to equality (1), is linear. Therefore, the coefficients a, b, c, d can be identified by the least squares method [22]. S, R, T are determined according to the data of the laboratory experiment. Then for the initial value S, the input values S*R, S*T, R of linear dependence (3) are calculated. The coefficients a, b, c, d of the linear dependence (3), calculated from the experimental data by the least squares method, are substituted into the nonlinear formula (1). We obtain dependence (1) for a given type of soil, which is used to bind the readings of both sensors when assigning watering.

3 Research Results and Discussion 3.1 Results of a Laboratory Experiment for Heavy Loamy Chernozem Soil Monoliths of the undisturbed structure of heavy loamy chernozem soil were selected at the experimental production site of the State Enterprise “Experimental farming Andriyivske” of IWPLR in Odessa region from the arable (10–25 cm) and underarable (30–45 cm) horizons. In the process of laboratory research, there was a problem with the cracking of the monolith with the soil of the underarable horizon and, as a result, the constant loss of vacuum by the soil medium and the tensiometer. The physical properties of both soils are similar. Therefore, despite some loss of information content of the experiment, the results are given according to the studies of one monolith. The monoliths were tested in the laboratory at two temperature thresholds of about 20.6 °C and about 24.65 °C inside each soil monolith (value was averaged over Dallas DS18B20 readings [23], ± 0.5 °C temperature accuracy from −10 °C to + 85 °C). The monoliths were saturated close to full capacity, and then dried naturally. The vacuum pressure with a tensiometer (based on the BMP280 sensor element [24], pressure resolution is 0.16 Pa), the gypsum block resistance in the Watermark 200SS sensor and the temperature of the water around the soil monoliths were automatically recorded. A comparison of the dynamics of vacuum pressure according to the readings of the automatic tensiometer and the calculated values of the readings from Watermark according to formula (1) with the coefficients [11] for light soils at different values of the gypsum blocks electrical resistance of the sensor was carried out (Fig. 2). Section 2 has a temperature threshold of about 20.6 °C, and Sect. 3 has a temperature threshold of about 24.65 °C. Visual analysis of dependencies in Fig. 2 shows the systematic deviations between the readings of the automatic tensiometer (vacuum in kPa) and the calculated vacuum values (soil water potential) according to the electrical resistance R data from Watermark on all four sections of the curves. The relative error in Sects. 2 and 3 of the systematic deviations to calibration was calculated using the formula: i − Xi | 1  |Xact calc ∗ 100%, i | n |X act i=1 n

M [X ]relative =

(4)

i are the experimental values of the potential (vacuum pressure); X i where Xact calc are calculated values of potential (vacuum pressure) according to formula (1) with coefficients i [11]. The error between the Xact values for deep drying of the soil monolith and the Xcalc

610

V. Kovalchuk et al.

Fig. 2. Comparison of vacuum pressure dynamics based on (a) automatic tensiometer readings and (b) calculated values based on electrical resistance data from Watermark according to formula (1)

values is about 80%. This confirms the need to identification of the sensor Watermark readings according to the experimental data for soils of heavy mechanical composition. According to the experiment plan, after the initial unstable period and soil monolith maximum saturation reaching from 01/24/2022 15:37, its natural drying began at an average temperature of 20.6 °C, which was detected by the sensors (Fig. 2, curve’s Sect. 2). The maximum vacuum value was reached on 02/04/2022 12:06. Another drying cycle (Sect. 3) after reaching the maximum saturation of the monolith with water lasted from 02/07/2022 09:19 to 02/21/2022 07:31. The sensors recorded the average temperature for this cycle of 24.65 °C. Based on the obtained data, two curves of the dependence R(P) were plotted at a certain constant value of the soil temperature (Fig. 3). As can be seen from Fig. 3, in the area close to the full moisture capacity (P = 0…8–10 kPa.), there is an uncertainty in the electrical resistance of Watermark 200SS, equal to a certain constant value.

Fig. 3. Graphs of experimental dependences R(P) for further mathematical processing

Identification of Soil Water Potential Sensors Readings

611

3.2 Mathematical Modeling and Discussion of Results As a result of mathematical processing of the experimental data (Fig. 3) by the least squares method (Table 1), the coefficients of formula (1) for soils of heavy mechanical composition were obtained: a = 9; b = 2.7; c = 0.004484; d = 0.011. The root mean square error between the calculated values and the experimental values is RMSE = 10.77. The formula with the obtained coefficients, taking into account the satisfactory accuracy of the results, should be used when assigning irrigation for heavy soils. The dependence of vacuum pressure (soil water potential) was calculated for various values of soil temperature with the obtained coefficients. Dependence graphs are shown in Fig. 4.

Fig. 4. Graph of dependence between the Watermark 200SS electrical resistance, temperature and vacuum pressure (soil water potential)

Regarding the identification results of Watermark 200SS readings by other authors, the most well-known is the verification of formula (1) by the authors of [12] for soils of light mechanical composition. In the study, the authors used a number of temperature thresholds in the temperature range of 21–30 °C. In addition, the authors of [13] attempted to describe the dependence P(R,T) by several equations, dividing the R curve into several sections and comparing the obtained data at T = 24 °C with the calculated values of the potential according to the Van Genuchten-Mualem model by the USDA Rosetta program [10] without using the control measuring instrument. Studies on the identification of Watermark 200SS readings [15] were carried out at a temperature of 24 °C for the sand silty soil, which can be attributed to the middle mechanical class. In [25] calibration was carried out in the field by comparing the measured resistance of Watermark 200SS with the known electrical resistance of the resistors.

612

V. Kovalchuk et al.

Table 1. An example of applying the least squares method to calculate the coefficients of the model (3) N

S

R

T

Ai1 (S*R) Ai2 (S*T) Ai3 (-R) Bi (S) Bi’

1

−1

0

20,6

0

−20,6

0

−1

−0,87357

0,015985

2

−10

2

20,6

−20

−206

−2

−10

−9

1

3

−10

1,8

24,65 −18

−246,5

−1,8

−10

−10,691

0,477505

4

−20

4,6

20,6

−92

−412

−4,6

−20

−18,1129

3,561268

5

−20

4,2

24,65 −84

−493

−4,2

−20

−21,492

2,226012

6

−30

7

20,6

7

−30

6,5

24,65 −195

8

−40

6

20,6

9

−210 −240

−40

9,5

24,65 −380

10 −45

3,9

20,6

11 −45

10,9

24,65 −490,5

12 −50

11,8

20,6

13 −50

10,6

24,65 −530

14 −52

12,3

20,6

15 −52

11,1

−175,5 −590

(Bi-Bi’)2

−618

−7

−30

−27,2343

7,648925

−739,5

−6,5

−30

−32,3133

5,35136

−824

−6

−40

−35,867

17,08146

−986

−9,5

−40

−43,276

10,73236

−927

−3,9

−45

−39,9256

25,74981

−1109,25

−10,9

−45

−48,758

14,12287

−1030

−11,8

−50

−45,5823

19,5157

−1232,5

−10,6

−50

−53,976

15,8083

−639,6

−1071,2

−12,3

−52

−47,4281

20,90221

24,65 −577,2

−1281,8

−11,1

−52

−56,1635

17,33447

Therefore, we think that the use of an experimental equipment for several temperature thresholds with the ability to constantly maintain it, in comparison with the calibration examples described above, makes it possible to obtain experimental data at various temperature thresholds (similar to [12]), to highlight in detail the effect of temperature on the Watermark 200SS resistance and on soil water potential for any soil type. In particular, this is done for soils of heavy mechanical composition. Results of Watermark 200SS sensor calibration can be performed both for a single field soil and other soil types. They are extrapolated for other temperature values and for any soils of heavy mechanical composition. This makes it possible to effectively use the Watermark 200SS in irrigation management systems for watering assigning.

4 Summary and Conclusion As the analysis showed, in irrigation control systems many farmers use Watermark 200SS sensors for field networks for measuring soil moisture by indirect methods. However, in order to watering assigning through measurements by these sensors, it is necessary to identify their readings of the soil water potential. There is such dependence only for light mechanical composition soils. Therefore, in our study, the problem of developing a method for finding a functional dependence between the soil water potential, the Watermark 200SS gypsum block resistance and the soil temperature P(R,T) was formulated and solved. The method makes

Identification of Soil Water Potential Sensors Readings

613

it possible to obtain experimental data on the relationship between electrical resistance and soil water potential for an arbitrary soil sample on experimental equipment. On the basis of experimental data, a method of identifying the parameters of linear dependence is presented, which ensures the identification of a nonlinear model of the relationship P = P(R,T). The experimental approach and the identification of non-linear models provided a constructive implementation of the method for identifying the non-linear relationship between water potential, resistance and soil temperature P = P(R,T) for soils of heavy mechanical composition. In the future, it is possible to implement the method for creating a knowledge base of model’s coefficients in automated irrigation control systems during watering.

References 1. Maddah, M., Olfati, J.A., Maddah, M.: Perfect irrigation scheduling system based on soil electrical resistivity. Int. J. Veg. Sci. 20(3), 235–239 (2014). https://doi.org/10.1080/19315260. 2013.798755 2. Lozoya, C., Mendoza, C., Aguilar, A., Román, A., Castelló, R.: Sensor-based model driven control strategy for precision irrigation. Journal of Sensors (2016). https://doi.org/10.1155/ 2016/9784071 3. Payero, J.O., Mirzakhani-Nafchi, A., Khalilian, A., Qiao, X., Davis, R.: Development of a Low- Cost Internet-of-Things (IoT) System for Monitoring Soil Water Potential Using Watermark 200SS Sensors. Advances in Internet of Things 7, 71–86 (2017). https://doi.org/ 10.4236/ait.2017.73005 4. Okine, A., Appiah, M., Ahmad, I., Asante-Badu, B., Uzoejinwa, B.: Design of a green automated wireless system for optimal irrigation. Int. J. Comput. Netw. Inf. Secur. 12(3), 22–32 (2020). https://doi.org/10.5815/ijcnis.2020.03.03 5. Kovalchuk, V., Voitovich, O., Demchuk, D., Demchuk, O.: Development of Low-Cost Internet-of-Things (IoT) Networks for Field Air and Soil Monitoring Within the Irrigation Control System. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2020. AISC, vol. 1247, pp. 86–96. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-55506-1_8 6. Akwu, S., Bature, U.I., Jahun, K.I., Baba, M.A., Nasir, A.Y.: Automatic plant irrigation control system using Arduino and GSM module. Int. J. Eng. Manuf. (IJEM) 10(3), 12–26 (2020). https://doi.org/10.5815/ijem.2020.03.02 7. Anusha, K., Mahadevaswamy, U.B.: Automatic IoT based plant monitoring and watering system using Raspberry Pi. Int. J. Eng. Manuf. (IJEM) 8(6), 55–67 (2018). https://doi.org/ 10.5815/ijem.2018.06.05 8. Irrometer Inc. WATERMARK Soil Moisture Sensor. https://www.irrometer.com/200ss.html 9. Van Genuchten, M.T.: A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44(5), 892–898 (1980) 10. Rosetta Version 1.0 (Free downloaded program). U.S.Salinity Laboratory ARSUSDA. www. ussl.ars.usda.gov. Retrieved from: http://www.ussl.ars.usda.gov 11. Shock, C.C., Barnum, J.M., Seddigh, M.: Calibration of Watermark Soil Moisture Sensors for Irrigation Management. Malheur Experiment Station, Oregon State University (1998). Retrieved from https://www.researchgate.net/profile/Clinton-Shock/2 12. Chard, J.: Watermark soil moisture sensors: characteristics and operating instructions. Utah State University (2002). https://www.researchgate.net/publication/237805713 13. Radman, V., Radonji´c, M.: Arduino-based system for soil moisture measurement. Proc. 22nd Conference on Information Technologies IT. vol. 17 (2017)

614

V. Kovalchuk et al.

14. Kumar, J., Patel, N., Rajput, T.B.S., Kumari, A., Rajput, J.: Performance evaluation and calibration of soil moisture sensors for scheduling of irrigation in brinjal crop (Solanum melongena L. var. Pusa Shyamla). Journal of Soil and Water Conservation 19(2), 182–191 (2020). https://doi.org/10.5958/2455-7145.2020.00025.9 15. Vettorello, D.L., Marinho, F.A.: Evaluation of time response of GMS for soil suction measurement. In: MATEC Web of Conferences, vol. 337, p. 01014. EDP Sciences (2021). https:// doi.org/10.1051/matecconf/202133701014 16. Thalheimer, M.: A low-cost electronic tensiometer system for continuous monitoring of soil water potential. J. Agri. Eng. 44(3), XLIV: e16. (2013). https://doi.org/10.4081/jae.2013.e16 17. Matus, S.K.: Information and measurement system of data collection and control of soil moisture reserves. Bulletin of the National University of Water and Environmental Engineering. Technical Sciences (2), 198–208 (2014). (in Ukrainian) 18. Pereira, R.M., Sandri, D., Rios, G.F.A., Sousa, D.A.: Automation of irrigation by electronic tensiometry based on the arduino hardware platform. Revista Ambiente & Água 15 (2020). https://doi.org/10.4136/ambi-agua.2567 19. Jabro, J., Evans, R., Kim, Y.: Estimating in situ soil-water retention and field water capacity in two contrasting soil texture. Irrig. Sci. 27, 223–229 (2009). https://doi.org/10.1007/s00271008-0137-9 20. Romashchenko, M.I., Koriunenko, V.M.: Recommendations for operational control of the crops’ irrigation regime using the tensiometric method. K.: DIA Ltd. (2012). (in Ukrainian) 21. Kovalchuk, V.P., Voitovich, O.P., Demchuk, D.O.: Ukrainian Patent for Utility Model UA132271 (25 February 2019). https://sis.ukrpatent.org/en/search/detail/1223767/ 22. Lawson, Ch., Henson, R.: Numerical solution of problems by the method of least squares. — M.: Nauka (1986). (in Russian) 23. Maxim Integrated Products, Inc. DS18B20 Programmable Resolution 1-Wire Digital Thermometer: Data sheet. Retrieved from: https://www.analog.com/media/en/technical-docume ntation/data-sheets/DS18B20.pdf 24. Adafruit. Digital Pressure Sensor BMP280: Data sheet. Retrieved from: https://cdn-shop.ada fruit.com/datasheets/BST-BMP280-DS001-11.pdf 25. Fisher, D.K.: Automated collection of soil-moisture data with a low-cost microcontroller circuit. Appl. Eng. Agric. 23(4), 493–500 (2007)

Machine Learning Algorithms Comparison for Software Testing Errors Classification Automation Liubov Oleshchenko(B) National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Kyiv 03056, Ukraine [email protected]

Abstract. Automatic classification of software errors is an important tool for ensuring the quality of software being tested. By automating this process, testers can save time and effort, improve accuracy, gain a better understanding of errors, and improve communication between stakeholders. Software error clustering is an important concept in software testing that involves identifying and analyzing patterns in software errors or defects. The goal of error clustering is to understand the relationships between defects and identify the root causes of software errors, so that they can be prevented or corrected. The article provides a comprehensive survey of various techniques for software error clustering, including clustering algorithms. The author note that many of these techniques require significant expertise in software engineering and data analysis, and that there is a need for more user-friendly tools for software error clustering. This article presents an empirical research of software defect clustering, in which the author analyzes a large dataset of defects from a software development projects to identify several patterns in the defects, including clusters of related defects and common root causes. The proposed software method uses stack traces to cluster data about software testing errors. The method uses a kNN algorithm to analyze test results, allowing the user to assign the text of the software test result to specified categories. The kNN algorithm has a high accuracy rate of 0.98, which is better than other clustering methods like Support Vector and Naive Bayes. The proposed method has several advantages, including a single repository for test results, automatic analysis of software testing results, and the ability to create custom error types and subtypes for error clustering. The developed new software method allows multiple launches to be combined. If there are a large number of test suites, they are divided into smaller groups as they cannot be included in a single run. The test suites can then be combined into a single run to present the data on dashboards and generate reports. The method is integrated into software using a Docker container, which reduces the time and human resources required for error analysis. Overall, the method provides real-time monitoring of project status for managers and customers. Keywords: Machine learning algorithms · software defect · software errors classification · clustering · automation

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 615–625, 2023. https://doi.org/10.1007/978-3-031-36118-0_55

616

L. Oleshchenko

1 Introduction Automatic classification of software errors is an important aspect of software testing, as it helps ensure that all errors are identified and categorized correctly. Software error clustering refers to the phenomenon where software defects are not randomly distributed across software components, but rather tend to occur in clusters within specific modules or parts of the system. This can have important implications for software testing and quality assurance, as it suggests that testing efforts should be focused on these high-risk areas of the system. Consider the several key benefits to automating this process. 1. Improved testing efficiency: automating the classification of software errors can save time and effort that would otherwise be spent manually identifying and categorizing errors. This can help speed up the testing process and make it more efficient. 2. Enhanced accuracy: automated error classification can help reduce human errors that may occur during the manual classification of software errors. This can help ensure that all errors are correctly identified and categorized, which is important for ensuring the overall quality of the software. 3. Better understanding of errors: by automatically categorizing errors, testers can gain a better understanding of the types of errors that are occurring and the frequency with which they occur. This can help them identify patterns and trends that may be useful in improving the software and preventing similar errors in the future. 4. Improved communication: automated error classification can help improve communication between testers, developers, and other stakeholders. By categorizing errors in a standardized way, everyone involved in the software development process can more easily understand the issues that need to be addressed. There are several machine learning algorithms that can be used for software error classification during software testing. K-means is a popular clustering algorithm that can be used to group software errors based on their similarity. The algorithm partitions the errors into k clusters, where k is a user-defined parameter. Each error is assigned to the nearest cluster based on its features or characteristics. Hierarchical clustering is another popular algorithm that can be used to group errors based on their similarity. The algorithm builds a tree-like structure of clusters, where the leaf nodes represent individual errors and the internal nodes represent groups of errors that share some common characteristics. Density-based clustering algorithms such as DBSCAN (Density-Based Spatial Clustering of Applications with Noise) can be used to identify clusters of errors that are dense in a feature space. These algorithms can be particularly useful when the errors are distributed in a non-uniform manner. Fuzzy clustering algorithms such as Fuzzy C-means can be used to assign errors to multiple clusters with varying degrees of membership. This can be useful when errors have characteristics that overlap with multiple clusters. It’s important to note that the choice of clustering algorithm depends on the specific characteristics of the errors being classified, as well as the goals of the software testing process.

Machine Learning Algorithms Comparison

617

1.1 Analysis of Recent Research and Related Work The study [1] provides a systematic review of software error clustering techniques, focusing on recent research from the past decade. The authors identify several key challenges in software error clustering, including the lack of standardized data formats and the difficulty of interpreting clustering results. The authors also highlight several promising approaches for software error clustering, including machine learning techniques and graph-based methods. They suggest that these approaches can provide more accurate and comprehensive analysis of software errors, and can be applied to a wide range of software development contexts. Software error clustering is a phenomenon in software testing where a particular software defect or bug is found to occur in multiple instances. In other words, certain errors tend to cluster together, indicating a deeper underlying issue in the software code. The article [2] presents an empirical study of error clustering in commercial software applications. The study found that a small percentage of software errors were responsible for a large percentage of the total defects in the software. The authors recommend using error clustering to identify high-priority areas for software testing and debugging. The article [3] presents an analysis of software error clustering in open-source software projects. The study found that errors tended to cluster in certain areas of the software code, such as specific modules or functions. The authors recommend using clustering analysis to identify areas of the software that are particularly prone to errors. The article [4] presents a study of error clustering in large-scale software systems. The study found that errors tended to cluster in certain parts of the code, such as frequently executed functions or those with complex logic. The authors recommend using clustering analysis to prioritize testing and debugging efforts in large-scale software systems. By analyzing the distribution of errors in software code, software developers and testers can gain insight into the underlying causes of defects and improve the quality of their software. The authors of article [5] conducted a systematic review of research studies on software error clustering, analyzing the methods used to identify and quantify clustering, the types of software defects that tend to cluster, and the factors that contribute to clustering. The article highlights several key findings from the research studies reviewed. For example, the studies generally found that a small number of modules or components tend to contain the majority of software defects. The studies also found that different types of software defects tend to cluster in different ways; for example, logic errors may cluster differently than memory-related errors. The authors also discuss some of the challenges and opportunities associated with studying and addressing software error clustering. For example, they note that identifying and quantifying clustering can be difficult, particularly in large and complex software systems. They also note that addressing clustering may require a shift in focus from detecting and fixing individual defects to identifying and addressing underlying structural or design issues. The article [6] also reviews research studies on software error clustering, with a focus on the methods used to identify and measure clustering, as well as the techniques used to address clustering in software testing. The authors highlight several key trends and challenges in the research on software error clustering. For example, they note that different techniques for identifying and measuring clustering have been proposed, but it

618

L. Oleshchenko

can be difficult to determine which technique is most appropriate for a particular software system. They also note that addressing clustering may require a combination of testing techniques, such as static analysis, dynamic analysis, and mutation testing. Overall, the last articles [7–19] highlight the importance of understanding and addressing software error clustering in software testing. By identifying high-risk areas of the software system, testing efforts can be focused more effectively, leading to better software quality and more efficient use of testing resources. These articles provide valuable insights into the challenges and opportunities of software error clustering in software testing. They highlight the importance of using multiple techniques for software error clustering, and suggest that advances in machine learning and data analysis can help improve the accuracy and efficiency of software testing and quality assurance. 1.2 Research Data To solve the problem of identifying and fixing software errors in a timely manner the author proposes an automated solution that can achieve high accuracy at all stages of software development using the microservice architecture to compare errors found during software testing. The dataset of 100,000 stack traces from software companies in Ukraine was use and was develop the software method for classifying testing errors using Java programming language and MongoDB as a database. To analyze the test results, author used HTML, CSS, JS technologies, Angular framework, Node. Js and compared Naive Bayes, kNearest Neighbors, Support Vector Machines machine learning methods. The research aims to develop an automated software error analysis process that can reduce the time it takes to identify and fix errors during software development. 1.3 Objective The object of the research in this article is creating the software method to automatically classify software testing errors more accurately and quickly using machine learning algorithms. The goal is to improve the efficiency and accuracy of identifying errors in software testing.

2 Research Method 2.1 Clustering Algorithms Comparison for Software Errors Classification Naive Bayes (NB) is a probabilistic algorithm that can be used for software error classification during software testing. It is a simple and efficient algorithm that is often used for text classification tasks, but it can also be applied to software error classification. NB algorithm works by calculating the probability of each error belonging to a particular class (i.e., error category). It does this by using Bayes’ theorem, which calculates the conditional probability of a class given the error features, assuming that the features are conditionally independent. This assumption of independence is called “naive” and is often violated in real-world applications, but the algorithm still tends to perform well

Machine Learning Algorithms Comparison

619

in practice. To apply the NB algorithm to software error classification, we need to first define the error features (i.e., input variables) that we will use to train the model. These features could include the error message, stack trace, error type, error severity, and other relevant information. Once the features have been defined, we need to train the Naive Bayes model on a set of labeled error data. During the training phase, the algorithm learns the conditional probability of each feature given each error class. To classify a new error, the NB algorithm calculates the conditional probability of the error belonging to each class, using the learned probabilities from the training data. The class with the highest probability is then assigned to the error. Overall, NB algorithm can be a useful approach for software error classification during software testing, especially when the number of error classes is relatively small, and the features are well-defined and easily measurable. It is important to note that the performance of the algorithm depends on the quality and relevance of the error features, as well as the quality and size of the training data. Support Vector Machines (SVM) is a popular machine learning algorithm that can be used for software error classification during software testing. SVM is a supervised learning algorithm that is often used for classification tasks in which the input data is linearly separable. To apply SVM to software error classification, we first need to define the error features and create a labeled dataset of error instances. The features could include information such as error message, stack trace, error type, and severity. Next, we train an SVM model on the labeled dataset of error instances. During the training process, the SVM algorithm uses the error features to create a hyperplane that separates the errors into different classes. The goal of the algorithm is to find the hyperplane that maximizes the margin (i.e., the distance between the hyperplane and the closest error instances of each class). Once the SVM model has been trained, it can be used to classify new error instances. The SVM algorithm maps the error features to a point in a highdimensional space, and then uses the hyperplane to classify the error into one of the predefined classes. One advantage of SVM algorithm is that it can handle non-linearly separable data by using kernel functions to transform the input features into a higherdimensional space where the data becomes separable. Additionally, SVM is relatively insensitive to the presence of irrelevant features, which can be useful when working with error data that contains noise or unimportant variables. Overall, SVM algorithm can be a useful approach for software error classification during software testing, especially when the error data is well-defined and the classes are linearly or non-linearly separable. However, it is important to carefully select and engineer the features used for classification, as well as tune the hyperparameters of the model to achieve the best performance. The k-Nearest Neighbors (kNN) algorithm can be used for software error classification during software testing. The kNN algorithm is a non-parametric and lazy learning algorithm that can be used for both classification and regression tasks. To apply kNN to software error classification, we first need to define the error features and create a labeled dataset of error instances. The features could include information such as error message, stack trace, error type, and severity. Next, we need to choose a value for k, which represents the number of nearest neighbors to consider when classifying new error instances. Then, when a new error instance is presented, the kNN algorithm searches

620

L. Oleshchenko

for the k nearest neighbors in the labeled dataset based on the similarity of their error features to those of the new instance. Once the k nearest neighbors have been identified, the kNN algorithm uses the class labels of those neighbors to assign a class label to the new error instance. This is done by majority vote, where the class with the most instances among the k nearest neighbors is assigned to the new instance. One advantage of the kNN algorithm is that it is simple and easy to implement. It also doesn’t make assumptions about the underlying distribution of the data, making it applicable to a wide range of problems. However, the performance of the kNN algorithm can be sensitive to the choice of k and the distance metric used to calculate the similarity between error instances. Overall, the kNN algorithm can be a useful approach for software error classification during software testing, when the classes are well-defined. However, it is important to carefully select and engineer the features used for classification, as well as choose an appropriate value for k and distance metric. 2.2 Proposed Software Error Classification Method The new proposed method involves forming a stack trace after running tests to identify errors in software. The stack trace contains messages about software errors, and describes the reasons for failed tests and shows where the errors are located in the software files at different levels of the libraries. The proposed method repeats this process for each test, removing repeated text and irrelevant information like dates, and analyzes the unique information specific to each test. This allows for the classification of software errors using clustering algorithms. (Fig. 1).

Fig. 1. Code text for clustering

After gathering a significant amount of redundant text, the text is analyzed to collect information on the frequency of repeated words for each error category. This information is shown in Fig. 2. As more tests are performed and more text is collected, the accuracy of the kNN algorithm improves. After obtaining the metrics on the frequency of repeated words for each error category, the proposed method uses two metrics called Term Frequency (TF) and Inverse Document Frequency (IDF). These metrics help to determine the significance of words in documents and assign them to the appropriate class of software errors. In the dataset that was analyzed, the errors causing failed tests were initially categorized into three groups: Product Bugs, System Issues, and Automation Issues. The

Machine Learning Algorithms Comparison

621

Fig. 2. Frequency indicators example

text from the files was then represented as a mathematical vector in a multi-dimensional space. 2.3 The Software Architecture of the Proposed Method The software system for error classification was developed using a microservices approach. The client part of the software consists of Logger, Agent, and Client services. HTTP clients are used to send HTTP requests through the API. The proposed software system for organizing tests is structured in four levels: Launch, Test Suite, Test Case, and Test Step. The Launch level contains sets of tests that were performed during a launch. The Test Suite level groups test cases that are related to one testing module, functionality, priority, or type of testing. Each Test Suite includes multiple Test Cases and is typically performed as a whole in the testing process. The Test Case level describes the algorithm for testing a program to determine the occurrence of a particular situation and source data. The Test Step level describes the steps required to reproduce a bug, with the aim of keeping these steps as short as possible. This level also allows users to identify at which step the error occurred and view the launch history. The result of the test case execution is displayed, including all its steps and software errors that occurred during the launch. Users can switch between different runs and view the errors that occurred in the past. Attachments can also be added to each step, such as images of code fragments, which can be viewed by clicking on the corresponding attachment in the developed software system. The developed software method allows multiple launches to be combined. If there are a large number of test suites, they are divided into smaller groups as they cannot be included in a single run. The test suites can then be combined into a single run to present the data on dashboards and generate reports. The software implements both linear and deep merging (as shown in Fig. 3). If the user selects the “Linear Merge” option, a new launch is created that includes elements from the merged launches, while the item levels remain the same as in the original launch.

622

L. Oleshchenko

Fig. 3. Test launches visual scheme: a) before merging, b) after linear merger

This passage describes a feature of a software system that displays the result of a test case execution, including any errors that occurred during the launch. The system also saves the last 10 test runs, allowing users to switch between them and see previous errors. Additionally, there is a feature to attach images of code fragments to the corresponding step, which can be opened in a pop-up window. The system also has a Stack trace tab that shows the entire stack trace used in automatic analysis. The root cause of the test failure can be written in the first lines of the log lines, and the analyzer can be configured to consider only the necessary lines.

Fig. 4. Launch statistics chart in startup mode of proposed software method

The software system that has been suggested enables users to design and build their own widgets. These widgets have specialized graphical tools that are intended to make it simple to view and evaluate the outcomes of the test automation process. (Fig. 4). The widget presents an overview of the statistics for every test run, including the total number of tests, as well as the number of tests that passed, failed, or were associated with various types of errors. These types of errors include software product bugs, system issues, automation bugs, defects that require investigation, and those that are not considered defects. Each type of error is displayed as a percentage of the total. The

Machine Learning Algorithms Comparison

623

widget also has interactive sections; if a user clicks on a specific section in the widget, the system will redirect them to the corresponding display (Fig. 5).

Fig. 5. Statistics panel of proposed software method

3 Research Results The proposed software system developed in this research involved comparing the kNN, Naive Bayes, and SVM methods. After creating a software testing method using the kNN algorithm, a real software project was used to compare the proposed method with manual analysis and classification of failed tests. An analysis of 450 tests was conducted, with the kNN algorithm completing the task in just one hour, compared to the approximately 12 h it takes a human tester to manually read and categorize the errors. This indicates that the algorithm was 12 times faster than human testers, demonstrating the effectiveness of the proposed error classification method. Based on the research, the kNN algorithm has been shown to exhibit superior accuracy results in large data sets, with 500–1000 or more tests of real project software written in the Java programming language. On average, it is approximately 8% more accurate than the SVM method, and on average, it delivers 25% better results than the Naive Bayes method. Additionally, utilizing the kNN algorithm enables automated classification of errors from 450 tests in just 1 h, which is approximately 12 times faster than a human tester could do manually. Furthermore, the accuracy of clustering with 1000 failed tests using the kNN algorithm was found to be 0.98 (Table 1), which is the best result among the clustering methods considered in this study (kNN, Support Vector, and Naive Bayes). The kNN algorithm provides flexibility to choose the distance when building the kNN model. The Naive Bayesian classifier has low computational costs for training and classification. However, the SVM algorithm requires a lot of memory and has long training times for large datasets, making it computationally complex.

624

L. Oleshchenko Table 1. Clustering algorithms accuracy comparison in proposed software method

Machine learning algorithm

The failed tests number 50

100

200

250

500

1000

kNN

0,73

0,75

0,83

0,85

0,91

0,98

SVM

0,76

0,79

0,86

0,86

0,86

0,92

NB

0,83

0,81

0,77

0,77

0,76

0,75

4 Conclusions Automatic classification of software errors is an important tool for ensuring the quality of software being tested. By automating this process, testers can save time and effort, improve accuracy, gain a better understanding of errors, and improve communication between stakeholders. The proposed new software method for classifying software testing errors has a number of competitive advantages, including a central location for test results, integration with other systems, and the ability to analyze test results automatically. It also allows for the creation of custom error categories and subcategories to better organize test results, maintains a history of test launches, and is compatible across multiple platforms. Users can view self-test performance in real time, filter results as needed, and access automation statistics in the form of tables and graphs. However, there are several areas that require further research and analysis. The study did not consider data-driven features that could enhance the accuracy and dependability of automated systems. Additionally, the evaluation was limited to only three methods, and a broader range of programs should be examined before drawing conclusions. Multiple programs that operate on different operating systems often interact, which means that information from multiple sources may be necessary to identify the root cause of errors. Relatedly, hardware failures can cause errors that are difficult to differentiate from software errors, and this is an area that requires more analysis.

References 1. Kaur, H., Malhotra, R., Singh, H.: Software error clustering: a systematic review. IEEE Access 8, 98244–98261 (2020). https://doi.org/10.1109/ACCESS.2020.2996594 2. Ramalingam, S., Arumugam, S., Kannan, S.: An empirical study on error clustering in commercial software applications. Int. J. Comp. Sci. Netw. Secu. 14(7), 27–38 (2014) 3. Aggarwal, A., Williams, L., Nagappan, N.: An analysis of software error clustering in opensource software. Empir. Softw. Eng. 23(1), 582–625 (2018). https://doi.org/10.1007/s10664017-9526-y 4. Xu, P., Fu, J., Su, Z.: Identifying error clustering in large-scale software systems: a study of 23 million lines of code. IEEE Trans. Software Eng. 42(4), 349–367 (2016). https://doi.org/ 10.1109/TSE.2015.2482201 5. Jalali, R., Wohlin, C.: A systematic review of software error clustering: trends, challenges, and opportunities. J. Syst. Softw. 85(6), 1213–1227 (2012). https://doi.org/10.1016/j.jss.2012. 01.060

Machine Learning Algorithms Comparison

625

6. Sedighi, S., Rezaei, M.: Software error clustering: A systematic literature review. J. Syst. Softw. 165, 110591 (2020). https://doi.org/10.1016/j.jss.2020.110591 7. Truong, A., Hellström, D.: Clustering and Classification of Test Failures Using Machine Learning. Master’s thesis work. Department of Computer Science Faculty of Engineering LTH (2018) 8. Dang, Y., Wu, R., Zhang, H., Zhang, D., Nobel, P.: ReBucket: a method for clustering duplicate crash reports based on call stack similarity. In: 2012 34th International Conference on Software Engineering (ICSE), pp. 1084–1093. IEEE. Conference on Software Engineering (ICSE). IEEE (2012, June) 9. Moran, K., Linares-Vásquez, M., Bernal-Cárdenas, C., Vendome, C., Poshyvanyk, D.: Automatically discovering, reporting and reproducing android application crashes. In: 2016 IEEE international conference on software testing, verification and validation, pp. 33–44. IEEE (2016) 10. Abbineni, J., Thalluri, O.: Software defect detection using machine learning techniques. In: 2nd International Conference on Trends in Electronics and Informatics, IEEE. pp. 471–475 (2018) 11. Buchgeher, B.R.R., Klammer, G., Pfeiffer, M., Salomon, C., Thaller, H., Linsbauer, L.: Improving defect localization by classifying the affected asset using machine learning. In: International Conference on Software Quality, pp. 125–148 (2019) 12. Durelli, V.H.S., et al.: Machine Learning Applied to Software Testing: A Systematic Mapping Study. IEEE Transactions on Reliability, 1–24 (2019) 13. Karim, S., Leslie, H., Spits, H., Abdurachman, E., Soewito, B.: Software metrics for fault prediction using machine learning approaches. In: IEEE International Conference on Cybernetics and Computational Intelligence, pp. 19–23 (2017) 14. Alsmadi, I., Alda, S.: Test cases reduction and selection optimization in testing web services. I.J. Information Engineering and Electronic Business 5, 1–8 (2012) 15. Nasar, M.d., Johri, P., Chanda, U.: Software testing resource allocation and release time problem: a review. I.J. Modern Education and Computer Science 2, 48–55 (2014) 16. Mohapatra, S.K., Prasad, S.: Finding Representative Test Case for Test Case Reduction in Regression Testing. Int. J. Intell. Sys. Appli. (IJISA) 7(11), 60–65 (2015) 17. Ansari, G.A.: Detection of infeasible paths in software testing using UML application to gold vending machine.Int. J. Edu. Manage. Eng. (IJEME) 7(4), 21–28 (2017) 18. Mahmood, H., Sirshar, M.: A case study of web based application by analyzing performance of a testing tool. Int. J. Edu. Manage. Eng. (IJEME) 7(4), 51–58 (2017) 19. Nawaz, Z.: Proposal of Enhanced FDD Process Model. Int. J. Edu. Manage. Eng. (IJEME) 11(4), 43–50 (2021)

Machine Learning for Unmanned Aerial Vehicle Routing on Rough Terrain Ievgen Sidenko1(B) , Artem Trukhov1 , Galyna Kondratenko1 , Yuriy Zhukov2 , and Yuriy Kondratenko1,3 1 Petro Mohyla Black Sea National University, 68th Desantnykiv Street, 10, Mykolaiv 54000,

Ukraine [email protected] 2 C-Job Nikolayev, Artyleriyska Street, 17/6, Mykolaiv 54006, Ukraine 3 Institute of Artificial Intelligence Problems, Mala Zhytomyrs’ka Street, 11/5, Kyiv 01001, Ukraine

Abstract. The paper considers the main methods of machine learning for unmanned aerial vehicle (drone) routing, simulates an environment for testing the flight of a drone, as well as a model with a neural network for the unmanned routing of a drone on rough terrain. The potential use of unmanned aerial vehicles is limited because today the control of drone flight is carried out in a semi-automatic mode on the operator’s commands, or in remote mode using a control panel. Such a system is unstable to the human factor because it depends entirely on the operator. The relevance of the work is to use machine learning methods for drone routing, which will provide stable control of the unmanned aerial vehicle to perform a specific task. As a result of the work, a neural network architecture was developed, which was successfully implemented in a test model for routing an unmanned aerial vehicle on rough terrain. The test results showed that the unmanned aerial vehicle successfully avoids obstacles in the new environment. Keywords: Transport routing problem · Unmanned aerial vehicle · Machine learning · Reinforced learning · Artificial intelligence

1 Introduction Recently, the development of mobile robots has gained rapid momentum, especially the use of unmanned aerial vehicles in various fields of human activity [1–3]. Among the main reasons and advantages of using unmanned aerial vehicles for the following purposes: high efficiency; no threat to the life and health of staff; economic efficiency due to the relative cheapness of the drone; visual monitoring in real time with high quality images; quick search with target identification [4]. Machine learning, including neural networks, helps unmanned aerial vehicles to move more accurately and positionally on rough terrain, detect and classify objects with a camera and bypass them, adapt to changes in weather conditions based on training models and traffic scenarios, and deliver medicines, essential goods in locations with increased risk to human life, and more [2]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 626–635, 2023. https://doi.org/10.1007/978-3-031-36118-0_56

Machine Learning for Unmanned Aerial Vehicle Routing

627

The main problem in this task is the choice of the effective use of machine learning for unmanned aerial vehicles routing in rough terrain. In addition, the problem also lies in the need to create a training environment for training and testing the operation of the UAV, and then modeling obstacle avoidance situations in conditions close to reality. The existing works did not provide a comprehensive solution to the problems of UAV routing on rough terrain. In this work, the authors try to solve these problems using reinforcement learning and game engine Unity3D. The main purpose of this paper is to describe the process of machine learning and developing system to ensure efficient routing of unmanned aerial vehicles on rough terrain, which will allow more accurate movement of a certain route and increase flight stability.

2 Related Works and Problem Statement Most of the works are related to object detection and recognition using UAVs [5–7] In the proposed approach [5], firstly, the training for machine learning on the objects is carried out using convolutional neural networks, which is one of the deep learning algorithms. By choosing the Faster-RCNN and YoloV4 architectures of the deep learning method, it is aimed to compare the achievements of the accuracy in the training process. In this study [7], the methods of deep learning based detection and recognition of threats, evaluated in terms of military and defense industry, using Raspberry Pi platform by UAVs are presented. In the proposed approach, firstly, the training for machine learning on the objects is carried out using convolutional neural networks, which is one of the deep learning algorithms. The flight controller, based on the performance of sensors and software protection algorithms, can independently change the flight parameters of the drone, without resorting to the help of the operator [6]. The design of the device determines its possible use. Currently, there are many areas where unmanned aerial vehicles have proven themselves well: geodesy and cartography; agriculture in terms of field control and cultivation; farming; normal photo and video shooting; transportation of medicines and essential goods [7]. The huge potential of drones in transporting medications or vaccines helps in situations where the time factor is important. A number of large trade and postal companies have seriously considered investing in the development of drones engaged in the delivery of parcels, mail, medicines, food [1, 3]. Machine learning allows computers to draw conclusions based on data without following strict rules. In other words, the machine can find patterns in complex multiparameter problems (which the human brain cannot solve) to find more accurate answers. The result is a correct forecast [8, 9]. Depending on whether there is a teacher, training is divided into teacher training (controlled training), lack of teacher training (uncontrolled training) and reinforcement training [10]. In this study [11], authors applied reinforcement learning based on the proximal policy optimization algorithm to perform motion planning for UAV in an open space with static obstacles. The test of the trained model shows that the UAV reaches the goal with an 81% goal rate using the simple reward function suggested in this work [11]. The success of deep learning directly depends on the power of technology. At the time of the emergence of neural networks, the power of computers was low, which is why

628

I. Sidenko et al.

the networks themselves were quite weak. That is why at that time it was impossible to create a large number of layers of neural networks, namely the number of layers depends on the capabilities of the network. But with the advent of GPU and TPU, everything has changed. Modern Deep Learning is able to cope with large network sizes. In addition, for deep learning use special frameworks: Keras, Detectron, TensorFlow, PyTorch, CNTK and others [12–14]. All of the above publications are directly related to the objectives of the study in this work, while all of them lack a comprehensive solution to this problem in conditions close to reality.

3 Methods of Machine Learning for Unmanned Aerial Vehicle Routing Convolutional neural network (CNN) is one of the most influential innovations in the field of computer vision [11, 15]. Use reinforcement training where the machine is facing a task - there are many possible options for performing the task correctly in the external environment, such as computer games, trade operations and unmanned vehicles [16]. Reinforcement learning (RL) is almost the same as learning with a teacher, but the role of a “teacher” is a real or virtual environment [10, 11, 16–18]. SARSA (state - action - reward - state - action) is a special algorithm of machine learning with reinforcement. It was suggested by Rammeri and Niranjan in a technical note entitled “Modified Connectionist Q-Learning” (MCQ L) [19]. The name of the algorithm reflects the fact that the main function for maximizing the value of Q depends on the current state of the agent S, the action that the agent chooses A, the reward R that the agent receives for choosing an action, the state S2 to which the agent goes after this action, and, finally, the next action A2, which the agent chooses in his new state.   (1) Q(st , at ) ← Q(st , at ) + α rt + γ Q(st+1 , at+1 ) − Q(st , at ) The SARSA agent interacts with the environment and updates policies based on actions taken. The value of Q for a certain action and state is updated by the alpha score. The value of Q represents the possible reward obtained in the next time step for performing action A in state S, plus the future reward obtained in the next observation of the action of the state [20]. In comparison, Q-learning updates the estimate of the optimal state-action function based on the maximum reward for available actions, while SARSA learns the value of Q associated with adopting a policy that it itself follows [21–24]. The speed of learning (α) determines the extent to which new information overlaps the old. A coefficient equal to 0 will force the agent to not change anything, and a coefficient equal to 1 will force the agent to consider only the most recent information. The discount factor (γ) determines the importance of future rewards. A coefficient of 0 forces the agent to consider only current rewards, while a coefficient of 1 forces it to seek long-term high rewards. Q-learning is a learning algorithm used in the field of reinforcement learning. It does not require an environment, and it solves problems with stochastic transitions and

Machine Learning for Unmanned Aerial Vehicle Routing

629

rewards without requiring adaptation [21, 22]. Q-learning uses an action-value function for strategy   Q(s, a) = E R|st = s, at = a , (2) where s is a state, a is an action. Q-learning in its simplest form stores data in tables. This approach fails when the number of states/actions increases, as the probability that an agent will visit a certain state and perform a certain action becomes increasingly small [24]. Genetic algorithm (GA) is a search algorithm and heuristic method that simulates the process of natural selection, using methods such as mutation and crossover, to create new genotypes in the hope of finding good solutions to this problem [25–28]. Genetic algorithms are mainly used to find solutions in very large, complex search spaces [29]. Different approaches to vehicle route planning using fuzzy logic and fuzzy sets, IoT technologies, neural networks, genetic, evolutionary and heuristic algorithms are also given in [30–35]. Due to the limitations of genetic algorithms and other machine learning methods, it was decided to use reinforcement learning algorithms, because they give better results and allow influencing the training of agents when solving the UAV routing problem. Also, reinforcement learning algorithms have a wide range of settings, which allows controlling the behavior of the agent and determining the correct learning goals, for example, neglecting the short-term reward for the long-term one, which is important for solving this problem. In addition, models trained by reinforcement learning algorithms are more versatile and adapt well in a dynamic environment.

4 Practical Implementation of the Drone Routing The main libraries for creating artificial intelligence models supported by the Unity game engine are Acord.Net, CNTK, Tensorflow and unity ml-agents. Accord.NET Framework is a library for scientific computing in.NET [36–38]. The source code of the project is available under the terms of the Gnu Lesser Public License version 2.1. The project was originally designed to expand the capabilities of the AForge.NET Framework, but since then it has included AForge.NET. The new versions combine both frameworks called Accord.NET [38, 39]. Microsoft Cognitive Toolkit [40–46] is a standardized toolkit for designing and developing neural networks of various types, uses artificial intelligence to work with large amounts of data through deep learning, uses internal memory to process sequences of arbitrary length. The system is implemented using a game engine Unity3D and a plug-in unity-mlagents. The main components of the plugin are the brain, the agent and the academy [47– 50]. The academy is responsible for the learning process and the learning environment. An agent is actually an object that uses the brain to learn and coordinate its actions. It has observational vectors and available actions, and accumulates a reward. The brain is responsible for the decision made by the agent. This final element is being trained [51]. To solve the problem of drone routing, a training environment was created that will force the aircraft to perform movements on all axes. It is shown in Fig. 1.

630

I. Sidenko et al.

Fig. 1. The training environment

An important task was to force the agent not only to move sideways and forward or backward, but also to rotate the body when necessary, for which the first segment of the environment is responsible (Fig. 2a). Instead, to create the effect of rising and falling, a section with special vertical obstacles was created (Fig. 2b).

Fig. 2. Parts for testing: (a) drone hull turns, (b) drone lifting and lowering

With the help of vectors, the agent will observe the environment, a total of 17 vectors were used: 8 diagonals at an angle of 20 and 45 degrees, forward at an angle of 45, 65, 90, 110 and 155, up, down, left and right. The agent is shown in Fig. 3. The development of the system includes the creation of a neural network architecture that can be used to control an unmanned aerial vehicle in the environment. Develop an environment for learning and testing the neural network, which includes: creating an agent, determining the desired behavior for him; creating an environment, placing sensors in it, which are used to orient the agent in the environment and to train the model; development of a system of physical interaction of the agent with the environment. The advantages of the developed system are that the user can start the learning process again with the parameters defined by him, teach the model, if necessary, and transfer the model to another environment, provided new sensors are used to orient the agent in the environment. The created model is saved in ONNX format for use on a real drone. In addition, the created system is well optimized and designed to run on any platform. The proposed neural network structure marks 19 neurons in the input layer, 128 neurons in 2 fully connected layers and 4 neurons in the output layer. To speed up

Machine Learning for Unmanned Aerial Vehicle Routing

631

Fig. 3. Agent with observable vectors

learning and improve its quality, four agents have been created who use the same brain to make decisions, but two of them are moving in the opposite direction. The input are 17 observable vectors, end point position and drone position, the output are discrete numbers, one for each axis of motion. Since the playing time is relative, it was decided to increase it by 50 times, to speed up the process of training agents. The training process lasted about 4 hours, with 1,500,000 decisions made and the average reward of agents increased from 9.55 to 2441. System testing begins with planning the strategy by which this process will take place. During testing, the correctness of the drone operation on a certain segment must be checked. Testing should include checking the movement of the drone on all axes and the correctness of their use depending on the situation. It is also necessary to create and test the developed model in another environment where there was no training and where the agent will be for the first time. One of the most important steps is to test behavior in a new environment. In addition to checking the correct behavior of the agent is also tested his ability to achieve a given goal. To move to a more realistic environment, a prototype city with a central park in the middle was created. In order to specify the end point, an empty object was created and located at the end of the route, thus transmitting the exact coordinates of the destination. The next step is to set the drone to the starting position and start the routing with the Play button, the agent starts moving towards the end point and wraps around the obstacles (Fig. 4). In the test environment, the agent successfully reached the end point without encountering any of the obstacles. During the testing, the agent’s ability to navigate in space, perform all basic movements along 4 axes, reach the end point and bypass obstacles in different environments was tested. It should be noted that the model is designed for specific physical conditions, drone models and observation vectors. Therefore, such a system can be used as initial data to complete the model, but it can not be used for any conditions. Use specific conditions or adjust certain parameters must be done in an artificial environment, then refine the model and transfer to a real device that contains analogues of all devices that provide observation vectors.

632

I. Sidenko et al.

Fig. 4. Flight agent in a test environment

5 Conclusions During the work, a drone routing system was designed using machine learning methods, which provides high efficiency and stable flight in rough terrain. The result is a comprehensive system that allows you to perform neural network training with custom parameters for your own purposes or use an already trained and developed model of artificial intelligence, which is provided by default with the drone model. A neural network architecture with reinforcement learning algorithms was developed, which was successfully implemented in a test model for routing an unmanned aerial vehicle on rough terrain. The test results showed that the unmanned aerial vehicle successfully avoids obstacles in the new environment.

References 1. Sinulingga, L.S.A., Ramdani, F., Saputra, M.C.: Spatial multi-criteria evaluation to determine safety area flying drone. Proc. of the ISyG Int’l Symp. on Geoinformatics, pp. 55–61. Malang, Indonesia (2017) 2. Kharchenko, V., Matiychyk, D., Babenko, A.: Mathematical model of unmanned aerial vehicle control in manual or semiautomatic modes. Proc. of the 4th Int’l Conf. APUAVD, pp. 37–40. Kiev, Ukraine (2017) 3. Zhdanov, S., et al.: The Mathematical Model for Research of the UAV Longitudinal Moving. I. J. Computer Network and Information Security 5, 29–39 (2021) 4. Alshbatat, A.: Fire extinguishing system for high-rise buildings and rugged mountainous terrains utilizing quadrotor unmanned aerial vehicle. I.J. Image, Graphics and Signal Processing 1, 23–29 (2018)

Machine Learning for Unmanned Aerial Vehicle Routing

633

5. Bayhan, E., et al.: Deep learning based object detection and recognition of unmanned aerial vehicles. Proc. of the 3rd Int’l Conf. HORA, pp. 1–5. Ankara, Turkey (2021) 6. Timchenko, V.L., Lebedev, D.O.: Control systems with suboptimal models for stabilization of UAV. Proc. of the 5th Int’l Conf. APUAVD, pp. 117–122. Kiev, Ukraine (2019) 7. Ozkan, Z., et al.: Object detection and recognition of unmanned aerial vehicles using raspberry Pi platform. Proc. of the Int’l Conf. ISMSIT, pp. 467–472. Ankara, Turkey (2021) 8. Tania, M., et al.: Image recognition using machine learning with the aid of MLR. I.J. Image, Graphics and Signal Processing 6, 12–22 (2021) 9. Meena, G., et al.: Traffic prediction for intelligent transportation system using machine learning. Proc. of the 3rd Int’l Conf. ICETCE, pp. 145–148. Jaipur, India (2020) 10. Yasir, M., et al.: Machine learning based analysis of cellular spectrum. I.J. Wireless and Microwave Technologies 2, 24–31 (2021) 11. Kim, S., et al.: Motion planning by reinforcement learning for an unmanned aerial vehicle in virtual open space with static obstacles. Proc. of the Int’l Conf. Control, Automation and Systems (ICCAS), pp. 784–787. Busan, Korea (2020) 12. Wu, W., Liao, M.: Reinforcement fuzzy tree: a method extracting rules from reinforcement learning models. Proc. of the 18th Int’l Conf. on Computer and Information Science (ICIS), pp. 47–51. Beijing, China (2019) 13. Zinchenko, V., et al.: Computer vision in control and optimization of road traffic. Proc. of the 3rd Int’l Conf. DSMP, pp. 249–254. Lviv, Ukraine (2020) 14. Striuk, O., et al.: Generative adversarial neural network for creating photorealistic images. Proc. of the 2nd Int’l Conf. ATIT, pp. 368–371. Kyiv, Ukraine (2020) 15. Leizerovych, R., et al.: IoT-complex for monitoring and analysis of motor highway condition using artificial neural networks. Proc. of the 11th Int’l Conf. DESSERT, pp. 207–212. Kyiv, Ukraine (2020) 16. Chornovol, O., et al.: Intelligent Forecasting System for NPP’s Energy Production. Proc. of the 3rd Int’l Conf. DSMP, pp. 102–107. Lviv, Ukraine (2020) 17. Bidyuk, P., Gozhvi, A., Kalinina, I.: Modeling military conflicts using bayesian networks. Proc. of the 1st Int’l Conf. SAIC, pp. 1–6. Kyiv, UKraine (2018) 18. Tiwari, A.K., Nadimpalli, S.V.: Augmented random search for quadcopter control: an alternative to reinforcement learning. I.J. Information Technology and Computer Science 11, 24–33 (2019) 19. Tathe, P.K., Sharma, M.: Dynamic actor-critic: reinforcement learning based radio resource scheduling for LTE-Advanced. Proc. of the 4th Int’l Conf. ICCUBEA, pp. 1–4. Pune, India (2018) 20. Kiran, B.R., et al.: Deep reinforcement learning for autonomous driving: a survey. IEEE Trans. Intell. Transp. Syst. 23(6), 4909–4926 (2022) 21. Pandey, D., Pandey, P.: Approximate Q-Learning: An Introduction. Proc. of the 2nd Int’l Conf. on Machine Learning and Computing, pp. 317–320. Bangalore, India (2010) 22. Sun, C.: Fundamental Q-learning Algorithm in Finding Optimal Policy. Proc. of the Int’l Conf. ICSGEA. Changsha, pp. 243–246 (2017) 23. Chen, Y.-H., Wang, H., Chiu, C.-S.: Navigation Application of Q-learning Neural Network*. Proc. of the Int’l Conf. ICSSE, pp. 1–4. Kagawa, Japan (2020) 24. Strauss, C., Sahin, F.: Autonomous navigation based on a Q-learning algorithm for a robot in a real environment. Proc. of the Int’l Conf. on System of Systems Engineering, pp. 1–5. Monterey, CA, USA (2008) 25. Ziwu, R., Ye, S.: A hybrid optimized algorithm based on simplex method and genetic algorithm. Proc. of the 6th World Cong. on Intelligent Control and Automation, pp. 3547–3551. Dalian, China (2006)

634

I. Sidenko et al.

26. Guo, P., Wang, X., Han, Y.: The enhanced genetic algorithms for the optimization design. Proc. of the 3rd Int’l Conf. on Biomedical Engineering and Informatics, pp. 2990–2994. Yantai, China (2010) 27. Kondratenko, Y., et al.: Fuzzy and evolutionary algorithms for transport logistics under uncertainty. Proc. of the INFUS Int’l Conf. on Intelligent and Fuzzy Techniques: Smart and Innovative Solutions. Advances in Intelligent Systems and Computing, 1197, pp. 1456–1463. Springer, Cham (2021) 28. Linkens, D.A., Okola Nyongesa, H.: A distributed genetic algorithm for multivariable fuzzy control. Proc. of the Coll. on Genetic Algorithms for Control Systems Engineering, pp. 9/1– 9/3. London, UK (1993) 29. Jiang, J., Butler, D.: A genetic algorithm design for vector quantization. Proc. of the 1st Int’l Conf. on Genetic Algorithms in Engineering Systems: Innovations and Applications, pp. 331–336. Sheffield, UK (1995) 30. Kondratenko, G.V., Kondratenko, Y.P., Romanov, D.O.: Fuzzy models for capacitive vehicle routing problem in uncertainty. Proc. of the 17th Int’l DAAAM Symp. on Intelligent Manufacturing and Automation: Focus on Mechatronics & Robotics, pp. 205–206. Vienna, Austria (2006) 31. Werners, B., Kondratenko, Y.: Alternative fuzzy approaches for efficiently solving the capacitated vehicle routing problem in conditions of uncertain demands. In: Berger-Vachon, C., et al. (eds.) Complex Systems: Solutions and Challenges in Economics, Management and Engineering. Studies in Systems, Decision and Control, 125, pp. 521–543. Springer, Cham (2018) 32. Wang, X.: Vehicle Routing. Operational Transportation Planning of Modern Freight Forwarding Companies, pp. 7–23. Produktion und Logistik (2015) 33. Escobar, J.W.: Heuristic algorithms for the capacitated location-routing problem and the multi-depot vehicle routing problem. 4OR-Q Oper. Res. 12, 99–100 (2014) 34. Pereira, F.B., Tavares, J.: Bio-inspired Algorithms for the Vehicle Routing Problem[M]. Springer, Berlin, Heidelberg (2009) 35. Skakodub, O., Kozlov, O., Kondratenko, Y.: Optimization of Linguistic Terms’ Shapes and Parameters: Fuzzy Control System of a Quadrotor Drone. Proc. of the 11th Int’l Conf. on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications, pp. 566–571. Cracow, Poland (2021) 36. Levchenko, B., Chukhray, A., Chumachenko, D.: Development of Game Modules with Support for Synchronous Multiplayer Based on Unreal Engine 4 Using Artificial Intelligence Approach. Proc. of the Int’l Conf. on Integrated Computer Technologies in Mechanical Engineering, pp. 503–513. Springer, Cham (2020) 37. Koonce, B.: Convolutional Neural Networks with Swift for Tensorflow [M]. Apress, Berkeley, CA (2020) 38. Firdaus, M., Pramono, A., Fibri, M.R.: Optimizing Performance of Real Time Multiplayer 3D Games on Smartphone Using Unsupervised Learning (K-Means). Proc. of the 4th Int’l Conf. ICIC, pp. 1–8. Semarang, Indonesia (2019) 39. Castaño, A.P.: Support Vector Machines. Chapter in Practical Artificial Intelligence, pp. 315– 365. Berkeley, CA: Apress (2018) 40. Etaati, L.: Deep Learning Tools with Cognitive Toolkit (CNTK). Chapter in Machine Learning with Microsoft Technologies, pp. 287–302. Berkeley, CA: Apress (2019) 41. Seide, F.: Keynote: The computer science behind the Microsoft Cognitive Toolkit: An open source large-scale deep learning toolkit for Windows and Linux. Proc. of the IEEE/ACM Int’l Symp. CGO. Austin, TX, USA (2017) 42. Banerjee, D.S., Hamidouche, K., Panda, D.K.: Re-Designing CNTK Deep Learning Framework on Modern GPU Enabled Clusters. Proc. of the Int’l Conf. CloudCom. Luxembourg, pp. 144–151 (2016)

Machine Learning for Unmanned Aerial Vehicle Routing

635

43. Sugiyarto, A.W., Abadi, A.M.: Prediction of Indonesian Palm Oil Production Using Long Short-Term Memory Recurrent Neural Network (LSTM-RNN). Proc. of the 1st Int’l Conf. AiDAS, pp. 53–57. Ipoh, Malaysia (2019) 44. Rahhal, J.S., Abualnadi, D.: IOT Based Predictive Maintenance Using LSTM RNN Estimator. Proc. of the Int’l Conf. ICECCE, pp. 1–5. Istanbul, Turkey (2020) 45. Sidenko, I., et al.: Peculiarities of human machine interaction for synthesis of the intelligent dialogue chatbot. Proc. of the Int’l Conf. IDAACS. Metz, France (2019) 46. Siriak, R., Skarga-Bandurova, I., Boltov, Y.: Deep convolutional network with long shortterm memory layers for dynamic gesture recognition. Proc. of the Int’l Conf. IDAACS, pp. 158–162. Metz, France (2019) 47. Kuntsevich, V.M., et al. (eds.): Control systems: theory and applications. Book Series in Automation, Control and Robotics. Gistrup, Delft: River Publishers (2018) 48. Opanasenko, V., Kryvyi, S.: Synthesis of multilevel structures with multiple outputs. Proc. of the 10th Int’l Conf. UkrPROG. pp. 32–37 (2016) 49. Kondratenko, Y.P., Simon, D.: Structural and parametric optimization of fuzzy control and decision making systems. In: Zadeh, L., Yager, R., Shahbazova, S., Reformat, M., Kreinovich, V. (eds.) Recent Developments and the New Direction in Soft-Computing Foundations and Applications. Studies in Fuzziness and Soft Computing, 361, pp. 273–289. Springer, Cham (2018) 50. Gerasin, O., et al.: Remote IoT-based Control System of the Mobile Caterpillar Robot. Proc. of the 16th Int’l Conf. ICTERI, vol. I, pp. 129–136. Kharkiv, Ukraine (2020) 51. Youssef, A.E., et al.: Building your kingdom imitation learning for a custom gameplay using unity ML-agents. Proc. of the 10th Int’l Conf. IEMCON, pp. 0509–0514. Vancouver, BC, Canada (2019)

Improvement of the Dispenser of the Liquid Fertilizer Dosing Control System for Root Feeding of Crops Sergey Filimonov, Dmytro Bacherikov, Constantine Bazilo(B) , and Nadiia Filimonova Cherkasy State Technological University, Shevchenko blvd., 460, Cherkasy 18006, Ukraine {s.filimonov,n.filimonova}@chdtu.edu.ua, [email protected]

Abstract. Agriculture is rapidly using the latest technologies that are applied to manage and optimize agricultural production. Fertilizers are an essential element of agriculture, as their proper use increases yields. The pouring of liquid fertilizers is carried out using special systems. The main element of the pouring system are dispensers of various types. The satisfactory vegetation of the plant will depend on the accuracy of pouring liquid fertilizers, which will affect the future harvest. The main features of modern models of dispensers for pouring liquid fertilizers are determined. Their advantages and disadvantages are revealed. The new design of the dispenser with a screw piezoceramic motor is developed. New original solution for automatic control of the pouring norm is proposed. A stand for research and determination of the parameters of dispensers is developed. The norms of pouring out the minute flow rate for the developed design of the dispenser are experimentally tested and the hydraulic characteristics are obtained. The research results can be used in the design of dispensers for pouring liquid fertilizers. Keywords: agriculture · dispenser · dosing orifice plate · piezoelectric motor · bimorph piezoelectric element

1 Introduction Today, modern agricultural production and the agro-industrial complex as a whole is one of the most important sectors of the economy of each country. Advanced implementation of innovative technologies in the modern agricultural sector is not only autopilots, monitoring systems, but also navigation systems that increase the productivity of technological operations. Innovative technologies in the modern agricultural sector also include both modern approaches to tillage technology and equipment for the efficient use of fertilizers [1, 2]. The most common fertilizers are granular ones. Traditionally fertilizers are applied in the form of granules and, firstly, must be dissolved in water. After the transition from one state to another one they become available to plants. All these processes take a certain time, which will negatively affect the development and condition of the plant, and is not very effective. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 636–644, 2023. https://doi.org/10.1007/978-3-031-36118-0_57

Improvement of the Dispenser of the Liquid Fertilizer

637

A more effective fertilizers are liquid ones, namely fertilizers that are on the market in the form of anhydrous ammonia (which can rather be attributed to gaseous form), ammonia water, complex fertilizers (CF) and UAM (urea-ammonia mixture) [3–5]. The liquid form operates at the start of growth, so the plant receives nutrition earlier. Liquid fertilizers have a prolonged action that provides plant nutrition throughout the growing season. Also, the direction of precision farming is becoming relevant, which is currently technically advanced in systems of pouring and accurate dosing. The most famous manufacturers of special equipment and machinery for the application of liquid fertilizers are the following brands: Precision Planting, Raven, Hexagon, JohnDeere, Amazone, Jacto, Fast, Fliegl, Vredo, Hi-Spec, Boguslav. Almost all of the above manufacturers use in their dosing systems the principle of pouring liquids, which depends on the pressure control in the system, this causes certain problems in dosing accuracy. The deviation of the actual application rate from the specified one is 3%. The development and implementation of priority areas of precision farming will contribute to the development of agricultural production and the agro-industrial complex as a whole.

2 Formal Problem Statement At the moment, a large number of farmers from different countries are upgrading existing models of tillage and sowing complexes and purchasing new units with additional equipment for applying liquid fertilizers. The equipment provides for the installation on the frame of the unit or on the wheels of a special container for the operating solution. At the same time, hoses are connected to each share or coulter, which supply fertilizer using pumps. Thus, the application of liquid fertilizers occurs together with the sowing of seeds, or with another technological operation. These pouring systems allow to accurately dose and comply with the rate set by the operator. This increases the economic effect and the yield in the fields [6–9]. For accurate dosing of liquid fertilizers, the agricultural machinery manufacturers offer the use of special dispensers that are installed on fertilizer dosing units. Also, manufacturers of agricultural machinery offer different types of control. Figure 1 shows the classification of the main methods to control the liquid pouring system. The disadvantages of these control systems, widely used in applicators, trailers, mounted and self-propelled sprayers, are the control of electric motors and solenoid valves are energy consuming due to significant currents for self-holding, as well as high pressure in the system and the complexity of manufacturing, a significant step of regulation and the complexity of controlling these types of systems [10, 11]. A more promising direction in the field of modern agricultural production and the agro-industrial complex as a whole is the creation and use of motors, the principle of operation of which is based on the reverse piezoelectric effect. The reverse piezoelectric effect is the appearance of mechanical deformations under the influence of an electric field. Due to the absence of radiating magnetic fields and to resistance to their effects, wide ranges of rotational speeds and torques on the shaft (0.1… 1.0 Nm) the design has a number of technical advantages over electromagnetic motors. The piezoceramic motor has a high positioning accuracy [12, 13] of the order of 0.5 µm, and a large torque on the

638

S. Filimonov et al.

Fig. 1. Classification scheme of the main control methods for fertilizer pouring systems

shaft of 10 N. Such motors can be used in all industries, and especially in modern agricultural machinery and the agro-industrial complex [14, 15]. The great advantage of these motors is their unpretentiousness in operating and maintenance conditions [16–19]. One of the significant advantages of the piezoelectric motor is the start-stop characteristics, which are characterized by the ability of the motor to instantly start and stop, as well as hold a given position, without applying additional energy or effort. Thus, the actual task is the development of liquid fertilizer dispensers based on piezoelectric motors.

3 Materials and Methods The work is based on the new method of dosage control proposed by the authors, which consists not in maintaining a certain pressure in the system, but in changing the area of the orifice, which will allow you to accurately change the flow rate. To implement this method, the design of the developed dispenser uses piezoelectric motor [20, 21] based on bimorph piezoelectric elements [22], which is shown in Fig. 2.

Fig. 2. Design of a piezoelectric motor using bimorph piezoelectric elements: 1 – brass plates; 2 – piezoelectric elements; 3 – four-sided metal nut, 4 – running shaft

Table 1 shows the main parameters of the developed screw piezoceramic motor.

Improvement of the Dispenser of the Liquid Fertilizer

639

Figure 3 shows the design of the dispenser, in which the main component is a screw piezoceramic motor. Table 1. Main parameters of the developed screw piezoceramic motor Parameter

Value

Motor’s sizes

36 × 12 × 12 mm

Bimorph piezoelement’s sizes

31 × 6 × 0.4 mm

Movement speed

1 mm/s

Rotational speed

0.5 rps

Supply voltage

100 V

Resonant frequency

7000–7300 Hz

Piezoceramic type

PZT-5H

Fig. 3. The design of the dispenser with a screw piezoceramic motor: 1 – liquid supply; 2 – stuffing box; 3 – developed screw piezoceramic motor, 4 – dispenser housing, 5 – liquid outlet, 6 – replaceable dosing orifice plate, 7 – shaft

The total length of the structure is 100 mm, its diameter is 20 mm, and the diameter of the dosing orifice plate is 18 mm. In one version, the orifices of the dosing plate are of the same diameter 0.5 mm, in the other version there are four orifices of different diameters 0.5 mm, 0.75 mm, 1 mm and 1.5 mm. The operating principle of the developed dispenser is as follows. When voltage is applied to the piezoceramic motor, as a result, a rotational movement of the running shaft occurs, which is rigidly connected to the orifice plate. As a result, the dosing orifice plate rotates and sets its orifice against the liquid outlet hole. In this way, the pouring out of the fertilizer can be controlled. The developed stand for the study and determination of the parameters of the dispensers is shown in Fig. 4. This stand consists of a tank 3 with a capacity of 50 l, pump

640

S. Filimonov et al.

1, filters 2, rubber sleeves 11, pressure regulator 4, manometer 5, rods 8 made of polymer with a diameter of 32 mm, liquid dispenser 6, dispenser holder 7, contain-er for collecting liquid 9, precision scales 10.

Fig. 4. Functional scheme of the stand for the study of the dispenser: 1 – pump; 2 – filter; 3 – tank with liquid; 4 – manual pressure regulator; 5 – manometer; 6 – dispenser; 7 – dispenser holder; 8 – rod; 9 – container for collecting liquid; 10 – scales; 11 – sleeve; 12 – fluid movement

The operation of the stand is as follows: pump 1 sucks the solution from tank 3, and supplies it to pressure regulator 4, which regulates the pressure in rod 8, the liquid passes through filters 2. Pressure control is carried out thanks to the manometer 5. The liquid passes through the rod to the liquid dispenser 6 and is distributed into the tank 9. The distributed liquid in container 9 is weighed on precision scales, for a more accurate result during the experiments and after weighing it is poured into tank 3. Stand operates cyclically.

4 Experiments and Results With the help of the developed stand, experimental studies of the dispenser were carried out, where the minute flow rate of the liquid was determined. For this, the following limiting conditions were adopted. The flow is measured in liters per 1 min for each orifice of dosing plate, the pressure in the tests was (0.25 – 4 atm.), with an error of less than 1%. The duration of the measurement, determined by a stopwatch with an error of less than 1 s, must be greater than or equal to 60 s. For a more accurate determination of the amount of liquid, the tank with liquid was weighed on scales with an error of 1 g. Figure 5 shows the hydraulic characteristics for dosing plates with different orifices. Figure 5 clearly shows that if you rotate the shaft with dosing orifice plate in the developed dispenser, you can set the required pouring rate. From the obtained graphs, a function that describes the dependence of the liquid pouring rate at various values of pressure P and diameter d of the orifice of the dosing plate of the dispenser was constructed. V = a − be−cP

f

(1)

Improvement of the Dispenser of the Liquid Fertilizer

641

Fig. 5. Hydraulic characteristics of dosing plates with different orifice diameter:a) d = 0.5 mm; b) d = 0.75 mm; c) d = 1.0 mm; d) d = 1.5 mm

where V is the volume of the pouring liquid, P is the pressure in the system, a, b, c, f are the coefficients given in Table 2. Table 2. The coefficients of the dependence (1) The coefficients for dispenser 0.5 mm

The coefficients for dispenser 0.75 mm

The coefficients for dispenser 1.0 mm

The coefficients for dispenser 1.5 mm

a: 896.6892 b: 646.69384 c: 0.21584292 f: 1.1506088

a: 951.10074 b: 661.13363 c: 0.1761869 f: 1.433036

a: 1628.8107 b: 1336.7101 c: 0.17188192 f: 1.1332526

a: 5845.618 b: 5391.4818 c: 0.026710483 f: 1.4750516

This dispenser is versatile and contains a dosing plate with various orifices, which allows you to quickly change the pouring rate and allows you to use it with liquids with different liquid fertilizer densities. Also, a big plus of this dispenser is that it has practically no clogging of orifices. The next experiment was to compare two dosing orifice plate, one of which was in operation and the other was not in operation. Figure 6 shows the hydraulic characteristics

642

S. Filimonov et al.

of the dispensers, which was in operation (about 2000 motohours), and which was not in operation.

Fig. 6. Hydraulic characteristics of dosing plates with an orifice diameter d = 0.5 mm

Figure 6 shows that the daily flow through the dispenser that was in operation at a pressure of 2 atm is about 751 l, and the daily flow through the dispenser was not in operation is about 710 l. The difference between the values of the dispensers is 41 L. Each dispenser that has been in operation for about 2000 h and is installed on a row has significant deviations from the true value. This means that if it is used later in technological operations, it will lead to non-compliance with the specified pouring rates, and excessive consumption of liquids, which in turn increases the consumption of the tank mixture. If such a dispenser is installed on a liquid fertilizer system that have 24 rows, then the total daily excessive consumption will be 984 L. Thus, from the experiments presented above, it can be seen that the use of a dispenser with one orifice (for one pouring rate) has negative consequences, which con-sist in over-consumption of liquid fertilizers or non-compliance with the specified pouring rate. Therefore, the use of the developed dispenser with the ability to regulate different pouring rates will reduce these negative factors. In the future, field studies can be carried out, as well as the creation of an equivalent circuit using the method of electromechanical analogies to match with the control circuit.

5 Conclusions The design features of modern models of dispensers for pouring liquid fertilizers were analyzed. Their advantages and disadvantages were revealed. A new design of the dispenser with a screw piezoceramic motor has been pro-posed and developed. The design proposed a new original solution for automatic regulation of the pouring norm.

Improvement of the Dispenser of the Liquid Fertilizer

643

The stand for research and determination of the parameters of dispensers was developed. The pouring norms of the minute flow rate for the developed design of the dispenser were experimentally tested and the hydraulic characteristics were obtained. The research results can be used in the design of dispensers for pouring liquid fertilizers.

References 1. Anusha, K., Mahadevaswamy, U.B.: Automatic IoT Based Plant Monitoring and Watering System using Raspberry Pi. Int. J. Eng. Manuf. (IJEM) 8(6), 55–67 (2018). https://doi.org/ 10.5815/ijem.2018.06.05 2. Bhagawati, K., Bhagawati, R., Jini, D.: Intelligence and its Application in Agriculture: Techniques to Deal with Variations and Uncertainties. Int. J. Intell. Sys. Appli. (IJISA) 8(9), 56–61 (2016). https://doi.org/10.5815/ijisa.2016.09.07 3. Parent, S.É., Dossou-Yovo, W., Ziadi, N., et al.: Corn response to banded phosphorus fertilizers with or without manure application in Eastern Canada. Agron. J. 112(3), 2176–2187 (2020). https://doi.org/10.1002/agj2.20115 4. Kablan, L.A., Chabot, V., Mailloux, A., et al.: Variability in corn yield response to nitrogen fertilizer in Eastern Canada. Agron. J. 109(5), 2231–2242 (2017). https://doi.org/10.2134/agr onj2016.09.0511 5. Scharf, P.C., Wiebold, W.J., Lory, J.A.: Corn yield response to nitrogen fertilizer timing and deficiency level. Agron. J. 94(3), 435–441 (2002). https://doi.org/10.2134/agronj2002.4350 6. Bermudez, M., Mallarino, A.P.: Yield and early growth responses to starter fertilizer in no-till corn assessed with precision agriculture technologies. Agron. J. 94(5), 1024–1033 (2002). https://doi.org/10.2134/agronj2002.1024 7. Galpottage Dona, W.H., Schoenau, J.J., King, T.: Effect of starter fertilizer in seed-row on emergence, biomass and nutrient uptake by six pulse crops grown under controlled environment conditions. J. Plant Nutr. 43(6), 879–895 (2020). https://doi.org/10.1080/01904167. 2020.1711945 8. Mallarino, A.P., Bergmann, N., Kaiser, D.E.: Corn responses to in-furrow phosphorus and potassium starter fertilizer applications. Agron. J. 103(3), 685–694 (2011). https://doi.org/10. 2134/agronj2010.0377 9. Kaiser, D.E.M., Mallarino, A.P.M., Bermudez, M.: Corn grain yield, early growth, and early nutrient uptake as affected by broadcast and in-furrow starter fertilization. Agronomy. Journal 97(2), 620–626 (2005). https://doi.org/10.2134/agronj2005.0620 10. Taghizadeh, M., Ghaffari, A., Najafi, F.: Modeling and identification of a solenoid valve for PWM control applications. Comptes Rendus Mec. 337(3), 131–140 (2009). https://doi.org/ 10.1016/j.crme.2009.03.009 11. Xu, X., Han, X., Liu, Y., et al.: Modeling and dynamic analysis on the direct operating solenoid valve for improving the performance of the shifting control system. Appl. Sci. 7(12), 1266 (2017). https://doi.org/10.3390/app7121266 12. Sharapov, V.M., et al.: Improvement of piezoceramic scanners. In: IEEE XXXIII International Scientific Conference Electronics and Nanotechnology (ELNANO). Kiev, vol. 2013, pp. 144– 146 (2013). https://doi.org/10.1109/ELNANO.2013.6552063 13. Bobtsov, A.A., Bojkov, V.I., S.V. Bystrov, V.V., et al.: Actuating devices and systems for micromovements. SPb, ITMO University (2017)

644

S. Filimonov et al.

14. Song, S., Shao, S., Xu, M., et al.: Piezoelectric inchworm rotary actuator with high driving torque and self-locking ability. Sens. Actuators, A 282, 174–182 (2018). https://doi.org/10. 1016/j.sna.2018.08.048 15. Kiziroglou, M.E., Temelkuran, B., Yeatman, E.M., Yang, G.Z.: Micro motion amplification – a review. IEEE Access 8, 64037–64055 (2020). https://doi.org/10.1109/ACCESS.2020.298 4606 16. Spanner, K., Koc, B.: Piezoelectric Motors, an Overview. Actuators 5(1), 6 (2016). https:// doi.org/10.3390/act5010006 17. Hunstig, M.: Piezoelectric inertia motors – a critical review of history, concepts, design, applications, and perspectives. Actuators 6(1), 7 (2017). https://doi.org/10.3390/act6010007 18. Brahim, M., Bernard, Y., Bahri, I.: Modelling, design, and real time implementation of robust H-infinity position control of piezoelectric actuator drive. Int. J. Mechatronics and Automation 6(4), 151–159 (2018). https://doi.org/10.1504/IJMA.2018.095516 19. Spanner, K., Koc, B.: Piezoelectric Motor Using In-Plane Orthogonal Resonance Modes of an Octagonal Plate Actuators 7(1), 2 (2018). https://doi.org/10.3390/act7010002 20. Bazilo, C., Filimonov, S., Filimonova, N., Bacherikov, D.: Determination of Geometric Parameters of Piezoceramic Plates of Bimorph Screw Linear Piezo Motor for Liquid Fertilizer Dispenser. In: Hu, Z., Petoukhov, S., Yanovsky, F., He, M. (eds.) ISEM 2021. LNNS, vol. 463, pp. 84–94. Springer, Cham (2022). https://doi.org/10.1007/978-3-031-03877-8_8 21. Li, J.: Design of active vibration control system for piezoelectric intelligent structures. Int. J. Edu. Manage. Eng. (IJEME) 2(7), 22–28 (2012). https://doi.org/10.5815/ijeme.2012.07.04 22. Bazilo, C.: Modelling of Bimorph Piezoelectric Elements for Biomedical Devices. In: Hu, Z., Petoukhov, S., He, M. (eds.) AIMEE 2019. AISC, vol. 1126, pp. 151–160. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-39162-1_14

A System of Stress Determination Based on Biomedical Indicators Lesia Hentosh1(B) , Vitalii Savchyn1 , and Oleksandr Kravchenko2 1 Rtificial Intelligence Department, Lviv Polytechnic National University, Lviv 79013, Ukraine

[email protected] 2 National Technical University Kharkiv Polytechnic Institute, Kharkiv 61002, Ukraine

Abstract. This paper investigated methods of determining stress in humans and developed of smart service system in medicine to automate this process. This paper evaluates existing research in this area. We conducted a study of the method of determining stress levels based on biomedical indicators. Also, we have developed a system that is relevant for use in many different areas. For example, such a system is convenient to use in office firms to prevent overexertion of workers, to prevent emergencies in jobs with a high level of human impact, where human life is endangered, and also for daily use in health care. The smart service system works with input data based on heart rate variability indices. Neural network training has been launched in 100 epochs. In each epoch, the results of accuracy and loss were recorded. Also, for better reliability of the results, we recorded the data obtained not only from training data but also from variation data. As a result, the classification problem is solved. We get 99% accuracy. Keywords: Stress level · Psychophysical state · Computer technologies · Machine- learning method · Accuracy

1 Introduction After studying the phenomenon of stress and its symptoms, it is advisable to find out what are the ways to determine stress. One of the most common methods, which was used before and is still actively used today is to conduct a survey. Based on the answers to the questions posed in the questionnaire, conclude the emotional and stressful state of the person. Another accurate but more complex method than the previous one is the detection of stress using various sensors in the laboratory. For example, electromyography, electrocardiography, etc. are performed, which examine the work of the human musculoskeletal and cardiac systems. So making decisions under uncertainty and lending strategies are also associated with the occurrence of stress. Another particularly interesting way to diagnose stress is to diagnose it with a smartwatch and fitness bracelets. Almost all of these devices have built-in heart rate sensors, but not all have a function for detecting stress and a function for determining HRV (heart rate variability). Usually, only more expensive or newer versions of such devices have this functionality, and for many other users the feature is not available. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 645–656, 2023. https://doi.org/10.1007/978-3-031-36118-0_58

646

L. Hentosh et al.

Also, often use the method of diagnosing stress using Schulte tables [1]. But psychologists do not recommend using Schulte tables in the electronic version to determine stress. This is because their idea is to use peripheral vision, and the computer mouse cursor will distract from the center square of the table. Therefore, researchers recommend using only paper Schulte tables. Devices with a touch screen do not have a mouse cursor, but will distract the finger that clicks on the screen. The latest in the list and the most innovative way to diagnose stress are intelligent systems for recognizing signs of stress using human biometrics [2]. Typically, such systems work using machine learning algorithms [3–6]. The input data for such a system are various biometric indicators, but more popular and accurate is heart rate variability. To classify a person’s stress level, the dynamics of his pulse can be analyzed using, for example, the Random Forest algorithm, Logistic Regression, or other classification methods. Smartwatches and fitness bracelets also use similar smart systems, but there these systems are directly integrated into the device. Such a system can be designed and used separately from smart devices, but you will have to enter the biomedical data yourself, as they will not be obtained automatically by the device. The purpose of this study is to develop a method for diagnosing human stress using computer technology and machine learning. To achieve this aim, we need to perform the following tasks: 1) to study the phenomenon of stress and all available methods of diagnosing it in humans; 2) develop a method for diagnosing stress; 3) to develop a smart service system for the diagnosis of stress in humans based on biomedical indicators. The object of research: stress and psychophysical condition of a person. The subject of research: smart service system for digital determination and prediction of stress. Therefore, the main contribution of this chapter can be summarized as follows: 1) it is designed a universal method for diagnosing human stress using machine learning methods, which is based on the use of computational intelligence tools; 2) algorithmic implementations of the designed universal method with the use of both: machine learning algorithm and artificial neural networks are presented; procedures for their training and application within the designed method are described; 3) it is designed a smart service system in medicine; an experimental investigation of the efficiency of work of proposed algorithm; it gets accuracy – 99%.

2 Related Works Interest in this topic is growing every year, because the timely diagnosis of stress and taking precautions can help people avoid nervous disorders, overwork or negative health consequences. Recently, stress detection systems have begun to integrate into smartwatches and fitness bracelets, and new scientific articles and research in this area have been published. This indicates that at the moment the research topic is still relevant.

A System of Stress Determination Based on Biomedical Indicators

647

The article [7] describes the development of a method for automatically determining stress on two popular scales of depression: the scale of stress, anxiety, and depression – DASS-21 and the scale of psychological stress Kessler – K-10. Six regression models were used for this: normal regression, least squares, generalized linear models, betabinomial regression, fractional logistic regression model, MM-estimation and censored smallest absolute deviation. This article is useful and interesting because it describes popular stress assessment scales and machine learning algorithms that have been adapted so that they can measure stress on these scales. The article [8] describes the construction of an information system not only to determine the level of stress, but also to track cognitive impairment. Another feature of this work is that studies have been conducted for infirm elderly people with mild cognitive impairment. To do this, it analyses the collected data from physiological sensors, which for some time were worn by the elderly. The paper also describes a stress detection system based on several machine learning algorithms, and tested their effectiveness on a real data set. It has been found that stress detection algorithms usually provide only the identification of a stressful/non-stressful event. Researchers suggest constantly monitoring physiological parameters in real-time to assess stress levels. In a study [9], the authors conducted a study of stress in the military or people who participated in hostilities. The research was conducted in Ukraine and the Ukrainian military took part in the survey and testing. Classification and clustering of data were performed using machine learning algorithms. The k-means algorithm was used to identify the number of clusters, the optimal number of clusters was found using the elbow algorithm, and the Random Forest algorithm was used to test the model and its subsequent use. The article [10] also investigates methods for determining the level of human stress using modern technologies. Researchers say the study is relevant because high levels of stress are a major factor that has a negative impact on heart health, and there is a relationship between stress levels and heart rate. This paper presents a technique where the heart rate of patients is recorded by IoT devices. Using an analytical model, the proposed system shows better accuracy in shorter processing times compared to other approaches to machine learning, and thus it is a cost-effective solution in the IoT system than medical research. The study [11] investigated the risk of high blood pressure in humans through machine learning. The article also describes the methods of processing human biomedical data using machine learning. To predict the risk of high blood pressure, the algorithm uses additional data such as age, anger, the general level of anxiety or stress in humans, and indicators of obesity and cholesterol in humans. The Random forest algorithm showed a prediction accuracy of 87.5%. Ways to use classification methods to identify key predictors of the patient, which are considered the most important in the classification of the worst cognitive abilities, which is an early risk factor for dementia, were studied in [12]. During the study, three classifiers were used: decision trees, Bayesian algorithm and random forests. The impact of the global pandemic COVID-19 on the emotional state and level of stress in humans was analyzed in [13]. A random forest algorithm was used to detect

648

L. Hentosh et al.

psychological stress during a pandemic. The study found that the prevalence of clinical levels of anxiety, depression and post-traumatic stress was higher during a pandemic. The article [14] investigates the potential of machine learning in predicting health status after the body has transmitted certain types of infections or diseases. A matrix of percentage similarity between helper genes and Escherichia coli genomes was created, which was later used as input for machine learning algorithms. The paper also compared the efficiency of such algorithms as random forest, vector support machine algorithm (radial and linear core) and gradient gain. This article does not describe how to determine stress, but it describes the application of machine learning algorithms for human biomedical parameters. Research [15] investigates a method called ensemble classification, which is used to increase the accuracy of weak algorithms by combining several classifiers. Experiments with this tool were performed using a data set on heart disease. A comparative analytical approach was performed to determine how ensemble techniques can be used to improve prediction accuracy in heart disease. The main focus of this paper is not only on improving the accuracy of weak classification algorithms, but also on the implementation of an algorithm with a medical data set to show its usefulness for predicting the disease at an early stage. This article also does not describe the definition or prediction of stress, but here is a useful study of ways to improve classification algorithms based on biomedical data. An indicator that quantifies the stress experienced by a cyclist on certain sections of the road has been studied in the article [16]. The study proposes a classification that consists of two parts: clustering and interpretation. At the heart of this methodology is a classifier combined with a predictive model to make the approach scalable to massive datasets. A hybrid approach to stress clustering at the personal level, using basic stress selfreports to increase the success of human-independent models without requiring significant amounts of personal data is described in [17]. Machine learning algorithms have been used to solve this problem. The study is developing a method to quickly detect persistent daily stress to prevent serious health consequences. The article [18] describes a comparative study of classifiers of tumors and heart disease. A variety of medical methods based on deep learning and machine learning have been developed to treat patients, as cancer is a very serious problem in this era. The researchers compared machine learning algorithms such as logistic regression, K-NN and decision trees. Researchers argue that stress in patients can also cause heart disease. The paper does not describe the topic of stress or ways to determine it, but a comparison of machine learning algorithms, which are trained on biomedical indicators. In a study [19], the authors developed a mathematical model for the problem of classification and analysis of the psychophysical state of man using a modification of the tapping test as a data source. To conduct the experiment, an application was developed for a smartphone that is a generalization of the classic tapping test. The developed tool collects much more data related to motility. This allowed us to collect a sample, to analyze the structure of each individual experiment, and the sample as a whole. The data were investigated for anomalies and a method of reducing their number was developed on the basis of methods of descriptive statistics and methods of optimization

A System of Stress Determination Based on Biomedical Indicators

649

of hyperparameters. The machine learning model was used to predict some features of the corresponding data set. As a result, the above-mentioned literature sources were considered and analyzed. Some articles have examined the level of stress for a certain category of people or under certain circumstances. For example, the articles analyzed situations for the elderly or the military or in a pandemic. In contrast to the existing approaches, this paper proposes a system that will determine the level of stress for anyone, by entering certain biometric indicators. The proposed smart service system will not require special devices for their measurement. Diagnosis will be performed with HRV or heart rate data obtained in any convenient way. The obtained results have a special practical value for diagnosing stress in students during training or for office workers.

3 Materials and Methods The approach used in this work focuses on the parallel encryption of the blocks into which the file is divided. This is possible because the encryption process for each block is independent of the others. For example, suppose you need to encrypt a file of 1048576 bytes. This file will be divided into 16-byte blocks. Thus, the AES algorithm will be applied to 65536 blocks. Here you can use parallelization to distribute the work between the available threads depending on the technology used. Neural networks were chosen to implement the stress diagnosis system. The main components of an artificial neural network are nodes, which in a certain hierarchy transmit to each other information, during the passage of each such node partially changes and the last layer of the neural network is already processed source data. To solve the problem, you need to classify stress into three categories: no stress, time pressure and interruption. To do this, we will use a serial-type neural network that will have an input layer that will input data, one hidden layer, and the last output layer, which will already return the data divided into three categories. So we need to develop a Sequential model with three layers, choose the activation function for these layers, and configure the model training. In the developed model, on the first two layers, we used the ReLU function, and on the last layer, we used the softmax function. The ReLU function is determined by the formula: f (x) = max(0, x), where x – is the input value of the neuron. This function leaves the input unchanged if it is greater than zero and converts all negative argument values to zero. This feature is popular because it is simple and fast to calculate. There are also variations and improvements to this feature in cases where zeroing negative elements are not suitable for some models. An example of such a function is the Leaky ReLU function, which is calculated by the formula: f (x) = max(0.1x, x). This type of function is used in the work on the input and hidden layers of the model, because it perfectly copes with the task and is quickly calculated, so it optimizes the

650

L. Hentosh et al.

neural network. But the third layer cannot be used because the effect of themultilayer will be lost. Therefore, for the last layer, we used Softmax [20]: e xj f (x) = K  i

i

e xi

,

where xj – the value of the input data, K – the number of classes in the classifier, j = 1, . . . , K. The advantage of this activation function is that it can already handle several classes. At the output, these classes take the form from 0 to 1. If we know the number of layers and the number of neurons on each layer, then we can calculate the number of weights that will be counted by the designed neural network. We chose to have three layers. On the first layer – 34 neurons, because it is the input layer, and the studied dataset has 34 input parameters. The last layer is the output layer, so its number of nodes is also known, there will be 3, because the system will classify three states of people. And only the hidden layer remains available for experiments with the number of nodes. For example, we first set the number of nodes to 30. In this case, the number of weights can be calculated as follows: 34 x 30 = 1020 – the number of weights between the first and second layer; 30 x 3 = 90 – the number of weights between the second and third layers; 1020 + 90 = 1110 – the total number of weights. Also, a necessary step in the design of a neural network is the choice of metrics. Metrics are a thing that will help determine how accurately the network has been trained. Metrics are of the following types: accuracy, recall, precision, and F-measurement. After analyzing the advantages and disadvantages of each of the metrics, for the implementation of the neural network, we chose the accuracy metric: TP + TN , TP + TN + FP + FN where TP – True Positive, TN – True Negative, FP – False Positive, FN – False Negative. Another important part of the design of neural networks is the choice of the function by which the losses will be calculated. We chose the categorical cross-entropy loss function: accuracy =

−

N C 1  1yi ∈Cc log pmodel [yi ∈ Cc ], N i=1 c=1

where N – is the number of observations (records in the dataset), C – is the number of categories of the classifier (classes). belonging to In the above formula, the entry 1yi ∈Cc indicates the i− th observation  class C among all N observations. Part of the formula pmodel yi ∈ Cc is the probability of the model predicting that the i− th observation belongs to class C. The Keras API was used to develop the neural network. This high-level API is part of TensorFlow. TensorFlow has developed this library to simplify and speed up the process of building neural networks. Thanks to Keras, we can design a neural network using intuitive abstraction. Keras has most of TensorFlow’s capabilities but speeds up the network development and design process.

A System of Stress Determination Based on Biomedical Indicators

651

4 Results We took the SWELL dataset [21] as input for training the neural network. This dataset contains 369289 records for network training and 41033 records for testing. So in total the data set has 410322 records with data. The data set contains 34 biometric indicators that will be included in the network input. These biometric indicators describe the work of the human heart, namely the indices of the internal heart rate. All data were collected from 25 office workers, who were temporarily monitored and their heart rate variability and other indicators were monitored using special sensors. During the study, these workers were given a variety of tasks and periodically created stressful situations to monitor the body’s response to such stressful stimuli. The condition of workers was divided into three states: no stress, time pressure and interruption. These three states are the source data for the network. The first condition is without stress. In this case, employees were given tasks and allowed to work on them for as long as they needed. In the dataset, it is called “no stress”. The second type of condition of employees is already stressful for them and was caused by the limited time to complete the task. In the dataset, it is called “time pressure”. In the third case, employees were distracted from the task by sudden e-mails. Sent eight letters with different content, with one that is relevant to the task, and one that was not related to the task. In a dataset, this type of stress is called “interruption”. During the development of the neural network, we experimented with and trained a neural network with a different number of nodes on the hidden layer. We also recorded the accuracy and number of losses in each of the epochs. We launched neural network training in 100 epochs. In each epoch, the results of accuracy and loss were recorded. Also, for better reliability of the results, we recorded the data obtained not only from training data, but also from variation data. Table 1 shows the values of losses and accuracy of the neural network as a result of 100 epochs. For convenience and compactness, the results in the table are given in increments of 10 epochs. The first column in the table indicates the number of epochs for which the results are recorded. The second and third columns indicate the loss and accuracy when using training data. The fourth and fifth columns also indicate loss and accuracy, but for validation data. We used validation data for better independence and reliability of results. Analysing Table 1, we can observe how the accuracy of model training improved and losses decreased as the number of training epochs increased. Figure 1 shows a graph of the model learning speed. This is a graph of the dependence of losses on past epochs. Analyzing this graph, we can see that until about the 20th century, the speed of learning was very high, because the values of losses were rapidly declining. This indicates that the model was not yet trained and was just beginning to train. With almost every era, the speed of network training has decreased, which is a sign of a good training process. If at some point in time, the loss function began to increase sharply, it would mean the effect of retraining and training the model at this stage would have to stop. Analysing Fig. 1 in more detail, we can see that almost along its entire length there are small fluctuations in the values of the function, similar to noise. This is because validation data is used to plot the graph. Due to the validation data in the figure, many

652

L. Hentosh et al. Table 1. The results of neural network training

№ epoch

Training data

Validation data

Loss

Accuracy

Loss

Accuracy

1

0.8105

0.6360

0.7059

0.6760

10

0.2549

0.9067

0.2431

0.9117

20

0.1302

0.9565

0.1302

0.9555

30

0.0804

0.9734

0.0739

0.9757

40

0.0534

0.9835

0.0514

0.9843

50

0.0388

0.9885

0.0371

0.9897

60

0.0301

0.9911

0.0261

0.9932

70

0.0243

0.9931

0.0241

0.9933

80

0.0204

0.9941

0.0228

0.9934

90

0.0177

0.9951

0.0189

0.9939

100

0.0156

0.9955

0.0143

0.9964

Fig. 1. Changing the learning speed of the model depending on the epoch

of these small local minima and maxima arise. And if you look at the whole chart, it is generally declining. This means that the effect of retraining does not occur. Very similar to the graph of learning speed, but inverted, is the graph of the accuracy of the model on the number of epochs passed (see Fig. 2). In contrast to the graph of learning speed, the graph shown in Fig. 2 increases almost throughout.

A System of Stress Determination Based on Biomedical Indicators

653

Fig. 2. Dependence of learning accuracy on epochs

One of the important stages in the development of a smart service system is also to check the operation of the system on tests and new data. In order to run the model on the new test data, you must pass an array with this data to the input of the predict() method and run it on execution, as shown in Fig. 3.

Fig. 3. The results of the prediction

The predict() function returns the probability that each record belongs to different classes. There are three classification classes, so the function returns three columns with probabilities. Figure 3 shows that for each record the probability of one of the classes is one, or a number close to one, and for other classes, the probability is a number close to zero. To verify the reliability of the results, look at the original values of these classes (Fig. 4). Figure 3 and Fig. 4 show that in all cases the prediction coincided with the original data, so it can be argued that the system works correctly with an accuracy of more than 99%. It should be noted that the first column in the output in Fig. 3 and Fig. 4 indicates the likelihood of stress due to employee distraction (interruption). The second column

654

L. Hentosh et al.

Fig. 4. Original values

indicates a state without stress (no stress). The third column indicates the probability of stress due to pressure in time constraints (time pressure). We have presented the data in Table 2 for the convenience of their analysis. Table 2. Probabilities of emotional states Interruption

No stress

Time pressure

2.0427158e-13 1.0000000e + 00 1.8924628e-21 3.8080251e-11 5.4651661e-08

1.0000000e + 00

8.2618625e-12 1.0000000e + 00 1.2235136e-08

It is convenient to analyze the results in the form of a table because we can see the probabilities of all three states at once. If the user is superfluous or not interested in this information, we can immediately get the name of one state, the probability of which is the highest. To do this, we need to use the predict_classes() function instead of the predict() function.

5 Summary and Conclusion In this research, we have developed a smart service system to identify signs of human stress based on its biomedical indicators. To do this, we first analyzed the various literature sources in which the study similar problems. We proposed a method for determining stress based on the neural network. We chose the neural network because neural networks have many parameters for tuning and optimization and allow new experiments. As a result of the analysis and research of the subject area, we chose a model and developed an algorithm based on which we implemented a system for determining the signs of stress in humans. The system works with input data based on HRV indices. At the input, it accepts 34 parameters based on which the prediction takes place. The source data for the network are from three different states. The first condition is not stress. The

A System of Stress Determination Based on Biomedical Indicators

655

second condition is stress due to time constraints. The third stage is also stress due to the distraction of the person from the task on which he focused. Surveys for recruitment have been conducted among office workers, and a network trained in such a dataset will also be better able to detect the stress of office work. There is a good reason to monitor stress in employees because early detection of stress helps to take precautions, maintain a high level of productivity, and prevent damage to health. The advantage of this system is that it is not necessary to wear special sensors to use human HRV to use it, as such sensors already have some smartwatches. Our proposed method for diagnosing stress showed an accuracy of 99%. The developed smart service system is tested by test data and validation data. Further research will be carried out in two main areas for making decisions under uncertainty [22] and for risk analysis in lending [23] to identify signs of human stress based on her biomedical indicators. 1) the use of CUDA technology [24]; 2) the use of OpenMP technology [25].

References 1. Maksimenko, V.A., et al.: Human personality reflects spatio-temporal and time-frequency EEG structure. PLoS ONE 13(9), 1–13 (2018) 2. Mochurad, L., Yatskiv, M.: Simulation of a Human Operator’s Response to Stressors under Production Conditions. Proceedings of the 3rd International Conference on Informatics & Data-Driven Medicine, November 19–21, pp. 156–169. Växjö, Sweden (2020) 3. Deo, R.C.: Machine Learning in Medicine. Circul. Bas. Sci. Clini. 132, 1920–1930 (2015) 4. Gao, J., Yang, Y., Lin, P., Park, D.S.: Computer vision in healthcare applications. J. Healthc. Eng. 2018, 1–5 (2018) 5. Izonin, I., Trostianchyn, A., Duriagina, Z., Tkachenko, R., Tepla, T., Lotoshynska, N.: The combined use of the wiener polynomial and SVM for material classification task in medical implants production. Int. J. Intelli. Sys. Appli. (IJISA) 10(9), 40–47 (2018) 6. Santra, A., Dutta, A.: A comprehensive review of machine learning techniques for predicting the outbreak of Covid-19 cases. Int. J. Intell. Sys. Appli. (IJISA) 14(3), 40–53 (2022) 7. Gamst-Klaussen, T., Lamu, A.N., Chen, G., Olsen, J.A.: Assessment of outcome measures for cost-utility analysis in depression: mapping depression scales onto the EQ-5D-5L. BJPsych Open 4(4), 160–166 (2018) 8. Delmastro, F., Martino, F.D., Dolciotti, C.: Cognitive training and stress detection in MCI frail older people through wearable sensors and machine learning. IEEE Access 8, 65573–65590 (2020) 9. Pavlova, I., Zikrach, D., Mosler, D., Ortenburger, D., Gora, T., Wasik, J.: Determinants of anxiety levels among young males in a threat of experiencing military conflict-applying a machine-learning algorithm in a psychosociological study. PLoS ONE 15(10), 1–14 (2020) 10. Akanksha, E.: Framework for propagating stress control message using heartbeat based IoT remote monitoring analytics. Int. J. Elec. Comp. Eng. 10(5), 4615–4622 (2020) 11. Nimmala, S., Ramadevi, Y., Sahith, R., Cheruku, R.: High blood pressure prediction based on AAA++ using machine-learning algorithms. Cogent Engineering. 5(1), 1–12 (2018) 12. Rankin, D., et al.: Identifying key predictors of cognitive dysfunction in older people using supervised machine learning techniques: Observational study. JMIR Med. Inform. 8(9), 1–23 (2020)

656

L. Hentosh et al.

13. Prout, T.A., et al.: Identifying predictors of psychological distress during COVID-19: a machine learning approach. Front. Psychol. 11, 1–14 (2020) 14. Njage, P.M.K., Leekitcharoenphon, P., Hald, T.: Improving hazard characterization in microbial risk assessment using next generation sequencing data and machine learning: predicting clinical outcomes in shigatoxigenic escherichia coli. Int J Food Microbiol. 292, 72–82 (2019) 15. Latha, C.B.C., Jeeva, S.C.: Improving the accuracy of prediction of heart disease risk based on ensemble classification techniques. Info. Medi. Unloc. 16, 1–9 (2019) 16. Huertas, J.A., et al.: Level of traffic stress-based classification: a clustering approach for Bogotá, Colombia. Transp. Res. Part D: Transp. Environ. 85, 1–17 (2020) 17. Can, Y.S., Chalabianloo, N., Ekiz, D., Fernandez-Alvarez, J., Riva, G., Ersoy, C.: Personal stress-level clustering and decision-level smoothing to enhance the performance of ambulatory stress detection with smartwatches. IEEE Access. 8, 38146–38163 (2020) 18. Khan, T.A., Kadir, K.A., Nasim, S., Alam, M., Shahid, Z., Mazliham, M.S.: Proficiency assessment of machine learning classifiers: an implementation for the prognosis of breast tumor and heart disease classification. Int. J. Adv. Comput. Sci. Appl. 11(11), 560–569 (2020) 19. Mochurad, L., Ya, H.: Modeling of psychomotor reactions of a person based on modification of the tapping test. Int. J. Comp. 20(2), 190–200 (2021) 20. Du, K.-L., Swamy, M.N.S.: Neural Networks and Statistical Learning, pp. 1–24. SpringerVerlag, London (2014) 21. Dataset SWELL [Electronic resource] // Access mode: https://www.kaggle.com/qiriro/swellheart-rate-variability-hrv 22. Nehrey, M., Hnot, T.: Data science tools application for business processes modelling in aviation. In: Cases on Modern Computer Systems in Aviation. IGI Global, pp. 176–190 (2019). https://www.igi-global.com/gateway/chapter/222188 23. Kaminskyi, A., Nehrey, M.: Information technology model for customer relationship management of nonbank lenders: coupling profitability and risk. In: 11th International Conference on Advanced Computer Information Technologies (ACIT), pp. 234–237 (2021) 24. Kalaiselvi, T., Sriramakrishnan, P., Somasundaram, K.: Performance of medical image processing algorithms implemented in CUDA running on GPU based machine. Int. J. Intell. Sys. Appli. (IJISA). 10(1), 58–68 (2018). Jan 25. Umbarkar, A.J., Rothe, N.M., Sathe, A.S.: OpenMP Teaching-learning based optimization algorithm over multi-core system. Int. J. Intelli. Sys. Appli. (IJISA) 7(7), 57–65 (2015)

Sub-Gigahertz Wireless Sensor Network for Smart Clothes Monitoring Andrii Moshenskyi1 , Dmitriy Novak2 , and Liubov Oleshchenko3(B) 1 National University of Food Technologies, Kyiv 01601, Ukraine 2 Kyiv National University of Technologies and Design, Kyiv 01011, Ukraine 3 National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”,

Kyiv 03056, Ukraine [email protected]

Abstract. Wireless sensor system is based on a set of sensor modules and custom devices that are installed on clothing to achieve the goal of collecting micrometeorological data for interlayer spaces research. Last time clothes testing is usage short distance data transfer systems or usage of cellular networks for the coverage extension. Sub-gigahertz systems integration into existed data networks in the research was proposed. Module consists of an Atmel microcontroller, a battery power supply, a battery management module, radio module and a set of temperature and relative humidity sensors. These modules are integrated into the clothing in interlayer spaces. The microcontroller firmware is written in C programming language. The software client side for displaying data collection and statistics processing is written in the Python programming language with the use of Numpy and Matplotlib libraries for the serial port. It was established that for the first subject an increase of the average temperature and relative humidity is due to physical activity and type of uniform. The second subject results were more stable than for the first one. Parameters of changes in temperature and air humidity can be used to predict comfortable conditions and optimize clothing in relation to climatic conditions and the load on the human body. Proposed adopted licensing and data modes make possible to build networks with coverages, which can concurrent with national cell networks but with much lower data rates. Usage ISM on the sub-gigahertz bands data links produce data rates less than 10 Kb/s but the radius increase more then on hundreds of percent (>200%). Amateur radio and the same speed limits produce extension of coverage radius in 10 and more times versus ISM 433. Testing of proposed system is perform in thousands of data transfer sessions. Coverage simulation was very close to the real data transfer area. Proposed network can be recommend for the similar researches of smart systems of local data collecting. Keywords: Sub-gigahertz · wireless sensor network · software · microcontroller · sensor · radio module · temperature · humidity

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 657–669, 2023. https://doi.org/10.1007/978-3-031-36118-0_59

658

A. Moshenskyi et al.

1 Introduction Recently, there has been intensive work on the development of so-called smart textiles. Most textile parameters depend on the fact that it is produced, namely on fibers and filler. In order to achieve the above optimization, there are actual tasks related to the collection of micro-meteorological parameters, namely the temperature and humidity of the air in the interlayer spaces between the human body and clothing, between layers of clothing and between clothing and external spaces. The integration of radio electronics and textiles enables the development of new materials for industry. Smart textiles are dividing into several subgroups. The first group includes the so-called passive clothing, which allows analyzing the environment or the actions of the user with the help of sensors. Active smart textiles belonging to the second group can react to external conditions. The third group includes the so-called very smart textiles, they can not only react and analyze the conditions and adapt to them. Sometimes a fourth group is distinguished, the so-called intelligent textiles, which react not only to external conditions and adapt, but also allow be programmed. The integration of conductive and non-conductive materials from which clothes are made allows to build, as it were, integrated microcircuits, but not according to integral, but according to the so-called hybrid technology. In the form of clothing, integrate some types of sensors, some various devices in the form of electronic circuits that make up elements of construction and design. The main problem of smart clothes testing is usage short distance data transfer systems or usage of cellular networks for the coverage extension. Sub-gigahertz systems integration into existed data networks was not consider. Sub-gigahertz technology operates in a frequency band less than 1 GHz and can be used for wireless applications requiring low power consumption and long-range transmission, which reduces the locating intermediate base stations cost. 1.1 Analysis of Recent Research Smart clothes researchers have developed materials that integrate various sensors to study such physiological parameters as temperature, humidity, heart rate, concentration of gases, namely carbon and carbon dioxide. Leading manufacturers use the monitoring of biological parameters of the human body and international gas spaces, namely such manufacturers as Adidas and Numetrex [1, 2] smart clothes in the form are used to detect so-called dangerous situations for the human body and even the environment [3, 4]. In certain works, the temperature limit of 37.80 °C means that monitoring is carried out not every 10 s, but every second. The upper limit of 65.6 degrees Celsius, which leads to the activation of an alarm, looks rather strange. Usually, for a biological object, a temperature exceeding 43 degrees Celsius is dangerous, and a temperature exceeding 53 degrees Celsius leads to painful symptoms. Work [5] provides a set of methods and materials that allows developing clothing intended for industrial purposes and sports purposes. In papers [6–8], detailed monitoring of the inter-layer parameters of clothing, namely temperature and air humidity, is carried out.

Sub-Gigahertz Wireless Sensor Network

659

In work [9] emergency data network with unstandardized usages of digital radio modules was described, which allows using human based ghost analytics to show the coverage and levels. Ad-Hoc networks [10] can be use in the similar research, but this architecture is too difficult for our subject area, peer-to-peer will be enough. Materials usage [11, 12] and interlayer microclimatic features were describe, but nothing about sensors that can be uses in the interlayer areas of clothes. WSN security and key generation [13–15] for not secured network for clothes studying can be use in future research, but not on this step. Attack detection and models actuality [16] is urgent task, but real coverages of WSN not described. Adaptive antenna systems for radar and telecommunications applications [17] not adopted for chosen ISM bands. In review of 15 papers [18] the reconfigurable antennas for body centric communication were focused on wide used band 2.4 GHz, but nothing about long-range subgigahertz systems. Robots really can help in lifesaving tasks, but radio control data network is described poor in work [19]. Cloud of Things is very close to IoT Cloud, especially for research of smart things and clothes [20] and unfortunately, that TEQIP-III project decryption is too slow for our task. Sleep-mode energy management [21] in cellular networks management technic cannot be use in peer-to-peer network. In [22] compared measured and calculated signal losses on the low distances, but again only for popular 2.4GHz band only. In the analyzed works was not described a sub-gigahertz wireless sensor network and a computer system for processing statistical data collected from this network. 1.2 Objective The objective of the research in this article is sub-gigahertz wireless sensor network development that allows increase coverage area for the manufacture of military troop shirt with the zonal arrangement of temperature and relative humidity sensors with the possibility of their real-time remote monitoring and cloud storing of obtained data.

2 Materials and Research Method The proposed experimental setup consists of a power supply lithium-ion batteries module with a small capacity of approximately 500 milliamps per hour, microcontroller of the Atmel family of 8 -bit architecture, which can be from mega 88 to 328, temperature and humidity sensor ds18b20 for temperature measuring in hard-to-reach places and sensors temperature humidity, produced by Amsong company am2301 2302. Radio modules produced by Silabs company, namely si4432 or si4463, are used for data transmission over a wireless channel, which allow working in the unlicensed ISM band of 433 MHz with the use of a small power of approximately 10 mW. On the server side, there is a radio module also from the Silabs family, which is connected to a computer via UART (Universal Asynchronous Receiver Transmitter). Further data processing is done by reading the stream from the UART and processing the data using the built-in libraries of the Python language. Figure 1 shows the internal structure and construction of the sensors, with 1-wire bus in sensor Dallas for use in hard-to-reach areas of clothing.

660

A. Moshenskyi et al.

Fig. 1. Temperature and relative humidity sensors that were used in the research

Figure 2 shows the structural diagram of the module with the microcontroller in the center and shows which buses and ports the corresponding sensors and peripheral devices are connected to. It shows that the analog- to- digital converter of the controller is used to monitor the battery status.

Fig. 2. Scheme of the mobile module

Sub-Gigahertz Wireless Sensor Network

661

The diagram in Fig. 3 shows a graphical interface for monitoring the temperature and humidity between the air layers of clothing.

Fig. 3. Measurement of object parameters in real time

Figure 4 shows the optimal model of the sensors location on clothes according to the authors. The microcontroller module is located on the back of the garment in the upper part and this is optimal from the point of view of ergonomics and the use of overall modules during wearing.

Fig. 4. Scheme of the sensor’s placement in the experimental sample

In Fig. 5, part a) shows the appearance of the sensors and the receiving device, part b) shows another sensor module, which consists of an accelerometer, sensors for microclimatic local conditions, a navigation module and a radio module for GSM communication and Wi-Fi communication connected, and built on another microcontroller of the esp8266 family. Using Wi-Fi or GSM module wireless communication technology, it is possible to directly connect the sensor unit to the global Internet using a router or cellular station as shown in Fig. 6. The research of the authors is related to the development of smart clothing for military purposes, therefore the range of wireless and Wi-Fi technology was insufficient, and the coverage of cellular networks in some areas, namely military training grounds, was also

662

A. Moshenskyi et al.

Fig. 5. a) Modifications of sensors for the remote monitoring system of changing in the internal microclimate in the air gap between cloth layers: 1 – base module – control station 1SI4463 based radio module, 2 – USB convertor ch340, 3 – powerbank 1 X 18650 USB port of PC mobile module, sensors on the clothes, 4– AM2302 temperature and humidity sensor, wire connected, 5 –DS18b20 thermometer, 6 – 100mah LiIon battery, 3 h of supply without sleep mode and with cyclic transmissions on 100 mW, 7 – ADXL345 3 axis digital accelerometer, 8 – Atmega328 MCU for sensor request, data processing, 9 – SI4463; b) GSM module diagram

Fig. 6. The block diagram of the moving objects survey using a WiFi network

insufficient, therefore the technology of data transmission over a radio channel on the VHF and UHF was used, which are shown in more detail in the structural diagram of Fig. 7. The left part of Fig. 7 shows a sensor module consisting of a microcontroller, a set of sensors, a power supply module and a radio module with an integrated portable antenna. In the right part of Fig. 7, a radio module connected to a serial interface of a computer connected to the Internet is shown, and an example of using the APRS (Automatic Packet Reporting System) network is given. Figure 8 shows the structural diagram of the data transmission channel. Data from the sensors are processed by microcontrollers and sent to the transmitter radio module of the portable device. Through the air, data is transmitted to the receiving radio module, processed by a computer and sent to a data storage in the Internet or a local server. The computer transfers data to the cloud database where it is stored for usage.

Sub-Gigahertz Wireless Sensor Network

663

Fig. 7. Structural scheme of the system using the VHF radio network (range up to 100 km)

Fig. 8. Scheme of data transmission channels

3 Research Results For monitoring hard-to-reach places between layers of clothing, the ds18b20 sensor is recommended due to the use of a convenient one-wire bus and the ability to work in 12-bit mode, which allows assess temperature trends with a fairly high resolution. The unique 64 bit address of each sensor cannot impose a limit on the number of sensors on one item of clothing within reason. From the Fig. 9 we could see that average temperature 1 slightly increases from 27.4 to 27.9 °C and relative humidity 1 increases from 44.8 to 46.5% when subject do physical activities (walking for 10 min) after some rest average temperature decreases to 24.1 °C and relative humidity decreases to 37.5%, after another portion of physical activities (running for 15 min) average temperature increases to 26.2 °C and relative humidity to 43%. Without physical activity temperature in inter-layer clothing space decreases up to 23.1 °C and relative humidity almost does not changes it slightly decreases to 40.8 and increases to 43%. The average temperature 2 almost does not changes and account for about ~ 26 °C and relative humidity decreases from 43.7 to 29.8% for 2 h on the street in the winter with temperature ~ - 10 °C. The research subjects were in different types of troop uniform. Temperature and humidity sensors manufactured by the Amsong company can be removed from the outer case and deeply integrated into thin interlayer spaces. In this research we obtained the following indicators: operating range of humidity is 0–100%

664

A. Moshenskyi et al.

Fig. 9. Dependence of average temperature and relative humidity in inter-layer clothing space

RH; operating range of temperature -40 ~ 80 deg. Celsius; accuracy of related humidity + -2% RH in this range; accuracy of temperature + -0.5 °C in this range. Sensor initialization and library usage is on the Wiring language for replication simplicity. Figure 10 shows radio coverage parameters modeling of the research. Figure 11 shows calculation of the radio link losses in the free space. On UHF band we can use Friis equation for free space losses plus 20 dB for compensation of the “green” and surface losses in the first Frenel zone on distances for mile less. Figure 11 shows the calculation of attenuation in the radio channel described by the ideal radio communication formula, the Friis formula. In the band of 430 MHz, at the permissible power of the radio module of 10 mW and the maximum sensitivity of the radio-receiving device, respectively, for a data transfer rate of 9600 bits per second, a communication distance of more than 1 km can be easily achieve. The Fig. 11 shows the calculation of the path losses and influence the choice of data modem modes. Radio coverage area for the specified frequency, power and the antenna location height of 2 m above the ground in Ukraine, Cherkasy region in the northern part in Fig. 12. To legalize the conduct of experiments, the call sign of the author ut5uuv [9] was used and the power was increased to 100 mW for the permissible limit of 5000 mW of the CEPT (European Conference of Postal and Telecommunication administration) limits. Testing of proposed system is perform in thousands of data transfer sessions. Coverage simulation was very close to the real data transfer area. Proposed network can be recommend for the similar researches of smart systems of local data collecting. Experiment result comparison shows the great increasing of data transfer coverage areas (Table 1). Proposed adopted licensing and data modes make possible to build networks with coverages, which can competitive with national cell networks but with much lower data rates. As shown, usage ISM on the sub-gigahertz bands data links produce data rates

Sub-Gigahertz Wireless Sensor Network

665

Fig. 10. Radio coverage parameters modeling

less than 10 Kb/s but the radius increase more then on hundreds of percent (>200%). Amateur radio and the same speed limits produce extension of coverage radius in 10 and more times versus ISM 433.

666

A. Moshenskyi et al.

Fig. 11. Link path calculation and speed comparison in proposed system.

Fig. 12. Radio link path loss and wireless sensor network coverage

Superposition of proposed usage of public APRS service and Amateur CEPT license adaptation for low speed data links extends the range over the areas with allocated gates of proposed services to hundreds of kilometers. Digipiter (Digital repeater) on the International Space Station and another LEO Satellites also can be used for proposed system extension. For increase of the link budget, author uses the LoRa technology on the local and LEO satellites [23]. Some problems has a CEPT license now, so transmission can be

Sub-Gigahertz Wireless Sensor Network

667

Table 1. Radius of average coverage comparison for different technologies Research technologies

Standard Wi-Fi

Cellular

ISM 2.4 GHz baseband

Proposed ISM 433 MHz

Amateur CEPT 433 MHz (5W)

Amateur CEPT 433 MHz (5W) + APRS

Radius of average coverage [km]

0.1

35(71 extended range)

0.3–0.5

>1

>>10

>>100 across the gates

done not under the amateur call sign in the Ukraine, but only like ISM object. Local emission class tables not changed yet.

4 Conclusions Wireless sensor network developed and tested by the authors, was used to test military clothing in Ukraine. This made it possible to monitor interlayer spaces of clothes in real time in an area that did not have cellular communication and the ability to connect to local Wi-Fi networks. The accuracy of temperature sensors of 12 bits made it possible to measure the temperature with a resolution of 0.0625 degrees Celsius, which allows determine clearly the tendency to increase or decrease in temperature. At the same time, a real accuracy of plus or minus 0.5 degrees Celsius was ensure in the excessive temperature range from minus 10 to + 85 degrees Celsius. New combination of proposed equipment and usage of sub-gigahertz band and ISM licensed network extended the radius of WSN to more than 1 km. Usage the same band in combination with amateur radio license and output power 5 watts less increase the coverage to limits of line-of-slight. First time data transfer protocols and emission classes was adopted to ISM and amateur radio service (CEPT) limits. Integration of the above wireless sensor network into APRS networks made it possible to conveniently use publicly available services and extend the coverage by public receiver gates and tens kilometers radius across them. The new technology of textile materials modification for smart clothes with the use of sensors and microcontroller equipment were proposed. New dependencies of temperature and relative humidity in inter-layer clothing space from the time were determined. It was establish that for the first subject an increase of the average temperature and relative humidity is due to physical activity and type of troop uniform. The second subject results were more stable than for the first one. Proposed WSN have chances for the real life cycle due the ISM limitations. Future work will be integration of the LoRa networks for the subject areas and similar research. Chirp modulation and frequency spreading increase the link budget, but the local license limitation for amateur radio did not allow to use wide band modulations, so ISM power limitation create problems for the authors research.

668

A. Moshenskyi et al.

References 1. Langereis, G.R., Bouwstra, S., Chen, W.: Sensors, Actuators and Computing Architecture Systems for Smart Textiles. In: Chapman, R. (ed.) Smart Textiles for Protection, vol. 1, pp. 190–213. Woodhead Publishing; Cambridge, UK (2012) 2. Custodio, V., Herrera, F.J., López, G., Moreno, J.I.: A review on architectures and communications technologies for wearable health-monitoring systems. Sensors. 12, 13907–13946 (2012) 3. Giddens, H., Paul, D.L., Hilton, G.S., McGeehan, J.P.: Influence of body proximity on the efficiency of a wearable textile patch antenna. Proceeding of the 6th European Conference Antennas & Propagation (EuCAP); Prague, Czech. pp. 1353–1357 (26–30 March 2012) 4. Viry, L., et al.: Flexible three-axial force sensor for soft and highly sensitive artificial touch. Adv. Mater. 2659–2664 (2014) 5. Zhang, L., Wang, Z., Psychoudakis, D., Volakis, J.L.: Flexible Textile Antennas for BodyWorn Communication. Proceedings of IEEE International Workshop on Antenna Technology, pp. 205–208. Tucson, ZA, USA (5–7 March 2012) 6. Yan, S., Soh, P.J.: Wearable dual-band composite right/left-handed waveguide textile antenna for WLAN applications. Electron. Lett. 50, 424–426 (2014) 7. Salvado, R., Loss, C., Gonçalves, R., Pinho, P.: Textile materials for the design of wearable antennas: A survey. Sensors. 12, 15841–15857 (2012) 8. Dursun, M., Bulgun, E., Senol, ¸ Y., Akkan, T.: A Smart Jacket Design for Firefighters. Tekstil ve Mühendis 26(113), 63–70 (2019) 9. Moshenskyi, A.: Private rescue echo beacon with FSK radiomodule. Scientific journal “Science-Based Technologies” 4(48), 478–483 (2020) 10. Oleshchenko, L., Movchan, K.: AODV Protocol Optimization Software Method of Ad Hoc Network Routing. Lecture Notes on Data Engineering and Communications Technologies 83, 219–231 (2021) 11. Zhdanova, O., Bereznenko, S., Bereznenko, N., Novak, D.: Textile materials manufacturing features with the use of antimicrobial additives. Fibres and Textiles 24(4), 41–46 (2017) 12. Kurhanskyi, A., Bereznenko, S., Novak, D., et al.: Effects of multilayer clothing system on temperature and relative humidity of inter-layer air gap conditions in sentry cold weather clothing ensemble. Fibres and Textiles 25(3), 43–50 (2018) 13. Salehi, M., Karimian, J.: A Trust-based security approach in hierarchical wireless sensor networks. Int. J. Wirel. Microw. Technol. (IJWMT) 7(6), 58–67 (2017) 14. Kaur, J., Randhawa, S., Jain, S.: A novel energy efficient cluster head selection method for wireless sensor networks. Int. J. Wirel. Microw. Technol. (IJWMT) 8(2), 37–51 (2018) 15. Pandey, A.K., Gupta, N.: An Energy Efficient Clustering-based Load Adaptive MAC (CLAMAC) Protocol for Wireless Sensor Networks in IoT. Int. J. Wirel. Microw. Technol. (IJWMT) 9(5), 38–55 (2019) 16. Nangue, A.: Elie Fute Tagne, Emmanuel Tonye, Robust and Accurate Trust Establishment Scheme for Wireless Sensor Network. Int. J. Comp. Netw. Info. Secu. (IJCNIS) 12(6), 14–29 (2020) 17. Shcherbyna, O., Zaliskyi, M., Kozhokhina, O., Yanovsky, F.: Prospect for using low-element adaptive antenna systems for radio monitoring stations. Int. J. Comp. Netw. Info. Secu. (IJCNIS) 13(5), 1–17 (2021) 18. Munawar,H.S.: An Overview of Reconfigurable Antennas for Wireless Body Area Networks and Possible Future Prospects. Int. J. Wirel. Microw. Technol. (IJWMT) 10(2), 1-8 (2020) 19. Punith Kumar, M.B., Sumanth, S., Savadatti, M.A.: Internet Rescue Robots for Disaster Management. Int. J. Wirel. Microw. Technol. (IJWMT) 11(2), 13–23 (2021)

Sub-Gigahertz Wireless Sensor Network

669

20. Bashir, A., Sholla, S.: Resource efficient security mechanism for cloud of things. Int. J. Wirel. Microw. Technol. (IJWMT) 11(4), 41–45 (2021) 21. Agubor, C.K., Akande, A.O., Opara, C.R.: On-off Switching and Sleep-mode Energy Management Techniques in 5G Mobile Wireless Communications – A Review. Int. J. Wirel. Microw. Technol. (IJWMT) 12(6), 40–47 (2022). https://doi.org/10.5815/ijwmt.2022.06.05 22. Ng, S.T., Lee, Y.S., Radhakrishna, K.S., Krishnan, T.M.: Study of 2457 MHz WIFI Network Signal Strength at Indoor and Outdoor Enviroment. Int. J. Wirel. Microw. Technol. (IJWMT) 11(6), 1–9 (2021). https://doi.org/10.5815/ijwmt.2021.06.01 23. UT5UUV Console. https://tinygs.com/station/UT5UUV@320362520

Paradoxical Properties Research of the Pursuit Curve in the Intercepting a Fugitive Problem Viktor Legeza and Liubov Oleshchenko(B) Igor Sikorsky Kyiv Polytechnic Institute, Kyiv 03056, Ukraine [email protected]

Abstract. The actual problem of building a mathematical model of the process of “flat” pursuit with the selection of the optimal angle of inclination of the line of escape in order to avoid capture by the fugitive on the longest segment along the given direction has been solved. The proposed new formulation of the problem makes it possible to determine new properties of the pursuit curve and characteristics obtain from it, which affect the fugitive‘s ability to avoid capture at the longest horizontal distance. A quantitative and qualitative analysis of the influence of the characteristics of the pursuit curve on the maximum movement of the fugitive before his capture was carried out. The best case shows the minimum extremum in result comparison, it can show the optimal path. The obtained optimized model makes it possible to choose the angle of inclination of the line of escape, which ensures exactly such an optimal movement of the escapee. It was established first that for a successful escape, the strategy of choosing an angle of inclination of the escape line equal to zero (that is, escape along the shortest segment connecting two parallel lines – the vertical axis of the ordinate and the “life” line) is not optimal. First time the dependence of the change in the distance (on the horizontal axis) to the point of capture of the fugitive, depending on the magnitude of the angular coefficient of his movement along an inclined line at a fixed angular coefficient, was established. It is shown that this dependence has a well-defined local maximum, which indicates the presence of a certain non-zero angle of inclination of the fugitive’s movement line, which provides him with the opportunity to reach the maximum movement towards the lifeline on the border between countries. New paradoxical phenomenon was founded: for a successful escape, the strategy of choosing the angle of inclination of a straight line equal to zero (that is, moving along the shortest segment connecting two parallel straight lines) is not correct.The results of the numerical experiment give reasons to recommend the proposed model for practical use. The future development of this research can be use in military task in the Southeast of Ukraine in torpedoes usage. Keywords: Pursuit curve · local maximum · angle coefficient · selection strategy · slope angle · escape line · “life” line

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 670–681, 2023. https://doi.org/10.1007/978-3-031-36118-0_60

Paradoxical Properties Research of the Pursuit Curve

671

1 Introduction The pursuit curve is the line along which the pursuer moves while pursuing the fugitive. The problem of persecution probably dates back to the time of Leonardo da Vinci. He was the first to investigate this problem when the fugitive moved in a horizontal straight line. The general case was studied by the French scientist Pierre Bouguer in 1732. The task was to find the curve of pursuit of a merchant ship by a pirate ship. At the same time, it was assumed that the speeds of the two vessels are always in the same ratio. The use of differential equations for the pursuit-evasion problem has recently become very relevant for the development of computer games. Multi variables vas reduced to only two variables in research [1] for the tasks of graphic in pursuit-evasion game. In research [2] authors described pursuit and evasion strategy and used only the deep neural network instruments, but nothing in determinate math. Works [3–5] describe the same game tasks by the deep reinforcement learning that can make instruments for the decision-making technology without proposed optimal routes for the moved object. In works [6, 7], the well-known problem of “chasing four mice” in various settings was considered. Suppose four mice are located in each of the four corners of a square table, and each mouse runs to the one to the right of it. We need to find parametric curves that describe the trajectory of each mouse. The solutions of this problem will be spiral trajectories that coincide in the center of the table. In research [8], deterministic continuous pursuit is considered, in which n ants chase each other in a circle and have predetermined variable speeds. Two discrete analogues are considered, in which a cricket or a frog is cyclically pursued with a constant and equal speed. The possible evolution of this motion as time approaches infinity is explored: collisions, limit points, equilibrium states, and periodic motion. The paper [9] presents a simple mathematical model of the local interaction of a colony of ants or other natural or artificial creatures with a great “sense of global geometry” to find a direct path from the anthill to food. This task was also considered within the framework of the general task of chasing n ants. The paper [10] investigates the movement of an arbitrary set of points (or beetles) on a plane chasing each other in cyclic pursuit. It is shown that for regular centrally symmetric configurations, analytical solutions are easily obtained by switching to the corresponding rotating frame of reference. Several cases of asymmetric configurations are discussed. For the case where all beetles have the same speed, a theorem is proved that whenever premature (i.e., non-reciprocal) capture occurs, the collision must be head-on. This theorem is then applied to the case of three- and four-beetle configurations to show that these systems collapse up to a point, i.e., the capture is mutual. Some aspects of these results are generalized to the case of systems with n beetles. The work [11] considers a well-known problem that continues to be analyzed today. Three (or more) beetles, which are initially located at the vertices of a regular polygon, begin to cyclically chase each other, moving together at the same speed. It is usually necessary to know the distance traveled by each beetle before mutual capture. The case of four beetles is the simplest. Due to symmetry, the four beetles always remain on the four vertices of the square. Since the fleeing beetle Z 2 always moves at

672

V. Legeza and L. Oleshchenko

right angles to the pursuing beetle Z 1 , the capture speed of these two beetles depends only on the speed of the beetle Z 1 . The case for n bugs is only slightly more complicated. This problem is also symmetric, since n bugs are always located at n vertices of a regular n-gon. The book [12] gives systematic methods of winning in differential pursuit and evasion games. The scope and application of game procedures have been studied. Numerous useful examples illustrate basic and advanced concepts, including capture, strategy, and algebraic theory. Also includes consideration of both linear and non-linear games. Chapters in the book include stroboscopic and isochronous target acquisition, algebraic theory, and selection strategies. The article [13] considers the pursuit problem with simple movement for the case when the maximum speeds of the players are the same, and the fleeing person moves along a strictly convex smooth n-dimensional surface. It is proven that the end of pursuit is possible from any starting position. It is established that evasion is possible from some initial positions if the hypersurface contains a two-dimensional flat part. The book [14] is a well-known textbook on differential games. It treats differential games as conflict situations with an infinite set of alternatives that can be described by differential equations. In the book, the main attention is paid to fundamental mathematical questions, the solution methods are illustrated by a large number of interesting examples, which are important in themselves. A number of unsolved problems are proposed. The technical report [15] provides a detailed description of the implementation of a clean pursuit curve tracking algorithm. Given the general success of the algorithm in recent years, it is likely that it will be used again in land navigation tasks. The report also includes a geometric version of the method and provides some information on the performance of the algorithm as a function of its parameters. The article [16] describes a third-order chase, a game of evasion in which both players have the same speed and minimum turning radius. A idiosyncratic game is first resolved for a barrier or shell of states that can be captured. When capture is possible, the game at some level is decided for optimal control of the two players as a function of relative position. It is found that the solution of the problem includes a universal surface for the pursuer and an acceleration surface for the evaders. The article [17] considers the coplanar problem of evading pursuit, in which two pursuers P1 , P2 and one fugitive participate E. The fugitive E evades with a constant speed w > 1 greater than that of the pursuers, and must pass between the two pursuers P1 and P2 with a single speed, the gain is the distance of closest approach to any of the pursuers. This is a typical two-player zero-sum game theory problem. Velocity directions P1 , P2 and E are chosen as controlling variables. A closed-loop solution is obtained through elliptic functions of the first and second kind. The closed-loop solution is shown graphically in several diagrams for different values of w. Research [18] uses differential equations with fractional variable coefficients that can describe our task, but the simplicity and the accuracy of the proposed numerical scheme not optimal for work on the surface. In [19, 20] collective described predicting rate of divorce tendency and based on it rate of human happiness in Nigeria. Differential equations and limits closely describe same task, but the correlation in results cannot describe binary result.

Paradoxical Properties Research of the Pursuit Curve

673

Economical bonuses by routing optimization [21] makes research task very attractive. Terminal nodes limitations for author’s solutions were chosen. This limit type cannot be use in our task without weapon usage in the model. Surface geometry tasks in [22] describes slow dynamical models with strong (stochastic) noise components. This type of prediction too hard for constant speed tasks. A review of literature sources showed that the problem of pursuit with a formulation in which the maximization of the fugitive’s movement towards a certain “saving” limit due to the selection of the angular coefficient of the escape line was chosen as the quality criterion was not considered before.

2 Problem Statement In some cases, with the passage of time, the old problems of finding the pursuit curves “one pursuer-one fugitive” on the plane are revised in new formulations. This may be related to the possibility of changing different strategies for avoiding interception by the fugitive due to the choice of one or another quality criterion. For example, as a criterion of quality in such tasks, the maximum advance of the fugitive towards a certain border, which is considered “life” line for the fugitive, can be considered. Therefore, from a practical point of view, solving the problem of pursuit “one pursuer-one fugitive” in a production with a new quality criterion will be relevant in such areas as transport, logistics, military affairs, sports events, computer games, etc. The object of research is the process of chasing a fugitive, who tries to move as far as possible in the direction of a certain given limit before the moment of interception, moving along an inclined straight line. The subject of the research is the properties and characteristics of the pursuit curve, which influence the choice of a fugitive’s successful escape strategy. The research objective. The task is: to find 1) the trajectory L : {y = y(x)} of the pursuit boat; 2) the point on the plane at which it will catch up with the boat-fugitive; 3) the angular coefficient that ensures the maximum displacement of the boat-fugitive along the horizontal axis in the direction of the border between the two countries. The flow rate of the reservoir is not taken into account. Consider the starting position of two moving points, one of which is chasing the other (Fig. 1). On one shore (marked as the OY axis in Fig. 1) of the reservoir, two boats stand at a distance y0 = 1 from each other – a coast guard boat U (hereinafter the pursuit-boat) and a border violator boat V (hereinafter the boat-fugitive). The reservoir separates the two countries, as a natural border between them. At the same time, the runaway boat V is trying to cross the body of water as quickly as possible in the direction of the border (“lifeline”)AB between the two countries. We believe that this border is parallel to the shoreline of the start of the two boats (AB ||OY ). The chasing boat U tries to overtake the boat-fugitive V before it reaches a certain point, which is on the border AB of two countries. So, at the initial time t = 0, the violator boat V begins to move along a straight line η = kξ (Fig. 1) with a speed of v. The boat U starts at the same time as the boat V and in the process of pursuit chooses the direction of movement to the current location of the boat V on the plane OXY . At the same time, the pursuit boat U has a constant speed

674

V. Legeza and L. Oleshchenko

Fig. 1. Pursuit curve

u, which is α times greater than the speed of the boat V : u = αv. The fugitive’s task is to move as far as possible in the direction of the border AB between the two countries, choosing for escape the optimal angular coefficient k of the inclined line with a constant α coefficient.

3 Construction of the Differential Equation of the Trajectory of the Pursuit We will call the L trajectory of the boat U the pursuit curve. Let us identify the location of the boats on the plane OXY with the points U and V . We denote by (x, y) the coordinates of the point U and by (ξ, η) are the coordinates of the point V at the current time t. Let’s draw a tangent at a point U (x, y) to the curve L connecting the points U and V and establish the relation for the derivative function y(x): y =

η−y kξ − y ⇒ y = . ξ −x ξ −x

(1)

Let‘s express the variable ξ from Eq. (1): ξ=

xy − y . y − k

(2)

Let’s find the time derivative of the variable ξ : ξ˙ =

y − kx  y x˙ . (y − k)2

(3)

Now let’s determine the ratio α for the speed modules of the boats, taking into account that u=

ds = dt



1 + (y )2 x˙ and v =

 1 + k 2 ξ˙

(4)

Paradoxical Properties Research of the Pursuit Curve

675

Taking into account (3) and (4), we obtain the equation of the trajectory L : {y = y(x)}:  1 + (y )2 (y − k)2 α= √ (5) 1 + k 2 (y − kx)y Let’s rewrite the differential Eq. (5) of the desired curve differently: 

1 + k 2 αy =



1 + (y )2

(y − k)2 . (y − kx)

(6)

Let’s add two boundary conditions to this equation: y(0) = 1; y (0) = −∞.

(7)

So, to find the pursuit curve, we have a boundary value problem with differential Eq. (6) and boundary conditions (7).

4 Solving the Boundary Value Problem of Finding the Trajectory Equation In Eq. (6), we will replace the variable: z = y − kx ⇒ z  = y − k; z  = y As a result, we obtain a nonlinear differential equation of the second order:   α 1 + k 2 ·z  · z = (z  )2 · 1 + (z  + k)2

(8)

Equation (8) allows the order to be reduced by one if the following substitution is made: d 2z dz dp = p(z); 2 = p(z) · . dx dx dz In the new variables, Eq. (9) will take the following form:   α 1 + k 2 ·pz · p · z = p2 · 1 + (p + k)2 .

(9)

(10)

In Eq. (10), we reduce by a factor p(z) = 0 and get the equation with separated variables:  dp dz (11) = . α 1 + k 2·  2 z p 1 + (p + k) After integrating the Eq. (11), we get the equation of the first order:    1 p tgα · arctg = C1 · z. 2 1 + k(p + k)

(12)

676

V. Legeza and L. Oleshchenko

To determine the first constant in Eq. (12), we use the relation z = y − kx and boundary conditions (7). As result of the limit transition, we obtain the first constant:    1 α 1 . (13) C1 = tg · arctg 2 k Let‘s express the function p = z  from (12) and (13) in terms of z:      1 1 1 p α = tg 2arctg z tg arctg . 1 + k(p + k) 2 k

(14)

We denote the right-hand side of relation (14) by f (z, k, α). As a result, we got equations of the first order (here p = z  ): f (z, k, α) · (k 2 + 1) dz = . dx 1 − f (z, k, α) · k Let’s write the second integral using (15):

z 1 − f (z, k, α) · k dz. x(z, k, α) = f (z, k, α) · (k 2 + 1) 1

(15)

(16)

Let’s supplement expression (16) with an expression for y(z, k, α), which together describe the pursuit trajectory y(x) in a parametric form of a pair of functions, where the variable z acts as a parameter, namely:  x = x(z, k, α); (17) y = z + k · x(z, k, α)

5 Research Results for Numerical Analysis of the Pursuit Trajectory and Its Characteristics So, two Eqs. (17) describe the trajectory L : {y = y(x)} of the pursuing boat U . We present a graph (Fig. 2) constructed for the case of the pursuit process, in which the following movement parameters are selected: k = −2 and α = 2. On the graph in Fig. 2, the blue curve represents the trajectory of the fleeing boat V , and the red curve represents the trajectory of the boat U . The point of the plane XOY in which the escape boat is apprehended has coordinates (0,431; −0,863). Geometrically, the fact of arrest means that the red curve is tangent to the blue line. The  time T required to detain a fugitive boat V in the process of pursuit is equal to: T = 0,4312 + (−0,863)2 /v = 0,965/v. Let’s calculate the distance S traveled by the pursuit boat U during the pursuit:S = T · u = 0, 965 · α = 1,93. Figure 3 presents a curve that reflects the dependence of the change in the distance x(0, k, α) (on the horizontal axis OX ) of the arrest of the runaway boat depending on the angular coefficient k of its direct movement at a fixed coefficient α.

Paradoxical Properties Research of the Pursuit Curve

677

Fig. 2. The schedule of the pursuit process, α = 2 and k = −2

Fig. 3. Graph of the dependence of x(0, k, α) on the coefficient k

The point of the plane XOY in which the escape boat is apprehended has coordinates (0,431; −0,863). Geometrically, the fact of arrest means that the red curve is tangent to the blue line. The  time T required to detain a fugitive boat V in the process of pursuit is equal to: T = 0,4312 + (−0,863)2 /v = 0,965/v. Let’s calculate the distance S traveled by the pursuit boat U during the pursuit:S = T · u = 0,965 · α = 1,93. Figure 3 presents a curve that reflects the dependence of the change in the distance x(0,k,α) (on the horizontal axis OX ) of the arrest of the runaway boat depending on the angular coefficient k of its direct movement at a fixed coefficient α. It has a well-defined local maximum, which indicates that there is a certain non-zero angle of inclination of the direct movement of the escape boat, which gives it the opportunity to achieve the maximum movement towards the border between the two countries. So, we observe a certain paradox: the curve in Fig. 3 establishes that the strategy of choosing the angle of inclination of a straight line equal to zero (that is, moving along the shortest segment connecting two parallel lines - which would be natural) is not correct for a successful escape. In this version of the calculation α = 2, k = −0, 4), this movement is as follows:x(0; −0,4; 2) = 0,734.

678

V. Legeza and L. Oleshchenko

Considering the indicated in Fig. 3, the paradoxical property of the pursuit process using (17), we establish the dependence y = y(x) (Fig. 4), where x = xmax (0,k,2) is the maximum value of the fugitive’s V movement along the OX axis with a variable coefficient k and a fixed coefficient α = 2. The graph of the curve (Fig. 4) also shows the point of maximum movement of the fugitive V along the OX axis with a non-zero angular coefficient k = −0, 4, which acts as a parameter of this curve.

Fig. 4. Graph of dependence y = y(x) on the condition that x = xmax (0, k, 2) and the coefficient k changes

Let’s find out the behavior of the quantity x(0, k, α) under the condition that the coefficient α goes either to infinity or to unity. Figure 5 shows the graph of the dependence of the value x(0, k, α) on the coefficient α for a fixed value of the angular coefficient k = −0.45. From the behavior of the curve in Fig. 5, it follows that if the coefficient α → ∞, then the distance x(0, k, α) at which the latter will be detained tends to zero. On the other hand, if α → 1(that is, the speeds of the boats are equal), then the indicated distance goes to infinity, that is, the arrest of the runaway boat will not take place. The Table 1 summarizes the results of numerical experiments to determine the movement of the fugitive x = x(0, k, α) along the axis OX and y = y(0, k, α)- along the axis OY before its interception, the time spent T = T (0, k, α) by the pursuer on delaying the fugitive, and the path S = S(0, k, α) traveled by the pursuer before the delay of the fugitive. Table 1. Research result comparison -62172563900k

0.00

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

x

0.667

0.696

0.718

0.730

0.734

0.730

0.719

0.703

0.683

0.661

0.638

y

0.00

0.07

0.144

0.218

0.294

0.365

0.431

0.492

0.547

0.595

0.638

T

0.667

0.700

0.732

0.762

0.790

0.816

0.838

0.858

0.875

0.890

0.902

S

1.334

1.400

1.464

1.525

1.581

1.632

1.676

1.716

1.750

1.779

1.805

Paradoxical Properties Research of the Pursuit Curve

679

In this case, the parameter is fixed, the speed of the fugitive is equal to one, and the parameter varies from zero to one. From the given table it follows that the best escape strategy of the fugitive is to choose a straight line with the tangent of the slope angle of the straight line of escape, equal to about. At the same time, the pursuer will spend units of time and travel units of the path to delay him. Figure 6 shows the corresponding graph of the dependence of the value on the coefficient, which is constructed for those angular coefficients of the escape line that provide the maxima of the curves on the graph of Fig. 3.

Fig. 5. Graph of dependence of on

Fig. 6. Graph of dependence of on

6 Summary and Conclusions The urgent problem of building a mathematical model of the process of “flat” pursuit with the selection of the optimal angle of inclination of the line of escape with the aim of avoiding the capture of the fugitive by the longest movement has been solved. The fugitive’s goal is to advance as far as possible to the lifeline on the border between the two countries, choosing the new optimal angular ratio of the straight line to escape.

680

V. Legeza and L. Oleshchenko

The equation of the pursuit optimized curve is obtained in a parametric form. Its numerical analysis was carried out, the influence of the parameters of the pursuit process on the form of the curve was investigated. The best case, the parameter is about 0.4, the time units and travel in result comparison show optimal path. First time the dependence of the change in the distance (on the horizontal axis) to the point of capture of the fugitive, depending on the magnitude of the angular coefficient of his movement along an inclined line at a fixed angular coefficient, was established. It is shown that this dependence has a well-defined local maximum, which indicates the presence of a certain non-zero angle of inclination of the fugitive’s movement line, which provides him with the opportunity to reach the maximum movement towards the lifeline on the border between countries. New paradoxical phenomenon was founded: for a successful escape, the strategy of choosing the angle of inclination of a straight line equal to zero (that is, moving along the shortest segment connecting two parallel straight lines) is not correct. The future development of this research can be use in military tasks in the Southeast of Ukraine in torpedoes usage. The mathematical results of this research may also be useful for computer game industry that needs movement engines for much realistic in game physic.

References 1. Patsko, V., Kumkov, S., Turova, V.: Pursuit-Evasion Games. In: Ba¸sar, T., Zaccour, G. (eds.) Handbook of Dynamic Game Theory, pp. 951–1038. Springer, Cham (2018). https://doi.org/ 10.1007/978-3-319-44374-4_30 2. Xu, C., Zhang, Y., Wang, W., Dong, L.: Pursuit and evasion strategy of a differential game based on deep reinforcement learning. Frontiers in Bioengineering and Biotechnology (March 2022). https://doi.org/10.3389/fbioe.2022.827408 3. Lopez, V.G., Lewis, F.L., Wan, Y., Sanchez, E.N., Fan, L.: Solutions for multiagent pursuitevasion games on communication graphs: finite-time capture and asymptotic behaviors. IEEE Trans. Autom. Control. 65, 1911–1923 (2019). https://doi.org/10.1109/TAC.2019.2926554 4. Zhou, Z., Xu, H.: Mean field game and decentralized intelligent adaptive pursuit evasion strategy for massive multi-agent system under uncertain environment with detailed proof. Proceedings of the Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications II; Online. (27 April–8 May 2020) 5. Wan, K., Wu, D., Zhai, Y., Li, B., Gao, X., Hu, Z.: An improved approach towards multi-agent pursuit-evasion game decision-making using deep reinforcement learning. Entropy (Basel). 23(11), 1433 (2021). https://doi.org/10.3390/e23111433. Oct 29 6. Malacká, Z.: Pursuit Curves and Ordinary Differential Equations. Komunikacie 14(1), 66-68. University of Žilina (March 2012). https://doi.org/10.26552/com.C.2012.1.66-68 7. Nahin, P.J.: Chases and Escapes: The Mathematics of Pursuit and Evasion. Princeton-New Jersey: (Princeton Puzzlers) 272 (2012) 8. Bruckstein, A.M., Cohen, N., Efrat, A.: Ants, crickets and frogs in cyclic pursuit, Preprint, Technion CIS Report, Haifa #9105 (July 1991) 9. Bruckstein, A.M.: Why the ant trails look so straight and nice. The Mathematical Intelligencer 15(2), 59–62 (1993). https://doi.org/10.1007/BF03024195 10. Behroozi, F., Gagnon, R.: Cyclic pursuit in a plane. J. Math. Phys. 20, 2212–2216 (1979). https://doi.org/10.1080/00029890.1975.11993941. November

Paradoxical Properties Research of the Pursuit Curve

681

11. Klamkin, M.S., Newman, D.J.: Cyclic pursuit or “The Three Bugs Problem”. Amer. Math. Monthly 631–639 (June 1971). https://www.jstor.org/stable/2316570 12. Hajek, O.: Pursuit Games: An Introduction to the Theory and Applications of Differential Games of Pursuit and Evasion. Academic Press, New York (1975) 13. Kuchkarov, A.S., Rikhsiev, B.B.: A pursuit problem under phase constraints. Autom. Remote. Control. 62(8), 1259–1262 (2001) 14. Isaacs, R.: Differential Games, p. 260. John Wiley and Sons, New York (1967) 15. Craig Conlter, R.: Implementation of the pure pursuit path tracking algorithm. CMU-RI-TR92–01. The Robotics Institute Carnegie Mellon University Pittsburgh, Pennsylvania 15213 (January 1992) 16. Merz, A.: The game of two identical cars. Journal of Optimization Theory and Applications 9, 324–343 (1972). https://doi.org/10.1007/BF00932932 17. Hagedorn, P., Breakwell, J.V.: A differential game with two pursuers and one evader. Multicriteria Decision Making and Differential Games 443–457 (1976). https://doi.org/10.1007/ 978-1-4615-8768-2_27 18. Falade, K.I., Tiamiyu, A.T.: Numerical solution of partial differential equations with fractional variable coefficients using new iterative method (NIM). Int. J. Mathem. Sci. Comp. (IJMSC) 6(3), 12–21 (2020). https://doi.org/10.5815/ijmsc.2020.03.02 19. David, O.O., et al.: A mathematical model for predicting rate of divorce tendency in nigeria: a study of taraba state, nigeria. Int. J. Mathem. Sci. Comp. (IJMSC) 6(5), 15–28 (2020). https:// doi.org/10.5815/IJMSC.2020.05.02 20. David, O.O., et al.: Mathematical model for predicting the rate of human happiness: a study of federal university Wukari community of Nigeria. Int. J. Mathem. Sci. Comp. (IJMSC) 6(6), 30–41 (2020). https://doi.org/10.5815/IJMSC.2020.06.05 21. Gnanapragasam, S.R., Daundasekera, W.B.: Optimal solution to the capacitated vehicle routing problem with moving shipment at the cross-docking terminal. Int. J. Mathem. Sci. Comp. (IJMSC) 8(4), 60–71 (2022). https://doi.org/10.5815/ijmsc.2022.04.06 22. Nazimuddin, A.K.M., Ali, S.: Application of differential geometry on a chemical dynamical model via flow curvature method. Int. J. Mathem. Sci. Comp. (IJMSC) 8(1), 18–27 (2022). https://doi.org/10.5815/ijmsc.2022.01.02

Oscillations at Nonlinear Parametric Excitation, Limited Power-Supply and Delay Alishir A. Alifov1(B) and M. G. Farzaliev2 1 Mechanical Engineering Research Institute of the Russian Academy of Sciences,

Moscow 101990, Russia [email protected] 2 Azerbaijan State University of Economics, Baku, Azerbaijan

Abstract. The work is devoted to the development of the theory of oscillatory systems with limited power-supply, the relevance of which has increased due to energy and environmental problems. A system with an energy source of limited power and a time delay in the elastic force under the action of nonlinear parametric excitation is considered. On the basis of the direct linearization method, the linearization of the nonlinear elastic force is carried out, the equations of unsteady motions are obtained. Stationary modes and their stability are considered. Using the Routh-Hurwitz criteria, the stability conditions of stationary oscillations are obtained. To obtain information about the effect of lag on stationary modes of motion, calculations were performed. The obtained graphical results clearly show the influence of nonlinear parametric excitation on the dynamics of the system in the presence of a delay in the elastic force. Depending on the value of the delay, different features appear in the dynamics of oscillations and their stability. The curves shift in amplitude and frequency, and a rather significant shift appears in the case of a nonlinear elastic force. Keywords: Oscillations · Nonlinearity · Parametric excitation · Energy source · Limited power · Delay · Method · Direct linearization

1 Introduction A machine in the modern definition is a device that performs some of the functions assigned to it and is able to adapt to environmental changes with the help of artificial intelligence. Any device uses an energy source to perform its assigned functions. As an object of control in an artificial intelligence system (robots, automatic machines, manipulators, etc.), there can be time-varying displacement, speed, electrical voltage, etc. They are included in the design schemes and models of the description of the device when it is created. Differential equations, including nonlinear ones, are very often used as an analysis of dynamic models of devices. To solve nonlinear differential equations, a number of approximate methods of nonlinear mechanics have been developed [1–8, etc.], which are characterized by high time and labor costs. In these methods, the form of the solution is first specified, and then © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 682–689, 2023. https://doi.org/10.1007/978-3-031-36118-0_61

Oscillations at Nonlinear Parametric Excitation

683

rather laborious calculations are performed. The complexity and laboriousness of mathematical relations increases sharply with an increase in the order or number of the approximation, which causes the problem of their applicability in practice. The same growth takes place with an increase in the degree of nonlinearity, which in most cases leads to the construction of only the first approximation. With reference to [9–12], in [13] this problem is noted in the study of coupled oscillatory networks that play an important role in electronics, neural networks, physics, etc. The method of direct linearization (MPL) described in [14–17, etc.] is fundamentally different from the commonly used known methods of nonlinear mechanics. In MPL, the nonlinear function is replaced by a linear one, the form of solving the original nonlinear equation is not required, and there are no approximations of different orders. It is quite simple to use, reduces labor and time costs by several orders of magnitude. Simplicity and low labor intensity of using any method are very important when calculating real devices (machines, equipment, electrical systems, etc.). In devices of various kinds (electronics, tracking systems, regulators, etc.), systems with a delay are widespread. The source of lag in autonomous and remote-controlled systems (mobile robots, manipulators, etc.) is the data transmission channel. For example, when managing a group of robots, neglecting them can lead to significant errors [18]. A large number of papers [18–22, etc.] are devoted to the problems of oscillations in systems with a delay without taking into account the properties of the energy source. And the work in which there is such a record is incomparably less. In this article, oscillations are considered under nonlinear parametric excitation, taking into account an energy source of limited power and a retarded elastic force. The purpose of the work is to development the theory of oscillatory systems with limited excitation to use the results obtained in the calculation of real objects and the selection of energy sources with its minimum consumption. The relevance of this direction of the theory of oscillations has increased due to the growing energy and environmental problems in the world. The ecological crisis is a crisis of human society. The total energy consumption on Earth is growing in proportion to the square of the population, humanity has very significantly exceeded the limit of its energy capacity and has gone beyond the threshold of self-destruction of the biosphere.

2 Model and Equations of the System Figure 1 shows the model of the system described in [23]. It represents a rod with a spring, which is connected to a crank driven by a limited-power motor. The system is nonlinear and, taking into account the nonlinearity of parametric excitation and the delay in the elastic force, is described by the equations y¨ + β1 y˙ + ω2 y + by3 sin ϕ = −m−1 f (y) − cτ yτ

(1)

J ϕ¨ = M (ϕ) ˙ − 0.5c2 y2 cos ϕ − 0.5c3 sin 2ϕ − c4 cos ϕ where ω2 =

c m, c β β1 = m ,

=

π 4 EIx (1 − PP01 ), 2l 3

P0 = f0 c1 , P1 =

b=

c2 m,

π 2 EIx , l2

c2 = − π

2r c 1 1

2l

, m=

c3 = c1 r12 , c4 = f0 r1 c1

ρl 2

684

A. A. Alifov and M. G. Farzaliev

Here c1 = const and β = const are the coefficients of stiffness and resistance, b = const, f (y) is the nonlinear part of elasticity, cτ = const, yτ = y(t − τ ), τ = const is time delay, J = const is the moment of inertia of the motor rotor rotating the crank with ˙ is the driving torque of the radius r1 , ϕ˙ is the speed of rotation of the motor rotor, M (ϕ) motor taking into account the forces of resistance to rotation.

Fig. 1. System model

Let us represent the nonlinear function f (y) as a polynomial f (y) =

 s

γs y s , s =

2,3,4,…. Using DLM [14–17], we replace it with a linear function f∗ (y) = Bf + cf y

(2)

where Bf and kf are the linearization coefficients defined by the expressions  Ns γs as , s = 2, 4, 6, . . . (s is even number) Bf = s

kf =



N s γs as−1 , s = 3, 5, 7, . . . (s is odd number)

s (2r+1) (2r+3) Here a = max|x|, Ns = (2r+1+s) , N s = (2r+2+s) , r is the linearization accuracy parameter which affects the accuracy of the method and is introduced into the functional for determining the linearization coefficients. Taking into account (2), Eqs. (1) take the form:

y¨ + β1 y˙ + ω02 y + by3 sin ϕ = −m−1 Bf − cτ yτ J ϕ¨ = M (ϕ) ˙ − 0.5c2 y2 cos ϕ − 0.5c3 sin 2ϕ − c4 cos ϕ where ω02 = ω2 + cf m−1

(3)

Oscillations at Nonlinear Parametric Excitation

685

3 Equation Solutions Solutions of Eqs. (3) can be constructed by two methods [14], of which the method of replacing variables with averaging allows us to study stationary and non-stationary processes. In [14] there is a standard form of the relation for a linearized nonlinear equation of a fairly general form. The first Eq. (3) is one of the special cases of such an equation. In [16], the averaging procedure is described, which is applicable to solving the second equation in (3). It replaces the velocity ϕ˙ with its averaged value . Based on the above, using y = a cos ψ, yτ = a cos(ψ − pτ ), ϕ˙ = , ψ = pt + ξ, p =

2

we obtain the following equations of unsteady motions da dt dξ dt

3

= − a2 (β1 − 2cτ −1 sin pτ ) + ba 4 cos 2ξ 2 − 2 2 c = 4ω4 + 4 f + c τ cos pτ − ba2 sin 2ξ d dt

=

1 J

M ( ) −

(4)

c2 a 2 8

The equations of stationary motions follow from the conditions a˙ = 0, ξ˙ = 0, ˙ = 0: E 4m2 B2 + E 2 = 4m2 b2 a4 , tg2ξ = − 2mB M ( ) − S(a) = 0

where B = 2 β1 − 4cτ sin pτ, E = m(4ω2 − 2 + 4cf + 4mcτ cos pτ 2 S(a) = c28a The expression S(a) defines the load on the energy source and the intersection point of the curves S(a), M ( ) gives a stationary velocity value .

4 Stability of Stationary Motions It is necessary to check the stability of stationary modes of movement. By composing the equations in variations for (4) and using the Routh-Hurwitz criteria D1 > 0, D3 > 0, D1 D2 − D3 > 0

(5)

where D1 = −(b11 + b22 + b33 ), D2 = b11 b33 + b11 b22 + b22 b33 − b23 b32 − b12 b21 − b13 b31 D3 = b11 b23 b32 + b12 b21 b33 − b11 b22 b33 − b12 b23 b31 − b13 b21 b32

686

A. A. Alifov and M. G. Farzaliev

we have 2a b11 = J1 Q, b12 = − c4J , b13 = 0 a ba3 b21 = − 2 cτ sin pτ − 4 2 cos 2ξ, b22 = − 21 (β1 − 2cτ −1 sin pτ ) + 3 b23 = − ba 2 sin 2ξ, b31 = 1 ∂cf ba b32 = m ∂a − sin 2ξ, Q = dd M ( )

When calculating calculating

∂cf ∂cf ∂ , ∂a

∂Bf ∂

,

∂Bf ∂a

ω2 −0.25 − 2 2 b33 = − ba

−

cf m 2

−

cτ 2

cos pτ +

ba2 2 2

3ba2 4

cos 2ξ

sin 2ξ

cos 2ξ

, only even powers of s are taken into account, and when

odd powers of s are taken into account.

5 Calculations To obtain information about the influence of delay on the dynamics of the system, calculations were carried out. The nonlinearity f (y) was chosen as γ y3 and the following parameters were used: ω0 = 1 s–1 , m = 1 kgf·s2 ·cm–1 , c2 = 0.07 kgf·cm–1 , β = 0.02 kgf·s·cm–1 , cτ = 0.05 kgf·cm–1 , γ = ±0.2 kgf·cm–3 The linearization coefficient N 3 is calculated with the linearization accuracy parameter r = 1.5 and has the value N 3 = 34 (for r = 1.5, the results of the direct linearization method completely coincide with the results based on the asymptotic method of Bogolyubov-Mitropolsky [2]). Figure 2 shows the amplitude-frequency dependences a( ), where the double-dashed curves correspond to pτ = π/2, dashed curves to pτ = π, dash-dotted curves to pτ = 3π/2, and solid curves to the absence of delay. Stationary solutions do not take place at γ = – 0.2 and pτ = 3π/2. The steepness of the characteristic of the energy source M ( ), which is within the shaded sector, corresponds to stable oscillations. These sectors should be indicated on the S(a) load curve, but for brevity they are shown on the amplitudefrequency curves. Criteria (criterion) of stability (5) in the black-filled parts of the sectors are performed in the form 0.000X > 0, where X ≤ 9, i.e. there is a rather weak stability. At pτ = π/2, the curves are unstable for both linear (γ = 0) and non-linear (γ = ± 0.2) elasticity. There is also the following interesting feature. For pτ = π, in the case of γ = 0.2, small amplitude values are stable, and in the case of γ = – 0.2, large ones. Moreover, in both cases, stability takes place in small sections of the amplitude curves.

Oscillations at Nonlinear Parametric Excitation

687

Fig. 2. Amplitude-frequency curves

6 Discussion and Conclusion We have considered the effect of the delay in the elastic force on the dynamics of a system with nonlinear parametric excitation at a limited power of the energy source supporting the oscillations. It turned out that depending on the magnitude of the delay,

688

A. A. Alifov and M. G. Farzaliev

different features appear in the system. These features relate both to the location of the amplitude-frequency curves of stationary oscillations in the frequency domain, and their stability. The curves shift in amplitude and frequency, and a rather significant shift appears in the case of a nonlinear elastic force. At some value of delay, stationary oscillations turn out to be unstable both for linear and nonlinear elasticity. At other value of delay and nonlinear elasticity, stability takes place in small sections of the amplitude curves, and in the case of “hard” nonlinearity, small values of the amplitude are stable, and in the case of “soft” – large.

References 1. Butenin, N.V., Neymark, Y., Fufaev, N.A.: Introduction to the Theory of Nonlinear Oscillations. Nauka, Moscow (1976). (in Russian) 2. Bogolyubov, N.N., Mitropolskii, Y.: Asymptotic Methods in Theory of Nonlinear Oscillations. Nauka, Moscow (1974). (in Russian) 3. Moiseev, N.N.: Asymptotic Methods of Nonlinear Mechanics. Nauka, Moscow (1981). (in Russian) 4. Tondl, A.: On the interaction between self-exited and parametric vibrations. National Research Institute for Machine Design Bechovice. Series: Monographs and Memoranda, p. 25 (1978) 5. Hayashi, C.: Nonlinear Oscillations in Physical Systems. Princeton University Press, New Jersey (2014) 6. He, J.H.: Some asymptotic methods for strongly nonlinear equations. Int. J. Modern Phys. B 20(10), 1141–1199 (2006) 7. Karabutov, N.: Structural Identification of Nonlinear Dynamic Systems. Int. J. Intell. Syst. Appl. 9, 1–11 (2015) 8. Qi, W., Fu, F.: Variational iteration method for solving differential equations with piecewise constant arguments. Int. J. Eng. Manuf. 2(2), 36–43 (2012). https://doi.org/10.5815/ijem. 2012.02.06 9. Acebrón, J.A., et al.: The Kuramoto model: a simple paradigm for synchronization phenomena. Rev. Mod. Phys. 77(1), 137–185 (2005) 10. Bhansali, P., Roychowdhury, J.: Injection Locking Analysis and Simulation of Weakly Coupled Oscillator Networks. Simulation and Verification of Electronic and Biological Systems. Springer Science+Business Media B.V., Dordrecht pp. 71–93 (2011) 11. Ashwin, P., Coombes, S., Nicks, R.J.: Mathematical frameworks for oscillatory network dynamics in neuroscience. J. Math. Neurosci. 6(2), 1–92 (2016) 12. Masoud Taleb Ziabari, Ali Reza Sahab, Seyedeh Negin Seyed Fakhari: Synchronization New 3D Chaotic System Using Brain Emotional Learning Based Intelligent Controller[J]. International Journal of Information Technology and Computer Science (IJITCS), 2015, 7(2): 80–87 13. Gourary, M.M., Rusakov, S.G.: Analysis of oscillator ensemble with dynamic couplings. In: AIMEE 2018. The Second International Conference of Artificial Intelligence, Medical Engineering, Education, pp. 150–160 (2018) 14. Alifov, A.A.: Methods of Direct Linearization for Calculation of Nonlinear Systems. M.Izhevsk: Regular and Chaotic Dynamics (2015) (in Russian) 15. Alifov, A.A.: Method of the direct linearization of mixed nonlinearities. J. Mach. Manuf. Reliab. 46(2), 128–131 (2017) 16. Alifov, A.A.: About some methods of calculation nonlinear oscillations in machines. In: Proc. Inter. Symp. of Mechanism and Machine Science, Izmir, pp. 378–381 (2010)

Oscillations at Nonlinear Parametric Excitation

689

17. Alifov, A.A.: About Direct Linearization Methods for Nonlinearity. In: Advances in Artificial Systems for Medicine and Education III. Advances in Intelligent Systems and Computing, vol. 1126, pp.105–114. Springer, Cham (2020) 18. Zolotukhin, Y., Kotov, K., Maltsev, A.S., Nesterov, A.A., Filippov, M.N., Yan, A.P.: Correction of transport lag in the control system of a mobile robot. Autometry 2(47), 46–57 (2011) 19. Than, V.Z., Dementiev, Y., Goncharov, V.I.: Improving the accuracy of calculating automatic control systems with a delay. Softw. Products Syst. 31(3), 521–526 (2018) 20. Garkina, I.A., Danilov, A.M., Nashivochnikov, V.V.: Simulation modeling of dynamical systems with delay. Mod. Prob. Sci. Educ. 1–1, 285 (2015). (in Russian) 21. Kashchenko, S.A.: Dynamics of a logistic equation with delay and diffusion and with coefficients rapidly oscillating over a spatial variable. Rep. Acad. Sci. 482(5), 508–512 (2018) 22. Chun Hua Feng: Oscillation for a mechanical controlled system with time delay. Adv. Mater. Res. 875–877, 2214–2218 (2014) 23. Kononenko, V.O.: Vibrating Systems with Limited Power-Supply. Iliffe, London (1969)

Fairness Audit and Compositional Analysis in Trusted AI Program Sergiy Gnatyuk, Pylyp Prystavka, and Serge Dolgikh(B) National Aviation University, 1 Lubomyra Huzara Ave, Kyiv, Ukraine [email protected]

Abstract. In this work, definitions and approaches in the analysis of fairness and bias in “black box” learning systems were examined from the perspective of unsupervised generative learning. Challenges in direct analysis of fairness and bias in “black box” systems that cannot provide suitable explanations for their decisions have several facets, including: sufficiency and representativity of marked trusted data sets for confident correlation analysis; dimensionality and processing challenges with complex real-world data; and trustworthiness of the trusted data itself. We investigated approaches to fairness audit from the perspective of generative compositional analysis that does not depend on significant amounts of prior data and for that reason can provide additional perspective to the analysis. With generative structure of informative low-dimensional representations that can be obtained with different methods of unsupervised learning, certain forms of bias analysis can be performed without prior data. As demonstrated on a practical example, the significance of the proposed method is that it can provide additional directions and valuable insights in the analysis of bias and fairness of learning systems. Keywords: Fairness and bias · Black box learning systems · Explainable AI · Trusted AI

1 Introduction While AI technology has been developing at an outstanding pace in the past decades, finding applications in many areas and domains of research, industry and societal functions, the progress has not been entirely and unconditionally positive. One area of concern and persistent research and discussions is understanding the reasoning of AI systems and ability to provide explanations or audit of their decisions (“black box” learning systems, explainable AI [1, 2]). Another, though closely related area of research is developing conceptual and ontological frameworks describing the fairness, trustworthiness and bias of AI systems [3, 4]. Many studies examined and described examples of bias in different functional applications of AI, including criminal justice, health care, human resources, social networks and others [5–7]. It was noted that the problems of explainable and trusted AI can be closely interrelated [8]: understanding the reasons why learned systems make certain decisions can be a key factor in determining whether it can be trusted; on the other hand, © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 690–699, 2023. https://doi.org/10.1007/978-3-031-36118-0_62

Fairness Audit and Compositional Analysis in Trusted AI Program

691

it is not obvious to imagine a mechanism or a process of confident and reliable determination of a trusted AI without some insight into the reasons of its decisions, though directions for exploration of this question exist and some of them are discussed in this paper. A number of different approaches in verification of learned intelligent systems and extraction of knowledge were developed in the recent years [9, 10]. To address the limitation of unknown, a priori, reasoning in decisions made by black box systems and limitations on available trusted prior data is some tasks and domains, in this work we attempted to outline a possible approach to these essential and actual questions that does not involve massive prior trusted data for examination, and in some cases, confident determination of trustworthiness and bias. This approach allows to untie the problem of “chicken and egg” in determination of trustworthiness, for example: could the system that produced prior decisions be itself trusted? while suggesting sound and practical methods of analysis of fairness and bias based on the natural conceptual structure of the data in the problem. In the authors’ view, methods of unsupervised generative concept learning that have been actively developed [11–13] can provide a both insightful and useful perspective for such analysis. Assuming that generative structure of the data can be identified by one or combination of known methods of dimensionality reduction preserving essential informative characteristics of the input distribution, an analysis of distributions of decisions produced by the audited learning systems across general types of input data can provide valuable insights about the system, including whether it can be trusted or produces biased decisions. Such analysis can directly address the objective formulated for this work, that is, examine and introduce methods of analysis of decisions of black-box learning systems that do not depend significantly and essentially, on large prior trusted knowledge about the system.

2 Methodology 2.1 Black-Box Learning Systems A black-box learning system L v , v = {v1 ,.. vk }, k: 1.. L: internal parameters of the system can be characterized by the condition that for a given set of inputs, I x = {i1 ,.. ik }, k: 1.. N described by parameters x in the input parameter space X, decisions Dy (I x , v) = {i1 ,.. ik }, k: 1.. N described by parameters y in the output (decision) space are produced based entirely from the parameters of the inputs and internal parameters of the system, without any additional context (“explanations”). In this definition, inputs and decisions are the only variables that can always be observed externally, while the internal parameters, nor the decision logic of the system that were essential for producing certain decision may not be known to the external observer. 2.1.1 Decisions and Correctness Following a process of learning, the parameters of a black-box system are “trained” to produce decisions that are considered to be correct or optimal. In other words, a system is trained to maximize the correctness, and correspondingly, minimize the error of the

692

S. Gnatyuk et al.

produced decisions, that can be measured by some metric (“cost” or error functional) in the output space, Y. The accuracy, and error are numerical measures of correctness of the outputs or decisions produced by a black-box system L on a certain set of inputs: A, E(I , D, w) : D → M

(1)

where M: a numerical space in which the accuracy, error take values; w: additional parameters of the accuracy functional. For example, in a common practice, the set of decisions produced by the learning system on the known, “training” inputs can be compared to prior known correct decisions (standard decisions, w = { DS }) with the resulting deviation of the produced decisions from the standard ones related to the error, and accuracy of the system. E(Is , D, Ds ) ∼ = dist(D, Ds )

(2)

where dist: a distance metrics in the output (decision) space Y. The following characteristics of the accuracy and error functions are essential for the subsequence analysis: • The measure of error (and accuracy) can be calculated for a single input ix , as long as the correct decision d c (ix ) is available or can be produced (for example, in the process of audit): E(d(ix ), d c (ix )), A(d(ix ), d c (ix )). • The measure of error (accuracy) is valid within the statistical margins of the standard set of known (input, decision) pairs. • While the measure of accuracy based on standard decisions (2) can be relatively easy to calculate, and is a standard practice in machine learning, its relevance for arbitrary inputs depends on a number of assumptions. Primarily, the standard (training) set has to be sufficiently representative to describe a wide range of realistic inputs, and the input on which the error is being evaluated, sufficiently similar to some of the inputs in the standard set. If the distance between certain real input ix and the nearest standard inputs iS is sufficiently large, estimation of the accuracy of the decision d(ix ) produced by the system can be more challenging. In conclusion of this section, it needs to be noted that while the error and accuracy represent a measure of correctness of the decisions produced by a black-box system that can be estimated by standard statistical methods [14], it cannot always produce essential conclusions about a) internal context or “motivation” of the system and b) ethical factors or fairness of the produced decisions. This essential aspect will be addressed in the next section. 2.1.2 Accuracy, Fairness and Bias An ideal system that after being trained produces a correct decision on any input by definition, has perfect accuracy and is perfectly fair, in the sense that no input is more or less preferred by the system than another. Such a system though, even if it existed, could not be recognized by any practical means because it would require testing candidate systems with every possible input in the

Fairness Audit and Compositional Analysis in Trusted AI Program

693

input space. Therefore, one has to consider more practical black-box learning systems that produce the output on a certain standard set of decisions DS with the accuracy a = (A, E). A question can be examined then, is or could such a system be fair or biased, and would the accuracy measures be sufficient for such determination? To advance in addressing it, one needs to define fairness formally. A learning system L that is a black-box system can be said to be “fair”, if its decisions satisfy, or are compatible with, certain fairness objectives (fairness goal, criteria, etc.), Of :   f L, I , Of = F(I , D, Of ) (3) It is presumed that a fair system has to satisfy the objective with any (realistic) set of its decisions. So, if a system fails the objective on a certain set of inputs I f , a determination of a “bias” is made:   ¬f L, If , Of : B(L) = True (4) where F(I, D, Of ), B(I, D, Of ): fairness, bias functionals. Some natural or basic fairness requirements can be defined: • Consistency (no arbitrary decisions): the system has to produce similar decisions on similar inputs: ix1 ∼ = ix2 → d (ix1 ) = d (ix2 ); • An extension of this rule is “group fairness”: if a subset of inputs belongs to the same general type G, characterized by essential similarity of inputs within the group, then the decisions produced on the inputs in the group have to be similar as well: d (ix ∈ G) ∼ const • Implications for the correctness: if a decision on an input is correct / wrong, the decision on a similar input has to be the same (correct or wrong, respectively): Now one can note that fairness objective is not immediately equivalent to the accuracy characteristics and has to constitute an additional input for the learning system. For clarification of this point, let us consider an example. 2.1.3 Example: Correctness vs. Fairness Consider a trained black-box learning system with the accuracy characteristics a = (A, E) measured on a set of inputs DS . Would such a system be fair or biased, relative to additionally defined fairness objective, Of ? Can the answer to this question be determined based only on characteristics of accuracy? Suppose the objective is defined as “egalitarian”: the decisions of the system should be distributed similarly, within statistical margins, for all inputs based on clearly defined criteria. Further, we consider a simpler case where the space of possible inputs is comprised of similarity domains containing inputs of the same general type, with significantly

694

S. Gnatyuk et al.

higher similarity within the type than across the entire space of possible inputs. Accuracy characteristics above can then be compiled by the general types of inputs, GS : a = {as } = {(As , Es )}, Gs ∈ Ds

(5)

Then, based on the general fairness principle and the fairness objective one has to conclude that the distributions of decisions based on the set criteria have to be the same (or similar, within statistical margins) for all known groups of inputs. That is, if the criteria for a certain decision are met, the resulting decision has to be the same, regardless of the general type of the input. Not in the least, verification of this requirement implies that all characteristic types of inputs have to be represented in the verification set DS , importantly, all inputs that can be encountered by the system in its operation, not only those seen by the system in training. Clearly, the conclusion that can be derived from this example imposes serious constraints on the accuracy results for determination of fairness that may not be met for all previously trained learning systems. Based on these considerations one has to conclude that correctness measurements alone, without additional information on composition of the observable input space, may not be sufficient for determination of fairness of a black-box learning system. 2.2 Unsupervised Concept Learning Methods of unsupervised concept learning, where successful, can identify a structure of characteristic patterns, general types or natural concepts in the input distributions by stimulating learning models to improve quality of generation from informative latent distributions, often of significantly reduced complexity. Unlike methods of conventional supervised learning, these approaches do not depend on specific assumptions about the distribution of the data, large sets of prior data and can be used in a general process with data of different types, origin and domains of application and for that reason, can be applied in the analysis of different types of data and learning models, tasks, domains and applications. Obtaining such structures of natural groups or types, with entirely unsupervised methods that do not depend on prior trusted data, can be instrumental, as demonstrated in the sections that follow, in the analysis of trust and fairness produced by a learning system. 2.3 Fairness Objective and Criteria Following the arguments presented earlier, designing learning systems, in particular those of machine intelligence for fairness and trust requires explicit and upfront definition of the fairness objective (FO). As the range of possible tasks and applications of learning systems can be very broad, formalization of all possible objectives can be a challenging, though a worthy in the authors view, task. Here, only some straightforward examples will be provided. – “Egalitarian” objective: any input satisfying clearly set decision criteria has to produce the same decision.

Fairness Audit and Compositional Analysis in Trusted AI Program

695

– “Protective” or non-discrimination: distributions of decisions for specific, identified groups of inputs (focus groups) have to be similar, withing statistical margins, to that of the general group, and certainly, many others. Thus, it can be essential for clear definition and interpretation of fairness audit to agree upon and formulate: • Clear, sufficient and robust definition of the Fairness Objective(s) explicitly and upfront, for each area or case of application of learning system • Definition of the criteria of passing the objective test and determination of fairness. Whereas defining design practices and guidelines can improve chances of the systems being developed being fair (in some implicitly understood sense), only explicit definition of the objective and criteria (of passing the objective test) will allow to perform fairness audit and make a formal determination of the fairness and trustworthiness of the resulting system. 2.4 Unsupervised Compositional Analysis in Fairness Audit Combining the definitions and analysis of the preceding sections one can propose an approach in the analysis of fairness, trust and bias based on the structure of general types (concepts) that in some / many cases can be resolved with methods of unsupervised learning, dimensionality reduction and similar. One can argue that many, if not most of practical cases and applications of learning systems satisfy the assumption of generally typed data, (5): The space of observable inputs is composed of a finite number of general types, with significantly higher similarity of inputs within the type subgroups, plus possible non-typed inputs and / or random noise. The challenge in the case of a general application of a learning system is that all characteristic types of inputs may not be known prior to, or in the process of development and training of the learning system. Yet, based on the defined fairness objective and criteria, a determination of fairness and trust would have to be made for such systems, within the program of Trusted AI. A possible direction to address this challenge was outlined in [15], based on the methods of unsupervised composition analysis of general data developed in the recent years. The aim of the analysis is to identify similarity groups or “general types” in the input data with entirely unsupervised means that do not require additional prior knowledge. The approach includes the following general stages of processing the data that can be used with different types of data in a range of tasks and applications: • Preprocessing observable data and producing a representative set of inputs described by observable or input parameters, that have to be sufficiently sensitive and informative to describe essential variation of observable inputs. • Application of methods of unsupervised dimensionality reduction to produce informative representations of the representative set of inputs. The resulting, commonly low-dimensional latent space describing the distribution of inputs is defined by the coordinates of informative features found by the methods. A sufficiently large array of methods that can be used to this end, including: linear dimensionality reduction

696

S. Gnatyuk et al.

(PCA); non-linear generative learning, including autoencoder-type neural networks; non-linear dimensionality reduction, manifold learning and others. • Application of clustering methods, specifically, non-parametric density clustering, in the resulting informative representation space to identify characteristic clusters, groups, or general types of inputs. An essential distinct feature of the described process and is that it is entirely unsupervised, and at no point requires additional, prior knowledge about the content, composition, assumptions about distributions or other. Consequently, the approach of Unsupervised Composition Analysis (UCA) can be applied to general types of data in a wide range of tasks and applications (Fig. 1), additionally and in parallel with conventional methods of development of learning system, with no dependence on trusted prior training data. For these reasons, there is a natural case for application of the methods of Unsupervised Composition Analysis in the area of fairness and trust audit of learning systems as they can provide critical information and insight into the composition of inputs, including description of general types of inputs.

Fig. 1. Unsupervised Composition Analysis, image data (left: geometrical shapes, generative neural model, [16]; right: printed digits, manifold learning, [17])

Combining the case for clearly defined fairness objective, FO and unsupervised composition analysis, UCA, one can propose an approach for auditing operational learning systems with respect to defined fairness and trustworthiness criteria. Importantly, the proposed approach does not depend on, nor require additional inputs or prior trusted data. As the input to the analysis, a matrix of distributions of decisions by general types of inputs in the axes of: decision criteria and general types of inputs in the data is compiled based on: • The recorded decisions of the system on provided inputs.

Fairness Audit and Compositional Analysis in Trusted AI Program

697

• The distribution of inputs across the general types, or clusters of the inputs identified with UCA methods. Based on compiled distributions of decisions across characteristic types of inputs, and the defined fairness objective a determination of compliance (i.e., a fair trusted system) or bias can be made.

3 Results 3.1 Unsupervised Compositional Fairness Audit: a Practical Case For an illustration of the proposed approach to fairness audit of pre-trained and operational black-box learning system consider an illustrational example. Assume that the system, L P has to produce distinct decisions in three different ranges of some observable parameter of significance, PS : R1 , R2 , R3 . According to the process of unsupervised composition analysis, suppose application of UCA produced three distinct groups or general types of inputs: A, B, C. As described in the preceding sections, distributions of decisions of the system across the subsets of input instances grouped by identified general types were compiled: DA , DB , DC as inputs to the unsupervised compositional fairness audit of the decisions produced by the learning system. 3.2 Application of Unsupervised Compositional Fairness Analysis Table 1 shows the Type / Decision matrix compiled for the illustrational learning system L P according to described assumptions. The fairness objective, FO has not yet been considered at this stage of the analysis. Table 1. Type / Decision matrix, practical learning system General type

R1 (PS )

R2

R3

A

Decision criteria 

DA (1)

DA (2)

DA (3)

B

DB (2) DC (2)

DB (3)

C

DB (1) DC (1)

Other

-

-

-

DC (3)

As discussed earlier, the next essential input in the fairness audit is the Fairness Objective (FO). We will assume that in the case under consideration, the FO for the system L P has been defined as “Egalitarian”, as described in Sect. 3.1, requiring that outside of defined decision criteria, other factors including general type, must have no influence on the decision of the system. Based on the clearly stated Fairness Objective and distributions of group decisions obtained with UCA methods, one can compare the distributions of decisions in the same

698

S. Gnatyuk et al.

criteria range, for example, R2 across different general types (A – C, Table 1) to make a substantiated, including statistically significant, determination of the fairness or bias in the decisions produced by the audited system, by calculating statistical variances between decision sets of different groups X, Y in the same criteria range, VR (X,Y ). The logical sequence of the audit is shown in Table 2. Table 2. Fairness audit: determination of fairness / bias, practical learning system Learning system

FO

Observed variance, general types (*)

Determination

L1

Egalitarian

No ((VR (X,Y ) ~ 0)  0)

Fair

L2

--

Yes

Biased

(*) Statistically significant variance of distributions of decisions in general type groups

In the first case (learning system L1 ), a determination of fairness was made because distribution of decisions across characteristic types of inputs satisfied the stated fairness objective within the margin of statistical analysis. In the second case (L2 ), the criterion of the FO was failed, resulting in the determination of bias. As can be concluded from this practical example, application of methods of Unsupervised Compositional Analysis in combination with explicitly defined fairness objective allowed to obtain a substantiated determination of fairness / bias in a practical learning system.

4 Summary and Conclusion Due to variety of methods, architectures, domains, tasks and applications, confident determination of trustworthiness or bias in artificial learning systems can be a challenging and time-consuming task. Principal insights proposed in this paper are first, explicit formulation of the Fairness Objective (FO) that is relevant to the task or application domain and definition of clear and functional criteria of compliance of the decisions produced by the system with the defined fairness objective. The second essential contribution proposed here is application of methods of Unsupervised Compositional Analysis in the audit of decisions produced by a learning system within the framework of the defined Fairness Objective. Methods of unsupervised composition analysis of general data developed in the recent years have proven both effective and efficient in identifying characteristic informative structure in the data such as characteristic clusters or general types, natural concepts, etc. These methods do not depend on nor require additional prior knowledge about the data and can be used additionally and in parallel to conventional methods of development of artificial learning systems. It was shown that a combination of methods of conventional statistical analysis with the methods of unsupervised composition analysis proposed in this work can be effective in producing substantiated and reasoned determination of fairness or bias in black-box learning systems, based on the formulated Fairness Objective and the structure of general types characteristic for the data and can significantly improve the outcomes of detection

Fairness Audit and Compositional Analysis in Trusted AI Program

699

of bias and confidence in the trustworthiness of intelligent systems compared to the status quo. Due to demonstrated wide and successful application of unsupervised composition analysis, it is expected that the approach proposed here can have general applicability in a wide range of tasks, domains and applications of artificial learning systems.

References 1. Longo, L., Goebel, R., Lecue, F., Kieseberg, P., Holzinger, A.: Explainable Artificial Intelligence: concepts, applications, research, challenges and visions. Proceedings of CD-MAKE, LNCS 12279, pp. 1–16 (2020) 2. Rai, A.: Explainable AI: from black box to glass box. J. Acad. Mark. Sci. 48(1), 137–141 (2019). https://doi.org/10.1007/s11747-019-00710-5 3. Schwartz, R., Vassilev, G.K., Perine, L., Bart, A.: Towards a standard for identifying and managing bias in Artificial Intelligence. National Institute of Standards and Technology, USA, Special Publication 1270 (2022) 4. Bartneck, C., Lütge, C., Wagner, A., Welsh, S.: Trust and fairness in AI systems. In: An Introduction to Ethics in Robotics and AI. Springer Briefs in Ethics. Springer, Cham (2020) 5. Bogen, M.: All the ways hiring algorithms can introduce bias, Harvard Business Review, 06.05 (2019) 6. Gianfrancesco, M.A., Tamang, S., Yazdany, J., Schmajuk, G.: Potential biases in Machine Learning algorithms using electronic health record data. JAMA Intern. Med. 178(11), 1544– 1547 (2018) 7. European Union Agency for Fundamental Rights (FRA). Bias in algorithms - Artificial intelligence and discrimination. European Union (2022) 8. Zhou, J., Verma, S., Mittal, M., Chen, F.: Understanding relations between perception of Fairness and Trust in algorithmic decision making. arXiv (2021). 2109.14345 [cs.CY] 9. Derakhshan, M.A., Maihami, V.B.: A review of methods of instance-based automatic image annotaton. Int. J. Intell. Sys. Appli. 8(12), 26–36 (2016) 10. Fotsoh, A., Sallaberry, C., Lacayrelle, A.LP.: Retrieval of complex named entities on the Web: proposals for similarity computation. Int. J. Info. Technol. Comp. Sci. 11(11), 1–14 (2019) 11. Benjio, Y., Courville, A., Vincent, P.: Representation Learning: a review and new perspectives. arXiv (2014). 1206.5538 [cs.LG] 12. Jing, L., Tian, Y.: Self-supervised visual feature learning with deep neural networks: a survey. IEEE Trans. Pattern Anal. Mach. Intell. 43(11), 4037–4058 (2021) 13. Liu, W., Wang, Z., Liu, X., et al.: A survey of deep neural network architectures and their applications. Neurocomputing 234, 11–26 (2017) 14. Li, J., Gao, M., D’Agostino, R.: Evaluating classification accuracy for modern learning approaches. Stat. Med. 38(13), 2477–2503 (2019) 15. Dolgikh, S.: Fairness and bias in Learning Systems: a generative perspective. In: BEWARE2022 Workshop, 21st International Conference of the Italian Association for Artificial Intelligence, Udine, Italy, CEUR Workshop Proceedings 3319, pp. 61–66 (2022) 16. Dolgikh, S.: Topology of conceptual representations in unsupervised generative models. In: 26th International Conference Information Society and University Studies, Kaunas, Lithuania, 2021, CEUR Workshop Proceedings, pp. 150–157 (2915) 17. Uniform Manifold Approximation and Projection (UMAP) (2018)

Determination of Brachistochronous Trajectories of Movement of a Material Point in a One-Dimensional Vector Field Viktor Legeza(B) , Oleksandr Neshchadym, and Lyubov Drozdenko National Technical University of Ukraine “KPI Named After Igor Sikorsky”, Kyiv 03056, Ukraine [email protected]

Abstract. A new approach to solving the Zermelo navigation problem using variational methods is proposed, which differs from methods based on the Pontryagin maximum principle. In this work, the motion of the ship is presented as the motion of a material point in a flat horizontal vector field of a moving fluid, for which the classic variational problem of finding brachistochronous trajectories is formulated. The purpose of the work is to establish the equations of brachistochronous trajectories of the movement of a material point, which ensure the minimum time of its movement between two given points The objective function of the time of movement of a material point between given points is constructed, which is then minimized. With its use, Euler’s equation was obtained, which allows us to directly write the differential equations of extreme trajectories. Algebraic equations of the brachistochronous motion of a material point have been established for a given version of the boundary conditions using numerical experiments. A comparative analysis of the time spent both on extreme trajectories and on an alternative straight path was carried out. It is established that the given variational problem generates two solutions of opposite sign, and only one of them delivers the minimum of the time functional. A comparative analysis of the calculation results showed that the brachistochronous trajectory of the movement of a material point is not a straight line connecting two given points, but is an oscillating curve. Keywords: Zermelo problem · variational problem · brachistochronous motion · objective time functional · euler equations · extremal trajectory

1 Instruction Let us briefly outline the review of scientific research carried out within the framework of the problem under consideration, related to the search for equations of extremal trajectories, in various problem formulations. I.Bernoulli was the first to set and solve the problem of finding an equation that describes the shape of the brachistochron curve. This is a curve along which a material point moves from a position of rest and without initial speed and resistance in a vertical gravitational field from one known point to another in the minimum time. As a result of the study of this problem in such a statement, it turned out © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 700–708, 2023. https://doi.org/10.1007/978-3-031-36118-0_63

Determination of Brachistochronous Trajectories of Movement of a Material Point

701

that its solution is a cycloid [1]. In the classic textbook [2], I. Bernoulli’s problem about the brachystochrone was solved by the methods of calculus of variations. The formulation of the problem of finding a brachistochrone within the framework of geometric optics (the problem of light beam propagation) was implemented in work [3]. In works [4, 5], the “brachystochronous” movement of a material point in a vertical gravitational field was considered by the methods of variational calculus, taking into account the forces of dry Coulomb friction. In [6], the same problem was solved using the methods of optimal control theory. The author chose the time derivative of the angle, which determines the direction of the brachistochronous movement of the point along the curve, as the controlling parameter in this formulation of the problem. The paper [7] provides a solution to the problem in which a geometric constraint in the form of a cylindrical surface is imposed on the brachistochronous motion of a material point, and dry friction acts between the point and this surface. The problems of finding brachistochronous curves in the process of movement of a material point in non-conservative force fields were considered in articles [8, 9]. In works [10, 11], the search for brachistochrone curves was considered in uniform force fields, and in [10] the movement of a material point was limited by a bond in the form of a cylindrical surface, and in [11] in the form of a spherical surface. The search for a brachistochronous curve during the movement of a material point in non-homogeneous force fields was implemented in works [12–14], and in [14] this problem was solved for the case when a material point moved in a linear radial force field. In works [13, 15, 16], solutions to the brachistochrone problem were found for the case when a point moves in a radial field with a force dependence that is inversely proportional to the distance between the interacting points. In works [17–19], the classical problem about the brachistochrone was generalized to the case of relativistic motion of a material point. In addition, work [19] also solved problems in which the brachistochronous motion of a point was considered on algebraic surfaces of rotation without energy loss. For this, Euler’s equations and a special approach developed for solving a number of geometrical optics problems were applied. The article [20] presents the following generalization of the problem of the brachistochronous movement of a material point without friction along a curve that connects a given point and a given curve or two given curves. This formulation of problems leads to a class of variational problems with free boundary conditions. In work [20], such problems were solved using variations at different end points. The next step in the generalization of brachistochrone problems was made in work [21] and related to the search for a brachistochrone curve for a body of finite dimensions that rolls on a cylindrical surface without friction. However, the author of the article [21] did not provide a mathematically correct solution to the brachistochrone problem in such a formulation. In the article [22], the brachistochronous motion of a thin homogeneous disk rolling on a horizontal plane without slipping under the action of three moments was considered. This task was set and solved within the framework of the theory of optimal control. The possibility of realizing the brachistochronous movement of the disc due to the forces of dry friction has been established. The influence of the value of the dry friction coefficient on the structure of extreme trajectories was also analyzed. The article [23] gives a generalization of the brachistochrone problem to the case of a cylindrical container filled with a viscous liquid, which plays the role of a material point in the classical Bernoulli problem. A brachistochronous curve is found that connects

702

V. Legeza et al.

two given points between which a container with liquid moves in the minimum time. It is proved that the brachistochron obtained for such a case differs from the cycloid. This effect is explained by the fact that the increase in the rate of change of the kinetic energy of the liquid container entails an increase in energy dissipation due to the movement of the viscous liquid. In articles [24, 25], the brachistochrone equation was found for a heavy cylinder that rolls along a concave cylindrical notch without slipping under the influence of the forces of a vertical gravitational field. The classical calculus of variations was used to find the brachistochrone equation. It is shown that the center of mass of the cylinder traces out a cycloid in the vertical plane, and the oscillations of the cylinder itself in the recess are isochronous. In a number of scientific studies, the tasks of finding extreme trajectories, which are related to various formulations of the Zermelo navigation problem, were considered. In works [26–29], the specified problem was solved within the framework of the theory of optimal control. In these studies, the task was set as follows: it is necessary to find such an optimal control that would ensure the ship’s arrival at a given point in the minimum time, i.e., optimal speed control. The time of the ship’s movement when closing these problems was determined from transcendental equations, which were solved by numerical methods. As examples, the results of specific solutions of problems for different forms of the profile of the velocity field are given. There are cases when the maximum principle does not give an optimal solution. Since in a number of the above-mentioned works, the problem of finding brachistochronous trajectories leads to certain boundary value problems, we present related works in which new approaches to their numerical solution are considered [30, 31]. The review of literature sources in the field of finding extreme curves showed that previously the Zermelo navigation problem was not solved by variational methods. In addition, other authors [26–29] noted that in some cases this problem cannot be solved using the Pontryagin maximum principle, since it does not provide an optimal solution. Another advantage of the direct variational approach is that it provides the possibility to directly obtain the differential equation of the brachistochrone curve from the target functional. Therefore, in this report, it is proposed for consideration the solution of the problem of finding brachistochronous trajectories of the movement of a material point in a horizontal vector field using the methods of variational calculus. First, we construct the target functional, then, based on the Euler equations, we obtain the brachistochrone differential equation, which is solved by numerical methods. 1.1 The Purpose of the Research The purpose of the research is 1). obtaining the equations of extreme (brachystochronous) trajectories of motion by variational calculus methods, along which a material point moves from a given starting point to a given final point in a flat vector field of a moving fluid in the shortest time; 2). numerical analysis of the features of brachistochronous trajectories. 1.2 Formulation of the Problem Let a material point move with variable velocity V = (u, v) in a flat one-dimensional vector field along a certain curve y = y(x) connecting two given points O(0, 0) and

Determination of Brachistochronous Trajectories of Movement of a Material Point

703

M (L, y(L)) in the minimum   time. At the same time, the modulus of the point velocity   vector is constant, i.e. V  = C :  u2 + v2 = C,

(1)

where u– is the projection of the speed V on the horizontal axis OX ; v− projection of speed V on the vertical axis OY u is the horizontal projection of the velocity; v vertical projection of speed. The vector V is always directed along the tangent drawn at the current point of the sought trajectory y = y(x) of the moving material point. The vector field considered in this problem is formed by a flat one-dimensional river flow. The speed of the river is variable depending on the horizontal coordinate and is determined by some known function f (x). The direction of the river velocity vector is opposite to the positive direction of the vertical axis OY .

2 The Method of Constructing the Time Functional and Determining the Equation of the Desired Trajectory Let us write expressions for the projections of the speed of a point on the axes OX and OY , taking into account the speed of the river flow: ⎧ dx ⎪ = u; ⎨ dt ⎪ ⎩ dy = v − f (x). dt

(2)

Let us define the objective time functional to be minimized: 

L

T= 0

dx → min. y(x) u

(3)

We use the system of Eqs. (2) to construct functional (3). After some transformations, we get:  L −f (x)y ± (Cy )2 − f 2 (x) − C 2  dx. (4) T= f 2 (x) − C 2 0 Let us denote by F(x, y ) the integrable function in expression (4). We use the Euler equation [2] to find the differential equation for the desired point trajectory: Fy −

d  F  = 0. dx y

(5)

The integrand in (4) does not depend on the variable y, so Fy = 0 and Eq. (5) are reduced. As a result, we obtain a first-order equation: Fy  = C1 ,

(6)

704

V. Legeza et al.

where C1 is the first arbitrary constant. After identical transformations from expression (6) we find the differential equation of the trajectory of a material point: y = ±h(x, C, C1 ), where

 1  h(x, C, C1 ) =  C 1−

(7)

f 2 (x) − C 2 [C1

C2 2 ·(f 2 (x)−C 2 )+f (x)

.

]

After integrating (7), we obtain the final equation of the extremal curve:  y(x) = ± h(x, C, C1 )dx + C2 ,

(8)

where C2 is the second arbitrary constant. Formula (8) establishes the final form of the extremal equation y(x), to which two boundary conditions should be added: y(0) = 0; y(L) = yL .

(9)

As we can see, the differential Eq. (7) and its solution (8) for the brachistochronous trajectory of a material point can be immediately obtained from the constructed time functional (4) without constructing other intermediate mathematical constructions, as it happens in works [27–29].

3 The Method of Solving the Boundary Value Problem Therefore, the differential Eq. (7) obtained above and the boundary conditions (9) form a boundary value problem for finding the brachistochron curve. Let the speed of the river flow be represented by a trigonometric function: πx . (10) f (x) = sin L In formula (8), we replace the function f (x) with the trigonometric function (10) and perform all the necessary identical calculations. After transformations similar to those carried out above, we obtain the function h(x, C, C1 ) corresponding to expression (10). Note that from the general form of the integral function in the integral (8), it can be stated that it does not have a primitive in the elementary functions. Hence, we conclude about the need to integrate this integral by numerical methods. To this end, let’s represent the function h(x, C, C1 ) using Taylor’s formula, taking into account the first six terms in order to achieve sufficient accuracy. The development of the function h(x, C, C1 ) was carried out by the variable x around the point x = 0. Based on the results of the

Determination of Brachistochronous Trajectories of Movement of a Material Point

705

calculations, we will first present all six derivative functions h(x, C, C1 ) to construct this development: π

C·C1 , h(1) (0, C, C1 ) = h(0) (0, C, C1 ) = − K(C,C 1)

(0, C, C ) =

h

,

h(3) (0, C, C ) = − h(4) (0, C, C1 )

=−

,

L·C·K 3 (C,C1 ) π 2 ·C·C13 [(C·C1 )2 −4] (2) 1 L2 ·K 5 (C,C1 ) π 3 ·[(C·C1 )4 −15·C 2 ·C14 −2·(C·C1 )2 +1] 1 L3 ·C·K 7 (C,C1 ) π 4 ·C·C13 [4(C·C1 )6 −3C 4 ·C16 −24(C·C1 )4 +36C 2 C14 +36C 2 ·C12 +72C12 −16] L4 ·K 9 (C,C1 ) π 5 [(C·C1 )8 −150C 6 ·C18 −4(C·C1 )6 +525C 4 C18 +300C 4 ·C16 +6(C·C1 )4 C·L5 ·K 11 (C,C1 ) 420C 2 ·C14 −150C 2 C14 −4(C·C1 )2 + 1] C·L5 ·K 11 (C,C1 )

,

,

h(5) (0, C, C1 ) =

+

,

where K(C, C1 ) =



[1 − (C · C1 )2 ].

Let’s represent the function h(x, C, C1 ) using Taylor’s formula in compressed form: h(x, C, C1 ) =

5 

h(n) (0, C, C1 )

n=0

xn . n!

(11)

Let’s substitute the expression (11) in the integral (8) and integrate it. As a result, we get the final brachistochrone equation: y(x, C, C1 , C2 ) = ±

5  n=0

h(n) (0, C, C1 )

xn+1 + C2 , (n + 1)!

(12)

where C1 , C2 are unknown constants, which should be determined from the boundary conditions (9).

4 Numerical Implementation and Analysis of Results Numerical analysis was performed for three cases. The first two cases concerned two extreme solutions of the variational problem, and the third case was chosen as an alternative one, which represents the motion of a material point along a straight line connecting two given limit points. The measure of the optimality of the chosen path will be the time spent on moving a material point between two given points on the plane. The following boundary conditions were chosen for all three cases: L = 1, C = 1, 1, y(0) = 0, y(L) = 0. 1. The first case. In expression (12), we choose a positive sign and find the unknown constants C1 and C2 using numerical methods and using boundary conditions. The result of calculating the unknown constants is as follows: C1 = 0, 385 and C2 = 0. After that, the final expression for the desired function y(x) for the first case can be obtained. To determine the time of movement of a material point along the brachistochrone, we use the functional (4). After calculating the functional, we get: T = 2, 46 (time units). Figure 1 shows the graph of the brachistochronic trajectory y(x) for the first case.

706

V. Legeza et al.

Fig. 1. Brachistochronic trajectory for the first case

2. The second case. In expression (12), we choose a negative sign and again find the unknown constants C1 and C2 by numerical methods and using boundary conditions. The results of the calculation of unknown constants coincide with the previous case: C1 = 0, 385 and C2 = 0. Next, we obtain the final expression for the desired function y(x) for the second case. To determine the time of movement of a material point along the brachistochrone, we will again use the functional (4). The calculation of the functional gives less value than in the first case: T = 1, 140 (time units). Figure 2 shows the graph of the brachistochronic trajectory y(x) for the second case. Note that the ordinates of the points of this graph on the same abscissa have opposite signs in relation to the graph in Fig. 1.

Fig. 2. Brachistochronic trajectory for the second case

3. The third case. Here we consider the movement of a point along a straight line connecting two given points: O(0,0) and M(L,0). So, the trajectory of the point is described by the function y(x) = 0. It would seem that this trajectory should be optimal in terms of time, but the results of the numerical experiment refute this assumption. The time functional (4) for the third case gives a greater result than for the second case: when substituting the function y(x) = 0 in (4), we have the following result: T ≈ 1, 344 (units of time).

Determination of Brachistochronous Trajectories of Movement of a Material Point

707

5 Summary and Conclusion In this report, the classical Zermelo problem is formulated and solved as a variational problem of determining the brachistochronous trajectory of a material point in a horizontal one-dimensional vector field of a moving fluid. The velocity module of the moving fluid is given as a function of the transverse coordinate (relative to the velocity vector of the moving fluid). A target functional was built, which determines the time of movement of a material point along the brachistochron curve. With its help, the Euler equation was found, which leads to the differential equation of extreme trajectories. Next, the boundary conditions of the boundary value problem are added to this equation, which allows obtaining a solution in quadratures. Based on the expansion of the integral function into a power series, an approximate equation of the extreme trajectories of the point’s motion is obtained, taking into account the given boundary conditions. Within the framework of the numerical experiment, specific examples of the implementation of the proposed approach are given. For the selected version of the boundary conditions, the equations of extreme (brachystochronous) trajectories were determined, along which the material point passes from the start to the finish in the minimum time. An estimate of the time of movement of a material point was made both by brachistochronic trajectories and by an alternative rectilinear path. It is shown that this variational problem has two different solutions that differ only in sign. A comparative analysis of the point movement time on two extreme trajectories was carried out. At the same time, only one of them provides a minimum of the target time function. An interesting scientific fact has been established: the brachistochronous trajectories of the point’s movement are not rectilinear, but have an oscillatory character. Thus, the “shortest path” of the ship’s motion is not the “fastest” in this variational problem.

References 1. Dunham, W.: Journey Through Genius, p. 304. Penguin Books, New York (1991) 2. Eltsgolts, L.P.: Differential Equations and Variational Calculus, p. 432. Nauka, Moscow, SU (1974) 3. Erlichson, H.: Johann Bernoulli’s brachistochrone solution using Fermat’s principle of least time. Eur. J. Phys. 20(5), 299–304 (1999). https://doi.org/10.1088/0143-0807/20/5/301 4. Ashby, N., et al.: Brachistochrone with Coulomb friction. Am. J. Phys. 43(10), 902–906 (1975). https://doi.org/10.1119/1.9976 5. van der Heijden, A.M.A., Diepstraten, J.D.: On the brachistochrone with dry friction. Int. J. Non-Linear Mech. 10(2), 97–112 (1975). https://doi.org/10.1016/0020-7462(75)90017-7 6. Lipp, S.: Brachistochrone with Coulomb friction. SIAM J. Control Optim. 35(2), 562–584 (1997). https://doi.org/10.1137/S0363012995287957 7. Covic, V., Veskovic, M.: Brachistochrone on a surface with Coulomb friction. Int. J. NonLinear Mechanics 43(5), 437–450 (2008). https://doi.org/10.1016/j.ijnonlinmec.2008.02.004 8. Hayen, J.C.: Brachistochrone with Coulomb friction. Int. J. Non-Linear Mech. 40(8), 1057– 1075 (2005). https://doi.org/10.1016/j.ijnonlinmec.2005.02.004 9. Vratanar, B., Saje, M.: On analytical solution of the brachistochrone problem in a nonconservative field. Int. J. NonLinear Mech. 33(3), 489–505 (1998). https://doi.org/10.1016/ S0020-7462(97)00026-7

708

V. Legeza et al.

10. Yamani, H.A., Mulhem, A.A.: A cylindrical variation on the brachistochrone problem. Am. J. Phys. 56(5), 467–469 (1988). https://doi.org/10.1119/1.15755 11. Palmieri, D.: The brachistochrone problem, a new twist to an old problem. Undergraduate Honors Thesis, Millersville University of PA (1996) 12. Aravind, P.K.: Simplified approach to brachistochrone problem. Am. J. Phys. 49(9), 884–886 (1981). https://doi.org/10.1119/1.12389 13. Denman, H.H.: Remarks on brachistochrone-tautochrone problem. Am. J. Phys. 53(3), 224– 227 (1985). https://doi.org/10.1119/1.14125 14. Venezian, G.: Terrestrial brachistochrone. Am. J. Phys. 34(8), 701 (1966). https://doi.org/10. 1119/1.1973207 15. Parnovsky, A.S.: Some generalisations of the brachistochrone problem. Acta Physica. Pol. Suppl. A 93, 5–55 (1998) 16. Tee, G.: “Isochrones and brachistochrones”, Neural. Parallel Sci. Comput. 7, 311–342 (1999) 17. Goldstein, H.F., Bender, C.M.: Relativistic brachistochrone. J. Math. Phys. 27(2), 507–511 (1986) 18. Scarpello, G.M., Ritelli, D.: Relativistic brachistochrone under electric or gravitational uniform field. Z. Angew. Math. Mech. 86(9), 736–743 (2006). https://doi.org/10.1002/zamm. 200510279 19. Gemmer, J., et al.: Generalizations of the brachistochrone problem. Pi Mu Epsilon J. 13(4), 207–218 (2011) 20. Mertens, S., Mingramm, S.: Brachistochrones with loose ends. Eur. J. Phys. 29, 1191–1199 (2008). https://doi.org/10.1088/0143-0807/29/6/008 21. Rodgers, E.: Brachistochrone and tautochrone curves for rolling bodies. Am. J. Phys. 14, 249–252 (1946). https://doi.org/10.1119/1.1990827 22. Obradovic, A., et al.: The brachistochronic motion of a vertical disk rolling on a horizontal plane without slip. Theor. Appl. Mech. 44(2), 237–254 (2017). https://doi.org/10.2298/TAM 171002015O 23. Gurram, S.S., et al.: On the brachistochrone of a fluid-filled cylinder. J. Fluid Mech. 865, 775–789 (2019). https://doi.org/10.1017/jfm.2019.70 24. Legeza, V.P.: Cycloidal pendulum with a rolling cylinder. Mech. Solids 47(4), 380–384 (2012). https://doi.org/10.3103/S0025654412040024 25. Legeza, V.P.: Brachistochrone for a rolling cylinder. Mech. Solids 45(1), 27–33 (2010). https:// doi.org/10.3103/s002565441001005x 26. Bryson, A.E. Jr., Ho, Y.-C. Applied Optimal Control: Optimization, Estimation and Control, pp. 496. Taylor&Francis Group, New York (1975). https://doi.org/10.1201/9781315137667 27. Pashentsev, S.V.: Zermelo navigation problem: an analytical solution for a vortex field. Bull. MSTU 3(1), 17–22 (2000). (in Russ.) 28. Pashentsev, S.V.: Zermelo problem: analytical solution for a linear velocity field. Vestnik MSTU 5(2), 187–192 (2002). (in Russ.) 29. Pashentsev, S.V.: Solution of the zermelo navigation problem for an arbitrary axial velocity field. Vestnik MSTU 13(3), 587–591 (2010). (in Russ.) 30. Belhocine, A.: Exact analytical solution of boundary value problem in a form of an infinite hypergeometric series. Int. J. Math. Sci. Comput. (IJMSC) 3(1), 28–37 (2017). https://doi. org/10.5815/ijmsc.2017.01.03 31. Falade, K.I.: A Numerical approach for solving high-order boundary value problems. Int. J. Math. Sci. Comput. (IJMSC) 5(3), 1–16 (2019). https://doi.org/10.5815/ijmsc.2019.03.01

A Fuzzy Based Predictive Approach for Soil Classification of Agricultural Land for the Efficient Cultivation and Harvesting Rajalaxmi Hegde and Sandeep Kumar Hegde(B) Nitte (Deemed to Be University), Department of Computer Science and Engineering, NMAM Institute of Technology, Nitte, Karkala Taluk, Udupi District 574110, India [email protected], [email protected]

Abstract. Soil classification is a crucial component for geotechnical engineering purposes. When compared to conventional methods, smart and soft computing technology can quickly and accurately classify various soil types with high precision. Based on earlier studies, a database of soil characteristics is compiled and created, and it is used to train and test machine learning classifier algorithms like Naive Bayes and decision tree induction. Depending on the attitude place of origin, or inherent qualities soils can be grouped in a variety of ways. The goal is to forecast soil classification based on soil characteristics and information gathered about soil mechanics parameters. In the recommended paper, the fuzzy C-Means classifiers are used to classify the soil. Several data mining techniques such as data preprocessing, data analysis techniques are used to analyze and preprocess the data. Soil classification techniques influences several properties such as use and management of land. Soil texture is an important part of soil classification it influences fertility tillage, holding capacity of water, drainage etc. The algorithm’s performance is contrasted with that of other conventional machine learning methods. The experiment’s findings indicate that the recommended technique outperformed the alternatives in terms of accuracy as well as efficiency. Keywords: Fuzzy C Means · Soil Morphology · Machine Learning · Agriculture · Decision Tree

1 Introduction India is primarily a farming nation. India is the second-largest country in terms of total agricultural land with over 60% of its land area being used for agriculture. Farmers still use conventional methods like manual field [1] monitoring, which involves frequently watering crops and pesticides to control pests, whether or not they are aware of the appropriate dosage. These conventional methods have drawbacks like flooded fields, overuse of pesticides and fungicides, and ignorance of the crops that should be developed for a specific type of soil. Low quality yield, decreased soil fertility, and decreased expected yield from agricultural land are the impacts of these drawbacks. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 709–720, 2023. https://doi.org/10.1007/978-3-031-36118-0_64

710

R. Hegde and S. K. Hegde

Soils can be categorised into different groups based on their behaviour, place of origin, or inherent characteristics [1]. The classification approach may be impacted by differences in the importance of physical characteristics to various land uses and disease theories. Contrarily, a technical system approach to classifying soils divides soils into groups based on their suitability for particular purposes and microclimatic characteristics. Geologists changes in soil structure, texture, and properties may take tens of thousands of years to occur [2]. The soil data must be analyzed based on the pH for effective cultivation and harvesting. Applications of machine learning algorithms have been utilized to categorize soil by both commercial and academic institutions [3]. These methods have been applied for commercial, scientific, and industrial objectives. For instance, machine learning techniques have been applied to evaluate enormous data sets and identify insightful categorizations and patterns. Research investigations in agriculture and biology have used a variety of data analysis approaches, including decision trees, statistical machine learning, and other analysis methods [4]. This study explores the potential of machine learning techniques to enhance the analysis and pattern detection of large experimental soil profile datasets. The goal of this effort is to comprehend potential processes underlying the aging effects in sands. This was achieved through a review of the available data from the literature and the creation of a lab testing program to categorize them according to the attributes [5]. The laboratory testing program was created so that the effects of several factors, including soil type, relative density, temperature, pH, and geographic locations, could be researched [6]. Finding a collection of fuzzy rules to address a particular classification problem from training data with uncertainty is the most crucial task in the design of fuzzy classification systems [7]. One of the key tasks in data mining investigations and statistical data analysis, which is widely applied in many fields, is clustering. Datasets are grouped into groups called clusters using the relationships between them as the basis for clustering. The relationship here between data in datasets and the effectiveness of the cluster formation determine the different clustering algorithms. The most common types of relationships between data are those defined by their distance from one another, their connectivity, their density in the data space, their interval or distribution, and other factors. Clustering is viewed as a multi-purpose optimization problem from the perspective of how to accomplish efficient cluster formation. In the proposed approach the soil data were classified using the fuzzy C-Means technique through an unsupervised learning approach. The complete training data set is transformed into fuzzy rules, and the relationships between the input variables are used to calculate the weights for each input variable that appears in the resulting fuzzy rules. The test dataset is fed into the proposed model, and the accuracy obtained with the test dataset is examined. In the proposed paper, C language was utilized to implement the fuzzy rules by first establishing the membership functions for the input attributes of the soil data, and then generating the initial fuzzy rules for the training data. The proposed approach’s accuracy is compared to that of conventional machine learning methods like SVM, random forest, naive Bayes classifier, and decision tree techniques. The results of the experiments show that the proposed approach achieved better accuracy when compared to conventional machine learning techniques.

A Fuzzy Based Predictive Approach for Soil Classification

711

2 Literature Review The following section summarizes the various research studies that have been proposed in the field of soil classification using various machine-learning approaches. In [8] Support vector machines were proposed for recognizing, mapping, and categorizing different types of soil. Hyper spectral data with a high spectral resolution is used in this method. When working with a tiny sample size of data, a support vector machine classifier produces accurate results. It has been found that high-dimensional datasets with few training samples benefit from the support vector machine technique. The investigated soil datasets were categorized using a technique that is proposed in [9]. By using the Naive Bayes and k-Nearest Neighbor algorithms, soils are divided into three categories: low, medium, and high. The Soil Testing Laboratory in Jabalpur, Madhya Pradesh, is where the data is gathered. The tuples in the dataset describe the number of nutrients and micronutrients present in the soil. By dividing these nutrients and micronutrients into two groups, it is possible to determine the soil’s capacity for yield. The medium group of soils shows good yielding capabilities, according to the JNKW Jabalpur department of soil science. The category high (H) and very high soils exhibit modern yielding capability, while the category low (L) and very low soils do not. A categorization approach for soil texture based on hyper spectral data was presented in [10]. A random forest classifier’s performance is compared to that of the CNN techniques. The soil dataset from the Land/Use and Cover Area Frame Statistical Survey (LUCAS) is used in this study. Hyper spectral and soil texture data are included. One drawback of the proposed technique is that atmospheric error, which is typically present in hyper spectral data, must be corrected. When compared to traditional statistical methods, in [11] the proposed methodology performs better which uses a limited amount of attributes from the dataset to analyze. The agricultural soil profiles were chosen to be comprehensive and to make soil classification easier. In [12] author proposed a convolutional neural network-based empirical model to investigate the daily soil moisture retrieval for passive microwave remote sensing. The soil moisture retrieval model based on deep learning can learn detailed features from a vast amount of remote sensing data. In [13] author developed a support vector machine-based model for the classification of soil samples with the help of numerous scientific factors. Several algorithms, characteristics, and filters are used to collect and process color photos of the soil samples. These algorithms extract numerous characteristics, including color, texture, and others. Support Vector Machine only uses a small subset of the training samples that are situated at the boundaries of the class distribution in feature space to fit an ideal hyper plane between the classes. The supervised categorization’s accuracy is influenced by the training data that was used. In [14] authors proposed an improved model with more features than the current method, such as crop lists, urea levels, and soil nutrients. To categorize soils where color, energy, and HSV may be visible, image processing techniques can be employed. In [15] authors applied J48, JRip algorithm, decision tree, and Naive Bayes classification techniques that are used on the soil data sets for the soil classification. The Naive Bayes algorithms obtained higher accuracy compared to other machine learning approaches when imposed on soil datasets. This paper introduces the fuzzy Sequence - to - sequence neural network clustering technique along with a mediated activation function to address combinatorial optimization problems. The fuzzy Hopfield neural network therefore uses a particular

712

R. Hegde and S. K. Hegde

numerical process. It can reduce an energy function to locate the membership grade. [16].The process of clustering entails dividing or classifying a collection of objects into distinct subsets or clusters. The assumption that there are particular predefined c clusters in the data sets is predicated on the clustering algorithm clustering. It is possible to locate the ideal group of clusters by minimizing an objective function.

3 Methodology The architecture of the proposed approach is shown in Fig. 1 below. As shown in the figure, initially soil data is given as input to the proposed model. The data are passed as input and undergo pre-processing stage where noise and irrelevant features are filtered from the given dataset.

Fig. 1. Proposed Architecture Diagram

3.1 Data Pre-processing To increase performance accuracy when classifying or forecasting time series data, data pre-processing is a crucial step that must be completed first. In order to transform large datasets into more complete, organized, and noise-free data, incomplete words and noise must be reduced or eliminated [18, 19]. This pre-processing stage aims to get the findings of extracting the features that is not widely used and minimize errors induced by bias from every dataset [19].This study’s pre-processing stage included a number of procedures for a wide range of data types, beginning with the steps for removing noise, cleaning incomplete data, and finishing incomplete data. The default value, 0, may be utilized in a mathematical operation if a value data is absent or incomplete [19].

A Fuzzy Based Predictive Approach for Soil Classification

713

Crop harvest and rainfall data were among the different types of data included in the dataset used in this study. Data on plant commodities obtained from the primary sources, specifically from the Selman Agriculture Service, are presented as a series of graphs for the last five years, starting with the volume of production from different plant varieties, pests that attack, and pest management data from 2014 to 2018. Data on rainfall were also gathered concurrently from secondary sources, specifically from the BPS (Statistics Indonesia) state website in Selman from 2005 to 2018.The data is still obtained in raw and unstructured form, so preprocessing is required to clean and normalize the data before further processing can be done with it.

Fig. 2. Flowchart of Fuzzy C means Algorithm

The process of forecasting rain is carried out using rainfall data from previous years after that the data is deemed normal. The classification of horticultural crops that are suitable for plants in particular months is based on the outcomes of rainfall forecasting and a number of parameters or dependent variables. The expectation is that the words or text produced by the normalization process can retrieve into features that impact their respective classes because normalization is one of the data transformation texts that works to process raw and disjointed data from the data mining process [17]. In this study, the normalization of the data is accomplished by grouping the data according

714

R. Hegde and S. K. Hegde

to the pertinent attributes and will be used as dependent variables. The next step is to remove the unnecessary information while keeping the variables that influence how plant varieties are determined untouched. The normalization of statistics in a database schema differs from the normalization of data in general. Once the dataset is pre-processed by removing noise and outliers, a fuzzy rule set is defined on the extracted dataset as per functional specification from which the classification of soil will be derived. In the end, the performance of the predictive model will be evaluated. The dataset required to experiment was gathered from the Kaggle website. The dataset is normalized using a min max normalizer and outliers are removed using the interquartile range method. The insignificant features are removed from the dataset using the recursive feature elimination technique. Figure 2 above depicts the flowchart of the proposed Fuzzy C means algorithm. Initially, a membership matrix is generated for the given dataset. From the generated membership matrix degree of the membership is generated using Eq. (1) µjk =

1 

d



ejk l=1 ekl



2 n−1



(1)

where µjk indicates the membership of the jth and kth cluster centroid, ‘d’ represents the centers of the cluster, ejk indicates the Euclidean distance between the jth and kth cluster centroid, and ‘n’ is the index of fuzziness. For each of the memberships, a fuzzy centroid is derived and membership has been upgraded using Eq. (2) below.     n u j=1 µjk Sj uk =     = 1, 2..d (2) n u , ∀ µ jk k j=1 uk represents the kth fuzzy center, Xj represents the data points and ‘k’ is the objective function. The fuzzy centroid computations will be performed until the difference in the centroid matrix between the new and previous iterations will be less than the threshold. The pseudocode of the proposed fuzzy c means algorithm is as below. Let S = {s1, s2, s3 …, sn} are the data points from the considered dataset and U = {u1, u2, u3 …, uc} are the cluster centre. 1) Randomly choose the centers of the cluster ‘c’. 2) Compute the fuzzy membership ‘µjk ‘ as µjk =

1 d



ejk l=1 ekl





2 n−1



3) Determine the fuzzy centroid ‘uk ‘ as     n u j=1 µjk Sj uk =     = 1, 2..d n u , ∀k j=1 µjk

A Fuzzy Based Predictive Approach for Soil Classification

715

4) Repeat stages 2 and 3 till the minimum value for ‘k’ is achieved. One of the advantages of the proposed algorithm is that it obtains better results with the overlapped dataset. The fuzzy c means algorithms are applied to the considered dataset and the result obtained with the test dataset is analyzed. The section below illustrates a detailed discussion of the result obtained with the proposed approach.

4 Results and Discussions The enormous amounts of information that are currently practically harvested alongside the crops need to be analyzed and used to the fullest extent possible. The analysis of a soil input data using data mining methods is the goal of this study. It focuses on classifying soil using the various available algorithms. Regression analysis is used to predict untested attributes, and automated sample soil classification is put into practice. A soil classification process groups soils with comparable characteristics or attributes together. The data from the soil series can be classified using machine learning techniques. To forecast the crop production that are best suited for a specific region’s soil series and climatic conditions, the results of this classification are merged with crop datasets. Finding a collection of fuzzy rules to address a particular classification problem from training data with uncertainty is the most crucial task in the design of fuzzy classification systems. The Values can be predictably determined by grouping soil forms. After going through the classification procedure, the database for future references has successfully been updated with soil data. The figure below indicates the result obtained with the proposed approach when it is applied to the soil dataset.

Fig. 3. Snapshot of Training Dataset Loading form

716

R. Hegde and S. K. Hegde

Fig. 4. Snapshot of Data Pre-processing Stage

As shown in Figs. 3 and 4, initially the collected soil dataset is loaded for the further visualization process. The dataset in its original form may contain noise and outliers which has to be pre-processed to obtain optimal accuracy. As shown in Fig. 5 feature extraction process is carried out to extract the data from the Biomass information. The dataset is classified based on different parameters such as cultivated area, Forest area, and Grass area. The implemented interface also has the option to display different information such as the number of countries, number of lands, and number of forests from which the soil dataset is taken from. As shown in Fig. 6, the forest area is classified using different parameters such as Temperate Coniferous Forest, Temperate Broadleaf Forest, and Tropical/Subtropical forest by applying the proposed algorithm. As shown in Fig. 7 above Cultivated lands are classified from the biome by applying Fuzzy rules. The grasslands are classified from the biome using different parameters such as tundra, glacier, wetland, Shrub, and Pasture. Figure 8 above illustrates the validation rate achieved using various machine learning algorithms. From the graph, it can be evident that the proposed fuzzy c means algorithms obtained higher accuracy and precision rate compared to the traditional machine learning approaches.

A Fuzzy Based Predictive Approach for Soil Classification

Fig. 5. Snapshot of Feature Extraction Process

Fig. 6. Snapshot of Forest Area Classification

717

718

R. Hegde and S. K. Hegde

Fig. 7. Snapshot of Cultivated Lands Classification

Fig. 8. Accuracy analysis of machine learning algorithms

5 Conclusion In the proposed paper new method to generate fuzzy rules from training data to deal with the soil classification problem was presented. The entire training data set is converted into fuzzy rules and then the weight of each input variable appearing in the generated fuzzy rules is derived by determining the relationships between input variables. Accordingly, the test dataset is passed to the proposed model, and the accuracy obtained with the test dataset is analyzed. The collected dataset contains four parameters including soil

A Fuzzy Based Predictive Approach for Soil Classification

719

mechanics properties such as cohesion and friction angle, and the physical characteristics such as water and dry density. The training data contain 70 data (70%) and the testing data includes the remaining values of 34 data (30%). The 30% remaining data were also entered into the modeling section without being used in the design section. Also, both sections contain 7 types of soil to test the flexibility of the models. The results showed that only 6 samples were not correctly identified among the total testing data (34 samples) in Naïve Bayes model and 7 samples were not correctly identified among the total testing data (34 samples) in other ML model. Thus, these models provide the most reliable results and can help the automatic classification of different types of soil. The performance of the proposed approach is compared with the traditional machine learning algorithms such as Accuracy, Precision, and Recall. The experimental results illustrate that the proposed weighted fuzzy rules generation approach obtained a higher classification accuracy compared to the traditional technique. Hence it can be concluded that fuzzy approaches are more suitable for soil classification through which the efficiency of agricultural cultivation and harvesting can be improved.

References 1. van Klompenburg, T., Kassahun, A., Catal, C.: Crop yield prediction using machine learning: a systematic literature review. Comput. Electron. Agric. 177, 105709 (2020) 2. Rahman, S.K., Zaminur, A., Chandra Mitra, K., Mohidul Islam, S.M: Soil classification using machine learning methods and crop suggestion based on soil series. In: 2018 21st International Conference of Computer and Information Technology (ICCIT), IEEE (2018) 3. Meier, M., et al.:Digital soil mapping using machine learning algorithms in a tropical mountainous area. Revista Brasileira de Ciência do Solo 42 (2018) 4. Wadoux, A.M.J.-C., Minasny, B., McBratney, A.B.: Machine learning for digital soil mapping: applications, challenges and suggested solutions. Earth-Sci. Rev. 210, 103359 (2020) 5. Kingsley, J., et al.: Using machine learning algorithms to estimate soil organic carbon variability with environmental variables and soil nutrient indicators in an alluvial soil. Land 9(12), 487 (2020) 6. Palanivel, K., Surianarayanan, C.: An approach for prediction of crop yield using machine learning and big data techniques. Int. J. Comput. Eng. Technol. 10(3), 110–118 (2019) 7. Khosravi, K., et al.: A comparative assessment of flood susceptibility modeling using multicriteria decision-making analysis and machine learning methods. J. Hydrol. 573, 311–323 (2019) 8. Pereira, G.W., et al.: Soil mapping for precision agriculture using support vector machines combined with inverse distance weighting. Precision Agric. 23(4), 1189–1204 (2022) 9. Taher, K.I., Abdulazeez, A.M., Zebari, D.A.: Data mining classification algorithms for analyzing soil data. Asian J. Res. Comput. Sci. 8, 17–28 (2021) 10. Pandith, V., et al.: Performance evaluation of machine learning techniques for mustard crop yield prediction from soil analysis. J. Sci. Res. 64(02), 394–398 (2020) 11. Zhigang, X., et al.: Multisource earth observation data for land-cover classification using random forest. IEEE Geosci. Remote Sens. Lett. 15(5), 789–793 (2018) 12. Xuebin, X., et al.: Applying convolutional neural networks (CNN) for end-to-end soil analysis based on laser-induced breakdown spectroscopy (LIBS) with less spectral preprocessing. Comput. Electron. Agric. 199, 107171 (2022) 13. Zhu, Q., et al.: Drought prediction using in situ and remote sensing products with SVM over the Xiang River Basin, China. Nat. Hazards 105(2), 2161–2185 (2020)

720

R. Hegde and S. K. Hegde

14. Abraham, S., Huynh, C., Huy, V.: Classification of soils into hydrologic groups using machine learning. Data 5(1), 2 (2019) 15. Wankhede, Disha S.: Analysis and prediction of soil nutrients pH, N, P, K for crop using machine learning classifier: a review. In: International Conference on Mobile Computing and Sustainable Informatics, Springer, Cham (2020) 16. Alzaeemi, S.A.S., Sathasivam, S., Velavan, M.: Agent-based Modeling in doing logic programming in fuzzy hopfield neural network. Int. J. Mod. Educ. Comput. Sci. 13(2), 23–32 (2021) 17. Putra, A.B.W., Gaffar, A.F.O., Suprapty, B.: A performance of the scattered averaging technique based on the dataset for the cluster center initialization. Int. J. Mod. Educ. Comput. Sci. 13(2), 40–50 (2021) 18. Kajol, R., Akshay Kashyap, K.: Automated agricultural fieldanalysis and monitoring system using IOT. Int. J. Inform. Eng. Electron. Bus. 12(2), 17 (2018) 19. Yudianto, M.R.A., et al.: Rainfall forecasting to recommend crops varieties using moving average and naive bayes methods. Int. J. Mod. Educ. Comput. Sci. 13(3), 23–33 (2021)

Sector-Independent Integrated System Architecture for Profiling Hazardous Industrial Wastes Wilson Nwankwo1(B) , Charles O. Adetunji3 , Kingsley E. Ukhurebor4 , Acheme I. David2 , Samuel Makinde2 , Chukwuemeka P. Nwankwo2 , and Chinecherem Umezuruike5 1 Department of Cyber Security, Delta State University of Science and Technology,

Ozoro, Nigeria [email protected] 2 Department of Computer Science, Edo State University, Uzairue, Nigeria 3 Department of Microbiology, Edo State University, Uzairue, Nigeria 4 Department of Physics, Edo State University, Uzairue, Nigeria 5 College of Computing, Bowen University, Iwo, Nigeria

Abstract. The accummulation of waste in urban societies has been christened the “industrialization and urbanization aftermath” in some quarters. In recent times, the climate change phenomenon is attributed to the generation and release of massive wastes from humanity’s economic activities globally. The move to restore the ecological integrity of the earth’s habitat is paramount and many programmes and policies are currently being designed to solve this global challenge. This paper reviews the current industrial waste management approaches and practices including legislation, machineries, and policies in Lagos Nigeria, a major urban area and commercial centre in sub-Saharan Africa. The study adopts a narrativeexpository posture with enquiries made in respect of all operations and machineries of waste management in the urban and surrounding areas. Findings show that while machineries and policies are substantially crafted to meet the changing socio-economic developments, there is a gap in the deployment of modern information technologies in fostering the policy provisions in such a manner that complements the efforts of the relevant government machineries while ensuring sustainable development. The current waste management practices do not reflect what is considered the minimum global best practices. Accordingly, this study proposes a technology-driven architecture that employs an intelligent integrated resource planning system to support collective and collaborative management of industrial wastes particularly in the area of risk management, profiling, and resource allocation. The proposed architecture is significant in that it provides a realistic coordinating single window to harness all relevant existing practices through the introduction of a computerized platform that includes risk and hazard profiling of various industry subsectors, industry operators, location of operation, citizen reporting, environmental impact, feedback, etc. Thus, this architecture promotes collective and citizen-driven waste management ecosystem which in turn enhance the operations of all environmental management stakeholders through data provisioning, messaging, real-time data analytics, and forecasting. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 721–747, 2023. https://doi.org/10.1007/978-3-031-36118-0_65

722

W. Nwankwo et al. Keywords: Hazard · Industrial Wastes · Computerized Risk Management · Sustainable development · Targeting

1 Introduction Waste management (WM) poses challenges to developing and developed countries. It has been observed that the daily increase in human population currently experienced globally has led to increase in waste generation. This could also be linked to high level technological advancement most especially in the industrial sector, consumption, lifestyle choices, which has reinforced the necessity to address this environmental concern [1]. WM is an essential part of industrialization programs directed at minimizing the impact of large quantities of wastes generated from the industrial activities. WM policy is imperative in every clime as it is designed to ensure sustainable economic, environment, social relations that promote sustainable development (SD). The following are the major stages involves in the process of waste management: generation, accumulation, collection, movement, processing and the eventual disposal of these wastes. All these stages involve numerous techniques that could be utilized at each stage in order to ensure adequate management of wastes [2]. Experience has shown that one of the most crucial strategies that could be applied for effective waste management is to ensure their eventual minimization thorough an effective developmental design and plan. In addition, it has been shown that if minimal waste were generated at the initial stage, definitely there would be minimal pressure on other later stages involved in the process of managing wastes [3, 4]. The later stages may factor in some evaluation techniques especially on the efficacy and efficiency of disposal plants in the WM process. Such evaluation techniques could reveal the points where due diligence or care was not adhered to by some persons, organizations or agencies, for instance, where unselective and poor waste disposal had led to numerous environmental maneuverings and complications. In many societies, it has become mandatory to carefully chose and adopt sustainable disposal approaches that are ecofriendly for effective WM. Such choices might be linked to some merits contemplated and some degree of certainty to which such measures could be adjudged to yield useful resources in addition to their primary intendment, for instance, where a chosen WM approach could aid translate wastes to necessary raw materials for other industries [5, 6]. Various researches had stated that waste disposal management includes the following: pyrolysis, burning, recycling, incineration, shredding, composting, reuse, open dumping, and landfilling [7–11]. Recycling entails the careful utilization of materials that had attained their end-of-life state for the creation, production, manufacturing, and/or rebranding of new products. The products so produced could afford users further uninterrupted uses [12]. The process of recycling involves the reduction of wastes materials most especially the finite resources while the significant resources are preserved [13–15]. Some forms of recycling may involve some other waste disposal methods: incineration, composting, landfilling, open dumping, burning, reuse, shredding, and pyrolysis landfilling [7–11]. However, recycling has its negatives, which include decline of product’s value and effectivenessp [16] and the tendency to recover inappropriate energy [17,

Sector-Independent Integrated System Architecture

723

18]. Incineration involves combustion at very high temperatures [19], on the site. The merits of incineration include decrease of waste, generation of electricity and heat energy [8], while the demerits entail generation of acid rain and fabrication of ash. Composting entails the application of beneficial microorganisms in the decomposition of organic waste [8]. This might lead to the incidence of global warming through generation of carbon dioxide, plastic and glass like residue [20]. Landfill reflects an allocated space on the earth’s surface where waste materials are buried. Landfills are usually made specifically for a certain period and there are two types which include natural and sanitary attenuation [10]. Some merits of landfills include: opportunities for scavengers to perform their role, cost effectiveness, and release of gases from which reasonable energy could be generated; while some of the demerits include high level of air and ground water pollution as well as air borne diseases and fire hazards [21, 22]. Burning is recognized to have similar features to open dumpsites where wastes are set on fire. The risks related to this include air pollution, explosion and fire. Burning produces huge amounts of greenhouse gases that readily predisposes the ozone layer to severe depletion. Reuse involves any operation through which components or products which are not wastes (in the true sense of the term but are embedded in wastes) are utilized again for the purpose they are designed for. Reuse is a disposal option that guarantees that supposed wastes are reused in an effective manner through indirect or direct application [10]. Some of the benefits of reuse include ecofriendiness, reduction of burden of disposal, and utility. Pyrolysis is adjudged an alternative to incineration. Pyrolysis guarantees the thermal disintegration of materials when subjected to very high temperatures in the presence of oxygen [11]. Pyrolysis has the potential to minimize the volume of waste, destroy toxic constituents, and produce gases that could be utilized as fuel. Nevertheless, pyrolysis generate ashes that could pollute the air, leading to pollution. One of the greatest disadvantages of pyrolysis is that it is not economical [11]. It follows that apart from recycling and reuse some other techniques are not ecofriendly, hence, are unsustainable. Unfortunately, not all categories of wastes could be recycled or reused. In recent times, there is a high demand for new efficient and integrative approaches to WM in developing nations including Nigeria where the production of material wastes is high owing to deficiencies in the existing waste disposal techniques. Research has affirmed that the high volumes of wastes generated in Nigeria is linked to absence of modern inexpensive disposal techniques which has placed a high demand on government to enforce organizational regulatory policy [23–25]. Similarly, it is submitted that the management of waste in Nigeria is generally poorly handled [26]. In this study, we focus on Lagos as a case study. Lagos is one of the largest cities in Africa, and it is widely believed that what works in Lagos would work in other parts of Africa especially in large African cities. In Lagos, the Lagos State Waste Management Authority (LAWMA) is the public agency overseeing all WM operations. According to LAWMA, it operates five landfill sites with two as temporary and three major sites in the state where different types of waste are disposed [27]. Similar postures are maintained by other waste management authorities across the 36 States in Nigeria. It is observed that the rate of recycling of waste in Lagos State is very low [28], while huge volumes of waste are generated from households, factories, and markets, an insignificant volume particularly papers and plastics are subjected to the process of recycling [29]. [28] notes that burying and burning are the major practices of WM in

724

W. Nwankwo et al.

Lagos State. Furthermore, [30] states that the reuse of some waste material is still feasible but [31] observed that these processes are under-utilized. The use of incineration for the management of waste is not commonly carried out in Lagos and in most cities in Nigeria [32]. It could be said that there is no functional machinery that targets the management of industrial wastes in the State. Wastes from industrial processes are often in three forms: solid, liquid water waste or effluents, gaseous. It is observed that most of these gaseous emissions leaches; and it is common knowledge that some liquid effluents pose health risks to the workers within and around the industries [33]. Majority of such industrial wastes (IW) contain hazardous wastes that pose threat and a source of environmental and health hazards when not properly managed. Prior to further discourse on hazardous wastes it would be ideal to recourse to necessary legislative provisions in determining what constitutes hazardous or harmful wastes. Accordingly, Section 15 of Nigeria’s Harmful Wastes (Special Criminal Provisions etc.) Act, rightly provides that “harmful wastes means any injurious poisonous, toxic or noxious substance and, in particular, nuclear wastes emitting any radioactive substance … as to subject a person to the risk of death, fatal injury or incurable impairment of physical or mental health”. The effect of this provision is that any waste whether or not industrial that could cause some harm is a harmful waste. Hazardous wastes could pollute the air, water, and soil. The contamination of these three environmental layers is akin to humankind intentionally creating a hazardous ecosystem antithetical to sustainable development and growth. Hazardous waste materials could route toxic chemicals through the soil, air, water thereby providing the platform for pollution, which predisposes man to various health anomalies such as infections, respiratory stress, visual impairment, allergic reactions, poisoning, blood-related problems, etc. Workers are exposed to numerous risks of varying magnitude depending on the jobs and work environments. A typical scenario is an asbestos plant where workers may have regular exposures to air pollutants such as benzene and toluene [34]. The result of the above scenario is very consistent with the position of [35] wherein he observed that approximately 2 million working hours is wasted during work – associated illness while about two million people testified that they are confronted with illness that could be associated with their workplace. In view of the aforestated gaps and challenges, this review proposes a technologydriven architecture that would employ a resource planning system to support collective and collaborative management of risks associated with industry wastes. The proposed architecture extends the existing practices through the introduction of computerized approaches that includes risk and hazard profiling of various industry subsectors, industry operators, location of operation, environmental impact etc.

2 Categorization of Industrial Wastes by Industry/Economic Sector Categorically, the main sources of poisonous waste are grouped into two: human and natural sources. Regardless of the category to which industrial wastes belongs, there is a common concern which is ‘harm’ to humans and the environment [36–38]. We shall discuss some of the categories briefly.

Sector-Independent Integrated System Architecture

725

2.1 Human Sources Mostly, poisonous wastes arise from industrial plants, firms, companies, refineries or industries. The quantity of wastes produced by such entities may be alarming owing to the cumulative degenerative effects they cause on the environment. The poisonous wastes produced by such organizations are often in solid or fluid (liquid or gases). Some culpable organizations [36–38] include: a. Chemical manufacturing industries and plants that produce chemical wastes such as spent solvents, strong acids and bases as well as other chemical reactive wastes, etc. b. Cosmetic/cleaning agents manufacturing industries that produce strong acids and bases, heavy metal dust, combustible waste and solvents, etc. c. Printing industries that produce heavy metal, wastes ink and spent electroplating, solvents, etc. d. Furniture and wood manufacturing and refinishing industries, which generate spent solvents, combustible wastes, etc. e. Metal manufacturing industries that generate heavy metals, cyanide, strong chemical (acids and bases), etc. f. Leather products manufacturing industries that release wastes such as toluene and benzene, etc. g. Paper manufacturing industries, which release heavy metal wastes, combustible solvents, strong chemicals (acid and bases), etc. h. Vehicle manufacturing and maintenance companies and plants that release huge amounts of heavy metal and flammable wastes, as well as spent solvents and used lead acid cells, etc. i. Petroleum exploration industries also generate enormous toxic organic and biochemical wastes. Presently, Nigeria does not have any nuclear power plant, firm, company, refinery or industry but the country is greatly exposed to high technology wastes (poisonous effluents, apprehensions and emissions) imported from the industrialized societies, some of which exhibit radioactive tendencies. 2.2 Natural Sources The major apparent natural sources of toxic wastes are volcanic eruptions (which produce several tons of poisonous gases and undesirable emissions as well as harmful lava), and some agricultural and chemical raw materials containing phytoxins, poisonous dust, lead, etc. which may be released during processing [36–38]. It is sad to note that more than 80% of the high technology wastes in the world are dumped directly or indirectly in landfills, rivers, seas or oceans in Africa and Asia [36–40] which when discovered many years after may appear as if such wastes are the result of nature’s activity. Nigeria is noted as one of the emerging dumping grounds for these toxic, organic, biochemical and electronic waste from the industrialized world [37, 39–43]. In addition to the aforementioned sources, oil spills had become a serious concern for decades now in the Niger delta area of Nigeria.

726

W. Nwankwo et al.

2.3 Industrial Health Hazards (IHH) Industrial infections and poisoning are a potential and possible risk to the health of persons that are unprotected in an insalubrious environment [36, 37, 44–46]. According to Occupational Health and Safety (OHS) provisions, occupational hazards and/or infection (industrial health hazard) occur following exposures to toxic and/or infectious biological, physical or chemical substances in the workplace, which could negatively affect the person’s normal physiochemical mechanisms leading to some impairments to their general wellbeing [44–47]. IHH could result to accidents or infections after such exposures in the workplace. Exposure could be to the immediate locality or to substances from a remote site. Some health challenges manifest after extensive exposure to working tools, equipment, apparatus and other devices, as well as infection-causing microorganisms, chemicals, waste products or even dust over a considerable period due to insanitary and other unhealthy situations at our various places of work [46–48], whereas some health problems may be immediate e.g. allergic reactions following inhalation of poisonous lead substances from a mining site. Some known diseases and/or anomalies arising from IHH include musculoskeletal difficulties, tuberculosis, blood borne infections, latex allergy, fierceness and some other work-associated trauma and stress [46–48]. IHH are not given appropriate attention and are mostly pronounced as public health problems amongst healthcare practitioners in most developing economies [38, 46, 47]. Complications amongst healthcare practitioners in health facilities is also within the contemplation of IHH and is one of the common issues that occur following exposure to substances and fluids from the body of patients, accidents and inadequate personal and other protective working devices. 2.4 Formal and Policy-Based IWM Practices In discussing formal policy-based practices reference must be made to industrialization and civilization in Nigeria. These two constitute the cornerstone on which the rise of several industries involved in production of goods and services to meet the demands of the growing national population, surely rests. Research has documented the continual release of poisonous effluents, and emissions to the surrounding environment by these industries [36, 38–40] regardless of the subsisting legislations and policies on industrial WM. It is interesting to submit that a conglomerate of socioeconomic concerns that tend to mar the effective waste management practices is contributing to the aforementioned pitiable state. These concerns are: a. b. c. d.

Uncontrolled consistent spread of disorderly and disorganized settlements Traffic congestion in major cities Insecurity and terrorism Ignorance and illiteracy

Given the astronomical increase in the country’s population, industrialization, and ongoing developmental projects, it would be interesting to submit that contemporary and operative waste management practices need be deployed to foster socioeconomic wellbeing of the general public [36, 38, 45]. Such practices and strategies should encompass

Sector-Independent Integrated System Architecture

727

all stakeholders, including the industries (that produce the wastes), the private sector, the public sector as well as the government regulatory agencies. That most societies are regularly confronted with some environmental problems resulting from industrial wastes as well as other human and natural phenomena is welldocumented [36–40]. It is trite that majority of the risks associated with these environmental dilapidations do flow from human activities and progression [50–52]. Thus, it may be safely held that the effective design and implementation of effective but efficient strategies is what distinguished a progressive society from a retrogressive sociopolitical system. While customary waste management practices do offer some reliefs there is no substitution to formal policies, legislations, procedures, guidelines and machineries directed at WM. Accordingly, it may be said that Nigeria has rich environmental laws, policies and regulations which provide valuable frameworks for the protection of the environment from dilapidations [39, 40, 53]. However, it is apt to state that these environmental laws and regulations have not yielded their intended results as there is little or no adherence to the provisions of these laws, policies and regulations [36–40, 49]. Notwithstanding the aforementioned shortfalls, it is instructive to submit that environmental and waste management practices in Nigeria are driven by legislations and policies. IWM practices in Nigeria are tightly connected to the environmental management (EM) laws, policies and agencies. Accordingly, our discussion would be dealing on both WM and EM and in some cases; the terms would be used interchangeably. Industrial wastes are the major cause of disorderliness in the environment’s ecosystem hence EM presupposes all practices directed at maintaining the integrity of the air, water, and land. EM therefore encompasses the management of emission to air and water, environmental impact assessments, waste and pollutions, land and environmental issues connected to wastes, etc. It also includes the management and mitigation of cross-border dealings [39, 40]. IWM practices in Nigeria in a broad sense, is chiefly based on the National Policy on Environment (NPE) and several legislative provisions (which are briefly discussed in Sect. 2.6 of this paper). The NPE came into effect in 1989 and has undergone reviews in 1999, 2007 and 2009 respectively. The NPE and other legal frameworks grant legitimacy to the various machineries of government including agencies, ministries and departments, and their operational guidelines and procedures all of which are made to guarantee the conservation and management of natural resources, safety of the environment and the wellbeing of the citizenry and other members of the ecosystem [37, 39, 40]. These machineries are discussed further in Sect. 2.5. It is vital to note that the operations of these machineries of Government are influenced by political, cultural and social mantra of Nigeria, which include political instability, policy somersaults, nepotism, and religio-political affiliations, etc. all of which are quite bewildering so to say. Such immense influence does pose some threats to the efficiency of these agencies. Consequently, compliance to guidelines, procedures and environmental laws are greatly marred [37, 39, 40].

728

W. Nwankwo et al.

2.5 Environmental Management Agencies Environmental agencies or environmental protection agencies are commissions, establishments or organizations which are established within the government or separate from the government agencies whose function is to protect the environment as well to manage and preserve the nation’s ecological, environmental, and biological resources [36, 38–40]. There are relevant laws to prevent or control any form of intervention that may hamper the integrity of the collective environmental resources. These laws empower these commissions, establishments, agencies or organizations. These agencies are responsible for protecting and managing the environment from dilapidation arising from climate change, environmental pollution, etc. Such agencies may have international, national, regional, state or local council parastatals or representatives as well as some non-governmental or private commissions, establishments, agencies or organizations [36, 38–40, 48]. It should be noted that Nigeria as a federal entity runs three tiers of government i.e. federal, state, and local hence environmental protection responsibilities is not the sole responsibility of any single tier but are shared across the various tiers. This accounts for the side-by-side existence of federal legislations and policies with state laws and environmental policies. Thus, federal ministries, agencies and departments perform environmental protection functions at the federal level whereas in the states, the various state ministries and agencies also perform their responsibilities. The same is true for local councils. However, there is a hierarchy in the complexity of responsibilities undertaken at each tier. The principal legislation on environmental standards, regulation and control is a federal legislation hence all other laws at the state level and bye-laws at the local council tier derive their legality and legitimacy from this principal legislation save where the principal legislation or other federal legislations in this context do not ‘cover the field’ and in this case the respective states may invoke the doctrine of not covering the field wherein they are constitutionally empowered to make laws that must not be inconsistent with any federal legislation. Essentially, the agencies, commissions and units at the State and Local council levels complement the federal government’s effort in the management and preservation of the integrity of the environment. The key environmental establishments in Nigeria are presented in Table 1. With respect to the regulation of major environmental protection activities including waste management, and enforcement of various environmental legislations in Nigeria, NESREA, a federal agency is the primary agency at the centre. NESREA operates the NESREA Act with the 33 subsidiary regulations attached to it as well as other vital environmental policies, legislations (e.g. Environment Impact Assessment Act, etc.), international conventions, treaties, and protocols to which Nigeria is a party. NESREA does not work in isolation; it collaborates with other federal agencies such as the NNRA, NEMA, etc. NESREA also collaborates with the following establishments: a. b. c. d. e. f.

State ministries and agencies International community and donor agencies Community-based organisations. Faith based organisations. Non-governmental organisations Civil society organisations.

Sector-Independent Integrated System Architecture

729

Table 1. Various Public Environmental Establishments No.

Establishment

Type

1

Forestry Research Institute of Nigeria (FRIN)

Agency

2

Federal Ministry of Health

Ministry

3

National Oil Spill Detection and Response Agency (NOSDRA)

Agency

4

National Emergency Management Agency (NEMA)

Agency

5

Nigeria Hydrological Services Agency (NIHSA) Agency

6

River Basin Authority (RBA)

Agency

7

Environmental Health Officers Registration Council of Nigeria (EHORCN)

Agency

8

Nigerian Conservation Foundation (NCF)

Agency

9

National Environmental Standards and Regulations Enforcement Agency (NESREA)

Agency

10

Directorat of Petroleum Resources (DPR)

Department

11

Nigerian Nuclear Regulatory Authority (NNRA) Agency

12

National Biosafety Management Agency (NBMA)

Agency

13

Energy Commission of Nigeria (ECN)

Agency

14

Drought and Desertification Agency (DDA)

Agency

16

Federal Ministry of Environment

Ministry

17

Federal Ministry of Water Resources

Ministry

18

Department of Climate Change

Department

19

Department of Planning, Research and Statistics

Department

20

Federal Ministry of Science and Technology

Ministry

21

Federal Ministry of Water Resources

Ministry

22

Erosion, Floods and Coastal Zone Management

Department

23

State and Local Council commissions and departments

Commissions and Departments

g. Other law enforcement agencies e.g. Nigeria Police, Nigeria Security and Civil Defence Corps, etc. The operations of NESREA are captured in Fig. 1. NESREA generally implements pre-emptive mechanisms in ensuring compliance with appropriate legislative necessities and licensing requirements [54]. However, in a situation where voluntary compliance is

730

W. Nwankwo et al.

not imminent, NESREA would use its enforcement machineries. The main elements in NESREA’s enforcement policies include [54]: a. b. c. d. e.

Inspection. Compliance monitoring. Negotiation. Legal action. Prosecution. The procedures of enforcement by NESREA are:

a. b. c. d. e. f.

Issue of permits and licences. Issue of prohibition and enforcement notices. Variation of license conditions. Implementing the “polluter pays” principle. Suspension and/or revocation of permits and licences. Injunction and carrying out of remedial works.

2.6 Environmental Laws in Nigeria Environmental laws in Nigeria have their roots in the Constitution of the Federal Republic of Nigeria [as amended]. Environmental remediation provision could be adduced from Section 33(1) of the Constitution. According to the said section, “every person has a right to life and no one shall be deprived intentionally of his life”. This provision presupposes a healthy environment in that healthy life is dependent on a healthy environment. It follows that the citizenry ought to be protected against hazards that may exist in degrading environments. Another provision that lends credence to constitutionality of environmental protection is in Section 2 of the Environmental Impact Assessment Act. This said provision requires that: “every company whether public or private must prior to embarking on any industrial project evaluate the environmental effects of such proposed projects”. The said assessment of proposed project impact on the environment is to be evaluated, approved/disapproved by a legitimate authority. This supervisory role is the mandate of NESREA. The content of environmental legislations in Nigeria is quite comprehensive [37–40, 42]. It covers the issues bordering on “land use, soil conservation, wildlife, forestry, protected natural area, water and marine resources, coastal areas, wastes and sanitation, environment/occupational health and safety, air quality, hazardous substances, pollution/pollutants land degradation and planning, afforestation, deforestation, desertification among others” [37–40, 42]. Apart from the constitution, other sources of environmental laws in Nigeria [37, 39, 40, 42] are: Nigerian judicial precedents/Nigerian case law; Subsidiary legislations; the received English law; Customary law, and the Islamic law. In the present dispensation not all these sources listed above are applicable in the strictest sense to the environmental management in Nigeria [37–42]. The major environmental legislations in Nigeria are listed in Table 2.

Sector-Independent Integrated System Architecture

731

Fig. 1. Operations of NESREA

Table 2. Major Environmental Legislations No

Legislation

In force

Main object

1

Constitution of the FRN 1999 (as amended)

Yes

Grundnorm and law of the land

2

Environmental Impact Assessment Act 2004

Yes

Techniques and rules of environmental impact assessment in different economic sectors

3

NESREA Act 2007

Yes

Principal environmental regulatory agency, enforces environmental laws; prosecutes offenders

4

Harmful Waste (Special Criminal Provisions etc.) Act 2004

Yes

Prohibits dropping, carrying, and dumping of injurious waste on land and in territorial waters

5

Oil in Navigable Waters Act

Yes

Regulates petroleum discharges from ships

6

Associated Gas re-injection Act

Yes

Controls gas flaring by petroleum firms. Bans, unlawful flaring by oil/gas companies

7

Water Resources Act

Yes

promote development, utilization and protection of water resources (continued)

732

W. Nwankwo et al. Table 2. (continued)

No

Legislation

In force

Main object

8

Endangered Species (Control of International Trade and Traffic) Act 2016

Yes

conservation and management of wildlife and the protection of endangered species

9

Nuclear Safety and Radiation Protection Act

Yes

Controls use of radioactive substances and apparatus that emit or produce ionizing radiation

10

National Oil Spill, Detection and Response Agency Act

Yes

Establish machinery to co-ordinate and implement National Oil Spill Contingency Plan to ensure safe, timely, effective and appropriate response to major or disastrous oil pollution

12

Hydrocarbon Oil Refineries Act

Yes

licensing and control of refining activities

13

National Park Services Act

Yes

conservation and protection of natural resources and plants in national parks

15

The Land Use Act

Yes

Use of land

16

Sea fisheries Act

Yes

Control, regulate, protect sea fisheries in Nigeria’s territorial waters

17

Nigerian Mining Corporation Act

Yes

Regulation of exploration of solid minerals

18

Environmental laws in 36 states

Yes

Environmental protection and waste management

2.7 Challenges of IWM Operations The loopholes in the operation of the policies and legislations on WM are not only documented by researchers but routine observation or surveillance in urban areas such as Lagos and Port Harcourt is enough to provide a convincing ground that the said legislative frameworks and machineries are not yielding the right results. Several pitfalls connected to the operations of these frameworks have been noted ranging from poor and crude infrastructure, poor funding, incoherent initiatives of government operatives, poor maintenance culture, corruption, nepotism, social stratification with the emergence of the ‘untouchables’ that include individuals and organizations which are covertly above the law, poor enforcement, to negligence and abdication of responsibilities on the part of the staff manning these agencies. One of the greatest blows to the principal legislation (NESREA Act) and its guidelines noncompliance and weak enforcement [50, 55, 56]. Noncompliance and weak enforcement had been shown to be connected to some sociopolitical and economic problems [37, 41, 52]. These problems are: a. Weak environmental awareness campaigns b. Deficiency of competent workforce, instruments and technology c. Corruption and embezzlement of environmental fund.

Sector-Independent Integrated System Architecture

d. e. f. g. h. i.

733

Levity on the side of the Government Inadequate databases Poor funding of environmental operations Poor maintenance culture Weak or absence of environmental pressure groups Weak enforcement strategies

These problems have negated the true objects of environmental laws, policies, programs and regulations in driving the safeguarding and conservation of the environment [36, 37, 41]. We submit that these challenges could be addressed through the use of a versatile technology framework or single window that provides a transparent platform for all stakeholders in the waste management ecosystem. The single window will provide planning/scheduling and monitoring for all necessary environmental awareness campaigns, workforce, budgets (allocations and expenditure), waste management databases, profiles, etc.

3 Why Technology-Based Waste Management? Coordinated and organized WM campaign is neither about collation and disposal of wastes nor about enforcement or prosecution of offenders. Efficiency, effectiveness, sustainability, and better approaches are factors that must be considered continually. Accordingly, the concept of technology-based WM is an inevitable part of modern IWM programs and campaigns. In recent times, the waste stream size and composition serve as determinants for the type of waste management system to be designed and/or implemented. Modern WM ideologies advocate integrated technology-based waste management approaches with consideration on resource recycling, recovery and energy generation facilities from industrial waste [57]. In fact, waste-to-energy conversion is now considered as one of the best ways of addressing the problem of waste management in a sustainable manner [58, 59]. Therefore, various thermochemical and biological waste-to-energy technologies are now being used to treat industrial wastes including harmful wastes. Thus, new approaches that could lead to efficient industrial WM have now extended the horizon of WM. Some of these are included in Fig. 2. Some of these approaches could provide a platform for transforming waste into useful energy [60]. Like other socio-economic sectors, waste management requires decision making at the tactical, operational and strategic levels. To achieve the objectives of such decisionmaking needed in the management of hazardous industrial wastes, information technologies (ITs) provide the right platform and resources for the 21st century waste management service delivery. Considering the growing need for large-scale complicated data storage, communication, analysis, and applications in connection with fast and parallel computational capabilities, IT is becoming more critical for IWM. In terms of monitoring, collection, transport, treatment and management, the scope of IT for IWM can be divided

734

W. Nwankwo et al.

Fig. 2. Various approaches to management of waste

into spatial technologies, identification technologies, data acquisition technologies and data communication technologies. 1. Spatial Technologies Spatial technologies are the most commonly used technologies in environmental modelling, and in many environmental studies, spatial analysis is of great importance. Such technologies are useful for managing complex spatial information and for providing mechanisms for the integration of various models, interfaces and subsystems—for instance, the integration of incineration and landfill. Spatial technologies may be categorized into three major groups: Geographic Information Systems (GIS), Global Positioning System (GPS) and Remote Sensing (RS) [61]. The essential functions of such technologies include the collection, storage, analysis and visualization of spatial data. A spatial dataset’s content can include attribute data, spatial topology, raster, features, and even network datasets. 2. Identification Technologies Researchers and organizations involved in WM have been studying different kinds of technology in recent decades to improve WM performance and automate bins collection. Several researchers have researched the possibility of implementing advanced WM systems that are based on identification technologies [62, 63]. WM systems are being improved through the use of identification technologies, such as barcode and RFID technologies as described in Tables 3 and 4 respectively.

Sector-Independent Integrated System Architecture

735

3. Data Acquisition Technologies With the advent and rapid growth of data acquisition technologies, manual data acquisition has been replaced by automated data acquisition due to its high performance, cheaper long-operating costs and lower demand for manpower [62]. Using advanced techniques, data collection could be enhanced by accurate and quantitative monitoring and interpretation of the targeted objects. Such information communication technologies are critical for applications where collecting data in real-time is an essential requirement [64, 65]). The data acquisition technologies are divided into two types: sensors, and imagery; and the scope of application depends on the particular WM application. Table 3. RFID-based WM System Authors

Strategies

Purpose

O’Connor [66]

RFID tag attached to bin floor, Collecting and transmitting tag truck-mounted RFID reader, information after emptying of Bluetooth enabled personal trash bins digital assistant (PDA), GPRS networking and remote server

Chowdhury and Chowdhury [67]

Intelligent RFID tag with an RFID reader in trash bin and Wi-Fi enabled Personal Digital Assistant (PDA)

Automatic capture of identity and weight of trash containers and promote identification of stolen containers

Pratheep and Hannan [68]

RFID tag built on bins and RFID reader with a vehicle-mounted antenna

Provides better visibility of the pick-up activities of drivers and track large trash receptacles

Gnoni et al. [69]

Trash bins and collection vehicles were fitted with an RFID tag and vehicle RFID reader

Facilitate growth of waste management services Pay as You Throw Billing

Nielsen et al. [70]

An RFID reader in a vehicle with a GPS receiver, tag in garbage bins and backend decision support device

Optimisation of waste disposal equipment and calculation of weight for billing purposes in some situations

Ali et al. [71]

RFID tag mounted in bin, Vehicle collection and solid black box mounted in vehicles waste bin control to simplify containing RFID receiver, waste management system camera, GPS receiver, GSM/GPRS module and remote server

Glouche et al. [72]

Smart bins fitted with an RFID Bin level effective waste reader, RFID tag with object handling for selective information about recyclable sorting and recycling materials for each waste object

736

W. Nwankwo et al.

4. Data Communication Technologies Rapid growth of networking systems and universal internet connectivity opens the possibilities for instant data transfer from remote locations. Wireless communication technologies that are used in WM applications include Global System for Mobile communications (GSM), General Packet Radio Service (GPRS), wireless fidelity (WiFi), Very High-Frequency Radio (VHFR for long-range communication) and Bluetooth (for short-range communication).

Table 4. Barcode-based WM System Authors

Strategies

Purpose

Li et al. [73]

Keeps track of materials and their status and combines with incentives to reduce waste

Reducing and handling the building of waste

Saar et al. [74]

Relates current waste material barcodes to websites via mobile phones to get information about disassembly

Gives the recyclers exact details about dismantling for automatic recycling

Greengard [75]

Barcodes are placed on waste items

Real-time monitoring of waste trajectory

Maruca et al. [76]

Scanning data from a waste item identifier and comparison with reference data to meet predetermined safety requirements

Determining and sorting banned waste products

3.1 Application of Artificial Intelligence (AI) in Managing Risks Associated with Industry Wastes The goal of AI in the management of risks associated with industrial wastes is to provide intelligent control by supporting the planning, scheduling, status monitoring, change initiation in components like sensors, analysis and reporting. These are done with the aid of the AI tools, which may include expert system, fuzzy logic, neural networks, decision support system and a hybrid expert network system [77–90]. The key opportunities for AI applications are in the area of multi-source data integration, including input data and timely and defensible recommendations. The use of AI in the management of risks with industrial wastes can generally be grouped into two areas: (a) decision support and (b) monitoring and reporting. The concept of AI in the decision support field will address the tools for burn plans and assist with the incinerator activities decision-making process. Whereas the monitoring and reporting system is aimed at providing monitoring status information as well as historical or predictive details and patterns. These components may be integrated into waste repositories, incinerators, scheduling and performance evaluation routines. The AI component could monitor pre-burn waste from inventories

Sector-Independent Integrated System Architecture

737

to incineration and secondary post-burn waste. Following are some of the AI applications in regards to risk management of industrial wastes: a. b. c. d. e. f. g. h.

Maintenance schedules Development and scheduling blends; waste feedstocks, etc. Approval/rejection of requests for incineration; Inventory control Scheduling industrial waste shipments; Evaluation of data processes and recommendations Data quality control Integrated performance assessments

3.2 Risk Management and Profiling in IWM Risk is the likelihood of obtaining unexpected or uncontrolled loss of value: this could be money, social status, environment variables, and a person’s physical health. [91]. Certain actions of man such as: rapid industrialization and urbanizations, are a trade-off for gaining or losing some of these values. Risks tend to exists in almost every area of human endeavor, hence the need for effective methods of risk management. Of substantial value to us in this study is health risk. Health risk reflects unforeseen chances of illnesses following direct and/or remote exposure to toxic environmental wastes. Closely related to health risks are safety and environmental risks which are consequences of hazardous biological and physico-chemical substances that negatively impacts the environment, consequently, environmental risk analysis (ERA) which seeks to study and understand actions and events that introduce risks to human health through the environment is a requirement during establishment of factories. Risk management involves the identification, evaluation, and prioritization of risks [92]. When every risk involved in a project has been identified and evaluated, then coordinated actions and wise utilization of resources can be applied to minimize and control the impact of unforeseen and undesirable occurrences. Effective risk management encompasses all strategies employed to manage all threats, which are uncertainties with negative consequences. The methods applied to risk management vary widely depending on the context of the risk as the strategies that would be employed in managing financial risk will be different from the industrial risks, however, generally, risk management involves avoidance of identified threats, reduction of the negative effect of the threat, reduction in the probability of the threat as well as transferring the majority of the threat to another party. Industrial and manufacturing activities result in the production of waste materials. Among several others are chemical solvents, industrial by-products, metals, and radioactive waste. Research has shown the direct relationship between population size and waste generation as a larger population tends to be associated with bigger industries and factories leading to the generation of larger amounts of industrial wastes. [93]. Solid wastes generated in the big manufacturing industries like lead and zinc batteries, detergent containers, and PVC contain cadmium which derives its toxicological properties from its chemical similarity to zinc is the most reported. These wastes especially those with cadmium content accumulate in the human body and causes harm to all the vital organs including the liver, kidneys, lungs, bones, the placenta, brain, and the central nervous system. Damage to human reproductive systems has also been reported as well as hematological and immunological damages [94].

738

W. Nwankwo et al.

These dire consequences and the threat to human lives and the environment have led to the need for the development of proper profiling and management of industrial wastes for efficient risk management as this will safeguard the environment and human lives by reducing the risk associated with industrial wastes [95]. In Nigeria, several legislations contain the objects of regulation and risk management for the protection of human lives through the reduction of pollution to air, water, land, and food [96]. Table 5. Risk Profiles of Industrial Wastes In Nigeria Industrial Sector

Waste Generated

Associated risks/hazards

Nigerian Oil and Gas Production

Drill mud, Glass, Metals, Plastic drums, Oil spills, Sludge, Chemicals

Inherently hazardous materials Water and Land Pollution Risk to Aquatic life Cancer causing substances from plant pollution

Nigerian Plastic and Polymer Producers

Accumulation of plastic objects, Polyethylene, Terephthalate (PET), High-Density Polyethylene (HDPE), Polyvinyl chloride (PVC), Polypropylene (PP)

Accumulation of plastic objects leading to land and water pollution Non-biodegradable ones causing drain blockages

Cement and Mining

Ore processing Slag, Phosphogypsum(from phosphoric acid production), Gasifier ash, Massive air pollution (from dust/sludge), Process wastewater, Calcium sulfate Chemical vapors

Air Pollution Respiratory diseases Lead Poisoning Ionizing radiation Increased Heat

Beverage Bottling Factories

Wastewater from, Left-over and discarded bottles, straws, etc

Land and water pollution

Pharmaceuticals

Cytotoxic waste materials, Land and water pollution Cytostatic waste materials, Expired pharmaceutical stock Kits for destroying controlled drugs, Recalled pharmaceutical stock

Paper and Pulp industry

Sludge from wastewater treatment plants Air emissions

Air Pollution Respiratory diseases

Sector-Independent Integrated System Architecture

739

3.3 Profiling Industries and Waste Generated Manufacturing and exploration industries at the point of manufacturing or delivering services accumulates wastes. The majority of these wastes pose adverse consequences [97–103]. Government policies and regulations have made it mandatory for these organizations to focus on the minimization of wastes in their production processes. Failures on part of these organization and the regulatory agencies have been clearly documented. It is submitted that without adequate mechanisms to ensure that every stakeholder contributes towards the objectives of such regulations would avail nothing. Hence, profiling risks/hazards prevalent in the environment and providing an “all-in-one” platform through which concerns may be communicated and addressed, is considered a worthy venture. Going forward, a sector-by-sector profiling of some industries in Nigeria as regards the wastes generated and the associated risks is presented in Table 5.

4 Integrated System for Risk and Hazard Profiling Having noted the various gaps in the workings of the various agencies, legislative and policy frameworks, we propose an integrated system for managing the risks associated with harmful wastes from industrial sources. This proposition is borne out of recent developments in AI, Big Data, and Internet of Things (IoT). These technologies are revolutionizing every facet of the socioeconomic sector. As have been noted, decision-making, efficiency, eco-friendliness, and sustainability are the pillars on which any system whether political or economic should anchor in these modern times. Authorities like NESREA and other agencies and organizations it collaborates with should be able to relate through a single window, which affords active collaboration unlike the existing passive system. Risk or hazard management is a collaborative effort and we submit that it is only a smart ICT-based system that can drive such a collaborative effort. All stakeholders must be recognized and taken into consideration while evolving the architecture for managing these hazards. The aforestated is entirely is true for risk management especially in the public sector where the majority of actions taken are on behalf and for the betterment of everyone. Transparency, accountability, and fairness must reflect in the architecture. Unlike the existing system wherein impact assessment is conducted by NESREA and the company/industry in question and approval granted without seeking inputs from the residents who live and do business around the area where such industry/company plans to implement a project, the proposed architecture is to be all-encompassing with all stakeholders involved and an unbounded platform provided for such stakeholders to relate directly with the operators on any development of mutual concern. The proposed architecture will require the documentation of all industries/companies registered with the corporate affairs commission whose businesses revolve around exploration, production, mining, fabrication, manufacturing, parts importation. As presented in Table 4 above, the risk profiles associated with the activities of the said industries would also be documented, as this would help during the modeling of a risk assessment

740

W. Nwankwo et al.

model that could be embedded in a more robust national integrated hazard management and planning system. Figure 3 presents the block structure of the proposed architecture. Four layers are recognized: a. b. c. d.

The Social layer Enterprise layer Logic and analytics layer Messaging layer

The Social layer would address the concerns of the ordinary people. For instance, this layer would present a platform for stakeholders (members of the public, experts, industry practitioners, etc.) to interact with the enterprise layer (government operatives). Inquiries, submissions, opinions, etc. could be made and tracked on this layer. The Enterprise layer would not exercise any powers as to deleting any event that occurs on this layer. The Enterprise layer incorporates the operations of the government operatives regardless of their placements in hierarchy. In other words, all collaborating agencies, ministries, departments would be incorporated into this layer with their identities maintained

Fig. 3. Proposed architecture of an Integrated System for Hazardous Wastes Management

Sector-Independent Integrated System Architecture

741

independently. This layer would afford real-time collaborations among these establishments. The supervising regulatory agency would be able to document and classify industries based on their risk potentials. The Logic and analytics layer would incorporate necessary risk management and profiling engine, robust database (of all companies as defined earlier). This layer provides intelligence to the entire system. It would receive inputs from stakeholders such as residents in areas where industries are located, expert opinions, etc. and would use such real-time information to re-compute the risk profiles of the industries involved and present such reports to all stakeholders. The Messaging layer would support seamless electronic data exchange between collaborating agencies. It would provide room for IoT applications wherein information from surveillance systems installed by the government agencies on various locations around high-risk industries could automatically relay such information to the central system over the Internet.

5 Conclusion In this paper, we present succinct details on IWM as it applies to a developing country like Nigeria with Lagos as a use case. We considered relevant literature applicable to the context of waste and environmental management. Sociological enquiries and real-life experiences also added to the plethora of materials utilized in this study. The various legislations and policy documents were examined. Documentation on the customary and more formal approaches to WM was articulated. The essence of the entire study is to present a coherent account on IWM as it obtains in Nigeria with a view to identifying the gaps and barriers to efficient, sustainable, and ecofriendly operations especially in respect of managing wastes and hazards therefrom [50, 52, 104–107]. Critical examination of WM practices revealed that managing wastes in urban cities poses a serious concern. Infrastructure deficits, maintenance culture, inadequate workforce, corruption, nepotism, were the major challenges identified. The aforementioned challenges, however, did not negate the existing submissions by other researchers that Nigeria has robust laws and policies on environmental management and protection. Nevertheless, the intended sustainability goal in respect of waste management is far from being realized [108, 109]. There appears to be a consistent decline in efficiency in the existing operational portfolio of the regulatory and enforcement agencies such as NESREA. There is absence of specialized infrastructure that provides a common platform to coordinate all stakeholders and activities in the waste management ecosystem. The authors conclude that while the identified gaps appear cumbersome, technological interventions could present a more sustainable and effective option. Going forward, we submit that all the challenges could be effectively addressed through an integrated system architecture that offers a data-driven single window platform. The proposed platform would be robust, support data provisioning from all stakeholders, real-time data analytics, real-time risk profiling of wastes at different locations, and forecasting. Accordingly, we recommend that the proposed architecture is necessary to drive the modernization of the EM and WM agencies in line with the sustainability development goals. More so, it will eliminate the existing skewness in decision-making (where every decision is almost taken by the

742

W. Nwankwo et al.

respective authority alone), present appropriate platform for community participation, and engender fairness in handling issues of collective concerns. Acknowledgment. We thank the assistance provided to us by officers of the lagos state waste management authority especially waste management reforms.

References 1. Asase, M., Yanful, E.K., Mensah, M., Stanford, J., Amponsah, S.: Comparison of municipal solid waste management systems in Canada and Ghana: a case study of the cities of London, Ontario, and Kumasi, Ghana. Waste Manage. 29, 2779–2786 (2009) 2. Rodgers, M.: Fundamentals of Development Administration. S.K. Publishers, London, United Kingdom (2011) 3. Osmani, M., Glass, J., Price, A.D.F.: Architects’ perspective on construction waste reduction by design. Waste Manage. 28, 1147–1158 (2008) 4. Ekanayake, L.L., Ofori, G.: Building waste assessments score: design based tool. Build. Environ. 39, 851–861 (2004) 5. Geng, Y., Fu, J., Sarkis, J., Xue, B.: Towards a national circular economy indicator system in China: an evaluation and critical analysis. J. Clean. Prod. 23, 216–224 (2012) 6. Preston, F.A.: Global Redesign: Shaping the Circular Economy. Chatham House, London, UK (2012) 7. Ghisellini, P., Cialani, C., Ulgiati, S.A.: Review on circular economy: the expected transition to a balanced interplay of environmental and economic systems. J. Clean. Prod. 114, 11–32 (2016) 8. Seo, S., Aramaki, T., Hwang, Y., Hanaki, K.: Environmental impact of solid waste treatment methods in Korea. J. Environ. Engineering 130, 81–89 (2004) 9. Kamalan, H., Sabour, M., Shariatmadari, N.A.: Review on available landfill gas models. J. Environ. Sci. Technol. 4, 79–92 (2011) 10. Den Hollander, M.C., Bakker, C.A., Hultink, E.J.: Product design in a circular economy: development of a typology of key concepts and terms. J. Ind. Ecol. 21, 517–525 (2017) 11. Ajie, U.E., Dienye, A.: Spatial data analysis of solid waste management system in Port Harcourt metropolis after 100 years of its existence. In: Proceedings of the FIG Congress (2014) 12. Van den Berg, M.R., Bakker, C.A.: A product design framework for a circular economy. Product. Lifetimes Env. 17, 365–379 (2015) 13. Shi, L., Xing, L., Bi, J., Zhang, B.: Circular economy: a new development strategy for sustainable development in China. In: 3rd World Congress of Environmental and Resource Economists, pp. 3–7 (2006) 14. Su, B., Heshmati, A., Geng, Y., Yu, X.: A review of the circular economy in China: moving from rhetoric to implementation. J. Clean. Prod. 1(42), 215–227 (2013) 15. Jun, H., Xiang, H.: Development of circular economy is a fundamental way to achieve agriculture sustainable development in China. Energy Procedia 1(5), 1530–1534 (2011) 16. Bartl, A.: Withdrawal of the circular economy package: a wasted opportunity or a new challenge? Waste Manag. (New York, NY). 44, 1–2 (2015) 17. Moreno, M., Braithwaite, N., Cooper, T.: Moving beyond the circular economy, p. 10 (2014) 18. Commission, E.: Directive 2008/98/EC of the European parliament and of the council on waste and repealing certain directives. Off. J. Eur. Union 312, 3–10 (2008)

Sector-Independent Integrated System Architecture

743

19. Wiles, C.C., Kosson, D.S., Holmes, T.: United States Environmental Protection Agency Municipal-Waste Combustion Residue Solidification/Stabilization Evaluation Program. Environmental Protection Agency, Cincinnati, OH (United States). Risk Reduction Engineering Lab. (1990) 20. Ezeigwe, C.: Appropriate solid waste disposal methods for developing countries. NSE Tech. Trans. 32(2), 33–34 (1995) 21. Ajani, O.I.: Determinants of an effective solid waste management in Ibadan metropolis, Oyo State, Nigeria. Int. J. Food Agric. Env. 6, 152–157 (2008) 22. Igoni, A.H., Ayotamuno, M.J., Ogaji, S.O., Probert, S.D.: Municipal solid-waste in Port Harcourt, Nigeria. Appl. Energy 84(6), 664–670 (2007) 23. Akanni, P.O.: An empirical survey of the effect of materials wastage on contractors’ profit level in construction projects. The Professional Builders. J. Niger. Instit. Build. 35–46 (2007) 24. Adewuyi, T.O., Idoro, G.I., Ikpo, I.J.: Empirical evaluation of construction material waste generated on sites in Nigeria. Civ. Eng. Dimension 16(2), 96–103 (2014) 25. Kareem, W.A., Asa, O.A., Lawal, M.O.: Resources conservation and waste management practices in construction industry. Oman Chapter Arab. J. Bus. Manag. Rev. 34(2603), 1–2 (2015) 26. Dania, A.A., Kehinde, J.O., Bala, K.: A study of construction material waste management practices by construction firms in Nigeria. In: Proceedings of the 3rd Scottish Conference for Postgraduate Researchers of the Built and Natural Environment, Glasgow, pp. 121–129 (2007) 27. LAWMA: Landfill https://www.lawma.gov.ng/inside-lawma/departments/landfill/ (2019) 28. Ogunmakinde, O.E., Sher, W., Maund, K.: An assessment of material waste disposal methods in the Nigerian construction industry. Recycling 4(1), 13 (2019) 29. LAWMA: Landfill—Monthly Dump Reports. http://www.lawma.gov.ng/insidelawma/dep artments/landfill/2014. Accessed on 20 May 2022 30. Obadina, A.: Solid Waste Management Livelihood on Lagos Dumpsite: Analysis of Gender and Social Difference. Ph.D. Thesis, Loughborough University, Loughborough, UK (2016) 31. Odewumi, S.G.: Appraisal of storage and collection strategies of municipal solid waste in Lagos State. J. Humanit. Soc. Sci. 10, 61–67 (2013) 32. Kofoworola, O.F.: Comparative assessment of the environmental implication of management options for municipal solid waste in Nigeria. Int. J. Waste Resour. 7(1), 1–5 (2016) 33. García, V., Pongrácz, E., Keiski, R.: Waste minimization in the chemical industry: from theory to practice. In: Proceedings of the Waste Minimization and Resources Use Optimization Conference, pp. 93–106, 10 Jun 2004 34. Ana, G.R., Sridhar, M.K., Olawuyi, J.F.: Air pollution in a chemical fertilizer complex in Nigeria: the impact on the health of the workers. J. Env. Health Res. 4(2), 57–62 (2005) 35. Armstrong, M.: A Handbook of Human Resource Management Practice. Kogan Page Publishers (2006) 36. Mallak, S.K., Elfghi, F.M., Rajagopal, P., Vaezzadeh, V., Fallah, M.: Overview of waste management performance of industrial sectors by selected Asian countries: current practices and issues. Int. Proc. Chem. Biol. Env. Eng. 99, 66–75 (2016) 37. Gana, A.J., Ngoro, D.: An investigation into waste management practices in Nigeria (a case study of lagos environmental protection board and Abuja environmental protection board). West Afr. J. Ind. Acad. Res. 12(1), 112–126 (2014) 38. Guerrero, L.A., Maas, G., Hogland, W.: Solid waste management challenges for cities in developing countries. Waste Manage. 33(1), 220–232 (2013) 39. Emejuru, C.T.: Human rights and environment: whither Nigeria. JL Pol’y Globalization 35, 108 (2015)

744

W. Nwankwo et al.

40. Gadzama, N.M., Ayuba, H.K.: On major environmental problem of desertification in Northern Nigeria with sustainable efforts to managing it. World J. Sci. Technol. Sustainable Dev. 13(1), 18–30 (2016) 41. Ijaiya, H., Joseph, O.T.: Rethinking environmental law enforcement in Nigeria. Beijing Law Rev. 05(04), 306–321 (2014) 42. Kankara, A.I., Adamu, G.K., Tukur, R., Ibrahim, A.: Examining environmental policies and laws in Nigeria. Int. J. Env. Eng. Manage. 4(3), 165–170 (2013) 43. Ladan, M.T.: Trend in Environmental Law and Access to Justice in Nigeria (2012) 44. World Health Organization: International Health Regulations, 3rd edn. Switzerland, Geneva (2016) 45. Etemire, U.: Insights on the UNEP Bali guidelines and the development of environmental democratic rights. J. Env. Law 28(3), 393–413 (2016) 46. Patwardhan, A.D.: Industrial solid wastes. The Energy and Resources Institute (TERI) (2013) 47. Pariatamby, A., Fauziah, S.H.: Sustainable 3R practice in the Asia and Pacific Regions: the challenges and issues. In: Municipal Solid Waste Management in Asia and the Pacific Islands: Challenges and Strategic Solutions, pp. 15–40 (2014). https://doi.org/10.1007/978981-4451-73-4_2 48. Zhang, D.Q., Tan, S.K., Gersberg, R.M.: Municipal solid waste management in China: status, problems and challenges. J. Environ. Manage. 91(8), 1623–1633 (2010) 49. May, J.R., Daly, E.: Global Environmental Constitutionalism. Cambridge University Press (2014). https://doi.org/10.1017/CBO9781139135559 50. Nwankwo, W., Olayinka, S.A., Ukhurebor, K.E.: Green computing policies and regulations: a necessity. Int. J. Sci. Technol. Res. 9(1), 4378–4383 (2020) 51. Ukhurebor, K.E., Aigbe, U.O., Olayinka, A.S., Nwankwo, W., Emegha, J.O.: Climatic change and pesticides usage: a brief review of their implicative relationship. AU eJ. Interdisc. Res. 5(1), 44–49 (2020) 52. Nwankwo, W., Ukhurebor, K.E.: An x-ray of connectivity between climate change and particulate pollutions. J. Adv. Res. Dyn. Control Syst. 11(8), 3002–3011 (2019) 53. Sengar, D.S.: Environmental Law Practices. Hall, New Delhi, India (2007) 54. NESREA: Establishment Act, CAP 164 Laws of the Federation of Nigeria (2007) 55. Naibbi, A.I., Mustapha, A.B.: Streamlining sustainability: environmental regulations in Nigeria: a mini review. Int. J. Env. Sci. Nat. Resour. 1(5), 142–144 (2017) 56. Akamabe, U.B., Kpae, G.: A critique on Nigeria national policy on environment: Reasons for policy review. IIARD Int. J. Geogr. Env. Manage. 3(3), 22–36 (2017) 57. Oh, J., Hettiarachchi, H.: Collective action in waste management: a comparative study of recycling and recovery initiatives from Brazil, Indonesia, and Nigeria using the institutional analysis and development framework. Recycling 5(1), 4 (2020) 58. Galeno, G., Minutillo, M., Perna, A.: From waste to electricity through integrated plasma gasification/fuel cell (IPGFC) system. Int. J. Hydrogen Energy 36(2), 1692–1701 (2011) 59. Zaman, A.U.: Comparative study of municipal solid waste treatment technologies using life cycle assessment method. Int. J. Environ. Sci. Technol. 7(2), 225–234 (2010) 60. Czajczy´nska, D., et al.: Potential of pyrolysis processes in the waste management sector. Thermal science and engineering progress. 1(3), 171–197 (2017) 61. Milla, K.A., Lorenzo, A., Brown, C.: GIS, GPS, and remote sensing technologies in extension services: where to start, what to know. J. Ext. 43(3), 1–8 (2005) 62. Palczynski, R.J., Scotia, W.N.: Study on Solid Waste Management Options for Africa. Project Report. Final Draft Version. Prepared for African Development Bank Sustainable Development and Poverty Reduction Unit, Abidjan (2002) 63. Nadaf, R.A., Katnur, F.A., Naik, S.P.: Android application based solid waste management. In: Pasumpon Pandian, A., Palanisamy, R., Ntalianis, K. (eds.) Proceeding of the International Conference on Computer Networks, Big Data and IoT (ICCBI – 2019), pp. 555–562.

Sector-Independent Integrated System Architecture

64. 65. 66. 67.

68. 69. 70.

71.

72. 73. 74. 75. 76.

77. 78. 79.

80.

81.

82.

745

Springer International Publishing, Cham (2020). https://doi.org/10.1007/978-3-030-431921_63 Faccio, M., Persona, A., Zanin, G.: Waste collection multi objective model with real time traceability data. Waste Manage. 31(12), 2391–2405 (2011) Ratnasabapathy, S., Perera, S., Alashwal, A.: A review of construction waste data and reporting systems used in Australia. 43RD AUBEA. 2019 Nov 6:396 O’Connor, M.: Greek RFID pilot collects garbage. RFID Journal, http://www.rfidjournal. com/article/view/2973. Accessed. 14:22 Jan 2007 Chowdhury, B., Chowdhury, M.U.: RFID-based real-time smart waste management system. In: 2007 Australasian Telecommunication Networks and Applications Conference, pp. 175– 180. IEEE, 2 Dec 2007 Pratheep, P., Hannan, M.A.: Solid waste bins monitoring system using RFID technologies. J. Appl. Sci. Res. 7(7), 1093–1101 (2011) Gnoni, M.G., Lettera, G., Rollo, A.: A feasibility study of a RFID traceability system in municipal solid waste management. Int. J. Inf. Technol. Manage. 12(1–2), 27–38 (2013) Nielsen, I., Lim, M., Nielsen, P.: Optimizing supply chain waste management through the use of RFID technology. In: 2010 IEEE International Conference on RFID-Technology and Applications, pp. 296–301. IEEE, 17 Jun 2010 Ali, M.L., Alam, M., Rahaman, M.A.: RFID based e-monitoring system for municipal solid waste management. In: 2012 7th International Conference on Electrical and Computer Engineering, pp. 474–477. IEEE 20 Dec 2012 Glouche, Y., Sinha, A., Couderc, P.: A smart waste management with self-describing complex objects. Int. J. Adv. Intell. Syst. 8, 63–70 (2015) Li, H., Chen, Z., Wong, C.T.: Barcode technology for an incentive reward program to reduce construction wastes. Comput.-Aided Civ. Infrastruct. Eng. 18(4), 313–324 (2003) Nabegu, A.B.: An analysis of municipal solid waste in Kano metropolis, Nigeria. J. Human Ecol. 31(2), 111–119 (2010) Iyer, M., et al.: Environmental survival of SARS-CoV-2–a solid waste perspective. Environ. Res. 1(197), 111015 (2021) Maruca, D., Leone, J., Cronin, N.E., inventors; Casella Waste Systems Inc, assignee. Systems and methods for identifying banned waste in a municipal solid waste environment. United States patent US 7,870,042. 11 Jan 2011 Nwankwo, W., et al.: Integrated fintech solutions in learning environments in the post-Covid19 era. IUP J. Knowl. Manage. 20(3), 22–43 (2022) Nwankwo, W., Nwankwo, C.P., Wilfred, A.: Leveraging on artificial intelligence to accelerate sustainable bioeconomy. IUP J. Knowl. Manage. 20(2), 38–59 (2022) Acheme, I.D., Makinde, A.S., Udinn, O., Nwankwo, W.: An intelligent agent-based stock market decision support system using fuzzy logic. IUP J. Inform. Technol. 16(4), 7–25 (2020) Nwankwo, W., Adetunji, C.O., Olayinka, A.S.: IoT-driven bayesian learning: a case study of reducing road accidents of commercial vehicles on highways. In: Artificial Intelligencebased Internet of Things Systems, pp. 391–418 (2022). https://doi.org/10.1007/978-3-03087059-1_15 Nwankwo, W., Ukhurebor, K.E.: Big data analytics: A single window IoT-enabled climate variability system for all-year-round vegetable cultivation. IOP Conf. Ser.: Earth Env. Sci. 655(1), 012030 (2021) Nwankwo, W., Ukhurebor, K., Ukaoha, K.: Knowledge discovery and analytics in process reengineering: a study of port clearance processes. In: 2020 International Conference in Mathematics, Computer Engineering and Computer Science (ICMCECS), pp. 1–7. IEEE, 18 Mar 2020

746

W. Nwankwo et al.

83. Chinedu, P.U., Nwankwo, W., Masajuwa, F.U., Imoisi, S.: Cybercrime detection and prevention efforts in the last decade: an overview of the possibilities of machine learning models. Rev. Int. Geo. Educ. Online 11(7), 956–974 (2021) 84. Nwankwo, W., et al.: The adoption of AI and IoT technologies: socio-psychological implications in the production environment. IUP J. Knowl. Manag. 19(1), 50–75 (2021) 85. Olayinka, A.S., Adetunji, C.O., Nwankwo, W., Olugbemi, O.T., Olayinka, T.C.: A study on the application of bayesian learning and decision trees IoT-enabled system in postharvest storage. In: Pal, S., De, D., Buyya, R. (eds.) Artificial Intelligence-based Internet of Things Systems, pp. 467–491. Springer International Publishing, Cham (2022). https://doi.org/10. 1007/978-3-030-87059-1_18 86. Anani, O.A., Adetunji, C.O., Olugbemi, O.T., Hefft, D.I., Wilson, N., Olayinka, A.S.: IoTbased monitoring system for freshwater fish farming: analysis and design. In: AI, Edge and IoT-based Smart Agriculture, pp. 505–515. Elsevier (2022). https://doi.org/10.1016/B9780-12-823694-9.00026-8 87. Adetunji, C.O., Anani, O.A., Olugbemi, O.T., Hefft, D.I., Wilson, N., Olayinka, A.S.: Toward the design of an intelligent system for enhancing salt water shrimp production using fuzzy logic. In: AI, Edge and IoT-based Smart Agriculture, pp. 533–541. Elsevier (2022). https:// doi.org/10.1016/B978-0-12-823694-9.00005-0 88. Adetunji, C.O., et al.: Machine learning and behaviour modification for COVID-19. In: Inuwa, H.M., et al. (eds.) Medical Biotechnology, Biopharmaceutics, Forensic Science and Bioinformatics, pp. 271–287. CRC Press, Boca Raton (2022). https://doi.org/10.1201/978 1003178903-17 89. Nwankwo, W., Ukhurebor, K.E.: Web forum and social media: A model for automatic removal of fake media using multilayered neural networks. Int. J. Sci. Technol. Res. 9(1), 4371–4377 (2020) 90. Makinde, A.S., Agbeyangi, A.O., Nwankwo, W.: Predicting mobile portability across telecommunication networks using the integrated-KLR. Int. J. Intell. Inform. Technol. (IJIIT) 17(3), 50–62 (2021). https://doi.org/10.4018/IJIIT.2021070104 91. Kiran, S., Kumar, K.P., Sreejith, B., Muralidharan, M.: Reliability evaluation and risk based maintenance in a process plant. Procedia Technol. 1(24), 576–583 (2016) 92. Jerie, S.: Occupational risks associated with solid waste management in the informal sector of Gweru, Zimbabwe. J. Environ. Public Health 21, 2016 (2016) 93. Sanders, A.P., Flood, K., Chiang, S., Herring, A.H., Wolf, L., Fry, R.C.: Towards prenatal biomonitoring in North Carolina: assessing arsenic, cadmium, mercury, and lead levels in pregnant women. PLoS ONE 7(3), e31354 (2012) 94. Aluko, O.O., Adebayo, A.E., Adebisi, T.F., Ewegbemi, M.K., Abidoye, A.T., Popoola, B.F.: Knowledge, attitudes and perceptions of occupational hazards and safety practices in Nigerian healthcare workers. BMC. Res. Notes 9(1), 1–4 (2016) 95. Nwufo. C.: Legal framework for the regulation of waste in Nigeria. Afr. Res. Rev. 4(2), 491-501 (2010) 96. Yunusa, U., Lawal, U.B., Idris, A., Garba, S.N.: Occupational health hazards among commercial motorcyclists in Ahmadu Bello University, Zaria. IOSR-JNHS. 3(17), 46–52 (2014) 97. Afube, G.C., Nwaogazie, I.L., Ugbebor, J.N.: Identification of industrial hazards and assessment of safety measures in the chemical industry, Nigeria using proportional importance index. Arch. Current Res. Int. 19(1), 1–5 (2019) 98. Agagu, O.: Threats to the Nigerian environment: a call for positive action. Nigerian Conservation Foundation (2009) 99. Kalu, C., Modugu, W.W., Ubochi, I.: Evaluation of solid waste management policy in Benin metropolis, Edo State, Nigeria. Afr. Sci. 10(1), 117–125 (2009)

Sector-Independent Integrated System Architecture

747

100. Oyeniyi, B.A.: Waste management in contemporary Nigeria: the Abuja example. Int. J. Politics Good Governance 2(2.2), 1–8 (2011) 101. Lu, J.W., Chang, N.B., Liao, L.: Environmental informatics for solid and hazardous waste management: advances, challenges, and perspectives. Crit. Rev. Environ. Sci. Technol. 43(15), 1557–1656 (2013) 102. Manu, P., Poghosyan, A., Mshelia, I.M., Iwo, S.T., Mahamadu, A.M., Dziekonski, K.: Design for occupational safety and health of workers in construction in developing countries: a study of architects in Nigeria. Int. J. Occup. Saf. Ergon. 25(1), 99–109 (2019) 103. Maphosa, V.: Sustainable e-waste management at higher education institutions’ data centres in Zimbabwe. Int. J. Inform. Eng. Electron. Bus. (IJIEEB) 14(5), 15–23 (2022) 104. Latif, S.D.: Technical improvement of air pollution through fossil power plant waste management. Int. J. Eng. Manuf. 10(4), 43–53 (2020). https://doi.org/10.5815/ijem.2020. 04.04 105. Nwankwo, W., Ukhurebor, K.E.: Data Centres: a prescriptive model for green and ecofriendly environment in the cement industry in Nigeria. Int. J. Sci. Technol. Res. 9(5), 239–244 (2020) 106. Nwankwo, W., Olayinka, A.S., Ukhurebor, K.E.: The urban traffic congestion problem in Benin city and the search for an ICT-improved solution. Int. J. Sci. Technol. 8(12), 65–72 (2019) 107. Nwankwo, W., Olayinka, A.S., Etu, M.: Management of poultry waste in Edo state: a case study of Etsako district. Niger. Res. J. Eng. Env. Sci. 4(2), 893–902 (2019) 108. Nwankwo, W., Kifordu, A.: Strengthening private sector participation in public infrastructure projects through concession policies and legislations in Nigeria: a review. J. Adv. Res. Dyn. Control Syst. 11(8), 360–1370 (2019). special issue 109. Nwankwo, W., Chinedu, P.U.: Green computing: a machinery for sustainable development in the post-covid era. In: Sabban, A. (ed.) Green Computing Technologies and Computing Industry in 2021. IntechOpen (2021). https://doi.org/10.5772/intechopen.95420

Design and Development of a Hybrid Eye and Mobile Controlled Wheelchair Prototype using Haar cascade Classifier: A Proof of Concept Yusuf Kola Ahmed1,2(B) , Nasir Ayuba Danmusa1 , Taofik Ahmed Suleiman3 , Kafilat Atinuke Salahudeen1 , Sani Saminu1 , Abdul Rasak Zubair4 , and Abdulwasiu Bolakale Adelodun5 1 Department of Biomedical Engineering, University of Ilorin, Ilorin, Nigeria 2 Department of Occupational Therapy, University of Alberta, Edmonton, Canada

[email protected]

3 Department of Medical Imaging and Computing, University of Girona, Girona, Spain 4 Department of Electrical and Electronic Engineering, University of Ibadan, Ibadan, Nigeria

[email protected] 5 ECWA College of Nursing and Midwivery, Egbe, Kogi State, Nigeria

Abstract. According to the wheelchair foundation, about 1.86% of the world’s population requires a functional wheelchair. Most of these wheelchairs have manual control systems which puts millions of people with total paralyzes (total loss of muscle control including the head) at a disadvantage. However, the majority of those who suffer from muscular and neurological disorders still retain the ability to move their eyes. Hence the concept of eye-controlled wheelchair. This paper focused on the design and development of a hybrid control system (eye and mobile interface) for a wheelchair prototype as a proof of concept. The system was implemented using the pre-trained Haar cascade ML classifier in open CV. Focus was shifted from high accuracy common to lab-based studies to deployment and power consumption which are critical to usability. The system consists of a motor chassis that takes the place of a wheelchair, a raspberry pi4 module which acts as a mini-computer for image and information processing, and a laser sensor to achieve obstacle avoidance. The Bluetooth module enables serial communication between the motor chassis and the raspberry pi, while the power supply feeds the raspberry pi and the camera. The system performance evaluation was carried out using obstacle avoidance and navigation tests. An accuracy of 100% and 89% were achieved for obstacle avoidance and navigation, respectively, which shows that the system would be helpful for wheelchair users facing autonomous mobility issues. Keywords: Paralysis · Wheelchair · Eye control · Obstacles avoidance · Camera

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 748–758, 2023. https://doi.org/10.1007/978-3-031-36118-0_66

Design and Development of a Hybrid Eye and Mobile Controlled Wheelchair Prototype

749

1 Introduction Paralysis is a total or partial loss of physiological functions in some part of the human body [1]. Paralysis is often characterized by either motor or sensory loss or both in the affected area. Common causes of paralysis include Stroke, Multiple sclerosis, Spinal cord injury, Cerebral palsy among others [1, 2]. An age-adjusted comparison of results from the 1980 and 1990, National Health Interview Survey (NHIS) found a 26% increase in the use of canes and a 57% and 65% increase in the use of walkers and wheelchairs, respectively, among all ages which increased to 83% by 1994 [3]. Besides, the latest study has reported that 1.86% of the world population need functional wheelchair [4]. The design of wheelchairs for paralyzed patients has witnessed advances from conventional manually powered to electrical wheelchairs [5]. The population of elderly citizens in the world is also increasing, with this increase, there is a need to design welfare robotic devices for such people. A conventional wheelchair tends to focus exclusively on manual use which assumes users are still able to use their hands and excludes those unable to do so. Many of those suffering from complete paralysis usually still retain the ability to control eye movements, which inspired the development of an eye-controlled electric wheelchair. There are in existence different modes of eye-controlled wheelchair such as EOG – control, gaze estimation control and video based control systems. However, most of them only have one control system (the eye) and few who incorporated an auxiliary control system employed manual gear steering (joy stick) which requires some form of energy and skill to use. In order to increase versatility and ease of use, a dual-controlled eye tracking system for wheelchair navigation is proposed, which uses eye gaze as the primary control system and mobile interface as an auxiliary. The proposed system uses an off-the-shelf webcam for eye image acquisition using Open CV/CV2 library to process the signal in a raspberry pi 4 device which has improved functions and uses Linux based operating system. Going by the ubiquity of mobile phone users, the auxiliary control module is implemented on a mobile phone interface such that instead of a joystick, the users’ phone is connected by Bluetooth to the Arduino Uno control board to steer the motors of the wheelchair. Hence, increasing the system’s usability for patients with either complete or partial paralysis and the elderly. A wheelchair prototype is employed in the form of a robotic car, and the obstacle avoidance mechanism is achieved using a laser sensor 1.1 Literature Review Over the years, research works have been using different control applications on wheelchair systems aside from manual control approaches. The common physiological signals often adopted include Electrooculography (EOG) signal, Electromyography (EMG) signal, and Electroencephalography (EEG) signal [2, 4]. The EOG signal and video-based control systems have been extensively studied for eye control applications in a wheelchair [6]. When the eye moves, the EOG signal is generated from the changing potential difference between the cornea and the retina. This change is registered as a potential difference useful in detecting the location or gaze of the eye. This system often requires the direct attachment of electrodes to the face/forehead of the user, as in Fahim et al. [1]. Some EOG-based studies developed better eye movement tracking algorithms

750

Y. K. Ahmed et al.

and wheelchair control [7–9]. [1, 9, 10] adopted EOG and EMG signals in the control system design, while Naeem and Qadar [11] used voice control. The major drawback of EOG-based design is that it often constitutes a physical burden on the user besides the heavy noise attributed to the signal [4] and the high level of cognition required. Voice control is often plagued with interference in a sound wave from the environment coupled with its high voice energy demands [11]. A few works on video-based eye control include the work of Mani et al. [12]. His work processed the image of the eye captured by a head-mounted camera to activate the direction of motion in 3 dimensions (left, right, and forward). However, the performance evaluation of the system was not wholly reported. Another design by Patel et al. [13] was built around Raspberry pi using an open CV library for image processing of the acquired face image for eye location. The system activates and deactivates in 3 seconds; eye closure and directional motion are initiated with the eye’s blinking. The procedure was adequately explained, but the system’s performance was not assessed using accuracy and response latency which are vital tools for such systems evaluation. Lately, machine learning algorithms/models [14] have been employed in the image processing of eye-controlled wheelchairs with a significant increase in accuracy [15, 16]. George et al. [16] implemented a convolution neural network (CNN) model using a webcam image for real-time gaze classification in seven directions: right, left, center, forward, upright, up left, down left, and down right. Localization of facial regions was carried out by a modified Viola-Jones algorithm. The CNN model for combined eyes (left and right eye) was observed to yield better accuracy than training each eye separately. It gave good accuracy, but it’s computationally complex. Krafka et al. [15] also proposed an eye tracking system using deep end-to-end CNN with a larger data set obtained from 1450 participants. To check computational time and high complexity, dark knowledge was used with a simpler network, however, the frame rate is less than the standard 20-25 FPS needed for smooth real-time application.

2 Methodology 2.1 Material Material selection was a key part of this project. Before the development of the project, certain factors were considered for the design to meet technical specifications while minimizing cost and increasing easy movement [17, 18]. The important materials selected are the Web camera, Raspberry pi board, Arduino Uno, H-bridge Motor driver (L298), Bluetooth module, and Motor chassis. 2.2 System Design and Description The system is divided into three modules: the image data acquisition module, the wheelchair control/steering module, and the wheelchair. Fig. 1 shows the block diagram of the eye-controlled wheelchair system. The web camera captures the eye image data which is sent to the Raspberry-pi. The Raspberry-pi serves as the mini processor

Design and Development of a Hybrid Eye and Mobile Controlled Wheelchair Prototype

751

for synthesizing the pupil location [19]. The location identified by the raspberry–pi is fed to the Arduino Uno board where different eye positions are programmed into left, right, and forward movement commands. These commands are fed to the motor driver of the wheelchair for navigation in the desired direction. The auxiliary control module involves the mobile App with a control interface connected by Bluetooth to the Arduino board [20]. This can also be used to navigate the wheelchair aside the eye control. The laser sensor feedback mechanism for obstacle avoidance takes care of system safety.

Fig. 1. System block diagram

2.2.1 Image Data Acquisition System An off-the-shelf raspberry pi camera is wired to the raspberry pi board with an extended cable. The camera is placed at a distance of 5cm from the face and at eye level such that it can capture the facial images from which eye/pupil location would be extracted. The dynamics of navigation proposed for the wheelchair includes forward, left, right, and backward movement with a start and stop process initiated by the blinking of the eye. The first left eye winking starts the image acquisition and subsequent movement of the wheelchair, while any second left eye winking would automatically stop data acquisition and, subsequently, the wheelchair stops, as implemented in [4]. Hence, four (4) eye states are employed in the navigation process. Haar cascade was used in the gaze tracking algorithm for pupil detection because of its simplicity and fast processing time, making it useful in real-time applications. This method uses the concept of features and classifiers in machine learning (ML) for its detection scheme. The fundamental idea behind Haar cascades is that they employ a collection of salient "features" that can be computed from an image, including the intensity, difference between neighboring pixels or tiny areas of the image [21–24]. The Haar-like features in Haar cascade are calculated using the following equation: f (x, y) =

x+w−1 y+h−1 i=x

j=y

I (i, j)

(1)

752

Y. K. Ahmed et al.

where I(i,j) represents the intensity of the pixel located at position (i,j). The Haar-like feature is calculated as the difference between the sum of intensity of pixels in two rectangular regions, defined as: f (x, y) =

x+w/2−1 y+h−1 i=x

j=y

I (i, j) −

x+w/2 y+h−1 i=x

j=y

I (i, j)

(2)

The Haar cascade uses an Adaboost classifier, which is a type of machine learning classifier that combines multiple simple classifiers to form a strong classifier. The Adaboost classifier makes its decision using the following equation:  T  1 t=1 ∝t ht (x) ≥ 21 Tt=1 ∝t (3) h(x) = 0 otherwise where T is the number of weak classifiers, ∝t represents the weight associated with each weak classifier, ht (x) represents the decision made by the tth weak classifier, and sign(x) is the sign function, which returns 1 if x is positive and -1 if x is negative A classifier is therefore trained using these characteristics to differentiate between the pupil and the background. The trained classifier is used by the Haar cascade to assess the characteristics of several eye regions, including the eye lid, iris, and cornea, and determine whether or not the eye frame is likely to contain the object of interest. This procedure is repeated for several eye regions and points most likely to contain the pupil is marked "detection". However, the frame illumination often changes with the nature of the environment; hence Histogram Equalization is used to change the probability distribution of the image to a uniform distribution for better image contrast [25]. 2.2.2 Wheelchair Control/Steering System The decision/output of the pupil data acquisition system is used to navigate the wheelchair. These decisions are coded in form of left, right, forward, backward, and start/stop modes such that when fed to the H-bridge motor driver, it steers according to the code in the received signal. 2.2.2.1 Mobile Control Interface The primary control system for the wheelchair prototype is the eye gaze detection system. However, a secondary control system is incorporated in an RC-controlled mobile interface to improve the versatility of the device for both tetraplegic and non-tetraplegic patients, such as the elderly. This uses a Bluetooth module to communicate between the app interface and the H-bridge Arduino motor driver to achieve motion control. The H-bridge motor driver is responsible for steering the wheels’ directions. An open-access RC controller program was adapted to achieve forward, backward, left, and right movements and other sub-directions (Fig.3). Two DC motors were installed in the prototype and linked with the front wheels. 2.2.3 Safety Subsystem: Laser Sensors For safety precautions, the wheelchair is designed to have an automated stopping mechanism that will disconnect the gaze-based controller and stop the chair as soon as it

Design and Development of a Hybrid Eye and Mobile Controlled Wheelchair Prototype

753

approaches any nearby object. A proximity sensor quantitively measures how close an object is to the sensor. It has a range, and a flag is raised whenever an object enters this threshold region. The sensor sends out electromagnetic waves, which are picked up after being reflected off nearby objects. The procedure is quite similar to how radar works. A laser proximity sensor is adopted in this work. 2.2.4 Mobile Application The Bluetooth RC controller is a mobile application that connects with the Bluetooth attached to the wheelchair (prototype) and is used to send commands serially to the Bluetooth, which in turn controls the motors, as an alternate method of control for partially paralyzed subjects (lower limbs) to aid easy control of the wheelchair. Fig. 2 (c) shows the Bluetooth RC controller mobile application interface. 2.2.5 Wheelchair A motor chassis serves as a prototype for a wheelchair and necessary components and modules are attached to it. 2.3 Overall Design Outlook The various modules are interconnected to form the complete system. Fig. 2 describe the complete system in more details. The device operational flow chart is shown in Fig. 3. The web camera picks the face and eye image frames and the algorithm checks if the eye is open or closed. If closed, the loop returns and check again for eye status. If open, the gaze direction is identified by template matching and accurate direction is detected before obstacle avoidance is checked. The obstacle avoidance is set at a threshold (stopping

Fig. 2. Pictorial description of the system

754

Y. K. Ahmed et al.

distance) of 1m as used in [4]. If an obstacle is detected at a distance ~ 90%) according to the metric results.

Method of Automatic Depersonalization of Databases for Application

783

Fig. 8. Data preparation

Fig. 9. Training and obtaining predictions

3.3 Determining the Price of Health Insurance (Regression Problem) The next task is to predict the price of health insurances. The dataset was taken from Kaggle [9]. This time we are dealing with continuous values since this is a regression problem. The data structure is presented below (Fig. 10).

784

A. Volokyta and V. Lipska Table 1. Gender classification problem results

Data (number)

Accuracy

Precision

Recall

F1 score

False positive rate

Original

0.940279

0.933133

0.881493

0.906578

0.118507

Generated 1 (+200)

0.978135

0.976032

0.890736

0.931435

0.109264

Generated 2 (+500)

0.968941

0.963491

0.871173

0.915009

0.128826

Generated 3 (+2000)

0.959172

0.958912

0.847345

0.899682

0.152655

Fig. 10. The task data structure

The data preparation algorithm is similar to the previous problem, therefore, the method, prediction of values and metrics are considered in more detail in the next image (because they differ from the previous example) (Fig. 11).

Fig. 11. Training and obtaining predictions

Method of Automatic Depersonalization of Databases for Application

785

The obtained results are recorded in a table and can be analyzed. Table 2. Results of the algorithm usage for regression problem Data

R2 score

Explained variance score Mean absolute percentage error

Original

0.854321 0.865893

0.284498

Generated 1 (+500)

0.873433 0.873496

0.263907

Generated 2 (+800)

0.865672 0.865871

0.267938

Generated 3 (+2000) 0.839087 0.841002

0.309629

Table 2 lists the metrics used for regression problems. By definition, the more successful the model is, the closer R2 and EVS are to 1 and MAPE to 0. The similarity of the indicators confirms the preservation of the nature of the data and their homogeneity. All datasets showed good metric values. Better performance of artificial data sets can occur due to the absence of exceptions or damaged objects as determined by the artificial data set construction algorithm. Thus, artificial sets constructed by the proposed method can be used for regression problems, thereby allowing depersonalization and data synthesis.

4 Summary and Conclusion This research proposes, in addition to known static extrapolations and data substitutions, to generate artificial data sets based on Bayesian models. It allows to automatically determine the nature of the data and generate artificial depersonalized data, the effectiveness of which has been practically confirmed. The steps in the algorithms were described as follows: 1) Remove noises. 2) Update values with static replaces (indicated values, minimum/maximum, changes using a regular expression pattern/certain delta, scale ranges, take into account the distribution and the step of values). 3) Select feature X. 4) Calculate probabilities for every parameter of X-feature. 5) Select feature Y (Y ! = X). 6) Calculate P(Y|X) for each value of Y-feature. 7) Add probabilities and values to the network as nodes and edges. 8) Return to step 5 and repeat for each feature except processed ones. 9) Return to step 3 and repeat for each feature except processed ones.

786

A. Volokyta and V. Lipska

References 1. Big & Personal: data and models behind Netflix recommendations. URL: https://xamat.git hub.io/pubs/BigAndPersonal.pdf 2. Geron, A.: Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems, 2nd Edition, p. 57 (2019) 3. GDPR. URL: https://en.wikipedia.org/wiki/General_Data_Protection_Regulation 4. If an app asks to track your activity. URL: https://support.apple.com/en-us/HT212025 5. Shai, S.-S., Shai, B.-D.: Understanding Machine Learning: From Theory to Algorithms (2014). URL: https://www.cs.huji.ac.il/~shais/UnderstandingMachineLearning/und erstanding-machine-learning-theory-algorithms.pdf 6. Ng, A.: Machine Learning by Stanford University, Coursera. URL: https://www.coursera.org/ learn/machine-learning 7. Deep Learning – An MIT Press Book. URL: https://www.deeplearningbook.org/ 8. Beyond Accuracy: Precision and Recall. URL: https://towardsdatascience.com/beyond-acc uracy-precision-and-recall-3da06bea9f6c 9. Kaggle (Gender Classification Dataset). URL: https://www.kaggle.com/hb20007/gender-cla ssification 10. Scale Synthetic. URL: https://scale.com/synthetic 11. Depersonalizing Your Test Data. URL: https://github.com/CleverComponents/Data-Depers onalizer 12. Synth. URL: https://www.getsynth.com/docs/getting_started/synth 13. Twinify. URL: https://github.com/DPBayes/twinify 14. What Is Synthetic Data? URL: https://blogs.nvidia.com/blog/2021/06/08/what-is-syntheticdata/ 15. OpenAI (ChatGPT). URL: https://openai.com/research/overview 16. ScienceDirect. URL: https://www.sciencedirect.com/topics/computer-science/caesar-cipher 17. Malani, R., Putra, A.B.W., Rifani, M.: Implementation of the naive bayes classifier method for potential network port selection. Int. J. Comp. Netw. Info. Secu. (IJCNIS) 12(2), 32–40 (2020) 18. Pandey, A., Jain, A.: Comparative analysis of KNN algorithm using various normalization techniques. Int. J. Comp. Netw. Info. Secu. (IJCNIS) 9(11), 36–42 (2017) 19. Zimba, A.: A bayesian attack-network modeling approach to mitigating malware-based banking cyberattacks. Int. J. Comp. Netw. Info. Secu. (IJCNIS) 14(1), 25–39 (2022) 20. Dasari, K.B., Devarakonda, N.: Detection of DDoSAttacks Using Machine Learning Classification Algorithms. Int. J. Comp. Netw. Info. Secu. (IJCNIS) 14(6), 89-97 (2022) 21. Bayes Server Learning Center. URL: https://www.bayesserver.com/docs/introduction/bay esian-networks/ 22. Horny, M.: Bayesian Networks N.5 (2014). https://www.bu.edu/sph/files/2014/05/bayesiannetworks-final.pdf 23. What are classification problems? URL: https://www.educative.io/answers/what-are-classific ation-problems 24. Oracle AI & Data Science Blog. URL: https://blogs.oracle.com/ai-and-datascience/post/asimple-guide-to-building-a-confusion-matrix 25. What is a good F1 score and how do I interpret it? URL: https://stephenallwright.com/goodf1-score/ 26. Lipton, Z.C., Elkan, C., Naryanaswamy, B.: Thresholding Classifiers to Maximize F1 Score. URL: https://arxiv.org/pdf/1402.1892.pdf 27. Metrics for evaluating regression models. URL: http://www.enlistq.com/top-5-metrics-eva luating-regression-models/

Method of Automatic Depersonalization of Databases for Application

787

28. Mean Absolute Percentage Error. URL: https://www.statisticshowto.com/mean-absolute-per centage-error-mape/ 29. de Myttenaerea, A., Goldena, B., Le Grand, B., Rossi, F.: Mean Absolute Percentage Error for Regression Models. URL: https://arxiv.org/pdf/1605.02541.pdf 30. What is MAPE? URL: https://www.indeed.com/career-advice/career-development/what-ismape 31. RandomForestClassifier. URL: https://scikit-learn.org/stable/modules/generated/sklearn.ens emble.RandomForestClassifier.html 32. RandomForestRegressor. URL: https://scikit-learn.org/stable/modules/generated/sklearn.ens emble.RandomForestRegressor.html 33. ChatGPT is a data privacy nightmare. URL: https://theconversation.com/chatgpt-is-a-dataprivacy-nightmare-if-youve-ever-posted-online-you-ought-to-be-concerned-199283 34. ChatGPT vs GDPR – what AI chatbots mean for data privacy. URL: https://www.inform ation-age.com/chatgpt-vs-gdpr-what-ai-chatbots-mean-for-data-privacy-123501570/

Mathematical Modeling of States of a Complex Landscape System with a Wind Farm Taras Boyko1 , Mariia Ruda2 , Elvira Dzhumelia3 , Orest Kochan4,5(B) , and Ivan Salamon6 1 Department of Intellectual Mechatronics and Robotics, Lviv Polytechnic National University,

Lviv 79013, Ukraine 2 Department of Ecological Safety and Nature Protection Activity, Lviv Polytechnic National

University, Lviv 79013, Ukraine 3 Department of Software, Lviv Polytechnic National University, Lviv 79013, Ukraine 4 Department of Measuring Information Technologies, Lviv Polytechnic National University,

Lviv 79013, Ukraine [email protected] 5 Department of Automation, Lublin University of Technology, 20618 Lublin, Poland 6 Department of Ecology, Faculty of Humanities and Natural Sciences, University of Presov, Presov 080 01, Slovakia

Abstract. The aim of this study is to model the processes of mutual influence of the wind turbine and the environment. An elementary structural element of an ecosystem is a compartment of a complex landscape system in which a wind turbine is considered during its life cycle. The task of creating a ‘cyber-twin’ of a wind turbine, as a technogenic object of the compartment of the complex landscape system, was realized by mathematical modeling of the states of the layers and subsystems of the compartment of the complex landscape system under integrated impact. Impacts are pollutants and harmful factors, primarily of anthropogenic origin, the values of which were estimated by simulation modeling in the experimental part of this study. As a result of mathematical modeling a system of differential equations was obtained, the input data for which were the values of environmental impacts, expressed by the specified indicators. The resulting model will act as ideal for a real system ‘wind turbine-environment’ and will allow predicting the consequences of harmful impact of a wind turbine on a complex landscape system. Keywords: environment · landscape complex · renewable energy sources · ecological indicators · harmful ecological influence · cyber-physical system

1 Introduction In recent times, there has been a surge in the adoption of innovative energy technologies that are aimed at developing efficient power grids (Smart Grid), advanced accounting and settlement systems (Smart Metering), managing energy demands (Demand Response), and energy storage solutions. These technologies are designed to address the increasing © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 788–802, 2023. https://doi.org/10.1007/978-3-031-36118-0_69

Mathematical Modeling of States of a Complex

789

demand for reliable and high-quality electricity, promote energy efficiency, and enhance the sustainability of the environment [1, 2]. However, despite the extensive research conducted by ecologists, the impact of wind energy facilities on the environment is not adequately explored and remains largely unaddressed [3–7]. When we consider the impact of human activities on the environment, we can view each component as a cyber-physical system. This perspective is possible when we use computer algorithms to monitor the integration mechanisms. The development of industrial cyber-physical systems typically involves five standard processes, one of which is the ‘cyber-implementation’ process. This process involves creating a ‘cyber-twin’ to describe the system’s state and comparing it with the real device before synthesizing the entire system. The purpose of the study is to create a “cyber-twin” for a wind turbine to simulate and predict the states of an individual complex landscape system (CLS), in which a wind farm operates. The object of the study is the process of modeling the man-made load of the wind turbine on the environment during its life cycle. The subject of the study is a model of a system that includes a complex landscape system – a biological system that combines biotic and abiotic components. For example, by definition the Eastern flysch Carpathians (the Borzhava Polonyna) belong to a CLS [8–19].

2 Literature Review There is a shortage of life cycle assessment (LCA) studies for wind turbines with a highrated power of 600 kW. Although some studies are available [20, 21], they have varying scopes, but they consistently highlight the significant impact of the manufacturing and usage of various materials on the ecological metrics of wind farms. These factors must be taken into account when assessing the adverse effects of wind turbines on the ecosystem. Some estimates also point to a large amount of indirectly generated waste [22], which must also be considered. In the works [8, 23] it is shown that the interaction between wind turbines and the CLS is advisable to consider based on the compartment concept of the CLS. Similar results are presented in the works [24]. The authors conclude that the construction of the corresponding model involves the interpretation of the CLS as a system characterized by a certain composition and structure which is a natural combination of natural components – landscape systems of different taxonomic rank [25, 26]. The CLS, or complex landscape system, in which wind turbines operate is characterized by the structural and functional unity of the interconnected components and the integrity of both biotic and abiotic elements [27–34]. The authors have demonstrated that the biotic elements are organized into compartments comprising hierarchically interconnected subsystems of varying levels of organization, with numerous distinct layers that have close hierarchical, material, and energetic relationships between them. This approach to the CLS allows identifying the purpose and objectives of the study as well as the initial conditions and limits of modeling. In work [32], the authors ‘proposed a new way to assessing the impact of the multicomponent composition of the power battery on the state of the environment, which

790

T. Boyko et al.

consists in determining the stability of ecosystems, and makes it possible to obtain quantitative indicators of the stability and loss of natural ecosystems, which can be used as indicators of the environment state, and therefore assessment of the environmental component, which is important for determining the real impact of the polyelement composition of power batteries’. The paper [34] proposes to use the SimaPro software to assess the impact of the wind power station on the landscape system.

3 Methodology This study presents a methodology for modeling the interaction between man-made and ecological systems in various regions. To accomplish this, an integrated approach is required that simulates the interaction of a wind turbine (WT) and the complex landscape system (CLS) throughout its life cycle, such as a cyber-physical system (CPhS). The CPhS is a complex, multi-component mechanism in which computer algorithms monitor and control processes. These algorithms must adequately describe the processes. It is widely known that the development and deployment of production CPhS can be implemented through five processes, which include communication, conversion, cyber-implementation, cognition, and configuration [35]. The process of ‘cyber-implementation’ involves creating a ‘cyber-twin’ to describe the system’s state, comparing it with a real device, and subsequently synthesizing the system. Creating a ‘cyber-twin’ for a wind turbine that interacts with the environment is a multi-stage process. This process can be based on the application of the compartment concept of the complex landscape system (CLS) for an ecosystem and the product life cycle concept for wind turbines [36]. This study utilized the 5C architecture, which is a common design for cyber-physicalhuman systems (CPhS) and was originally proposed by [35]. The International Organization for Standardization’s (ISO) 14040/44 standard defines LCA as a comprehensive assessment of a product’s environmental characteristics. The result of the LCA process is a quantitative measure of the product’s environmental friendliness [37–45]. In work [34], M. Ruda used an assumption for study the influence of elemental composition of batteries on the sustainability of ecosystems, that would be valid for assessing the impact of wind power plants.

4 Basic Theory of the Proposed Method As a result of interaction of the listed natural components and anthropogenic factors, the specific system of CLS is formed. In the presented system all values acquire mainly three indices: i1 – to denote the area of space (CLS compartment); i2 – to denote the geophysical environment; and i3 – to indicate the impact. Then the expression Qi1 i2 i3 will mean the influence of wind turbines at all stages of the life cycle –i3 , in the layers or subsystems –i2 , CLS compartment –i1 . According to, such indices are called multiindices and are denoted by one letter i = (i1 i2 i3 ), and the set of multi-indices in the model – by the letter . If the CLS consists of n compartments and m impacts are considered, then. 1 ≤ i1 ≤ n; 1 ≤ i2 ≤ f ; 1 ≤ i3 ≤ m

(1)

Mathematical Modeling of States of a Complex

791

Let Qi be the impact of i3 in the layers or subsystems i2 of the compartment growth area i1 that arising at time t. Denote by ki,j (t) the value of this impact, which after time τ will be in the layers or subsystems j2 of the growth region j1 , having carried out in the process of transforming the flow of matter f into the impact of j3 . Since the impacts in the process of transfer from the layers to the subsystems of the compartment and from the compartment to another compartment do not arise and do not disappear, then.  j=

ki,j =

n j1 =1

f j2 =1

m j3 =1

ki1 i2 i3 j1 j2 j3 = 1.

(2)

The coefficients ki,j (t) are the probabilities of the states Pi,j (t) of the corresponding matrix of states P and the transition intensity λi,j (t) of the corresponding matrix of transition intensities Λ of the concentration of harmful substances or compartments from state i to state j. Formalization of graphs of the interaction of the CLS compartment and wind turbine is carried out by means of the Kolmogorov system of differential equations. If, for example, we restrict to m, l = 1, 2, 3, 4 and m = l, then the system of differential equations will take the form (3), and solving them with the help of computer technology, it is possible to study the development of the CLS in dynamic and stationary modes: ⎧ dPi,j,S1 ⎪ = −λi,j,S1 ,S2 · Pi,j,S1 + λi,j,S2 ,S2 · Pi,j,S2 ⎪ dt ⎪ ⎪   ⎨ dPi,j,S2 = λi,j,S1 ,S2 · Pi,j,S1 − λi,j,S2 ,S1 + λi,j,S2 ,S3 · Pi,j,S2 + λi,j,S3 ,S2 · Pi,j,S3 dt (3)   dPi,j,S3 ⎪ = λi,j,S2 ,S3 · Pi,j,S2 − λi,j,S3 ,S2 − λi,j,S3 ,S4 · Pi,j,S3 + λi,j,S4 ,S3 · Pi,j,S4 ⎪ ⎪ dt ⎪ ⎩ dPi,j,S4 = λi,j,S3 ,S4 · Pi,j,S3 − λi,j,S4 ,S3 · Pi,j,S4 dt where i = 1, 2, . . . , N and j = 1, 2, . . . , M are ordinal numbers of multiindexes i and j, respectively; S1 ,…, Sl , Sm ,…, Sn are states of CLS compartments; Pi,j,Si is the probability of location the CLS compartment or the concentration of harmful substances in the state Sl or the state Sm ; λi,j,Si ,Sm is the intensity of the transitions of the CLS compartments or the concentration of harmful substances from the state Sl to the state Sm for each multiindex.

5 Research Results The intensity values of state transitions for each element of the hierarchical structure, including layers and subsystems of the compartment, are statistical information derived from studying the functioning of the complex landscape system (CLS). To evaluate and predict the states of these CLSs and their compartments, it is recommended to gather information about the impact of wind turbines at various stages of their life cycle and at different time intervals during the functioning of the CLS. At the same time such information can be obtained through simulation. Simulation computer modeling of the process of obtaining the results of experimental studies on wind turbines during their life cycle was carried out. The results of modeling the effects of harmful substances, considering the scenario of waste management, are presented in Table 1.

792

T. Boyko et al.

Table 1. Positive Impact of Waste Management Scenarios (WMS) on Reducing the Impact of Harmful Substances for each Category N

Impact categories

Harmful substances (effects)

1

Carcinogenic effects

electricity from coal for DALY steel and copper production

0.41

−100%

–

electricity from coal for DALY reinforcing steel production

0.38

−100%

–

arsenic in water and air

DALY

0.81 + 0.27

–

1.08

unidentified metals in water and air

DALY

0.028 + 0.342

–

0.37

cadmium in the air

DALY

0.06

–

0.06

−15.4%

5.91

+ 5.93%

33.2E-4

2

3

4

5

Effects of respiratory inorganics

Indicators

Value of Reduction Value of impact due to impact WMS after WMS

electricity from coal for DALY concrete production – dust

2.33

– nitrogen dioxide

1.65

– sulfur dioxide

1.51

Effects of respiratory organics

non-methane volatile organic compounds

DALY

25.2E-4

methane and unidentified aromatic hydrocarbons

DALY

7.45E-4

Climate change – global warming

electricity from coal for DALY reinforcing steel production

1.58

–

1.58

impact of the use of wind turbines

DALY

2.14

−41%

1.257

radiation effects

radon-222

DALY

2.55E-3

−16.5%

3,79E-3

carbon-14

DALY

1.22E-3 (continued)

Mathematical Modeling of States of a Complex

793

Table 1. (continued) N

Impact categories

Harmful substances (effects)

6

ozone layer depletion

Bromotrifluoromethane DALY

7

ecotoxicity

unidentified metals

9

Value of Reduction Value of impact due to impact WMS after WMS 5.02E-4

+ 7.34%

5.02E-4

PAF × m2 × year

1.25E6

−47.8%

2.32E6

nickel and zinc in the air PAF × m2 × year

6,33E5

PAF × m2 × year

9.64E4

acidification nitrogen oxides and eutrophication sulfur oxides

PAF × m2 × year

1.06E5

−4.07%

1.38E5

PAF × m2 × year

2.88E4

Land use

PDF × m2 × year

9.18E3

−32.3%

3.1E4

−79.1%

−1.01E5

–

–

lead in the air 8

Indicators

transformation of industrial zones

occupation of industrial PDF × m2 × year zones

10 Depletion of minerals

11 Depletion of fossil fuels

1.3E4

occupation of landfills

PDF × m2 × year

1.46E4

nickel in silicates – 1.98%

MJ

6.47E4

nickel in raw ore – 1.04%

MJ

copper in sulfide – 0.99%

MJ

pure copper – 0.36%

MJ

copper in raw ore – 8.2E-3%

MJ

coal, oil, gas

–

3.68E4

–

794

T. Boyko et al.

The typical ecological models are represents the impacts. In particular, the impact on human health is expressed by the indicator DALY (Disability Adjusted Life Years). The impact on the ecosystem quality (ecotoxicity) is expressed as an indicator of affected species of living organisms PAF (Potentially Affected Fraction) or an indicator of loss of living organisms PDF (Potentially Disappeared Fraction). PAF (Potentially Affected Fraction) and PDF (Potentially Disappeared Fraction) are determined by evaluating the toxicity of a substance to various organisms such as microorganisms, fish, crustaceans, mollusks, amphibians, algae, worms, and plants in both terrestrial and aquatic environments. The outcome of modeling the effects of hazardous substances is a compilation of potential impacts on the environment. This record of effect scores, designated for each category, is referred to as the environmental profile of the product or service. The Fig. 1 the environmental profile of entire wind turbine life cycle is shown. It shows the environmental load of the impact categories which were chosen for the analysis. It displays the sets of four scores, each one representing the impact on one of the four categories: 1. Resource depletion; 2. Global warming; 3. Acidification; 4. Ozone depletion.

Fig. 1. Entire wind turbine life cycle environmental profile

During the next step, the load exposure is calculated using the European average data. There are three categories of damage: 1. Human health; 2. Ecosystem quality; 3. Resource depletion. The first impact category is related to human health and includes radiation, ecotoxicity, carcinogens, respiratory substances, ozone depletion, and climate change. The second category covers acidification or eutrophication and land use. The third category, known as ‘resource depletion,’ encompasses natural resources, minerals, and fossil fuels. The values for each impact category are combined to provide a total impact score. The effect on human health is expressed as DALY, which stands for disability-adjusted life years. The impact on the ecosystem quality is expressed as PAF ×m2 ×year and PDF ×m2 × year. Resource depletion is expressed as the energy surplus required for the subsequent extraction of minerals and petroleum (fossil fuels).

Mathematical Modeling of States of a Complex

795

Fig. 2. Ecological indicators of the four stages of wind farms life cycle

Figure 2 shows the distribution of the values of environmental indicators obtained in the study for four stages of the wind farm life cycle: manufacture; dismantling and disposal; operation; transportation and installation. To predict the potential impacts of wind turbines on the CLS, it is necessary to study the dynamics of the states of hierarchical structures and their components. It can be achieved by solving systems of differential Eqs. (3) using the fourth-order numerical Runge-Kutta method under the given initial conditions. ⎞ ⎛ ⎞ ⎛ P00 P0 (0) ⎜ P (0) ⎟ ⎜ P ⎟ ⎜ 1 ⎟ ⎜ 10 ⎟ ⎟ ⎜ ⎟ ⎜ (4) P(0) = ⎜ P2 (0) ⎟ = ⎜ P20 ⎟, ⎟ ⎜ ⎟ ⎜ ⎝ ....... ⎠ ⎝ ... ⎠ Pn (0) Pn0 Using the methods of approximate calculations and computer technology, the solution of the system of Eqs. (3) can be written as Pi,k+1 = Pik +

H · (K1i + 2 · K2i + 2 · K3i + K4i ), 6

(5)

796

T. Boyko et al.

where K1i = K2i = K3i = K4i =

n  j=0 n  j=0 n  j=0 n 

aij · Pjk , i = 0, 1, 2, . . . , n;   H , i = 0, 1, 2, . . . , n; aij · Pjk + K1i · 2   H , i = 0, 1, 2, . . . , n; aij · Pjk + K2i · 2

(6)

  aij · Pjk + K3i · H , i = 0, 1, 2, . . . , n;

j=0

H = t. The study of the CLS as an object with a hierarchical structure in the stationary mode, when t → ∞ and dP dt = 0, is carried out using numerical methods based on the solution of a system of algebraic equations, whose matrix notation takes the form ·P =0

(7)

where  is the matrix of intensities of transitions from state to state and P is the matrix of states probabilities. Using the Khaletsky scheme of the Gauss numerical method [9], we find Pn , Pn−1 ,…, P0 , using recurrent formulas γi0 = ai0 , (i = 0, 1, 2, ..., n), α0k , (k = 1, 2, ..., n), α0k = α00 k−1  γik = aik − γij · αjk , (i, k = 1, 2, ..., n; n ≥ k),

(8)

j=0

⎞ ⎛ i=1  1 ⎝ αik = · aik − γij · αjk ⎠, (i, k = 1, 2, ..., n; i < k). γii j=k

in the following order P0 + a01 · P1 + a02 · P2 + ... + a0n · Pn = 0, P1 + a12 · P2 + ... + a1n · Pn = 0, ........................................ Pn = 0

(9)

Mathematical Modeling of States of a Complex

797

The methods proposed in this study for collecting input data, building, and solving differential and algebraic equation systems could serve as a fundamental framework for mathematically analyzing the CPhS. This approach can be used to investigate the state of individual compartments within the CLS and predict the potential environmental impacts of wind turbines. Below there are the results of the probable wind turbines’ impact on the simulated wind farms during the four stages of their life cycle (see Fig. 2, Table 2) on the subsystems and layers of the Complex Landscape System compartments, which is confirmed by the report on the research work ‘Study of the presence of plants and animals listed in the Red Book of Ukraine and plant groups that are protected by the legislation of Ukraine in the territory of the planned activities’. To ensure the reliability and accuracy of the theoretical results, we plan to conduct experiments in our future research. Table 2. The impact of wind turbine LC on the subsystems and compartments of the Complex Landscape System Stage of WT life cycle

Area & Substance

Compartment

Unit

Total

Formalized description of subsystems and layers of the compartment affected by the impact of wind turbine LC

Transport

Occupied area as industrial area

ha

0.2

Soil compaction

%

47

Pollution

Alfamel-naphthalene

Grass-shrub layer & soil subsystem of the forest compartment

%

53

CO

Air

kg

18

kg

17.5

kg

14.3

dBa

80.0

Changing migration routes

X

X

Vehicles crash. Intrusion of invasive species

g

21

Pollution

BOD

g

412

Calcium ions

g

25

Cl

mg

17

COD

g

14

Cx Hy

g

36

Cetane

CO2 Cx Hy Occupied area as industrial area

Fauna

Acid as H+

Wastewater

(continued)

798

T. Boyko et al. Table 2. (continued)

Stage of WT life cycle

Area & Substance

Compartment

Unit

Total

Formalized description of subsystems and layers of the compartment affected by the impact of wind turbine LC

Grass-shrub layer of the compartment

ha

~ 30.6

Extinction or damage to especially valuable species

aboveground phytomass of the Grass-shrub layer of the compartment

X

X

Destruction of or damage to a variety of herbaceous communities – Hypochoeridi uniflorae-Nardetum and Hieracio vulgati-Nardetum, according to the classification of biotopes Eur 28 (EUNIS 6230) – species-rich Nardus grasslands and fragments of relict phytocoenoses with dominance of arcto-alpine species Juncus trifidus L., which can be considered within the association Oreochloetum (distichae) juncosum (trifidi), listed in the Green Book of Ukraine

(land use)

X

X

12 types of settlements that are subject to protection (namely: ‘C2.12’, ‘C2.18’, ‘C2.19’, ‘C2.25’, ‘D2.226’, ‘D2.3’, ‘E1.71’, ‘E4.11’, ‘E4.3’, ‘E5.5’, ‘F4.2’, ‘G1.6’)

soil subsystem of the forest compartment

ha

~ 30.6

Removal of the top fertile layer of soil. Wind and water erosion Changes in the soil profile

(land use)

X

X

Change of hydrological regime

Fauna

X

X

Strengthening the edge effect Loss of natural habitats, nesting sites, food for birds and bats

Air

kg

27

Pollution

CO2

kg

17

Cx Hy

kg

14

Transport & Occupied area as Erection (Concrete industrial area foundation & Assembling) of the turbine on site

Occupied area as industrial area

CO

(continued)

Mathematical Modeling of States of a Complex

799

Table 2. (continued) Stage of WT life cycle

Area & Substance

Compartment

Unit

Total

Formalized description of subsystems and layers of the compartment affected by the impact of wind turbine LC

Operation & Maintenance

Occupied area as industrial area

soil subsystem of the forest compartment

ha

~ 30.6

Soil compaction

Fauna

X

X

Damage to bats and birds during a collision with wind turbines, power lines

(land use)

m

150

Influence on the visual perception of the environment

Fuel and lubricants

Grass-shrub layer of the compartment

ha

0.003

Possible spills when replacing component parts

BOD

Wastewater

g

456

Surface run-off

Occupied area as industrial area

grass-shrub layer & soil subsystem of the forest compartment

ha

~ 30,6

Extinction of or damage to especially valuable species

Acid as H+

Wastewater

Pollution

Decommissioning

g

48

BOD

g

850

Calcium ions

g

47

Cl

mg

37

COD

g

76

Cx Hy

g

26

Detergent/oil

mg

49

Occupied area as industrial area

Fauna

X

X

Hindering the free migration of animal species

CO

Air

kg

23

Pollution

CO2

kg

20

Cx Hy

kg

19

6 Summary and Conclusion It is obvious that any man-made energy-generating facility, even if it is classified as the so-called ‘green energy’, has a negative impact on the environment. A wind turbine, as an object of this kind, should be considered throughout all stages of its life cycle. It is assumed that the process of spreading and transformation of impact or pollutant in the layers and subsystems for a certain period of time can be described in simplified

800

T. Boyko et al.

form by a linear operator that connects the state of the system at the previous and the next point of time, and the process itself is investigated using the theory of the Markov chains. Then the state of the CLS compartments can be represented as a graph with vertices identifying these states. Formal representation of such graphs is proposed to be carried out by Kolmogorov’s differential equations which link the time change in the probability of finding a compartment with the intensity of the transitions of the concentration of harmful substances from the previous state to the next one. It was obtained an integrated system of categories and indicators of wind farm impacts on the layers and subsystems of the CLS compartment. The use of WMS made it possible to assess the potential reduction of harmful effects, but it was found that for the categories of ‘fossil fuels’ they cannot be used and for the categories of ‘ozone layer depletion’ and ‘respiratory non-organic effects’, the use of WMS has the opposite effect and increases the harmful impacts. According to the research, the dismantling and disposal stages of the wind turbine’s life cycle show the most potential for significant reductions in harmful environmental impacts when using WMS. Ecological profiles were developed, which made it possible to present the impacts during the life cycle of wind turbine system in the form of eco-indicators. An example is the system of eco-indicators, distributed over four stages of the life cycle of the wind farm: manufacture; operation; transportation and installation; dismantling and disposal. The greatest harmful environmental impact with an eco-indicator value of 14.1 occurs during the manufacturing stage of wind turbines, which is associated with the type of electricity used. The study of the dynamics of states of hierarchical structures and their elements is carried out on the basis of solving systems of differential equations using the fourth order Runge-Kutta numerical method.

References 1. Parsons, D.: The environmental impact of disposable versus re-chargeable batteries for consumer use. The International Journal of Life Cycle 12, 197 (2007). https://doi.org/10.1065/ lca2006.08.270 2. Sunil, K.K., Muniamuthu, S., Tharanisrisakthi, B.T.: An investigation to estimate the maximum yielding capability of power for mini venturi wind turbine. Ecological Engineering and Environmental Technology 23(3), 72–78 (2022) 3. Ginevicius, R., Trishch, R., Remeikiene, R., et al.: Complex evaluation of the negative variations in the development of lithuanian municipalities. Transform. Bus. Econ. 20(2), 635–653 (2021) 4. Chaudhary, A.S., Chaturvedi, D.K.: Efficient Thermal Image Segmentation for Heat Visualization in Solar Panels and Batteries using Watershed Transform. I.J. Image, Graphics and Signal Processing 11, 10–17 (2017) 5. He, Y., Qian, X.: Contemporary Development and Trend of Jiangsu Province Wind Power Generation Technology. I.J. Education and Management Engineering 2, 46–51 (2013) 6. Hui, Z., Jinhua, Z.: Landscape Pattern Evolvement in Mining Area: a Case of Liyuan Town in China. I.J. Education and Management Engineering 2, 48–56 (2011)

Mathematical Modeling of States of a Complex

801

7. Strušnik, D., Brandl, D., Schober, H., Ferˇcec, J., Avsec, J.: A simulation model of the application of the solar STAF panel heat transfer and noise reduction with and without a transparent plate: A renewable energy review. Renew. Sustain. Energy Rev. 134, 110149 (2020) 8. Bojko, T., Paslavskyi, M., Ruda, M.: Stability of composite landscape complexes: model formalization. Scientific Bulletin of UNFU 29(3), 108–113 (2019) 9. Korn, G., Korn, T.: Mathematical handbook for scientists and engineers: definitions, theorems, and formulas for reference and review. Dover Publications (2013) 10. Javorskyj, I., Yuzefovych, R., Lychak, O., Kurapov, P.: Hilbert Transform for analysis of daily changes of the earth magnetic field. Proc. of the 2021 IEEE 12th International Conference on Electronics and Information Technologies, ELIT pp. 181–185 (2021) 11. Javorskyj, I., Yuzefovych, R., Kurapov, P., Semenov, P.: The covariance and spectral properties of the quadrature components of narrowband periodically non-stationary random signals. Proc. of the International Scientific and Technical Conference on Computer Sciences and Information Technologies 2, 95–98 (2020). 9321964 12. Chen, J., Su, J., Kochan, O., Levkiv, M.: Metrological software test for simulating the method of determining the thermocouple error in situ during operation. Measurement Science Review 18(2), 52–58 (2018) 13. Hu, Z., Gnatyuk, S., Okhrimenko, T., Tynymbayev, S., Iavich, M.: Computer network and information security. Int. J. Comp. Netw. Info. Secu. 12(3), 1–10 (2020) 14. Jun, S., Kochan, O., Kochan, R.: Thermocouples with built-in self-testing. Int. J. Thermophys. 37(4), 37 (2016) 15. Melnyk, R., Hatsosh, D., Levus, Y.: Contacts detection in PCB image by thinning, clustering and flood-filling. Proc. of the International Scientific and Technical Conference on Computer Sciences and Information Technologies 1, 370–374 (2021) 16. Harrington, Jr., E.C.: The desirability function. Industrial Quality Control 21, 494-498 (1965) 17. Shkrabak, V., Kalugin, A., Averyanov, Y.: Assessing effectiveness of technical measures for improving working conditions of wheeled vehicle operators. Proc. of the 5th International Conference on Industrial Engineering (ICIE 2019), 1, pp. 29–39 (2020) 18. Schulte, P.A., Delclos, G., Felknor, S.A., Chosewood, L.C.: Toward an expanded focus for occupational safety and health: a commentary. Int. J. Environ. Res. Public Health 16, 4946 (2019) 19. Aurbach, D., et al.: On the surface chemical aspects of very high energy density, rechargeable li–sulfur batteries. J. Electrochem. Soc. 156, A694–A702 (2009) 20. Andersen, P.D., Bjerregaard, E.: Prospective life cycle assessment on wind power technology 2020. TA-Datenbank-Nachrichten 4, 10 (2001) 21. Elsam. Livscyklusvurdering af hav- og landplacerede vindmølleparker. 02–170261, Elsam Engineering A/S, Kraftværksvej 53, Fredericia, DK (2004) 22. Pombo, O., Allacker, K., Rivela, B., Neila, J.: Sustainability assessment of energy saving measures: a multi-criteria approach for residential buildings retrofitting. A case study of the Spanish housing stock. Energy and Buildings 116, 384–394 (2016). https://doi.org/10.1016/ j.enbuild.2016.01.019 23. Trishch, R., Cherniak, O., Kupriyanov, O., Luniachek, V., Tsykhanovska, I.: Methodology for multi-criteria assessment of working conditions as an object of qualimetry. Engineering Management in Production and Services 13(2), 107–114 (2021) 24. Matsuda, H., Nishioka, K., Nakanishi: High-throughput combinatorial screening of multicomponent electrolyte additives to improve the performance of Li metal secondary batteries. Scientific Reports 9, 6211 (2019) 25. Lu, Y., et al.: Ecosystem health towards sustainability. Ecosystem Health and Sustainability 1(1), 2 (2015) 26. Trishch, R., Nechuiviter, O., Dyadyura, K., Tsykhanovska, I., Yakovlev, M.: Qualimetric method of assessing risks of low quality products. MM Science Journal 4769–4774 (2021)

802

T. Boyko et al.

27. Martinez, E., Sanz, F., Pellegrini, S., et al.: Life cycle assessment of a multi-megawatt wind turbine. Renewable Energy 34(3), 667–673 (2009) 28. Dzhumelia, E., Pohrebennyk, V.: Methods of soils pollution spread analysis: case study of mining and chemical enterprise in Lviv region (Ukraine). Ecolog. Eng. Environ. Technol. 22(4), 39–44 (2021) 29. Hu, Z., Tereikovskyi, I., Chernyshev, D., Tereikovska, L., Tereikovskyi, O., Wang, D.: Procedure for processing biometric parameters based on wavelet transformations. Int. J. Modern Edu. Comp. Sci. (IJMECS) 13(2), 11–22 (2021) 30. Pohrebennyk, V., Koszelnik, P., Mitryasova, O., Dzhumelia, E., Zdeb, M.: Environmental monitoring of soils of post-industrial mining areas. J. Ecolog. Eng. 20(9), 53–61 (2019) 31. Kupriyanov, O., Trishch, R., Dichev, D., Bondarenko, T.: Mathematic model of the general approach to tolerance control in quality assessment. Lecture Notes in Mechanical Engineering 415–423 (2021) 32. Ruda, M., Bubela, T., Kiselychnyk, M., Boyko, O., Trishch, R.: Assessment of the impact of the elemental composition of batteries on the sustainability of ecosystems. 2022 IEEE 16th International Conference on Advanced Trends in Radioelectronics, Telecommunications and Computer Engineering (TCSET) 390–393 (2022) 33. Hu, Z., Tereykovskiy, I., Tereykovska, L., Pogorelov, V.: Determination of Structural parameters of multilayer perceptron designed to estimate parameters of technical systems. Int. J. Intell. Sys. Appli. (IJISA) 9(10), 57–62 (2017) 34. Ruda, M., Boyko, T., Chayka, O., Mikhalieva, M., Holodovska, O.: Simulation of the influence of wind power plants on the compartments of the complex landscape system. Journal of Water and Land Development 2(I–III), 156–165 (2022) 35. Lee, J., Bagheri, B., Kao, H.A.: A cyber-physical systems architecture for Industry 4.0-based manufacturing systems. Manufacturing Letters 3, 18–23 (2015) 36. Hamadea, R., Al, A.R., Bou, G.M., et al.: Life cycle analysis of aa alkaline batteries. Procedia Manufacturing 43, 415–422 (2020) 37. DSTU ISO 14040:2004. Ekologhichne keruvannja. Ocinjuvannja zhyttjevogho cyklu. Pryncypy ta struktura [Environmental management. Life cycle assessment. Principles and structure]. Derzhstandart Ukrajiny, Kyiv (2007). (In Ukrainian) 38. Muralikrishna, I.V.: Valli Manickam. Chapter Five - Life Cycle Assessment. Environmental Management, pp. 57–75 (2017) 39. Simensen, T., Horvath, P., Vollering, J., Erikstad, L., Halvorsen, R., Bryn, A.: Composite landscape predictors improve distribution models of ecosystem types. Biodiversity Research 26, 928–943 (2020) 40. Küçükkaraca, B., Barutcu, B.: Life cycle assessment of wind turbine in Turkey. Balkan Journal of Electrical & Computer Engineering 10(3), 230–236 (2022) 41. Chipindula, J., Sai, V., Botlaguduru, V.: Life cycle environmental impact of onshore and offshore wind farms in Texas. Sustainability 10(6), 2022 (2018) 42. Bonou, A.L., Olsen, S.I.: Life cycle assessment of onshore and offshore wind energy: from theory to application. Applied Energy 180, 327–337 (2016) 43. Hu, Z., Ivashchenko, M., Lyushenko, L., Klyushnyk, D.: Artificial neural network training criterion formulation using error continuous domain. Int. J. Modern Edu. Comp. Sci. (IJMECS) 13(3), 13–22 (2021) 44. Kent-Dobias, J., Kurchan, J.: Complex complex landscapes. Physical Review Research 3, 023064 (2021) 45. Cushman, S.: Calculation of configurational entropy in complex landscapes. Entropy 20(4), 298 (2018)

Development of Distributed System for Electric Personal Transporters Charging Roman Kochan1,2(B) , Nataliia Kochan3 , Nataliya Hots2 , Uliana Kohut2 , and Volodymyr Kochan4 1 University of Bielsko-Biala, 43-300 Bielsko-Biala, Poland 2 Lviv Plitechnic National University, Lviv 79013, Ukraine

[email protected]

3 Ivan Franko National University of Lviv, Lviv 79001, Ukraine 4 West Ukrainian National University, Ternopil, Ukraine

Abstract. The analysis of the market of Electric Personal Transporters (EPT) and charging infrastructure is presented in this paper. This analysis shows the positive trend of the EPT market and the lack of charging possibilities for EPT of existed charging infrastructure. This paper presents the development of the distributive systems that provide EPTs’ battery charging during their staying in public parking lots. It provides an operating range of EPT increasing. The presented system utilizes the Internet of Things methodology and provides charging process control via the mobile application. Its implementation provides increasing of mobility level EPT users. Keywords: Battery Charging · Distributive System · Structural Synthesis · IEEE 1451 Standard

1 Introduction The development of mechatronics and electrochemistry has provided the ability to build vehicles based on electric engine for different application, including land, air and voter transport [1]. Limitations of public transport functionality due to the COVID-19 pandemic have led to an avalanche increasing the number of users of light personal electric transporters (EPT) [2]. The forecast of the global EPT market obtained from open sources [3–18] is presented in Table 1. It shows a steady trend towards an increase in the number of EPT in the future. The existing system of specialized charging stations, which can be found using the query “charging points, , ” in the search engine of Google Maps for some largest cities of Poland and Ukraine is presented in Table 2. The total population of presented cities, according to the search query “population , ” in Poland cities is approximately 4.1 million, they use 458 charging stations. It means that an average number of users is approximately 9 thousand inhabitants per charging station. The total population of presented Ukrainian cities is approximately 4, 5 million persons. They use 381 charging stations. Therefore, the average number of users is approximately © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 803–814, 2023. https://doi.org/10.1007/978-3-031-36118-0_70

804

R. Kochan et al.

12 thousand inhabitants per charging station. It is necessary to mention that indicated population is based on Wikipedia and do not account the migration caused by the current war in Ukraine. Anyway, all these charging stations are classified in Google Maps as car charging stations. Approximately half of them are marked by electric power of the charging source. Most of them provide charging using source with a power 11… 50 kW. The number of channels is usually 2…4. It means that these charging stations can be classified as the level “SF4: DC high power charging infrastructure” of the model of electric vehicle support infrastructure presented in [19]. These charging stations are focused on high-capacity batteries that are installed in the car and are incompatible with EPT. Some of the charging stations listed in the Table 1 provide charging channels with a power less than 5 kW, but the total number of such stations does not exceed 20% of the total number of charging channels. These channels can be classified as the level “SF3: Public slow charging infrastructure” [19]. And only these charging channels are compatible with EPT. Therefore, the level SF3 is poorly developed, and the number of charging channels does not correspond to the number of available EPT. Table 1. Compound annual growth rate (CAGR) of the global EPT market №

Type of EPT

Years

CAGR

Source

1.

Bikes

2020–2030

10,5

[3]

2.

Bikes

2020–2027

7,9

[4]

3.

Motorcycles

2019–2025

10,35

[5]

4.

Motorcycles

2020–2025

10,35

[6]

5.

Trike

2018–2025

13,5

[7]

6.

Three wheeler

2020–2027

1,7

[8]

7.

Three wheeler

2020–2025

9

[9]

8.

Scooter & Bike & Skateboard

2019–2025

8,4

[10]

9.

Kick scooter

2021–2028

10,3

[11]

10.

Kick scooter

2020–2026

12,3

[12]

11.

Skateboard

2019–2025

3,1

[13]

12.

Skateboard

2019–2023

3,1

[14]

13.

Unicycle

2017–2021

26,9

[15]

14.

Hoverboard

2017–2022

6,1

[16]

15.

Onewheel

2019–2027

5

[17]

16.

Onewheel

2020–2027

5

[18]

One of the main factors that limits the development of EPT is the underdeveloped battery charging infrastructure, which, at least in Poland and Ukraine, is based on “Home charging infrastructure” – the level of “SF2” model of electric vehicle support infrastructure presented in [19].

Development of Distributed System

805

Table 2. Number of charging stations №

City

Country

Population, thousands*

Number of charging stations

1.

Warsaw

Poland

1765

121

2.

Krakow

Poland

766

89

3.

Lodz

Poland

696

50

4.

Gdansk

Poland

582

104

5.

Katowice

Poland

320

94

6.

Kyiv

Ukraine

2884

177

7.

Dnipro

Ukraine

966

98

8.

Lviv

Ukraine

721

106

* based on Wikipedia and do not account the migration caused by the current war in Ukraine

Solving the EPT charging problem by replacing separately pre-charged batteries has a few disadvantages, including the following: – Increasing the load on home electrical power networks, that is not oriented for such operation mode; – Complicates the construction of EPT; – Increasing the electrical losses on the connectors; – Decreasing the reliability of EPT; – The support of sharing services becomes more complicated. These disadvantages of existing approaches to solving the problem of battery charging limits the range of EPT effective using [20], and therefore inhibit the migration of users from traditional modes of transport to the EPT. Despite this, the forecast of the CAGR parameter of the global EPT market is optimistic with the average value 7.8% for the period 2021–2025 with a standard deviation 3.5%. The objective of this work is in development the system of charging stations that provides batteries of EPT recharging during mooring in public parking lots. It promises to increase the operating range of EPT. This solution will ensure the implementation of the level “SF3: Public slow charging infrastructure”, and probably the level “SF4: DC high power charging infrastructure” [19] for EPT. It provides maximization the distance of available trips using EPT and potentially will help to solve the transport and air pollutions problems in the cities by user migration from cars to EPT. The analysis of the charging station market, as well as modeling of the decision-making process by buyers [21] allows predicting that such a service will be in demand.

2 Concept of Distributed System for EPT Battery Charging The conception of proposed system of EPT batteries charging is based on Internet of Things technology [22, 23]. The developed system is distributed, and it provides the EPT battery charging by connection charger to electric power lines. It is oriented on batteries

806

R. Kochan et al.

with relatively small capacity. This system is oriented for implementation in public parking lots, for example near apartments, offices, malls, etc. The proposed system has a modular structure and open architecture, which ensures its scalability and high modernization potential. This system supports the set of following functions: – – – – – –

Personalized user accounting; Access and control using a mobile application; On-line monitoring of charging; Remote switch on/off of charging; Measurement of charge current and value of electric charge; Load balancing for all phases.

The developed distributed system has a hierarchical structure that corresponds to the structure presented in the IEEE 1451 family of standards [24] for smart distributed systems. Structure of this system is presented on Fig. 1. It consists of three levels: – Server; – Controllers of local area network – NC; – Charging Controllers – CC.

Fig. 1. Structure of developed distributed charging system

Server of this system provides connection the set of controllers of local area networks – NC. Each controller of local area network provides connection the set of Charging controllers – CC. Each charging controller provides the set of charging channel. Drivers of EPT that want to charge the battery of EPT are users of this system. They connect own charger of EPT battery to allowable charging channel of charging controller and run special mobile application. This mobile application connects to the server, provides user identification and personal account access. This mobile application provides scanning the ID tag of selected charging channel than user order electric charge value, that he wishes to charge or time of charger connection to electric power network. Then user prepays the charging service. After payment the Server sends the command of switching on the selected charging channel. This command is addressed to Charging controller that controls the selected charging channel. The parameters sent in this command are: charging time or charge value. Then charging controller switch on the selected

Development of Distributed System

807

charging channel and measure connection time as well as transmitted electric charge. The charging controller provides finishing the charging process by switching off the selected charging channel after achievement the predefined value of charging time or electric charge. Then the charging controller sends to the server the report about successful finishing the charging process. If EPT battery charger of some charging channel will be disconnected before finishing the charging process charging controller sends to the server special urgent query. Server informs user about charging process. This information consists of following fields: coordinates of charger, time of charging starting, time of charging finishing, charging time, value of transmitted electric charge, e.t.c. This information is presented to user by messages of mobile application or selected messenger.

3 Components of the Distributed Charging System 3.1 Server Server is the component that manages the system. It provides interaction with mobile applications of users by Internet communication channels, interaction with the lower hierarchical levels of the system by the communication channels of controllers of local area network and database maintenance. The server’s software is based on a web server, and provides interaction with mobile applications of all users and the set of charging controllers. It provides safe data communication protocols [25]. The prototype of the mobile application and methods of its interaction with the server software is the parcel monitoring system Nova Poshta [26]. The mobile application is the front end of the developed system, and the server software is its back end [27, 28]. Server software provides real-time functionality of developed distributed systems basing on the non-real time Internet communication channels and protocols. This combination is provided by some information redundancy during the operation of the system [29]. The following additional service functions of server are planed: communication with electric power distributor including reports generation for each point of the system connection to power grid [30] and testing of the system functionality [31] and error [32]. 3.2 Controller of Local Area Network The controller of local area network provides the division of the system into segments [33]. Its basic functionality corresponds to the network capable application processor of the IEEE 1451 standard [34, 35]. The specific functions of controller of local area network that are adapted to the target system are: – – – – – –

routing of messages between server and charging controllers; accounting of consumed electric energy; measurement the voltage values that is connected to chargers; measurement the charging current value; emergency shutdown of systems; charging controllers remote control;

808

R. Kochan et al.

– testing of modules of the system. The UML model of this controller is presented in [36]. Its connection with the server is provided using the stack of Internet protocols. Each controller of local area network provides connection to the set of charging controllers. This connection is provided by electric power lines using high frequency modulation of data signals. The structure of controller of local area network is presented on Fig. 2. This controller consists of four main subsystems that are cooperated between each other. These subsystems are presented lower, and they are marked by dotted lines: – – – –

Measuring; Control; Communication; User interface.

The measuring subsystem provides measurement of the values of consumed electric energy, and also rms values of voltage and current of each phase of three phase electric power network. The measuring subsystem consists of: – Counter of of consumed electric energy – Energy Counter – Instrumentation for voltage rms value measurement of all phases – Voltmeter; – Instrumentation for current rms value measurement of all phases (Ammeter);

Fig. 2. Structure of controller of local area network

The control subsystem provides the interaction of all components with each other, as well as the shutdown of the system in the event of an emergency, for example, when the parameters of the system exceed the permissible limits. It includes: microcontroller (MC) and protection switch (MUX). The communication subsystem consists of two interface controllers that provide data transmission between system hierarchical levels, particularly between the charging controllers and the controller of local area network (LAN Modem), and between the server and the controller of local area network (Server Interface). The controller of local area network is an invisible component of the system for users, and it provides automatic operation, therefore the user interface subsystem provides

Development of Distributed System

809

interaction with the controller of local area network only during system installation and maintenance. Users are then specialists of the relevant profile, and this subsystem displays the current operation parameters of the segment of charging system, which is controlled by the appropriate controller of local area network. 3.3 Charging Controller Charging controllers provides connection/disconnection of chargers of EPT batteries to electric power lines 230 VAC by the command of server. These modules provide: – – – – –

Multichannel charging process; Detection of the connection&disconnection of chargers; Switching on/off the charging current; Measurement the charging current of all channels; Measurement of charging time and accounting the electric charge transmitted to each EPT battery; – Load balancing of all phases; – Remote control and monitoring of the charging process; – Indication the parameters of charging process. The charging controller measures the charging current of all connected EPT batteries and switches them by such combination to three phase electric power network so that the load on all phases are balanced. This module operates with remote control from the server of the system with real-time Internet access. This imposes some requirements and limitations on the software functionality [37]. In addition, the protocols of server interaction with this module must take into account the delays associated with the response time for implementing server commands [29]. The block diagram of the charging controller is presented on Fig. 3. It contains the following interacting components: – – – – –

Charging channels – Channel 1, …, Channel n; Microkontroller with all components that provide its functionality – Microcontroller; Real time clock – RTC; Liquid crystal display – LCD; Modem – MOD. Charging channels provide:

– – – – – –

detection of chargers connection to the connector of the charging channel; connection of chargers to the electric power network; indication of the charging process; measurement the charging current & electric charge; fully automatic operation under control of server queries; scaling the number of channels. Each charging channel consists from:

– Connector for charger connection – CONN; – Multiplexer of charger to the selected phase A, B or C of three phase electric power network – MUX A, MUX B, MUX C;

810

R. Kochan et al.

Fig. 3. Structure of charging controller

– – – –

Pulse distortion filter – PF; Sensor of electric current – SI; Sensor of charger connection – SE; LED indicator – LED.

Microcontroller provides control of all components of charging module, server command execution and messages exchanging with server of this system. Real time clock provides measurement of the connected time of EPT batteries. LCD provides current status and parameters indication of charging controller. Modem provides connection to controller of local area network. We propose to use radio modules of radio transmitter/receivers with frequency 433 MHz, with electric power lines using as external antenna. All transmitters use the same phase as external antenna. All receivers also use the same phase as external antenna but this phase is another than transmitter’s phase. This solution provides trusty connection because the large antennas dimension and small distance between these antennas. Also it provides increasing the distance of connection in comparison with radio modules using and does not demand additional communication lines.

4 Prototype of the Developed System The prototype of the system is based on the Arduino microprocessor kit and Raspberry Pi single-board computers and is designed to debug algorithms of system components and data transmission protocols. The prototype of the controller of local area network is based on a module with remote reprogramming [38] connected to single-board computer Raspberry Pi 3 Model B. It partially implements the control and communication subsystems. The control subsystem additionally includes a three-phase solid state relay, which is controlled by generalpurpose input/output lines. The communication subsystem assumes that communication with the server is supported by the built-in WiFi adapter. Communication with

Development of Distributed System

811

the charging controler is provided by a modem based on 433 MHz wireless modules of the Arduino kit [39], which use the phase lines of the electrical power network as antennas. These modules implement the physical layer of bit transmission of the sevenlevel model of open systems interaction [40]. Channel and higher levels of this model are implemented by Raspberry Pi [41]. The measuring subsystem is connected to the serial ports of the general purpose bus. The user interface subsystem implements a standard USB keyboard and monitor that are connected only during system installation and maintenance. Archives with system parameters are stored on the SD-card that provides operating system loading. The prototype of charging controller is based on Arduino Mega 2560 module. It implements module Microcontroller of structure of charging controller. Multiplexers MUX A, MUX B and MUX C are based on 4-channel relay multiplexer of Arduino shields. It provides fast implementation of 4-channel charging system. Current sensors are based on current transformators that provides isolation of power lines from control lines. Sensor SE is based on reed switch. LED is based on two colors LED. It is installed together with sensor SE inside the socket box of charger connector. LCD is based on 16*2 positions module of LCD. RTC is based on chip DS1307. Modem MOD is based on wireless radiofrequency modules with frequency 433 MHz from Arduino set. The modem of Controllers of local area network uses the same approach. The distortion rejection is provided by connection of all transceivers’ antennas to one phase, and all receivers’ antennas to another.

5 Conclusions The market analysis of EPT and charging infrastructure show the positive trend of EPT market and lack of charging possibilities compatible with EPT. This paper presents synthesized the distributive systems that provides batteries f EPT charging. It is oriented on implementation on public parking lots of EPT. The developed distributive charging system is based on the network of charging controllers connected to electric power network and controlled by server. It provides on-line control of charging process using mobile application. The developed system is modular and based on distributive approach. It provides its easy installation and multiplication. The developed system implementation provides increasing the distance of EPT operation. This increasing is up to doubling in the condition of waiting enough time for charging the batteries (for example, during lessons in Universities or Schools, concerts, cinema, etc.). Generally proposed system implementation bring intensification of EPT implementation, promote solving transport and ecological problems of megalopolises.

References 1. Williamson, S.S., Rathore, A.K., Musavi, F.: Industrial electronics for electric transportation: Current state-of-the-art and future challenges. IEEE Transa. Ind. Electron. 62(5), 3021–3032 (2015). https://doi.org/10.1109/TIE.2015.2409052 2. 11 types of electric personal transportation vehicles: a comparison.https://www.hobbr.com/ types-of-electric-personal-transport-devices/

812

R. Kochan et al.

3. Electric Bike Market by Product (Pedelecs, Speed Pedelecs, Throttle on Demand, and Scooter & Motorcycle), Drive Mechanism (Hub Motor, Mid-Drive, and Others), and Battery Type (Lead Acid, Lithium-Ion (Li-ion), and Others): Global Opportunity Analysis and Industry Forecast, 2020–2030. https://www.alliedmarketresearch.com/electric-bikes-market 4. E-bike Market Growth, Industry Trends, and Statistics by 2027. https://www.marketsandma rkets.com/Market-Reports/electric-bike-market-110827400.html 5. Electric Motorcycles Market to Exhibit 10.35% CAGR by 2025 | Market Research Future. https://www.globenewswire.com/fr/news-release/2021/02/11/2173763/0/en/Electric-Mot orcycles-Market-to-Exhibit-10-35-CAGR-by-2025-Market-Research-Future.html 6. Electric Motorcycle Market Size, Share, Growth, Report, 2027. https://www.marketresearchf uture.com/reports/electric-motorcycles-market-8136 7. Electric Trike Market Showcase Approximately vigorous CAGR of Around 13.5% In Global Market Forecast Till 2025. https://www.mynewsdesk.com/us/news-world-center/pressrele ases/electric-trike-market-showcase-approximately-vigorous-cagr-of-around-13-dot-5-per cent-in-global-market-forecast-till-2025-2848684 8. Electric Three Wheeler Market Size, Share, Global Report, 2027. https://www.fortunebusin essinsights.com/electric-three-wheeler-market-105028 9. Electric Three-Wheeler Market – Growth, Trends, and Forecasts (2020–2025). https://www. globenewswire.com/news-release/2020/08/27/2084586/0/en/Electric-Three-Wheeler-Mar ket-Growth-Trends-and-Forecasts-2020-2025.html 10. Electric Transporters Market Size, Share & Trends Analysis Report By Vehicle (Bike, Scooter, Skateboards), By Battery (Sealed Lead Acid, NiMH, Li-Ion), By Voltage, And Segment Forecasts, 2019–2025. https://www.grandviewresearch.com/industry-analysis/electrictransporters-market 11. Electric Kick Scooters Market Size Worth $5.2 Billion By 2028 | CAGR: 10.3%: Grand View Research, Inc. https://www.prnewswire.com/news-releases/electric-kick-scooters-mar ket-size-worth-5-2-billion-by-2028--cagr-10-3-grand-view-research-inc-301210598.html 12. Global Electric Kick Scooters Market By Battery, By Voltage, By Region, Industry Analysis and Forecast, 2020–2026. https://www.globenewswire.com/news-release/2021/02/25/218 2331/0/en/Global-Electric-Kick-Scooters-Market-By-Battery-By-Voltage-By-Region-Ind ustry-Analysis-and-Forecast-2020-2026.html 13. Skateboard Market Size Worth $2.4 Billion By 2025 | CAGR: 3.1%. https://www.grandview research.com/press-release/global-skateboard-market 14. Electric Skateboard Market. https://askwonder.com/research/electric-skateboard-market-bud 11r96e 15. Electric Unicycle 2017 Global Market Expected to Grow at CAGR of 26.90% and Forecast to 2021. https://www.einpresswire.com/article/414374748/electric-unicycle-2017-glo bal-market-expected-to-grow-at-cagr-of-26-90-and-forecast-to-2021 16. Hoverboard Market Forecast, Trend Analysis & Competition Tracking – Global Market insights 2017 to 2022. https://www.factmr.com/report/219/hoverboard-market 17. One Wheel Electric Scooter Market by Product Type (Electric Unicycle, and Electric One wheel Hoverboard), Application (Off-road Activities, and Daily Commute), Sales Channel (Online Sales, and Offline Sales), and Speed Limit (Kmh) (20 Kmh – 30 Kmh, 30 Kmh – 50 Kmh, and More than 50 Kmh): Global Opportunity Analysis and Industry Forecast, 2020– 2027. https://www.alliedmarketresearch.com/one-wheel-electric-scooter-market-A08744 18. One Wheel Electric Scooter Market to Garner $146.7 Million by 2027: Allied Market Research, http://www.globenewswire.com/en/news-release/2021/01/18/2160005/0/en/ One-Wheel-Electric-Scooter-Market-to-Garner-146-7-Million-by-2027-Allied-Market-Res earch.html

Development of Distributed System

813

19. Funke, S.Á., Sprei, F., Gnann, T., Plötz, P.: How much charging infrastructure do electric vehicles need? A review of the evidence and international comparison. Transp. Res. Part D: Transp. Environ. 77, 224–242 (2019). https://doi.org/10.1016/j.trd.2019.10.024 20. Zagorskas, J., Burinskien˙e, M.: Challenges caused by increased use of e-powered personal mobility vehicles in European cities. Sustainability 12(1), 273 (2019). https://doi.org/10.3390/ su12010273 21. Kowalska-Stycze´n, A., Sznajd-Weron, K.: From consumer decision to market share – unanimity of majority? J. Artif. Soc. Soc. Simul. 19(4), 10 (2016). https://doi.org/10.18564/jasss. 3156 22. Greengard, S.: The Internet of Things. The MIT Press (2015). https://doi.org/10.7551/mit press/10277.001.0001 23. Jun, S., Kochan, O., Kochan, V., Wang, C.: Development and investigation of the method for compensating thermoelectric inhomogeneity error. Int. J. Thermophys. 37, 1–14 (2016) 24. Song, E., Lee, K.: Understanding IEEE 1451-Networked smart transducer interface standard – What is a smart transducer? IEEE Instrum. Meas. Mag. 11(2), 11–17 (2008). https://doi.org/ 10.1109/MIM.2008.4483728 25. Zhengbing, H., Dychka, I., Onai, M., Zhykin, Y.: Blind payment protocol for payment channel networks. Int. J. Comput. Netw. Inform. Secur. (IJCNIS) 11(6), 22–28 (2019). https://doi.org/ 10.5815/ijcnis.2019.06.03 26. Urgent and express delivery, logistic service in Ukraine and abroad. https://novaposhta.ua/ 27. Jun, S., et al.: A cost-efficient software based router and traffic generator for simulation and testing of IP network. Electronics 9(1), 40 (2019). https://doi.org/10.3390/electronics9010040 28. Xu, H., Przystupa, K., Fang, C., Marciniak, A., Kochan, O., Beshley, M.: A combination strategy of feature selection based on an integrated optimization algorithm and weighted knearest neighbor to improve the performance of network intrusion detection. Electronics 9(8), 1206 (2020) 29. Hrusha, V., Kochan, R., Kurylyak, Y., Osolinskiy, O.: Development of measurement system with remote access based on Internet. In: 4th IEEE Workshop on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications, pp. 126–128. Dortmund (2007). https://doi.org/10.1109/IDAACS.2007.4488388 30. Chen, X., et al.: Forecasting short-term electric load using extreme learning machine with improved tree seed algorithm based on Lévy flight. Eksploatacja i Niezawodno´sc´ 24(1), 153–162 (2022) 31. Fang, M.T., Chen, Z.J., Przystupa, K., Li, T., Majka, M., Kochan, O.: Examination of abnormal behavior detection based on improved YOLOv3. Electronics 10(2), 197 (2021) 32. Xiong, G., et al.: Online measurement error detection for the electronic transformer in a smart grid. Energies 14(12), 3551 (2021) 33. Yatsuk, V., Bubela, T., Pokhodylo, Y., Yatsuk, Y., & Kochan, R.: Improvement of data acquisition systems for the measurement of physical-chemical environmental properties. Paper presented at the Proceedings of the 2017 IEEE 9th International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications, IDAACS 2017, pp. 1 41–46 (2017). https://doi.org/10.1109/IDAACS.2017.8095046 34. Kochan, V., Lee, K., Kochan, R., Sachenko, A.: Approach to improving network capable application processor based on IEEE 1451 standard. Comput. Stan. Interfaces 28(2), 141–149 (2005). https://doi.org/10.1016/j.csi.2005.01.015 35. Kochan, V., Lee, K., Kochan, R., Sachenko, A.: Approach to improvement the network capable application processor compatible with IEEE 1451 standard. In: Proceedings of Second IEEE International Workshop on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications, pp. 437–441. Lviv (2003). https://doi.org/10.1109/IDAACS. 2003.1249602

814

R. Kochan et al.

36. Stepanenko, A., Lee, K., Kochan, R., Kochan, V., Sachenko, A.: Development of a minimal IEEE 1451.1 model for microcontroller implementation. In: Proceedings of the 2006 IEEE Sensors Applications Symposium, pp. 88–93. Houston, Texas (2006). https://doi.org/10.1109/ SAS.2006.1634243 37. Hrusha, V., et al.: Distributed Web-based measurement system. In: 2005 IEEE Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications. pp. 355–358. Sofia (2005). https://doi.org/10.1109/IDAACS.2005.283002 38. Kochan, R., Kochan, O., Chyrka, M., Vasylkiv, N.: Precision data acquisition (DAQ) module with remote reprogramming. In: 2005 IEEE Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications, pp. 279–282 (2005) 39. The RC link and ASK environment. https://www.electroschematics.com/rc-link-ask-enviro nment/ 40. Saxena, P.: OSI reference model–a seven layered architecture of OSI model. Int. J. Res. (IJR) 1(10), 1145–1156 (2014) 41. Maykiv, I., Stepanenko, A., Wobschall, D., Kochan, R., Kochan, V., Sachenko, A.: Software–hardware method of serial interface controller implementation. Comput. Stan. Interfaces 34(6), 509–516 (2012). https://doi.org/10.1016/j.csi.2011.10.009

Environmental Pollution from Nuclear Power Plants: Modelling for the Khmelnytskyi Nuclear Power Plant (Ukraine) Igor Vasylkivskyi1 , Vitalii Ishchenko1 , Orest Kochan2,3(B) , and Roman Ivakh3 1 Vinnytsia National Technical University, Vinnytsia, Ukraine 2 Lublin University of Technology, Lublin, Poland 3 Lviv Polytechnic National University, Lviv, Ukraine

[email protected]

Abstract. It is known that there is emission of radionuclides during normal operation of nuclear power plants. The paper analyses sources of radioactivity and environmental pollution for the case study of Khmelnytskyi nuclear power plant in Ukraine. Based on the radionuclides concentrations in air emissions, the highest concentrations are found for Nitrogen-16, Krypton-83m, Argon-41, Tritium, and Xenon-131m. Among non-radioactive chemicals, only nitrogen dioxide concentration in the air is found to be below the permissible limit. The highest concentrations of suspended particles in the air near the Khmelnytskyi power plant were measured in the range 3.4–7.7 mg/m3 . Up to 90% of non-radioactive air emissions are generated in the boiler room. The main contribution to the gas-aerosol emission of Khmelnytskyi nuclear power plant is made by inert radioactive gases, which do not directly migrate in ecosystem along food chain. The expected contribution of iodine radioisotopes and activated corrosion products to the total emission is very small. Crop contamination with aerosol radionuclides from nuclear power plant is analysed. Contamination of agricultural products with aerosol radionuclides in case of maximal projected accident at Khmelnytskyi nuclear power plant (Ukraine) is calculated. Keywords: Modelling · Crop Contamination · Radionuclide · Nuclear Power Plant

1 Instruction Reliable electricity sources are necessary in modern world [1] to maintain critical infrastructure [2, 3]. Heat power plants generate most of electricity in many countries. Despite many efforts intended to improve their operation [4–7] (because their efficiency has both economic and ecological effect [8, 9]), there are still many environmental issues. Under this circumstance, nuclear power becomes the priority electricity source for many countries. However, nuclear power plants also emit environmental pollution like any other industry [6]. There are gaseous, liquid, and solid emissions from nuclear power plants. Obviously, such emissions include radioactive elements [10]. According to the energy strategies of many countries, construction of more nuclear power plants is planned. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 815–826, 2023. https://doi.org/10.1007/978-3-031-36118-0_71

816

I. Vasylkivskyi et al.

In Ukraine, more nuclear units at existing nuclear power plants are projected with the total capacity of up to 22 GW [11]. Similar strategies are accepted in USA, China, India, and many other countries [12–16]. National safety, including reliable energy supply, is always of the highest priority. But, environmental risks should not be underestimated. There are studies [17, 18] revealing radionuclides emission from nuclear power plants. Radionuclides can be disseminated and precipitated at large areas leading to soil and crop pollution. Modelling of radionuclide emission for the Fukushima nuclear power plant was done by Leel˝ossy et al. [19]. Morino et al. [20] have found 22% of cesium-137 and 13% of iodine-131 distributed over 500,000 km2 area after the Fukushima accident. Other studies [21–24] also reported about long distances of radionuclides dissemination from nuclear power plants. Besides, impact on humans and environmental issues resulting from air emissions of nuclear power plants are analysed in [25]. There is a research [26] dedicated to the nuclear power contribution to climate protection due to decreased gaseous emissions in comparison to energy production from fossil fuels. Eisma et al. [27] have revealed that methane isotope 14 CH4 is emitted from nuclear power plants in higher amounts than generally assumed (known sources: pressurized and boiling water reactors). Also, there is a risk of transboundary transferring of radionuclides. Kim et al. [28] have calculated effective dose to Korea considering all China’s nuclear power plants. The results showed that the effective dose should be expected to increase to 1.1E-08 Sv/year, about a 500-fold increase from the 2000 level and a 2.6-fold increase from the 2010 level. One of the most common radionuclides emitted from nuclear power plants is xenon [29], but other elements are also important to measure. This research analyzes radionuclides dissemination from nuclear power plant for the case study of Khmelnytskyi nuclear power plant in Ukraine. The objectives of this study are as follows: assessment of radionuclides and non-radioactive chemicals concentration in the air near Khmelnytskyi nuclear power plant; modeling of agricultural products contamination during the maximal projected accident at Khmelnytskyi nuclear power plant.

2 Materials and Methods Radionuclides emission from Khmelnytskyi nuclear power plant (Netishyn, Khmelnytskyi region, Ukraine) was measured in the laboratory of the Department of Environmental Protection at the power plant aforementioned. Air pollution at the area of the power plant was analysed using high-altitude radio-sensing of the atmosphere. Air emissions from the turbine condenser ejectors was registered at Shepetivka research station, some 25 km from Khmelnytskyi nuclear power plant. The concentration of radionuclides was measured by radiometers RAC-02, PGG-03, PGB-101–01, PGB-101–02 combined in Automatic System of Radiological Control. The results used in this paper are average for 2014–2016 years. Besides, contamination of agricultural products with aerosol radionuclides for a maximal projected accident at Khmelnytskyi nuclear power plant was calculated based on radionuclides emission and sedimentation. Radionuclides concentrations in agricultural products were calculated in following way [30]: q = qa + qr ,

(1)

Environmental Pollution from Nuclear Power Plants

817

where qa is direct intake of radionuclides to the plant from atmosphere; qr is intake of radionuclides to the plant through roots. We can define qa as follows: qa = (0.25/ρ) · r · c · V · t,

(2)

where ρ is the amount of aboveground plant biomass, kg/m2 ; r is the ratio of radionuclide concentration in the commercial and aboveground part of plant (average 0.1); c is average annual radionuclide concentration in atmosphere (Bq/m3 ); V is the rate of radionuclide sedimentation (m/s); t is the time passed since the beginning of growing season (s). Thus, qr can be defined as follows: qr = (1/m) · cs · k,

(3)

where m is the mass of soil layer of 1 m2 square, where radionuclide uptake occurs, (average 300 kg/m2 ); cs is soil contamination with radionuclide, Bq/m2 ; k is the factor of radionuclide accumulation in plant, can be obtained from [31]. Radionuclide concentration in atmosphere (c) and soil contamination with radionuclide (cs ) can be assessed according to the methodology presented in [32] and based on the Gaussian air pollutant dispersion and sedimentation equations. The input data include: • • • • • • •

category of stability – F; wind velocity – 1 m/s; distance from the nuclear power plant – 2–25 km; soil roughness – 10–100 cm; altitude of layer mixing – 200 m; season – summer; rate of aerosol sedimentation – 0.001 m/s.

3 Results 3.1 The Sources of Radionuclides Emission from the Khmelnytskyi Nuclear Power Plant For a typical nuclear power plant, radionuclides can be found in products of nuclear fuel decomposition, activation products, and products of constructions corrosion [22]. Even if a nuclear power plant operates as usual, some fission products can be transferred into the heat carrier if damage of the fuel element coverage happens [10]. One of the main fission products in the heat carrier of the primary circuit is Tritium (around 12 years half-life). It is a radioactive isotope of hydrogen [11]. Tritium is generated in the reactor of a nuclear power plant as a result of following processes: i) radioactive decay of nuclear fuel; ii) combination of the neutron and deuterium nuclei; iii) interactions of the fast neutrons and reactor constructions; iv) boric acid activation in the heat carrier of the primary circuit. Water leakages from the first circuit into the water tanks can lead to tritium emission from the water of the first circuit. The dissolved fission and activation products are extracted from the heat carrier using ion exchange. This results in the ion-exchange resin generation

818

I. Vasylkivskyi et al.

in water treatment equipment [26]. These resins are periodically replaced, which leads to solid and liquid radioactive waste generation. Leakages from the steam generator of the primary circuit lead to radioactive pollution of water in the secondary circuit. There are also gaseous emissions generated in the primary circuit: aerosols, tritium water vapour, noble gases, and many other gases [22]. When the reactor is stopped for a revision, the pressure of the cooling system is decreased, the reactor lid is removed, and some fuel elements are moved to the storage pool. When removing spent fuel elements, air emissions and liquid radioactive waste may be generated in the storage pool, construction revision mine, and revision mine of protecting pipes. Radioactive aerosols are emitted into the air mainly from the following equipment: i) ventilation tubes of the reactor (emission height is over 100 m); ii) turbine ejector [11]. Such aerosols contain shortlived radionuclides (up to 3 h half-life) and long-lived radionuclides (over 3 h half-life) [11]. The half-life influences the consequences for the human and exposure time. The radioactive aerosols may be inhaled with air or consumed along with water and food. Table 1. Average pollution with radionuclides from the Khmelnytskyi nuclear power plant [33] Isotope

Half-life time

Emission [Ci/day]

Isotope

Half-life time

Emission [Ci/day]

Antimony-129

4.4 h

1.25 × 10^(−8)

Niobium-101

7.1 s

3.04 × 10^(−8)

Argon-41

1.82 h

1.05

Nitrogen-16

7.13 s

2.14

Barium-137 m

2.552 m

1.02 × 10^(−5)

Nitrogen-17

4.17 s

2.98 × 10^(−4)

Bromine-83

2.39 h

3.34 × 10^(−6)

Potassium-42

12.36 h

1.00 × 10^(−5)

Carbon-14

5730 y

1.12 × 10^(−7)

Praziodim-144m

7.2 m

1.62 × × 10^(−11)

Cerium-143

33.0 h

2.36 × 10^(−8)

Rhodium-103m

56.11 m

1.87 × 10^(−7)

Cesium-137

30.20 y

2.74 × 10^(−6)

Rubidium-88

17.8 m

7.96 × 10^(−2)

Chrome-51

27.7 d

6.72 × 10^(−8)

Ruthenium-103

39.25 d

2.06 × 10^(−9)

Cobalt-60

5.27 h

4.68 × 10^(−9)

Selenium-83

22.4 m

2.76 × 10^(−8)

Ferrum-55

2.68 y

2.34 × 10^(−9)

Sodium-24

14.97 h

3.34 × 10^(−7)

Iodine-131

8.01 d

1.91 × 10^(−4)

Strontium-89

50.62 d

3.68 × 10^(−8)

Ittree-90

64.26 h

4.12 × 10^(−11)

Technetium-101

14.2 m

9.84 × 10^(−7) (continued)

Environmental Pollution from Nuclear Power Plants

819

Table 1. (continued) Isotope

Half-life time

Emission [Ci/day]

Isotope

Half-life time

Emission [Ci/day]

Krypton-83m

1.83 h

2.66

Tellurium-129m

33.6 d

1.55 × 10^(−10)

Lanthanum-141

3.92 h

2.14 × 10^(−7)

Tin-130

3.7 m

9.44 × 10^(−8)

Manganese-54

312.2 d

1.66 × 10^(−9)

Tritium

12.33 y

3.22 × 10^(−1)

Molybdenum-99

66.02 h

5.88 × 10^(−10)

Xenon-131m

11.97 d

8.28

Niobium-95m

3.61 d

8.04 × 10^(−11)

Zirconium-95

64.02 d

4.76 × 10^(−7)

The data on radionuclides emission from the ventilation tubes of Khmelnytskyi nuclear power plant (average in 2010–2015 years) can be found in Table 1. Suspended aerosol particles may also emit into the air during transferring the ash and dust. The concentration depends on the surface and meteorological parameters. The highest concentrations of suspended particles in the air near the Khmelnytskyi power plant were measured in the range 3.4–7.7 mg/m3 (Netishyn, 4 km to the north from the plant) and 1.65 mg/m3 (Komarivka, 5 km to the east from the plant) [33]. 3.2 Non-radioactive Emissions of the Khmelnytskyi Nuclear Power Plant Non-radioactive emissions may be released into the atmosphere by the objects and the equipment generating harmful gases during technological processes. At the Khmelnytskyi nuclear power plant, up to 90% of non-radioactive air emissions are generated in the boiler room [33]. Emissions from other sources are low due to the low capacity of gas treatment equipment. Concentrations of some chemicals measured in the air (just over the surface) near the Khmelnytskyi nuclear power plant may be seen in Table 2. At the Khmelnytsjyi nuclear power plant, the monitoring of meteorological and aerial characteristics of the atmosphere is carried out in order to assess the impact of air emissions. These parameters directly influence the radionuclides transportation. Some meteorological parameters reduce the air self-cleaning mechanisms and contribute to the chemicals accumulation in the air. These parameters are as following: temperature stratification of the atmosphere, the velocity and the direction of wind; regime of precipitation, clouds and fogs formation [23]. 3.3 Crop Contamination with Aerosol Radionuclides from Nuclear Power Plant The level of radioactive contamination of crop products and its influence on elements of food chain (animals and humans) are determined by the following [22]: total amount of

820

I. Vasylkivskyi et al.

Table 2. The maximum concentrations of some chemicals measured in the air near the Khmelnytskyi nuclear power plant [33]. Chemical

Limit, [mg/m3 ]

Concentration at the border of the protection area [mg/m3 ]

abrasive-metal dust

0.4

0.05). Emotional intelligence could also significantly positively predict self-efficacy (β = 0.52, p < 0.001, R2 = 0.38). Both mindfulness, emotional intelligence and self-efficacy could significantly positively predict academic achievement (β = 0.17, p < 0.001; β = 0.36, p < 0.001; β = 0.51, p < 0.001; R2 = 0.76).

Fig. 2. The Chain Mediation Model (***means p < 0.001)

Further test of the mediation effect (Table 2) shows that the total effect of mindfulness on university students’ academic achievement is 0.38, of which the direct effect accounts for 42.08%. The total indirect effect of emotional intelligence and self-efficacy in the impact of mindfulness on university students’ academic achievements is 0.22, accounting for 57.92% of the total effect. And its bootstrap 95% CI is [0.12, 0.33], which excludes 0, indicating that emotional intelligence and self-efficacy are the mediating variables in the impact of mindfulness on university students’ academic achievements. This mediation effect is mainly composed of the following two paths: 1) mindfulness → EI → academic achievement (95% CI = [0.08,0.17], SE = 0.02). The mediation effect is 0.12, accounting for 31.82% of the total effect. H4 is supported; 3) mindfulness → EI → self-efficacy → academic achievement (95% CI = [0.05,0.13], SE = 0.02). The mediation effect is 0.09, accounting for 23.61% of the total effect. H9 is supported. Note that the 95% CI of the mediation effect path 2) mindfulness → self-efficacy → academic achievement is [-0.06,0.08] (SE = 0.04), including 0, which indicates that

882

Y. Gao and Y. Chen

the null hypothesis cannot be rejected. This mediation effect path is not significant. H7 is not supported. Table 2. Bootstrap analysis of the mediation effect test Dependent variable

Paths

Effect

Boot SE

Boot CI

Relative effect

Direct effect

1→4

0.16

0.04

[0.08,0.24]

42.08%

Indirect effect1

1→2→4

0.12

0.02

[0.08,0.17]

31.82%

Indirect effect2

1→3→4

0.01

0.04

[−0.06,0.08]

/

Indirect effect3

1→2→3→4

0.09

0.02

[0.05,0.13]

23.61%

Total indirect effect

0.22

0.05

[0.12,0.33]

57.92%

Total effect

0.38

0.06

[0.26,0.50]

100.00%

5 Conclusion and Future Directions By linking mindfulness of university students with their academic achievements, we confirmed that mindfulness, as a trait, relates to students’ academic achievements, which has certain innovation. Different from other personal traits, mindfulness can be cultivated and strengthened through mindfulness training. This means that university students can actively carry out systematic mindfulness training to improve their mindfulness and then improve their academic achievements. This study also confirmed the correlation between mindfulness, EI and academic achievement. We proved the intermediary role of EI in the impact of mindfulness on academic achievement. This conclusion reminds universities to pay attention to the cultivation of students’ non-intelligent abilities. Some students may be unable to get high academic achievements, not because of intellectual defects, but because of bad habits or emotional problems. Universities can introduce mindfulness training to help the worst performers adapt to learning and improve their self-confidence and self-management ability. One conclusion of this study is that there is no significant correlation between university students’ mindfulness and self-efficacy, which is different from the research results of other scholars. The reasons for the difference may be some variables limited by other studies, such as the situation of academic failure or mindfulness training, which is not limited in this study. Therefore, the interpretation of the result is that we can’t prove the relevance between the mindfulness level and academic self-efficacy of usual students who do not receive any mindfulness training. Furthermore, this study tests the mediation effect of self-efficacy, which is also not significant. Anyway, this study verifies the significance of the chain mediation effect of path mindfulness → emotional intelligence → self-efficacy → academic achievement, hoping this conclusion can enlighten future research. This result also implies that though mindfulness has no direct effect on self-efficacy, it may have an indirect effect through emotional intelligence, which needs to be confirmed by further studies.

Mindfulness and Academic Achievements of University Students

883

Although this study has made some achievement, there are still many deficiencies. Firstly, in terms of research methods, this study uses the self-report scale to measure university students’ academic achievements, which gets more from university students’ self-perception. If possible, we would like to use more comprehensive evaluation methods in the future, such as 360°Feedback. Secondly, in terms of research type, the data collected in this study is static data at a certain time. But individual’s mindfulness is changeable and can be improved through mindfulness training. It is hoped that the future research can be dynamic and explore the effect of mindfulness training by collecting time series data. Finally, in terms of data collection methods, the respondents of this study participated voluntarily, which may lead to the problem that the sample cannot effectively represent the whole. It will be better if future research could reduce the error in data collection by stratified sampling.

References 1. Kabat-Zinn, J.: Wherever You Go, There You are: Mindfulness Meditation in Everyday Life. Hyperion, New York (1994) 2. Kirk-Warren, B., Richard-M, R.: The benefits of being present: mindfulness and its role in psychological well-being. J. Pers. Soc. Psychol. 4(84), 822–848 (2003) 3. Rusadi, R.M., Sugara, G.S., Isti’adah, F.N.: Effect of mindfulness-based cognitive therapy on academic grit among university student. Curr. Psychol. 42, 1–10 (2021) 4. Gallego, J., Aguilar-Parra, J.M., Cangas, A.J., Langer, Á.I., Mañas, I.: Effect of a mindfulness program on stress, anxiety and depression in university students. Span. J. Psychol. e109(17), 1–6 (2014) 5. Wang, Y., Li, Y., Huang, Y.: A study on the relationship between university students’ psychological capital, achievement goal orientation and academic achievement. Explor. High. Educ. 06, 128–136 (2011). (in Chinese) 6. Zhong, Y.: An empirical analysis of the influence of class background on university students’ academic achievements. Dev. Eval. High. Educ. 02(28), 108–115 (2012). (in Chinese) 7. Na, Y.: A Study on the Relationship Between Emotional Intelligence, Self-Efficacy and Academic Achievements of University Students. Qufu Normal University, Rizhao (2016). (in Chinese) 8. Li, X., Yang, N.: A study on the structure of the academic achievement scale for university students: its development, reliability and validity. Da Xue (Res. Ed.) 03, 41–53 (2016). (in Chinese) 9. Chapman-Clarke, M.: Mindfulness in the Workplace. Kogan Page, London (2016) 10. Salovey, P., Mayer, J.D.: Emotional intelligence. Imagin. Cogn. Pers. 3(9), 185–211 (1990) 11. Schutte, N.S., Malouff, J.M.: Emotional intelligence mediates the relationship between mindfulness and subjective well-being. Personality Individ. Differ. 7(50), 1116–1119 (2011) 12. Wong, C.-S., Law, K.S.: The effects of leader and follower emotional intelligence on performance and attitude: an exploratory study. Leadersh. Quart. 13(3), 243–274 (2002) 13. Law, K.S., Wong, C.-S., Huang, G.-H., Li, X.: The effects of emotional intelligence on job performance and life satisfaction for the research and development scientists in China. Asia Pac. J. Manage. 25(1), 51–69 (2008) 14. Heidari, M., Morovati, Z.: The causal relationship between mindfulness and perceived stress with mediating role of self-efficacy, emotional intelligence and personality traits among university students. Electron. J. Biol. 4(12), 357–362 (2016)

884

Y. Gao and Y. Chen

15. Jenaabadi, H.: Studying the relation between emotional intelligence and self esteem with academic achievement. Procedia. Soc. Behav. Sci. 114, 203–206 (2014) 16. Bandura, A.: Self-efficacy: The Exercise of Control. Worth Publishers, Incorporated, New York (1997) 17. Chen, X., Ryung, K.J.: The effects of academic self-efficacy and academic emotion regulation on academic achievement among Chinese students in Korea: mediating effect of achievement goal orientation. Korean Assoc. Learner-Centered Curriculum Instr. 18(15), 361–382 (2018) 18. Hanley, A.W., Palejwala, M.H., Hanley, R.T., Canto, A.I., Garland, E.L.: A failure in mind: dispositional mindfulness and positive reappraisal as predictors of academic self-efficacy following failure. Personality Individ. Differ. 86, 332–337 (2015) 19. Lu, C., Pan, F., Fang, F.: The relationship between active and passive procrastination, mindfulness and self-efficacy of university students. J. Shandong Univ. (Med. Ed.) 10(59), 108–113 (2021). (in Chinese) 20. Udayar, S., Fiori, M., Bausseron, E.: Emotional intelligence and performance in a stressful task: the mediating role of self-efficacy. Personality and Individ. Differ. 156(C), 109790 (2020) 21. Saeed, W., Ahmad, R.: Association of demographic characteristics, emotional intelligence and academic self-efficacy among undergraduate students. J. Pak. Med. Assoc. 3, 70 (2020) 22. Cheng, X.-L., Zhang, H., Ma, Y.: The relationship between kindergarten teachers’ teaching mindfulness and job burnout: the chain mediation effect of emotional intelligence and selfefficacy. Res. Preschool Educ. 03, 65–78 (2022). (in Chinese) 23. Chen, S., Cui, H., Zhou, R., Jia, Y.: Revision of mindful attention awareness scale (MAAS). Chin. J. Clin. Psychol. 20, 148–151 (2012). (in Chinese) 24. Liang, Y.: Study on Achievement Goals, Attribution Styles and Academic Self-efficacy of College Students. Central China Normal University, Wuhan (2000). (in Chinese)

Evaluation System of College Students’ Key Abilities Development in the Context of “Integration Education of Specialization, Creation and Morality” Liwei Li(B) , Dalong Liu, and Yanfang Pan Transportation College of Nanning University, Guangxi 530200, China [email protected]

Abstract. In recent years, with the rise of reform in China’s education industry, it has become a major trend to create "golden courses" and eliminate "defective courses". All kinds of curriculum reform models and high-level academic evaluation methods came into being. In this context, this paper deeply analyzes the current situation of academic evaluation research under the teaching innovation reform of "integration of professional education and innovation", "ideological and political curriculum", and combines the development trend of high-level academic evaluation to build a developmental evaluation principle and evaluation system based on the three integration of "professional education, innovation and entrepreneurship education, and ideological and political curriculum". Finally, using the analytic hierarchy process to quantify the weight of each index in the evaluation system, a high-level academic evaluation system suitable for the current curriculum innovation model is established, which provides a certain reference for the training and assessment of comprehensive talents in colleges and universities. Keywords: Specialized and creative integrated education · Ideological and political education · Development evaluation system · Academic assessment

1 Introduction Improving the quality of higher education is a lofty task and historical responsibility entrusted to educators by the times [1]. Since the Ministry of Education of China specially issued the Notice on Implementing the Spirit of the National Undergraduate Education Conference in the New Era in August 2018, "high-level, innovative and challenging" has become the theme vocabulary of undergraduate education in Chinese universities [2]. Under the east wind of this reform, curriculum innovation models such as "integration of expertise and innovation" and "ideological and political curriculum" have become research hotspots, and effective academic evaluation has become a new pain point of curriculum reform in such colleges and universities [3–5]. Academic evaluation is a feedback on students’ development and an important part of teaching quality feedback [6]. The key to academic evaluation is to make a good © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 885–900, 2023. https://doi.org/10.1007/978-3-031-36118-0_76

886

L. Li et al.

diagnosis of the quality of education and teaching, guide students to develop in their comprehensive quality and personality, and provide evidence for the development of in-depth education and teaching. The developmental evaluation of college students’ key abilities, as the core link of high-level academic evaluation, will become the mainstream trend of research. Thereupon, this paper aims to study the principles and indicator system of the developmental evaluation of college students’ key abilities in the context of the integration of "professional education, innovation and entrepreneurship education, and ideological education", simply called three integration. It uses the analytic hierarchy process to design the indicator weights to form an effective evaluation system, which provides some ideas for diagnosing students’ growth and ensuring teaching quality.

2 The Evaluation of College Learning and Its Development in the Context of Teaching Innovation 2.1 Current Situation of Education Reform under the Three Integration In recent years, seeking for the coordinated development of innovative education and professional education, implementing the implementation details and evaluation mechanism of the integration of professional and creative education, has gradually become the focus and difficulty of higher education. From the current research literature, there is little difference between its assessment model and professional education assessment model. Among them, the evaluation model with strong innovation is as follows: Yang Qiuling, Huang Gaoyu and others proposed to establish a credit flexibility system. For example, students who start businesses abroad or study entrepreneurship and innovation courses abroad can convert their entrepreneurial achievements into innovation and entrepreneurship credits [7]; Ni Yu put forward an innovation incentive evaluation mechanism, and carried out incentive evaluation on innovative products, patent applications, awards in competitions, participation in exhibitions, etc. [8]. Gao Yixia and others proposed to combine short-term evaluation with long-term evaluation. The short-term evaluation is based on the spirit of innovation, entrepreneurial awareness and innovation and entrepreneurship ability. The long-term evaluation needs to evaluate the students’ working status, ability and performance in their posts [9]. 2.2 The Status Quo of Academic Evaluation under the Background of “Moral Education” "Curriculum Ideological" refers to taking curriculum as a carrier and moral cultivation as a fundamental educational activity and teaching subject. At present, Xiong Xiaoyi and others have built a quantifiable whole process assessment and evaluation system based on curriculum ideological and political education, and inspected students’ thinking ability through stage tests, research results, case studies, group discussion contribution rate, cooperation and cooperation, and acceptance of new knowledge [10]. Wang Hui and others discussed the significance and design principles of formative assessment applied

Evaluation System of College Students’ Key Abilities Development

887

to the teaching effect evaluation of "curriculum ideological and political education" [11]. Lu Daokun and others solved the problems in the organization of evaluation activities, the construction of evaluation standards and the construction of evaluation method system in the ideological and political curriculum, and used the form of "text evaluation + t eaching observation + customer evaluation" to conduct ideological and moral evaluation through anecdotal records, investigation reports, classroom records, symposiums, logs and other methods related to ideological and political elements [12]. 2.3 Development of High-Level Academic Assessment In the context of continuous innovation in higher education, the curriculum reform has gradually set up high-level goals, while the traditional learning evaluation mechanism will not be able to effectively detect high-level goals. This makes it a new trend of research to focus on the direction of students’ academic development, help them formulate development goals, and achieve self-recognition, self-education and self-development of high-level academic assessment [13–17]. To sum up, there are few studies on the evaluation system of the existing innovative courses of the "integration of professional innovation" and "ideological and political curriculum" types, mainly focusing on teaching design and teaching methods. More innovation was made on the basis of the traditional professional curriculum academic evaluation method, which failed to break through the critical point of the subversive reform of academic evaluation. Looking forward to the future research trend, the curriculum innovation combining "specialty, creativity and thinking" will therefore become an important form of curriculum innovation in China [18, 19], and its high-level academic evaluation system will also become an effective way to test the quality of such courses.

3 Evaluation Principles of Key Abilities of College Students in the Context of “Three Integration” The traditional evaluation method mainly based on "transcripts" leads to "high scores but low abilities" and "incomplete quality" [20]. In particular, in the face of the background of teaching reform such as the integration of specialty and creativity and the ideological and political education of curriculum, the traditional evaluation principles are no longer applicable, and the student evaluation system must be reformed to highlight the developmental function of evaluation and the cultivation of professional ability. Therefore, the new evaluation system should comply with the following basic principles. 3.1 Principle of Diversified Evaluation Contents What deserves full attention is the students’ professional ability, quality and innovation and entrepreneurship ability, while the general memory content can be diluted. In terms of evaluation content, it is necessary to try to incorporate professional knowledge and skills, methodological ability and social ability, emotional attitude and values, innovation awareness and entrepreneurial ability into the evaluation system [20]. The evaluation

888

L. Li et al.

of students should abandon the worship of knowledge, not blindly pursue academic achievements, but also focus on the cultivation and evaluation of students’ comprehensive quality, pay attention to students’ learning interest, emotional experience, psychological quality, values, innovation spirit and practical ability, and return to the essence of seeking college education [21]. 3.2 Principle of Diversity of Evaluation Methods In the teaching of higher education, appropriate testing methods must be adopted according to the characteristics of specialties, disciplines and courses. In addition to the traditional written examination, for the assessment of theoretical knowledge, classroom performance, quick answer and discussion, reports, small papers, structured five question reflection reports, symposiums and other methods can be used. For practical courses, skills operation assessment, virtual simulation, evaluation scale, awards and other methods can be used. Textual assessment methods such as self-evaluation log and personal deeds can be used for moral quality that is difficult to quantify. 3.3 Principle of Open Evaluation Subject It is a desirable method to combine self-assessment with his assessment. The academic assessment covers a wide range of areas and has a long time span. The evaluation of students requires the joint participation of teachers, individuals, classmates, parents, schools and enterprises [22]. 3.4 Principle of Dynamic Evaluation Process The result of teaching is static, while the teaching process and students’ growth are dynamic. Only dynamic process evaluation can effectively evaluate and regulate students’ learning behavior. The evaluation process must combine process evaluation, formative evaluation and summative evaluation, run evaluation through daily teaching, objectively and fairly reflect students’ ability performance in various periods, and give students the opportunity to evaluate many times. 3.5 Principle of Effectiveness of Evaluation Feedback In order to understand the real ability level of students through the original score of the representation, the adjustment model based on latent variables can be used to form a phased feedback mechanism with the ability level as a reference, adjust the scores obtained by students based on the ability factors, and analyze their real ability level through the score representation [23, 24]. Through the understanding of students’ real ability level, immediate feedback and academic intervention are carried out to form an effective and high-quality staged feedback mechanism.

Evaluation System of College Students’ Key Abilities Development

889

4 Construction of the Developmental Evaluation System of College Students’ Key Abilities 4.1 Evaluation System The developmental education evaluation system was proposed by scholars from the School of Education of the British Open University in the 1980s [25]. They believe that the fundamental purpose of evaluation is to promote the development of students. Table 1. Developmental evaluation system of key abilities of college students in the context of three integration Level I indicators

Working ability

Level II indicators Professional knowledge Related knowledge Humanities and social knowledge Professional skill Information collection and processing capacity Problem solving ability

Assessment subject Indi-vidule √ √

√

√

Interpersonal skills

Responsibility Morality Cultivation Humanistic accomplishment Professional ethics Professional norms

enterprise √

√

√

Evaluation carrier score/comments comment comment

√

√

Communication ability

Ideolo-gical literacy

teacher √

√

Learning ability

Team cooperation skills

classmate √

√

√

√

score / Video evaluation score/ comment

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

score / comment comment Video evaluation Video evaluation Video evaluation comment

√

√

√

√

√

√

√

√

√

√

comment comment score / comment score / comment

(continued)

890

L. Li et al. Table 1. (continued)

Level I indicators

Level II indicators Realistic and pragmatic Rigorous and meticulous Self-control Compressive capacity

Innovation and Entrepre-neurship

Critical thinking Observation and judgment ability Divergent thinking Entrepreneurship awareness Innovation and entrepreneurship achievements

Assessment subject Indi-vidule √

classmate

√

√

√

√

teacher √

enterprise

score / comment

√

√

score / comment

√

√

√

√

√

√

√

√

√

√

√

√

Evaluation carrier

Video evaluation Video evaluation comment comment Video evaluation score / comment score / comment

Based on the goal mentioned above, a developmental evaluation system for students’ key abilities in the context of three integration will be developed, as shown in Table 1. 4.2 Index System and Weight Design Based on Analytic Hierarchy Process After the establishment of the developmental evaluation system for students’ key abilities, it is necessary to further scientifically plan the weight of evaluation indicators [26] in order to clarify the evaluation priorities and highlight the emphasis of the evaluation system on talent cultivation under the new curriculum reform model. Based on the characteristics that AHP can deal with complex decision-making problems with multiple objectives, multiple criteria or no structural characteristics, this method is selected to calculate the index weight and verify its effectiveness [27]. The specific steps are described below. The first step is to establish a ladder structure according to the characteristics of talents’ abilities under the curriculum reform mode of "specialty, creativity and thinking", as shown in Fig. 1. The second step is to establish the judgment matrix A of each step through expert scoring method, as shown in Formula (1). def

A = (aij )

(1)

Evaluation System of College Students’ Key Abilities Development

891

Fig. 1. Ladder structure of student ability evaluation system in the context of three integration

In Formula (1) aij refers to the importance scale of element i compared with element j. The 1, 3, 5, 7, 9, 2, 4, 6, 8 and their reciprocal in the scale are used to judge the importance of adjacent elements respectively. The larger the value is, the higher the importance of the former is than that of the latter. Table 2. Basic Level Judgment Matrix Z

A1

A2

A3

A1

1

3

3

A2

1/3

1

1

A3

1/3

1

1

The scoring experts were scored by 30 students, 8 teachers and 4 enterprise technicians, totaling 42 people. Finally, the following four judgment matrices were obtained, as shown in Tables 2, 3, 4 and 5.

892

L. Li et al. Table 3. A1 Judgment Matrix

A1

A11

A12

A13

A14

A15

A16

A17

A18

A19

A20

A11

1

3

3

A12

1/3

1

1

1/5

1/3

1/5

1/5

1

1/7

1/5

4

2

2

1/5

2

2

1

A13

1/3

1

1

1/5

1

1/7

1/5

1

1

1

A14

5

5

5

1

3

1/5

1

3

3

3

A15

3

1

1

1/3

1

1/5

1/5

1

1

1

A16

5

7

7

5

5

1

2

3

3

3

A17

5

5

5

1

5

1/2

1

4

4

4

A18

1/4

1/2

1

1/3

1

1/3

1/4

1

1

1

A19

1/2

1/2

1

1/3

1

1/3

1/4

1

1

1

A20

1/2

1

1

1/3

1

1/3

1/4

1

1

1

Table 4. A2 Judgment Matrix A2

A21

A22

A23

A24

A25

A26

A27

A28

A29

A21

1

1/3

1

1/5

1/3

1

1

1

1

A22

3

1

2

1

1

2

2

3

2

A23

1

1/2

1

1/5

1/3

1/3

1/3

1

1

A24

5

1

5

1

1

1

2

4

2

A25

3

1

3

1

1

2

1

1

1

A26

1

1/2

3

1

1/2

1

1/3

1

1

A27

1

1/2

3

1/2

1

3

1

3

2

A28

1

1/3

1

1/4

1

1

1/3

1

1

A29

1

1/2

1

1/2

1

1

1/2

1

1

Table 5. A3 Judgment Matrix A3

A31

A32

A33

A34

A35

A31

1

2

1

1/3

1/3

A32

1/2

1

1/3

1/3

1/5

A33

1

3

1

2

1/5

A34

3

3

1/2

1

1/3

A35

3

5

5

3

1

Evaluation System of College Students’ Key Abilities Development

893

The third step is to use SPSSPRO software to complete the calculation of the subsequent AHP, and calculate the eigenvector, weight value, maximum eigenvalue, CI value, CR value, etc. The calculation results are shown in Tables 6, 7, 8 and 9. Table 6. AHP hierarchy analysis and consistency test results of the first criterion layer Items

Eigenvector

Weight (%)

Maximum characteristic root

CI

RI

CR

Consistency inspection

Working ability A1

2.08

60

3

0

0.525

0

Pass

Ideological literacy A2

0.693

20

Innovation and Entrepreneurship A3

0.693

20

Note: The calculation result of the analytic hierarchy process shows that the maximum characteristic root is 3.0. According to the RI table, the corresponding RI value is 0.525, so CR = CI/RI = 0.0 < 0.1, passing the one-time test

Table 7. AHP hierarchy analysis and consistency test results of the second criterion level (working ability) Items

Eigenvector

Professional knowledge A11

0.909

6.986

Related knowledge A12

0.614

4.721

Humanities and social knowledge A13

0.535

4.11

Professional skill A14

2.141

16.461

Information collection and processing capacity A15

0.725

5.572

Problem solving 3.564 ability A16

Weight (%)

Maximum characteristic root

CI

RI

CR

Consistency inspection

10.987

0.11

1.486

0.074

Pass

27.4 (continued)

894

L. Li et al. Table 7. (continued)

Items

Eigenvector

Weight (%)

Learning ability A17

2.692

20.697

Communication ability A18

0.568

4.364

Interpersonal skills A19

0.608

4.677

Team cooperation skills A20

0.652

5.013

Maximum characteristic root

CI

RI

CR

Consistency inspection

Note: The calculation result of the analytic hierarchy process shows that the maximum characteristic root is 10.987. According to the RI table, the corresponding RI value is 1.486, so CR = CI/RI = 0.074 < 0.1, passing the one-time test Table 8. AHP level analysis and consistency test results of the second criterion level (ideological literacy) Items

Eigenvector Weight Maximum CI (%) characteristic root

Responsibility A21

0.655

6.61

Morality Cultivation A22

1.737

17.526

Humanistic 0.537 accomplishment A23

5.416

Professional ethics A24

1.946

19.633

Professional norms A25

1.379

13.91

Realistic and pragmatic A26

0.857

9.572

RI

CR

Consistency inspection

0.072 1.451 0.049 Pass

8.649 (continued)

Evaluation System of College Students’ Key Abilities Development

895

Table 8. (continued) Items

Eigenvector Weight Maximum CI (%) characteristic root

Rigorous and meticulous A27

1.335

13.473

Self-control A28 0.672

6.775

Compressive capacity A29

8.008

0.794

RI

CR

Consistency inspection

Note: The calculation result of the analytic hierarchy process shows that the maximum characteristic root is 9.572. According to the RI table, the corresponding RI value is 1.451, so CR = CI/RI = 0.049 < 0.1, passing the one-time test Table 9. AHP level analysis and consistency test results of the second criterion level (Innovation and entrepreneurship capability) Items

Eigenvector Weight Maximum CI (%) characteristic root

Critical thinking A31

0.74

11.896

Observation and judgment ability A32

0.407

6.534

Divergent thinking A33

1.037

16.667

Entrepreneurship 1.084 awareness A34

17.428

Innovation and entrepreneurship achievements A35

47.475

2.954

5.37

RI

CR

Consistency inspection

0.093 1.11 0.083 Pass

Note: The calculation result of the analytic hierarchy process shows that the maximum characteristic root is 5.37. According to the RI table, the corresponding RI value is 1.11, so CR = CI/RI = 0.083 < 0.1, passing the one-time test

The fourth step is to calculate the total weight. Fill the weight value calculated by AHP into Table 10, and calculate the final specific weight value of each secondary index to get the calculation result in Table 10.

896

L. Li et al.

Table 10. The weight of developmental evaluation indicators of college students’ key abilities in the context of three integration Target layer

Criterion layer I

Criterion layer II

Weight

vocational ability Z

working ability A1 (0.6)

Professional knowledge A11 (0.06986)

0.042

5

Related knowledge A12 (0.04721)

0.028

10

Humanities and social knowledge A13 (0.0411)

0.025

13

Professional skill A14 (0.16461)

0.099

3

Information collection and processing capacity A15 (0.05572)

0.033

8

Problem solving ability A16 (0.274)

0.164

1

Learning ability A17 (0.20697)

0.124

2

Communication 0.026 ability A18 (0.04364)

12

Interpersonal skills A19 (0.04677)

0.028

10

Team cooperation skills A20 (0.05013)

0.030

9

Responsibility A21 (0.0661)

0.013

18

Morality Cultivation A22 (0.17526)

0.035

7

Humanistic 0.011 accomplishment A23 (0.05416)

19

Ideological literacy A2 (0.2)

Weight ranking

Professional ethics A24 (0.19633)

0.039

6

Professional norms A25 (0.1391)

0.028

10 (continued)

Evaluation System of College Students’ Key Abilities Development

897

Table 10. (continued) Target layer

Criterion layer I

Innovation and entrepreneurship A3 (0.2)

Criterion layer II

Weight

Weight ranking

Realistic and pragmatic A26 (0.08649)

0.017

15

Rigorous and meticulous A27 (0.13473)

0.027

11

Self-control A28 (0.06775)

0.014

17

Compressive capacity A29 (0.08008)

0.016

16

Critical thinking A31 0.024 (0.11896)

14

Observation and 0.013 judgment ability A32 (0.06534)

18

Divergent thinking A33 (0.16667)

0.033

8

Entrepreneurship awareness A34 (0.17428)

0.035

7

Innovation and entrepreneurship achievements A35 (0.47475)

0.095

4

4.3 Analysis on the Results of Developmental Evaluation of Students’ Key Abilities It can be seen from the weight value of the first level indicators that, for the evaluation of college students’ professional ability, “work ability A1”, as the latest basic ability in the workplace, has been widely recognized and is still the main parameter and indicator to measure students’ ability development. However, the requirements for “ideological and political literacy A2” and “innovation and entrepreneurship ability A3” are relatively low. On the one hand, there are few requirements for this from the job functions of enterprises. On the other hand, colleges and universities have only begun to pay attention to how to integrate moral education and mass entrepreneurship education into professional education in recent years. The implementation path is still in the process of exploration, and the training mechanism for these two types of abilities is not yet mature. With the

898

L. Li et al.

development of connotation construction of higher education and the change of enterprises’ demand for employees’ ability, the proportion of ideological and political literacy and mass entrepreneurship and innovation ability should be gradually increased. From the perspective of secondary indicators, for the 10 indicators under “working ability A1”, the most concerned ones are “problem-solving ability A16”, “learning ability A17” and “professional skills A14”, which indicates that in terms of working ability, higher education is mainly recognized to cultivate students’ ability to deal with problems independently, followed by professional skills. In addition to the above three most important indicators, “Innovation and Entrepreneurship Achievement A35”, “Professional Knowledge A11”, “Professional Ethics A24” and “Morality A22” are the most concerned. This also poses a new challenge to the development of higher education. The essence of education urges us to gradually abandon the worship of knowledge, and upgrade education evaluation from the once narrow knowledge evaluation to “meaning generation”. The diversified and comprehensive value evaluation standard has become the mainstream trend.

5 Conclusion This paper mainly studies the developmental evaluation system of students’ key abilities under several new curriculum reform models, and introduces the analytic hierarchy process to design the index weight. The main research work includes the following aspects. (1) This paper analyzes the current situation of academic evaluation in colleges and universities in the context of teaching innovation, such as "special integration" and "ideological and political curriculum". (2) According to the characteristics of the curriculum reform of "professional education, innovation and entrepreneurship education, and ideological and political curriculum", the evaluation principles of key abilities of college students are constructed. (3) This paper constructs the evaluation system of college students’ key abilities, and uses the analytic hierarchy process to scientifically define the index weights. In order to promote the development of students’ comprehensive professional ability, a targeted evaluation system has been developed. There are still some deficiencies in this study. The background is relatively limited, mainly focusing on the curriculum innovation model of "specialization, creativity and thinking". And the indicator system constructed only focuses on key capabilities. Acknowledgment. This paper is supported by (1) Guangxi Higher Education Undergraduate Teaching Reform Project “Research and Practice of High-Level Academic Evaluation based on the integration of “specialty, creativity and thinking”” (2021JGB431) and (2) Nanning University’s College Level Specialty Innovation Integration Teaching Reform Project “the Teaching Reform and Practice of Modern Logistics Equipment Course Combining the Work Process Integration and Specialty Innovation Integration Under the Mixed Teaching Mode” (2020XJZC05).

Evaluation System of College Students’ Key Abilities Development

899

References 1. Wang, F.: The course design for algorithm analysis and research on improving the students’ comprehensive ability. Int. J. Educ. Manage. Eng. (IJEME) 2(2), 42–46 (2012) 2. Hu, T., Li, X.: The value orientation, blocking factors and promotion strategies of the “integration of expertise and innovation” of application-oriented undergraduate universities. Heilongjiang High. Educ. Res. 40(12), 127–131 (2022). (in Chinese) 3. Dai, Y.: Research on innovative basic case design path of “thinking, expertise and innovation” integration. Dalian Cadre J. 38(11), 47–52 (2022). (in Chinese) 4. Boysen, M.S.W., Jansen, L.H., Knage, M.: To share or not to share: a study of educational dilemmas regarding the promotion of creativity and innovation in entrepreneurship education. Scand. J. Educ. Res. 64(2), 211–226 (2020) 5. Hameed, I., Irfan, Z.: Entrepreneurship education: a review of challenges, characteristics and opportunities. Entrepreneurship Educ. 2(3), 135–148 (2019) 6. Zhang, Z.: Research on the construction of “accounting foundation” golden curriculum from the perspective of integration of thinking, professional training and innovation. Sci. Technol. 31, 97–99 (2022). (in Chinese) 7. Liu, X., He, S., Liu, G.: The construction and practice of the talent training mode of higher vocational education with the triple integration of “thinking and innovation, professional innovation and scientific innovation” in the context of smart cities - taking computer application technology as an example. J. Guangdong Commun. Vocat. Tech. Coll. 21(03), 53–57+61 (2022). (in Chinese) 8. Da, L., Chen, T., Wang, L.: The construction and practice of “thinking, expertise and innovation” integrated innovation and entrepreneurship education system based on OBE concept. Comput. Educ. 08, 128–132 (2022). (in Chinese) 9. Ji, D., Liu, H.: Research on the design of the teaching work page of metal materials, which integrates professional, creative and ideological education. Ind. Technol. Forum 21(15), 165– 166 (2022). (in Chinese) 10. Wang, J.: The development and practice of the loose leaf teaching work page of the integration of specialty, creativity and education – taking the course “Welding Structure Production” as an example. Sci. Technol. 21, 41–43 (2022). (in Chinese) 11. Ji, D., Zhang, H.: Research on the micro organization teaching design of metal materials based on the integration of professional, creative and ideological education. Ind. Technol. Forum 21(14), 232–233 (2022). (in Chinese) 12. Zhu, L., Duan, Z.: From knowledge worship to meaning generation: a new interpretation of students’ academic evaluation. Contemp. Educ. Sci. 05, 27–33 (2022). (in Chinese) 13. Yan, B., Zhang, X.: Research on characteristics of applied undergraduate education quality evaluation system – based on the theoretical background of developmental education evaluation. Teachers’ Expo 36, 4–7 (2021). (in Chinese) 14. Jiang, S.: Green evaluation: transformation and exploration of academic evaluation. Jiangsu Educ. 87, 61–63 (2021). (in Chinese) 15. Kuang, S., Lan, Y., He, M., Lu, Y., He, Z.: Assessment of academic intelligence: current situation and trend. Educ. Inf. Technol. Z2, 8–14 (2021). (in Chinese) 16. Wang, J.: Core literacy oriented academic assessment framework: Australia national science literacy assessment project. China Exam 04, 60–67 (2021). (in Chinese) 17. Tan, L., Zhang, Y.: Preliminary exploration and practice of scientific modeling ability examination in large-scale academic assessment. Shanghai Educ. Res. 03, 27–32 (2021). (in Chinese)

900

L. Li et al.

18. Dong X.: Research on the integration of curriculum ideological and political education, innovation and entrepreneurship education into professional curriculum system in higher vocational colleges under the new engineering background. In: MATEC Web of Conferences. EDP Sciences, p. 355 (2022) 19. Zhao, X., Zhang, J.: The analysis of integration of ideological political education with innovation entrepreneurship education for college students. Front. Psychol. 12, 610409 (2021) 20. Liu, H., Zhang, P., Pan, J.: How to make the results of academic evaluation more effective - Research on the adjustment model based on latent variables. J. East China Normal Univ. (Educ. Sci. Ed.) 36(03), 87–98+169 (2018). (in Chinese) 21. Huang, J.: The new trend of international basic education science assessment – a comparative study based on PISA and TIMSS. Training Prim. Secondary Sch. Teachers 09, 74–78 (2017). (in Chinese) 22. Feistauer, D., Richter, T.: How reliable are students’ evaluations of teaching quality? A variance components approach. Assess. Eval. High. Educ. 42(8), 1263–1279 (2017) 23. Sun, G.: Research on academic assessment of Australian Higher Education – a case study of the Royal Melbourne University of Technology. Explor. High. Vocat. Educ. 16(04), 11–17 (2017). (in Chinese) 24. Annabestani, M., Rowhanimanesh, A., Mizani, A., et al.: Fuzzy descriptive evaluation system: real, complete and fair evaluation of students. Soft Comput. 24(6), 3025–3035 (2020) 25. Zeng, Y., Qu, Y., Chen, T., Liu, R.: The academic assessment system of logistics management professional ability evaluation and key ability training. China Market 41, 163–166 (2010). (in Chinese) 26. Shen, L., Yang, J., Jin, X., et al.: Based on Delphi method and analytic hierarchy process to construct the evaluation index system of nursing simulation teaching quality. Nurse Educ. Today 79, 67–73 (2019) 27. Shi, J., Dong, C., Hou, J.: Research and design of teaching evaluation system based on fuzzy model. Int. J. Educ. Manage. Eng. (IJEME) 2(10), 45–51 (2012)

Exploration and Practice of the Evaluation System of College Teaching Quality Based on MSF Mechanism in OBE Perspective Yanzhi Pang and Jianqiu Chen(B) School of Transportation, Nanning University, Nanning 530200, Guangxi, China [email protected]

Abstract. In response to the requirements of “all-staff, whole-process and allround” education, colleges and universities have increasingly higher requirements for the quality of classroom teaching. From the former “teaching” on the standpoint of teachers to the present student-centered, improve students’ autonomous learning ability, and pay increasing attention to the quality of curriculum teaching. The data of supervision and evaluation of teaching, peer assessment and student evaluation of teaching are also receiving increasingly more attention. Based on the OBE concept and focusing on the development of students, this paper constructs the evaluation system of college curriculum teaching quality, and optimizes the evaluation indicators. It should not only show solicitude for the achievement of knowledge and ability objectives, but also pay attention to quality objectives, emphasize the combination of teaching with the development of industry and enterprises, optimize the curriculum design of talent training programs, and reflect and improve students’ learning interest and enthusiasm in the classroom. Keywords: OBE · MSF mechanism · Teaching quality · Evaluation system

1 Introduction Quality is the guarantee of education success [1]. It is not enough to call teaching quality as the lifeline of schools, which has been agreed by all countries in the world [2–4]. Teaching quality plays an important role in ensuring the sustainable development of the school, shaping the successful life of students, and strengthening the comprehensive national strength of the country [5, 6]. Focusing on the growth of students, the OBE concept has been widely accepted and promoted in different disciplines [7–9]. The so-called “golden class” construction requirements of “high-level, innovative and challenging - abbreviated as HIC” need to be completed by teachers and students. It is urgent to establish a multi-dimensional evaluation system of college curriculum teaching quality [10–12]. At present, the teaching quality evaluation system pays more attention to teachers’ teaching, neglects students’ learning, emphasizes knowledge teaching and ignores ideological guidance, lacks logic rigor, teachers’ teaching characteristics are not obvious, and the evaluation system is not sustainable. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 901–912, 2023. https://doi.org/10.1007/978-3-031-36118-0_77

902

Y. Pang and J. Chen

In order to deal with the above problems, this paper intends to optimize the undergraduate course teaching quality evaluation table by establishing the framework of the course teaching quality evaluation system based on the OBE concept, introduce the MSF mechanism, and continuously improve the teaching quality evaluation indicators from both teachers and students. Through the improvement of teaching mode, teaching method and teaching design, improve the quality of teachers’ “teaching” and the degree of students’ “learning”. According to the strategic goal of “students are satisfied with the learning effect, and employers are satisfied with the quality of training talents”, the students’ practical ability, innovation ability, and the ability to serve enterprises in the industry and the society as the main criteria for evaluating the teaching quality, so as to set the evaluation indicators and control the classroom teaching quality in an all-round way [13–15]. According to the specific requirements for the specific measures of teaching quality evaluation in colleges and universities in the Overall Plan for Deepening the Reform of Education Evaluation in the New Era issued in 2020, as the basis for revising the teaching evaluation indicators, improving the efficiency of education and teaching evaluation, and reforming the current management of teaching quality in higher education [16–18].

2 Main Problems and Cause Analysis of Current Classroom Quality Evaluation With the specialty and curriculum construction as the core, and the integration of production and education as the center, we should reform the training mode of applicationoriented talents and cultivate high-quality application-oriented talents. Therefore, the main indicator of the evaluation of classroom teaching quality in colleges and universities is the quality of talent training. For curriculum construction, the Ministry of Education proposed that colleges and universities should eliminate “water courses” and build “gold courses”. And high order, innovation and challenge. To ensure the teaching quality of the “Golden Class” project, it is necessary to propose quantifiable evaluation criteria, optimize the evaluation indicators, and reflect the complementary process of teacher and student growth [19]. At the same time, the evaluation system of teaching quality is related to the quality of students, teachers, orientation of running a school and characteristics of talent cultivation in colleges and universities. With the gradual deepening of teaching reform, the existing evaluation system also highlights some common problems. 2.1 Main Problems in Current Classroom Quality Evaluation The main problems in the evaluation of classroom quality currently are summarized as follows. (1) Focus on “teaching” rather than “learning” The traditional teaching process and methods have been gradually changed. Despite the teaching reform in recent years, cramming teaching has been reduced, but in the selection of classroom teaching quality evaluation indicators, too much

Exploration and Practice of the Evaluation System

(2)

(3)

(4)

(5)

903

attention is still paid to teachers. The selection of indicators also mainly depends on the preparation of teaching materials, the implementation of teaching plans, the relationship between teaching content and the forefront of the industry, and other aspects, which does not fully reflect the “student development centered” teaching philosophy, The real sense of teaching students according to their aptitude has not been realized. The selection of classroom quality evaluation indicators is one-sided, ignoring the impact of teaching objectives, teaching content, teaching methods, etc. on students’ learning interest and enthusiasm, and lacking the measurement of students’ effectiveness. Value knowledge over thought In the process of teaching, teachers usually define knowledge objectives, but with the introduction of “HIC”, the requirements for students’ quality objectives and emotional objectives are also on the agenda. In terms of knowledge objectives, the teaching content is required to have certain depth and extensibility; The ability goal requires a breakthrough in career selection and ability expansion; Emotional objectives For engineering students, it is required to add moral elements to the teaching content, including the integration of patriotism, craftsmen of big countries, etc. At present, there is little consideration of the ideological and political content of the curriculum in the evaluation indicators of classroom teaching quality, and the effectiveness is basically lack of examination points. Insufficient layering and logic The establishment of classroom teaching quality evaluation system indicators needs certain scientific basis. Most of the previous evaluations were based on previous experience, and there was no clear basis for setting the weights of indicators. At the same time, the correlation and level of indicators were not obvious. The teaching characteristics of teachers are not prominent Some teachers rely too much on PPT when teaching, and the teaching content is not updated in time. They can not actively find, find and fully reflect new ideas, new ideas, new methods and new technologies in the classroom. The content of teaching materials is relatively lagging behind, resulting in some teaching content being out of touch with the reality of social development in varying degrees. There are also some teachers who are not active enough in using modern information technology, resources and means and have weak operational ability. The research on new classroom teaching modes such as “Internet plus”, “micro class” and “flipped classroom” is not deep enough, and some high-quality online teaching resources introduced by the school do not give full play to their effectiveness. At the same time, teachers have different teaching styles because of their different professional backgrounds. The improvement of teaching quality needs to guide and encourage teachers to form their own teaching characteristics. Among the current evaluation indicators of classroom teaching quality, there is less evaluation on teachers’ teaching method innovation. Lack of long-term, standardized and scientific evaluation and improvement system At the end of the semester, most colleges and universities will require students to evaluate the teaching of all courses anonymously, which has a certain promotion effect on the improvement of teaching quality, but because the evaluation indicators are not accurate enough, the process is relatively lagging behind, and the students’

904

Y. Pang and J. Chen

subjectivity is too large, the results can not objectively reflect the teaching effect, and the follow-up improvement measures are not continuous and closed loop, and there is a lack of feedback on the improvement results. 2.2 Cause Analysis Although the teaching quality evaluation system of colleges and universities has been continuously improved, the degree of attention needs to be strengthened. The evaluation indicators of classroom teaching quality in colleges and universities are closely related to the school’s top objectives, professional training objectives, and curriculum achievement objectives, but the degree of fit and goal achievement is not high enough, so it can not fully support the achievement of learning objectives. The evaluation object is still mainly teachers, and the results can only be fed back to relevant teachers after finishing the evaluation at the end of the semester. The course of the next semester is different from that of the current semester, which leads to the open-loop teaching evaluation. Problems found in the teaching process cannot be fed back and improved in time, and the second evaluation result of a certain course by the same group of students cannot be obtained. Such evaluation of teaching is not sustainable. Without the interaction between teachers and students, it is difficult to really play its role.

3 Construction of Teaching Quality Evaluation System Based on OBE Concept 3.1 Significance of Construction of Teaching Quality Evaluation System To build the evaluation system of classroom teaching quality, we must follow the concept of OBE, adhere to the student-centered principle, timely communicate the information of teaching and learning, give full play to the role of students in the evaluation of classroom teaching effect, establish a scientific view of quality, and form a more scientific and effective evaluation system [20]. At the same time, it can urge teachers to find the deficiencies in the teaching process in time, pay more attention to teaching, research teaching, constantly optimize the teaching process, and improve the teaching quality [21]. By optimizing the evaluation indicators and meeting the requirements of higher education teaching quality management policy guidelines in the new era, the implementation of the system can be improved, so that the key factors and key links affecting the teaching quality are always under control in the whole process of talent training, so as to survive with quality and seek development with quality. Build and gradually improve the teaching quality evaluation system, and form a long-term mechanism to continuously improve the quality of talent training [22–24].

Exploration and Practice of the Evaluation System

905

3.2 Formation Process of Classroom Teaching Quality under OBE Concept The OBE concept emphasizes output-oriented, and based on the orientation of the school, requires the development of a reasonable talent training program. Each course should serve the professional training objectives. The teaching objectives of each class and the learning objectives of students need to support the achievement of professional talent cultivation. The formulation and implementation of teaching plans must also be consistent with them. The selection of classroom teaching evaluation indicators should be consistent with the teaching implementation, and the established requirements should be finally completed, as shown in Fig. 1.

Fig. 1. Formation process of classroom quality under OBE concept

3.3 Evaluation Indicators of Undergraduate Classroom Teaching Quality Optimize the existing undergraduate course teaching quality evaluation form from the aspects of teaching concept, teaching content, teaching mode, teaching effect, teaching characteristics, etc. The design principles of evaluation indicators are shown in Table 1.

906

Y. Pang and J. Chen Table 1. Evaluation form of undergraduate class teaching quality

Teaching quality evaluation Teaching philosophy

Meet the requirements of disciplines and courses Pay attention to the ideological education in curriculum, and reflect the cultivation of morality and talents Keep “student-centered, output-oriented and continuous improvement”

Resources material

Carry teaching materials; Master teaching hour to consistent with the schedule

Teaching attitude

Take classes on time; Lecture with full spirit; Be strict with students and keep the classroom in good order

Content of courses

With HIC characteristics, in line with the requirements of the syllabus The teaching content has depth and breadth, reflects the frontier of the discipline, and penetrates professional ideas Highlight key and difficult points, reasonable logic, clear structure and complete teaching links Clear learning objectives, reflecting Bloom’s educational ideas, and multi-level curriculum design The goal of moral education is clear, guiding students to establish ideals and beliefs as well as correct outlook on world, life and values

Teaching implementation Effectively integrate ideological and moral content into the curriculum Clear teaching, combined with industry development trends The teaching methods are rich, and various media and means are used to stimulate students’ interest in learning, with good results (continued)

Exploration and Practice of the Evaluation System

907

Table 1. (continued) Observe students’ reactions and carry out interactions, discussions, exercises, activities, etc.; Cultivate students’ autonomous learning ability Reasonable and effective use of modern information technology to support teaching innovation Teaching effective-ness

The teaching is attractive, the classroom atmosphere is harmonious, the students are active in thinking and deeply participate in the classroom Students have a high attendance rate, listen carefully, take notes, and actively participate in learning activities such as interaction and practice Realize the preset learning objectives and improve students’ knowledge, ability and quality; The assessment method is reasonable

Teaching features

Effectively reflect the OBE concept Has obvious innovative characteristics Make full use of modern information technology to carry out course teaching activities and learning evaluation

4 Test and Evaluation of Teaching Quality Under MSF Mechanism 4.1 Meaning and Role of MSF MSF is the abbreviation of Midterm Student Feedback. It is a service provided by the teaching management department for teachers in the whole school. It aims to help teachers find problems and find solutions from the perspective of students, so as to improve teachers’ teaching level and classroom teaching effect. MSF is a tool and method for teaching evaluation and diagnosis, which is generally carried out in the early or middle stage of the course. The main body of participation is teachers. Through the teaching consultation activities, the relevant professional teaching consultants and teachers communicate one-on-one, and participate in the teaching activities of teachers synchronously. They collect the information of teachers and students in a multi-dimensional way, collect the real data of students, collect the concerns

908

Y. Pang and J. Chen

of students, help teachers reflect on teaching, improve teaching strategies and methods, and optimize the quality of classroom teaching. The flow of MSF is shown in Fig. 2.

Fig. 2. MSF process

4.2 MSF Process and Steps The implementation process of MSF mainly includes four steps: early talks, course observation, late talks and late tracking. 4.2.1 Preliminary Talks (30 m) The counselor understands the basic information of the course and class, informs the teacher of the specific practice of MSF, lets the teacher understand the whole process, and establishes a relationship of mutual trust. This process takes about 30 min. 4.2.2 Class Observation (55–90 m) The counselor will go to the classroom for 1–2 lessons (perhaps 50–90 min) at the appointed time. The teacher will set aside about 15 min. The counselor will organize students to talk and collect students’ feedback on the course and class (anonymous) when the teacher leaves. 4.2.3 Later Talks (30 m) During the 30-min conversation, the counselor will submit an analysis report to the teacher on the classroom observation and students’ feedback, and discuss with the teacher how to improve the teaching and classroom effect. 4.2.4 Post Tracking At the end of the semester, the counselor will further understand the classroom improvement from teachers and students, evaluate the satisfaction of MSF service, and put forward suggestions for improvement.

Exploration and Practice of the Evaluation System

909

4.3 Construction of Classroom Teaching Quality Evaluation System Model Based on the existing data and combined with the required basic functions, the overall framework of the classroom teaching quality evaluation system is formed as shown in Fig. 3.

Fig. 3. Model of course teaching quality evaluation system

The curriculum teaching quality evaluation system model mainly includes four parts: user information, data management, teaching evaluation module and history query. Query and modify the data of students’ evaluation of teaching, peer evaluation of teaching and supervision evaluation of teaching in order to help teachers improve the quality of curriculum teaching. Through research and material collection, analyze the problems raised in Sect. 2.1, organize experts to discuss and deduce, and propose the implementation plan of education reform based on OBE concept. Introduce the MSF mechanism, implement it in the classroom and promote it in an all-round way to improve the degree of achievement of teachers’ “teaching” and students’ “learning”. The specific implementation process is shown in Fig. 4. 4.4 Case Analysis Optimize the curriculum system from the following four aspects: (1) Combine the current informatization characteristics, add courses related to transportation informatization, learn the frontier knowledge of disciplines, and reflect the interdisciplinary characteristics, such as intelligent transportation system. (2) Combined with the development trend

910

Y. Pang and J. Chen

Fig. 4. Specific implementation process

of the industry, add comprehensive transportation courses, such as comprehensive transportation. (3) Integrating geographical features, Thai courses are offered to prepare for serving ASEAN and the the Belt and Road, such as applying Thai and ASEAN culture. (4) A template for achieving carbon peak and carbon neutrality, and green transportation courses, such as traffic safety, are added. Combined with curriculum optimization, the introduction of MSF mechanism has greatly improved teachers’ teaching level, mobilized students’ learning interest and initiative, and improved students’ evaluation scores for project quality, as shown in Table 2. Table 2. Data of students’ evaluation of teaching in recent three years Academic year

Total number of courses

Number of “excellent” courses

Number of “good” courses

Number of “medium” courses

Number of “poor t” courses

Excellence rate

2018–2019

62

51

11

0

0

82.26%

2019–2020

59

50

9

0

0

84.75%

2020–2021

219

195

20

4

0

89.04%

Exploration and Practice of the Evaluation System

911

5 Conclusion Classroom teaching in colleges and universities is an important link in cultivating highquality college students. To evaluate the quality of classroom teaching is to find out the advantages that need to be adhered to and developed in the teaching process, and find out the shortcomings that need to be improved. Using MSF mechanism and method to carry out teaching evaluation diagnosis in the early or middle stage of the course is to help teachers find and solve problems from the perspective of students. This paper summarizes the existing problems in the evaluation system of classroom teaching quality in colleges and universities, analyzes the reasons, constructs the formation process of classroom teaching quality based on the OBE concept, optimizes the evaluation indicators of classroom teaching in undergraduate courses, encourages teachers to participate in the MSF teaching diagnosis evaluation, establishes the evaluation system of classroom teaching quality, and gives the specific implementation process. According to the requirements of HIC, the optimized evaluation dimensions and indicators are used to guide the improvement of teaching quality. Improve the overall quality of teaching and learning through the optimization of the curriculum system, and verify it through the teaching evaluation data. At the same time, in compliance with the requirements of the education and teaching reform in the new era, the implementation of moral education and the cultivation of people, the teaching quality evaluation system, combined with ideological and political education, has achieved progress with the times and continuously improved the effect of education. Acknowledgment. This research is supported by 4 projects: (1) The Sub-Project of the Construction of the China-ASEAN International Joint Laboratory for Integrated Transportation (Phase I): “Research on the Technology of Fully Automatic Operation Training” (GK-AA21077011-7); (2) Guangxi Higher Education Undergraduate Teaching Reform Project: “Research and Practice on the Evaluation System of Curriculum Teaching Quality in Applied Universities from the OBE Perspective” (2022JGA390); (3) Nanning University Curriculum Ideological and Political Demonstration Specialty Construction -Transportation Specialty- Project” (2022SZSFZY03); and (4) Core Undergraduate Course of Nanning University: “Train Operation Control System” (2020BKHXK16).

References 1. Hamada, H.: Action research to enhance quality teaching. Arab World Engl. J. 1, 4–12 (2019). Conference Proceedings 2. Zhiqin, L., Jianguo, F., Fang, W., Xin, D.: Study on higher education service quality based on student perception. Int. J. Educ. Manag. Eng. (IJEME) 2(4), 22–27 (2012) 3. Kong, L.A., Wang, X.M., Yang, L.: The research of teaching quality appraisal model based on AHP. Int. J. Educ. Manag. Eng. (IJEME) 9(29), 29–34 (2012) 4. Weiss, K.A., McDermott, M.A., Hand, B.: Characterising immersive argument-based inquiry learning environments in school-based education: a systematic literature review. Stud. Sci. Educ. 58(1), 15–47 (2022) 5. Nawaz, R., Sun, Q., Shardlow, M., et al.: Leveraging AI and machine learning for national student survey: actionable insights from textual feedback to enhance quality of teaching and learning in UK’s Higher Education. Appl. Sci. 12(1), 514 (2022)

912

Y. Pang and J. Chen

6. Lindgreen, A., Di Benedetto, C.A., Brodie, R.J., Zenker, S.: Teaching: how to ensure quality teaching, and how to recognize teaching qualifications. Ind. Mark. Manag. 100, 1–5 (2021) 7. Shi, G., Niu, D., Zhou, X.: Construction of curriculum teaching quality evaluation system based on AHP in OBE perspective. J. Changchun Inst. Eng. (Soc. Sci. Edn.) 22(04), 81–84 (2021). (in Chinese) 8. Zhu, H., Cao, X., Cao, N., et al.: Exploration and practice of teaching quality evaluation system based on OBE concept. Financ. High. Educ. Res. 6(02), 117–128 (2021). (in Chinese) 9. Meng, F., Wen, X., Gao, L., Yang, Y.: Research on teaching quality evaluation model based on OBE concept. Sci. Technol. Wind 16, 37–38 (2021). (in Chinese) 10. Wu, Z., Cao, J., Fan, D.: Construction of teaching quality evaluation model in universities based on Kano model. Heilongjiang Educ. (High. Educ. Res. Eval.) 12, 21–24 (2021). (in Chinese) 11. Wang, Y.: Research on the evaluation index system of teachers’ connotative teaching in colleges and universities. J. Shenyang Agric. Univ. (Soc. Sci. Edn.) 22(06), 720–725 (2020). (in Chinese) 12. Zhuang, W., Xing, F., Fan, J., Gao, C., Zhang, Y.: An integrated model for on-site teaching quality evaluation based on deep learning. Wirel. Commun. Mob. Comput. 2022, 1–13 (2022) 13. Skedsmo, G., Huber, S.G.: Measuring teaching quality, designing tests, and transforming feedback targeting various education actors. Educ. Assess. Eval. Account. 32(3), 271–273 (2020) 14. Chen, B., Liu, Y., Zheng, J.: Using data mining approach for student satisfaction with teaching quality in high vocation education. Front. Psychol. 12, 1–8 (2022) 15. Wang, C.: Research and analysis of teaching quality evaluation system of Ningbo Dahongying College. Yunnan University, Kunming (2015). (in Chinese) 16. Chen, S., Chen, M.: Research on the optimization and upgrading of the operating mechanism of the university teacher development center - based on the perspective of the “Internet plus” era. J. Jimei Univ. (Educ. Sci. Edn.) 20(05), 1–8 (2019). (in Chinese) 17. Yang, J.: Cultivating the ability of scientific inquiry and evidence reasoning in experimental teaching. Minnan Normal University, Zhangzhou (2021). (in Chinese) 18. Yin, S., Ren, Y.: Research on the “3-all” teaching quality monitoring model of local colleges and universities. Inner Mongolia Educ. 02, 103–104 (2020). (in Chinese) 19. Lee, H., Lee, G.: Reconstructing education and ‘knowledge’ in the twentieth and twenty-first century: scientific management, educational efficiency, outcomes based education, and the culture of performativity. Educ. Soc. 32(2), 63–96 (2014) 20. Priya Vaijayanthi, R., Raja Murugadoss, J.: Effectiveness of curriculum design in the context of outcome based education (OBE). Int. J. Eng. Adv. Technol. 8(6), 648–651 (2019) 21. Lin, M.: Curriculum reform of employment-oriented “design of machinery” based on OBE education concept. Front. Educ. Res. 5(9), 55–59 (2022) 22. Wei, Y., Xu, Y.: Research on the function and operation mechanism of the teacher development center in higher vocational colleges. Career 28, 98–99 (2016). (in Chinese) 23. Hao, S., Liu, R., Xie, X.: Optimization and exploration of training programs for transportation majors based on OBE - taking application-oriented undergraduate universities as an example. Educ. Teach. Forum 51, 137–140 (2022). (in Chinese) 24. Kaviyarasi, R., Balasubramanian, T.: Exploring the high potential factors that affects students’ academic performance. Int. J. Educ. Manag. Eng. (IJEME) 8(6), 15–23 (2018)

Research on Teaching Application of Course Resources Based on Bloom’s Taxonomy of Educational Objectives Hao Cai(B) , Jing Kang, Qiang Zhang, and Yue Wang Wuhan Railway Vocational College of Technology, Wuhan 430205, Hubei, China [email protected]

Abstract. With the continuous construction of digital campus in Chinese universities, campus information technology and classroom teaching are increasingly integrated, and there are fundamental changes in teachers’ teaching methods and students’ learning methods and various teaching methods and theories have also been popularized and applied on a larger scale. With the goal of improving teaching quality and relying on the construction and application of campus information platform, Wuhan Railway Vocational and Technical College has applied the concept of student-centered teaching and practiced student-centered teaching applications and research. This paper puts forward a multi-factor regression analysis method of teaching objectives to analyze and evaluate teaching effects by using students’ learning cognitive process data. This method firstly uses container technology to quickly deploy course resources suitable for teaching needs, and collects student learning sample data of classroom teaching. Then, Bloom’s taxonomy of educational objectives is used for reference, which sums up the relationship of students’ learning cognitive process. Finally, this paper fits the sample data and proves the effectiveness and feasibility of achieving the “optimal” teaching objects classification relationship between different courses and classes based on the multiple regression mathematical model. Keywords: Taxonomy of educational objectives · K8S · Digital campus · Course resources · Teaching effects

1 Introduction The continuous development of network information technology has expanded more space for the application of education and teaching, and provided more convenience for the sharing of educational resources. At present, in the classroom teaching of Chinese universities, the total amount of information transmitted by teachers to students is often greater than the learning development ability students can match. As a result, classroom teaching focuses on knowledge transfer, neglects students’ subjective initiative, and doesn’t pay attention to the growth of students’ learning ability, and doesn’t match with the achievement of teaching objectives. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 913–926, 2023. https://doi.org/10.1007/978-3-031-36118-0_78

914

H. Cai et al.

With the in-depth development of education information reform and the increasingly integration of network information technology and classroom teaching, the teaching methods of teachers and learning methods of students in higher vocational colleges in Hubei Province have also changed fundamentally [1]. Taking the opportunity for building the “Double High-levels Plan” colleges and universities of the Department for Education, Wuhan Railway Vocational College of Technology has launched Information construction of digital campus in an all-round way. Relevant application systems involve all aspects of daily teaching, scientific research, management, and life services of the college. In the construction and development of campus informatization, our college has fully considered the needs of teachers and students for educational informatization integration. In the classroom teaching reform, with the goal of improving the teaching quality, based on the student-centered teaching concept, the promotion of the application of digital campus system platform has made a beneficial attempt to promote students to effectively aggregate the fragmented time of campus life. Since 2019, our college has promoted and used the Vocational Education Cloud Platform of Higher Education Press in teaching resource management and classroom teaching services. In the application of course resources teaching based on information technology, our college guides teachers to carry out teaching method reform activities, evaluates teaching effects and summarizes teaching experience based on Bloom’s taxonomy of educational objectives. However, due to various factors such as course resources, class members, teachers’ personalized differences in teaching etc., it is difficult for those methods and theories to be used to analyze and evaluate students’ learning effects in a short time, and to help teachers make teaching adjustments suitable for such courses, as well as to form “rules” of effect evaluation that are easy to adjust and observe. It is also difficult to “precipitate” teaching methods and experiences that can be shared and used for reference. At the same time, there are few studies on using multi-factor regression analysis techniques to find a method of Bloom’s taxonomy of educational objectives. Therefore, through the key activities of students’ learning cognitive process, we recorded quantifiable phased evaluation data item by item, and proposed a multi-factor regression analysis method of educational objectives to explore the data correlation of cognitive activities and evaluate teaching effects and students’ learning ability. This method firstly uses container technology to quickly deploy course resources suitable for teaching needs, and collects student learning sample data in classroom teaching. Then, it rationally draws on Bloom’s taxonomy of educational objectives, and concludes the relationship of students’ learning cognitive process. Finally, this paper fits the sample data of learning cognitive process, and verifies the feasibility of achieving the “optimal” taxonomy relationship of educational objectives for different courses and classes based on the multiple regression mathematical model.

2 Theory and Technology With the help of K8S container technology, our college has widely implemented and applied the digital campus platform among all teachers and students, continuously paying attention to the experience of teachers and students’ course application, actively

Research on Teaching Application of Course Resources

915

responding to the teaching needs of teachers and students for course resources, timely adjusting the arrangement and configuration of various resources in classroom teaching, and rapidly realizing the deployment of course resources. In the teaching of course resources, the basic theory of teaching methods based on Bloom’s taxonomy of educational objectives is practiced. With the help of information technology, students’ learning sample data has been effectively collected in all teaching links, providing data support for subsequent data analysis and evaluation using multiple regression models. 2.1 Bloom’s Taxonomy of Educational Objectives Bloom, a famous American psychologist, and his team published the book Taxonomy of educational objectives: The Cognitive Domain in 1956. They put forward the theory of classified teaching, and divided educational objectives into three parts, namely, cognitive domain, emotional domain and motor skill domain [2, 3]. In 2001, Anderson and Krathowohl revised Bloom’s Taxonomy theory and divided the cognitive field into two dimensions: knowledge dimension and cognitive process dimension. Namely, verbs are used to describe the cognitive process; Use nouns to describe the knowledge that learners are required to master. According to the development law of personal cognitive ability from simple to complex, from low to high, cognitive fields are divided into six basic categories: remembering, understanding, applying, analyzing, evaluating and creating [4, 5] (Table 1). Table 1. Bloom’s taxonomy table The Knowledge Dimension

The Cognitive Dimension Remembering

Understanding

Applying

Analyzing

Evaluating

Creating

Factual knowledge

Listing

Summarizing

Classifying

Ordering

Ranking

Combining

Conceptual knowledge

Describing

Interpreting

Experimenting

Explaining

Assessing

Planning

Procedural knowledge

Tabulating

Predicting

Calculating

Differentiating

Concluding

Composing

Meta-cognitive knowledge

Using Appropriately

Executing

Constructing

Achieving

Taking action

Actualizing

In 2015, Nancy E. Adams also believed that teachers can encourage students to learn by themselves through setting teaching goals with verbs, and helped them to carry out high level cognitive thinking activities. It highlights the characteristics of the integration of learning, teaching and evaluation, and is conducive to improving the classroom teaching effect [6].

916

H. Cai et al.

2.2 K8S Container Management Platform Kubernetes is a portable, extensible, open source platform. The name Kubernetes originates from Greek, meaning helmsman or pilot. K8S as an abbreviation results from counting the eight letters between the “K” and the “S”. Google open-sourced the Kubernetes project in 2014 [7] (Fig. 1).

Fig. 1. K8S Global Architecture

In early time, organizations ran applications on physical servers. There was no way to define resource boundaries for applications on a physical server, and this caused resource allocation issues. Because each application runs on a different physical server, it doesn’t scale as resources were underutilized, and it is expensive for organizations to maintain many physical servers. Virtual Machine (VM for short) technology can make good use of physical server resources, and because it can easily add or update applications, it can have higher scalability and lower hardware costs. Containers are similar to VMs, but they have relaxed isolation properties to share the Operating System (OS) among the applications. Therefore, containers are considered lightweight. Similar to a VM, a container has its own file system, share of CPU, memory, process space, and so on. As they are decoupled from the underlying infrastructure, they are portable across clouds and OS distributions (Fig. 2).

Fig. 2. Evolution process of application deployment mode

Research on Teaching Application of Course Resources

917

In the previous promotion and use experience of classroom teaching, our college usually constructs a business system based on virtual machine resources to carry out teaching activities. However, this mode is often faced with problems such as the delayed response to the actual application of teachers and students, poor scalability, and waste of hardware resources. Based on the realistic situation of digital campus, our college selects Vocational Education Cloud Platform and adopts the method of “container + micro service” for system deployment to replace the traditional construction method of “virtual machine + single server”. We gradually eliminate the non-synchronization between construction and application of information system, and start the practice of integrating school system construction, management and use (abbreviated as DevOps) [8]. In the construction of digital campus, our college has built a K8S container management platform, and continued to pay attention to the user experience of the digital campus platform, timely paid attention to the arrangement and configuration of various information resources in classroom teaching, quickly implemented the deployment of applications, and actively responded to the teaching needs of teachers and students for course resources [9] (Fig. 3).

Fig. 3. K8S container platform

2.3 Vocational Education Cloud System of Digital Campus Platform Our college has deeply promoted the integration of education and information technology, widely implemented the digital campus platform among teachers and students, applied the Vocational Education Cloud platform based on the K8S container systems, and carried out a series of information upgrading and transformation for existing online course resources including team training for teachers and students, basic operation of the learning system, online teaching and learning exercises for teachers and students, familiarizing the course design function of the system, etc.

918

H. Cai et al.

2.4 Multi-factor Regression Analysis Method of Education Objective 2.4.1 Educational Objective Factors Analysis Model In the teaching work for students, we gradually realize that there is a correlation and restriction between various “verbs” in the dimension of students’ learning cognitive process. However, there will be some differences in the teaching of each course. It often takes dozens or even hundreds of times of experimental teaching to get the teaching experience that is more in line with expectations. In order to achieve the goal of “quickly” aggregating the optimal taxonomy relationship of teaching objectives in relevant courses and minimizing the workload of experiments, we adopt the method of Multi-factor regression analysis, select a multiple system that can handle a large number of random data, and establish a reasonable regression model using function transformation. With the popularization of computer application, the Multi-factor regression analysis, which has a lot of calculation work, becomes easy to realize. Multi-factor regression analysis method can determine whether there is a correlation between specific variables. For many independent variables that jointly affect a dependent variable, this method can find out which are important factors, which are secondary factors, and what is the relationship between these factors. Moreover, it can predict or control the value of another variable according to the value of one or several variables, and can predict or control the possible accuracy. Multivariate linear regression considers the linear relationship: Y = β0 + β1 X1 + β2 X2 + · · · + βm Xm + ε

(1)

Between dependent variables Y and multiple independent variables X1 , X2 , · · · , Xm . Among them β0 , β1 , · · · , βm are unknown parameters, X1 , X2 , · · · , Xm are m known variables that can be accurately measured and controlled, and ε are random errors. The following data relationships can generally be assumed: E(ε) = 0, Var(ε) = σ 2

(2)

In many cases, such as significance test or Bayes analysis, stronger assumptions can be made: ε ∼ N (0, σ 2 )

(3)

In order to estimate the regression coefficients β0 , β1 , · · · , βm , Observe variable n for times, Get n groups of observation data items (Yi , Xi1 , Xi2 , · · · , Xim ), i = 1, · · · , n. In general, n > m is required. It is expressed in matrix form: ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ε1 β0 Y1 ⎡ ⎤ 1 X X · · · X 11 12 1m ⎢ ε2 ⎥ ⎢ β1 ⎥ ⎢ Y2 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 1 X21 X22 · · · X2m ⎥ ⎢ ⎥ ⎥ ⎢ ⎥, β = ⎢ ⎥, ε = ⎢ ⎥ (4) Y =⎢ ⎢ .. ⎥ ⎢ .. ⎥ ⎢ .. ⎥, X = ⎣ · · · · · · ⎦ ⎣. ⎦ ⎣. ⎦ ⎣. ⎦ 1 Xn1 Xn2 · · · Xnm Yn β0 εn Then the multiple linear regression model: Y = Xβ + ε

(5)

Research on Teaching Application of Course Resources

919

Among them n × (m + 1) matrix is called regression design matrix X . Generally, X is assumed to a full column rank, namely rk(X ) = m + 1. Assumptions about errors keep up with (2): E(ε) = 0, Var(ε) = σ 2 In

(6)

Among them In is the unit matrix. It corresponds to the above data association (3): ε ∼ N (0, σ 2 In )

(7)

All data relationships (5), (6) and (7) together are called multivariate linear models. 2.4.2 Solving Model Main steps of the method are listed: 1) 2) 3) 4)

Arrange and adjust the course resources through the K8S container systems. Collect the phased data of students’ cognitive process generated by the course. Refer to Multi-factor regression analysis method. Through the fitting and calculation, obtain the factor data of various cognitive activities of educational objectives on the dependent variables of evaluation results. 5) Repeat steps 1), 2), 3) and 4) until the factor data converges.

3 Research on Teaching Application of Course Resources 3.1 Process Implementation and Method Summary This research takes the teaching of Career Planning as an example, and the implementation methods and processes are as follows: 3.1.1 Determine Teaching Objectives The teaching objectives of this course are defined according to Bloom’s taxonomy of education objectives. For example, conceptual knowledge can guide students to enumerate and explain the policies and situations for our college students’ employment; Procedural knowledge can guide students to complete an interview with professionals by using the occupational data collection method. 3.1.2 Compile Course Outline Design the overall teaching objectives of the course, sort out the teaching objectives of each learning unit, and then complete the writing of the course. Through the Vocational Education Cloud Platform, the syllabus is presented to teachers and students in the form of handouts, including: course introduction, introduction of teachers, course learning objectives, requirements for students’ knowledge and skills, course teaching content, reference materials, and assessment methods, etc.

920

H. Cai et al.

Before the first course starts, teachers will issue the course outline. Guide students to learn about the course objectives, learning methods, learning activities, assessment methods and other course requirements by previewing, so as to reach a consensus on teaching and learning. 3.1.3 Improve Course Design In order to carry out teaching activities, teachers design courses according to the teaching requirements and teaching contents of each unit’s knowledge points and skills points in the course outline. Through the three stages, namely before class, during class and after class, the online learning activities of students are arranged by adopting Bloom’s taxonomy of education objectives. 3.1.4 Carry Out Teaching Activities Before class, students can carry out independent learning, complete the task of topic preview and put forward learning problems through videos, rich texts and other relevant learning materials provided by teachers on the Vocational Education Cloud system. In class, teachers can carry out online and offline teaching procession through the network, based on the problem analysis of students’ preview before class; Students can follow teachers to participate in classroom teaching activities and complete learning tasks through laptops and intelligent terminals. Each classroom learning task takes 15 min as a cycle to carry out a presentation activity. Teachers arrange group discussion, brainstorming and other teaching interactions according to the classroom teaching objectives. Students submit the project learning materials, including learning summary, grouping conclusion and test results to the system. 3.1.5 Teaching Feedback and Evaluation Through the system, teachers evaluate the classroom learning effects, including homework, tests, group discussions, topic messages, etc., and guide students to continue to pay attention to their personal learning status to achieve the course teaching goals (Figs. 4 and 5). Teachers will feedback the learning effect to students, and students can intuitively obtain the current course learning situation. 3.1.6 Learning Effect and Data Analysis Students can view the course scores and corresponding grades through the system, and teachers can guide students to gradually achieve the course teaching goals in the process of continuous correction through it. At the end of the semester, teachers will evaluate students’ learning effect of this course. The evaluation of students’ course learning consists of 20% of their usual performance (class attendance, participation in discussion, question messages, etc.), 40% of their practical performance (learning tasks completed in class and after class), and 40% of their final course design.

Research on Teaching Application of Course Resources

921

Fig. 4. Teaching activities

Fig. 5. Classroom test

4 Simulation Case According to The Latest Theory of taxonomy of educational objectives, the important goals of classroom teaching include “retention” and “transfer” [10, 11]. 1) Many methods of learning during a course can be remembered and reused by students in the later learning process; 2) Students can use what they have learned to master new knowledge, answer new questions and solve new situations [12]. Generally, there are 19 cognitive processes that students use to enhance learning effects: 1) “Remember” includes recognizing and recalling; 2) “Understand” includes interpreting, exemplifying, classifying, summarizing, inferring, comparing and explaining; 3) “Apply” includes executing and implementing; 4) “Analyze” includes differentiating, organizing and attributing; 5) “Evaluate” includes checking and critiquing; 6) “Create” includes generating, planning and producing [13–15].

922

H. Cai et al.

4.1 Teaching Statistics In 2021, 216 students from four parallel classes taught in two semesters of this course were selected for tracking and analyzing the teaching effect. For two classes, 110 students in total, we carried out classroom teaching activities dominated by this research method, and learned about learning effects and follow-up feedback from students. Through the platform data analysis, the class participation of the students who adopted the research method was close to 88%, and was 13% higher than the other ones of the students who did not carry out the research method (Table 2). Table 2. Tracking table of students’ learning cognitive process No. Cognitive process Dimension

Cognitive Levels Traditional Data Application Data Standard Data

1

Creating

6

2

Evaluating

5

3

Analyzing

4

Applying

70%

96%

100%

4

2 * 30%

2 * 46%

>80

3

75%

84%

100

5

Understanding 2

75%

88%

100

6

Remembering

72%

122%

100%

1

84% of the surveyed students said that they had achieved the learning objectives described in this course; In the traditional method, the former statistical proportion was about 75%. 96% said they used the learning methods mastered by this course in the follow-up study of other courses, and maintained a high interest and state of learning; In the traditional method, the former statistics accounted for about 70%. According to the statistics of the platform, for the classes that use research methods to carry out teaching, the number of students’ after class learning hours every week is 122% of the required after class learning hours in this course. Students who did not use this research method invested about 72% of the required after-school learning hours in the course every week (Fig. 6). According to previous teaching experience, the outstanding rate of students’ performance after the course teaching is the number of people with a course evaluation score greater than 80 divided by the total number of students. Under the same performance evaluation method, the outstanding rate of student classes without research method is about 30%; the outstanding rate of final results of the student class using the research method is 16% higher than that of other classes. When carrying out the research on the taxonomy of educational objectives, We have considered the score data of students’ comprehensive analysis of questions when they usually participate in classroom tests, group discussions and assessments, and made a quantitative treatment of the cognitive processes of “analysis”, “evaluation” and “creation”.

Research on Teaching Application of Course Resources

923

Fig. 6. Radar Chart of Students’ Cognitive Process

4.2 Establishment of Regression Model We randomly selected 50 samples from the tracking data of students’ teaching effects for analysis. At the same time, according to the theoretical results of the classification of educational goals, the important goals of classroom teaching are “retention” and “transfer”. The important cognitive processes used to strengthen learning achievements are “creation” and “evaluation”. To facilitate analysis and reduce regression error, we set “creation & evaluation” as the dependent variable Y, and other cognitive process activities as the independent variable: X1 (Learning time), X2 (Understanding activities score), X3 (Application effect), X4 (Analysis effect), and add quadratic items. The regression model is: Y = β0 + β1 Xi1 + β2 Xi2 + β3 Xi3 + εi , i = 1, · · · , 50 εi ∼ N (0, σ 2 ) And the variables ε1 , ε2 , · · · , ε50 are independent of each other [16]. 4.3 Experimental Data Acquisition According to the taxonomy of educational objectives, we selected the teaching effect data of 50 students randomly from the student classes, and analyzed them using the research method (Table 3). 4.4 Result Analysis in MATLAB Through Matlab and Excel fitting, we obtain the regression linear equation of the teaching objective classification method of this course [17]. As follow: y = −0.5978 + 0.18155x1 + 0.02101x2 + 0.04177x3 + 0.41188x4. It shows in the blue line in the figure below.

924

H. Cai et al. Table 3. 50 groups of measured data

No.

X1

X2

X3

X4

Y

No.

X1

X2

X3

X4

Y

1

8.9

89

94

TRUE

TRUE

26

9.2

87

89

TRUE

TRUE

2

4.8

90

95

TRUE

TRUE

27

7.2

85

67

FALSE

FALSE

3

3.2

62

74

FALSE

FALSE

28

6.5

79

99

TRUE

TRUE

4

6.5

90

98

TRUE

TRUE

29

6.4

95

98

TRUE

FALSE

5

5.8

95

69

FALSE

TRUE

30

6.8

94

99

TRUE

FALSE

6

6.9

88

71

FALSE

TRUE

31

7.9

93

73

FALSE

TRUE

7

7.5

84

100

TRUE

TRUE

32

9

91

97

TRUE

FALSE

8

6.3

85

75

FALSE

TRUE

33

6.1

87

77

FALSE

TRUE

9

7.9

80

77

FALSE

TRUE

34

6.2

89

78

FALSE

TRUE

10

7.7

78

79

FALSE

TRUE

35

6.3

90

76

FALSE

TRUE

11

5.8

89

86

TRUE

TRUE

36

9.7

86

69

FALSE

TRUE

12

8.8

83

74

FALSE

TRUE

37

7.2

94

74

FALSE

TRUE

13

9.5

69

61

FALSE

TRUE

38

8.5

78

98

TRUE

TRUE

14

6.2

93

100

TRUE

TRUE

39

9.4

92

79

FALSE

TRUE

15

7.7

84

76

FALSE

TRUE

40

11.5

96

76

FALSE

TRUE

16

3.5

75

88

TRUE

TRUE

41

8.8

97

79

FALSE

TRUE

17

5.8

85

95

TRUE

TRUE

42

7.3

86

76

FALSE

FALSE

18

7.1

88

78

FALSE

TRUE

43

8.9

82

89

TRUE

TRUE

19

4.7

90

64

FALSE

TRUE

44

7.9

95

98

TRUE

FALSE

20

6.3

92

65

FALSE

TRUE

45

7.4

95

78

FALSE

TRUE

21

3.9

89

100

TRUE

TRUE

46

9.5

87

87

TRUE

TRUE

22

5.5

87

95

TRUE

TRUE

47

7.5

84

96

TRUE

TRUE

23

7.7

91

78

FALSE

TRUE

48

7.2

98

77

FALSE

TRUE

24

8.4

95

97

TRUE

TRUE

49

6.3

92

97

TRUE

FALSE

25

10.6

98

62

FALSE

TRUE

50

9.4

86

100

TRUE

TRUE

∧

∧

∧

∧

∧

Those variables β 1 , β 2 , β 3 , β 4 , in this equation have clear meanings. β 1 = 0.18155 means that for each additional unit of student learning time, the contribution to “cre∧

ation & evaluation” effect is equal to 0.18155; β 2 = 0.02101 means that the contribution ∧

∧

to “creation & evaluation” effect is 0.02101. Similarly, other situations as β 3 , β 4 can be understood (Fig. 7).

Research on Teaching Application of Course Resources

925

Fig. 7. Measured value, estimated value and residual analysis

5 Summary and Outlook Based on the classification method of teaching objectives, this paper proposes a multifactor regression analysis scheme of educational objectives to analyze and evaluate the teaching effect by using students’ learning cognitive process data. It provides an operable tool for analyzing course resources, teaching activities, and students’ cognition to evaluate the teaching effect. The scheme is as follows: 1) Evaluate the teaching effect from the perspective of students’ learning cognition. The multi-dimensional teaching of learning cognitive activities based on Bloom’s taxonomy of educational objectives which is carried out and the corresponding data is collected. 2) The multi-factor regression analysis formula is used to link teachers’ teaching, students’ learning and evaluation, highlighting the consistency of teaching effect evaluation. 3) According to the fitting and calculation of the phased impact factors, teachers can choose teaching strategies and evaluation methods more targeted. Teachers and students in teaching activities can know more clearly about how to invest in learning to achieve better learning results. With the help of information technology, this research based on the theory of Bloom’s taxonomy of educational objectives shifts the focus of classroom teaching from teachers’ teaching to students’ learning and mastering. It has made a beneficial attempt in promoting students’ learning interest, improving students’ learning engagement and participation, and improving teaching effects. The application of this research in the early stage of classroom teaching will increase the workload of teachers. While, after five semesters of running in and application, teachers generally feedback the application

926

H. Cai et al.

scheme, providing convenience for teachers to carry out student-centered teaching, plan classroom teaching content activities (including correcting course assignments, evaluating students’ learning process, etc.), and reduce teachers’ subsequent workload. This application research scheme can be used to promote teaching in different disciplines and different courses in various colleges and universities that have well solidified digital campus achievements and have developed online course teaching. Acknowledgment. This project is supported by Projects of China University IUR innovation fund (2020ITA01008).

References 1. Zhang, X., Huang, X., Li, D.: The investigation and study on learning style and behavior habits of college students. Educ. Res. Mon. 2021(1), 92 (2021) 2. Bloom, B.S.: Taxonomy of Educational Objectives: Handbook I: The Cognitive Domain. David Mckay Co., Inc., New York (1956) 3. Bloom, B.S.: Taxonomy of Educational Objectives Book 1: Cognitive Domain, 2nd edn. Addison Wesley Publishing Company, New York, London (1984) 4. Anderson, L.W., Krathowohl, D.R.: A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom’s Taxonomy of Educational Objectives Complete. Allyn & Bacon, Boston, MA (2001) 5. Chang, J., Lan, W.: New development of bloom’s taxonomy of educational objectives. J. Nanyang Normal Univ. 7(05), 84–85 (2008) 6. Adams, N.E.: Bloom’s taxonomy of cognitive learning objectives. Med. Lib. Assoc. 103(3), 152 (2015) 7. The Kubernetes Authors [EB/OL]. https://kubernetes.io/docs/concepts/overview 8. Sankaranarayanan, H.B., Rathod, V.: Airport merchandising using micro services architecture. Int. J. Inf. Technol. Comput. Sci. 6, 52–59 (2016) 9. Mamo, R.: Service-oriented computing for effective management of academic records: in case of Debre Markos University Burie Campus. Int. J. Mod. Educ. Comput. Sci. 4, 48–53 (2020) 10. Krathwohl, D.R., Bloom, B.S., Masia, B.B.: Taxonomy of Educational Objectives, the Classification of Educational Goals. Handbook II: Affective Domain. David McKay Company, New York (1973) 11. Simpson, E.J.: The Classification of Educational Objectives in the Psychomotor Domain. Gryphon House, Washington, DC (1972) 12. Pohl, M.: Learning to Think, Thinking to Learn: Models and Strategies to Develop a Classroom Culture of Thinking. Hawker Brownlow, Cheltenham, Vic. (2000) 13. Gunaratana, B.H.: Mindfulness in Plain English: Revised and Expanded Edition. Wisdom Publications, Boston (1996) 14. Karunananda, A.S., Goldin, P.R., Talagala, P.D.: Examining mindfulness in education. Int. J. Mod. Educ. Comput. Sci. 12, 23–30 (2016) 15. Kabat-Zinn, J.: Mindfulness-based interventions in context: past, present, and future. Clin. Psychol. Sci. Pract. 10(2), 144–156 (2003) 16. Larsen, R.J., Marx, M.L.: An Introduction to Mathematical Statistics and Its Applications. China Machine Press, Beijing 17. Yasir, M., Shah, Z.S., Memon, S.A., Ali, Z.: Machine learning based analysis of cellular spectrum. Int. J. Wirel. Microwave Technol. 2, 24–31 (2021)

Reform and Practice of Virtual Simulation Practice Course for Logistics Engineering Specialty Based on OBE Concept and School-Enterprise Linkage Yanfang Pan, Dalong Liu(B) , and Liwei Li School of Transportation, Nanning University, Nanning 530200, China [email protected] Abstract. The talent training of logistics engineering majors focuses on cultivating the ability of students to solve complex engineering problems. The students should adapt to the rapid development of industry and technology under the economic transformation, be able to collect and process the basic data of logistics economic operation, find scientific methods and ways to solve problems through data processing and analysis, and carry out systematic optimization, demonstration and evaluation. This puts forward higher requirements for the logistics virtual simulation practice course. Based on the concept of “results-oriented” + “university-enterprise linkage”, relying on the multi-field, cross-industry, crossregional off-campus production and teaching integration base, and combining the “cross-curriculum” + “multi-stage” + “in and off-campus and class” curriculum system, the article builds an integrated platform of “virtual reality integration” + “innovation integration” logistics virtual simulation teaching system, and uses AHP analysis method to analyze the evaluation indicators, Strive to find the most critical factor for the construction of virtual simulation practice course in the cooperation mechanism. Keywords: Virtual simulation · Logistics engineering specialty · OBE · School-enterprise linkage

1 Introduction Outcome-based education (OBE) is an advanced educational concept, and also a construction philosophy of curriculum system that focuses on the requirements of society for talents and adopts the way of reverse thinking [1]. In order to adapt to the new normal development of higher education and meet the needs of the development of “Industry 4.0” and “Internet plus”, China’s logistics industry has ushered in the development of intelligent upgrading and transformation, and has also driven the undergraduate education of logistics engineering to “smart”, “creative” and other fields [2]. As an enabling and value-added entry for the development of industries in the new era, logistics enterprises require logistics professionals to be able to adapt to the form of the times and provide enterprises with the ability and quality of the ecological industry chain of automated, intelligent and intelligent whole-process integrated solutions [3]. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 927–938, 2023. https://doi.org/10.1007/978-3-031-36118-0_79

928

Y. Pan et al.

With the rapid changes in the world today, virtual simulation teaching came into being [4]. It is the product of the deep integration of modern education and virtual simulation technology. With the help of “AR/VR”, intelligent big data, Internet of Things and other cutting-edge information technologies, it builds a highly simulated training environment and experimental conditions [4, 5]. Students can carry out training and experiments in the virtual simulation environment, and can operate from multiple angles and repeatedly. It is urgent to solve the problems of “high investment, high consumption, high risk, difficult to implement, difficult to observe, and difficult to reproduce” in traditional training and experimental teaching [6, 7]. In the teaching of logistics courses, virtual modeling and simulation is also an important content [8]. In the virtual laboratory environment, it is undoubtedly of obvious value to use simulation experiments as the basis of production capacity acquisition [9].

2 The Crux of Logistics Virtual Simulation Course In the construction of logistics specialty, the construction of virtual simulation course group and its experimental teaching platform can realize the sharing of experimental teaching resources inside and outside the school, in the region and in a wider range, and establish a sustainable virtual simulation experimental teaching service support system [10–12]. The study found that the crux affecting the construction of the curriculum system and the quality of talent training is mainly reflected in the following three aspects. 2.1 The Virtual Simulation Course is Difficult to Learn Due to the influence of textbooks and the limitation of course hours, teaching is inevitably a mere formality. In the training program for logistics engineering professionals, virtual simulation courses such as Logistics System Modeling and Simulation, Logistics System Planning and Design, and Supply Chain Optimization Program Design need to focus on training students’ ability to solve complex engineering problems. Students should not only master solid professional theoretical knowledge, have logistics technical ability, but also understand certain basic knowledge such as computer language, engineering drawing, data statistics and analysis, It also needs to have certain data processing ability and text expression ability [13]. Virtual simulation course is a comprehensive, difficult to operate, and highly applicable course category, which is very difficult for students to learn. The existing textbooks lack innovative guidance and application value of drawing inferences from one example. 2.2 The Training Results are Not Ideal and Lack of Practical Significance In the teaching process of the virtual simulation course, students complete the establishment of the whole model according to the experimental steps. Although the model can be built according to the fixed procedures, it cannot cultivate students’ ability to find, analyze and solve problems, which is not conducive to cultivating students’ creative thinking and pragmatic spirit.

Reform and Practice of Virtual Simulation Practice Course

929

In essence, the virtual simulation experiment teaching course of logistics engineering undergraduate specialty should focus on comprehensive design, with the goal of cultivating students’ ability to independently design and solve practical problems through practical training, and pay attention to the degree to which students can use modern analysis tools to carry out independent design and independent analysis, so as to meet the requirements of talent training of logistics engineering specialty [14]. Considering the support and importance of professional core courses for professional talent training, we should take the virtual simulation experiment teaching course as the carrier to cultivate students’ realistic, scientific and rigorous learning and working style [15]. However, due to the influence of the teacher-oriented concept, the teaching content of the virtual simulation course needs to be further adjusted and improved. 2.3 The Assessment is Not Objective and Cannot Reflect the Level of Students The traditional virtual simulation course teaching adopts the course assessment mode of “model” + “experiment report”. The students’ understanding and mastery of knowledge reflected in the experiment report is fragmented, and cannot form a systematic cognition. The causes and consequences of simulation optimization, overlap and conflict with reality, and inadequate consideration of model simplification and hypothesis have greatly reduced the goal and significance of model establishment [16].

3 Reform Ideas of Virtual Simulation Practice Course for Logistics Engineering Specialty Based on the OBE concept, the development and reform of virtual simulation courses for logistics engineering specialty are comprehensively carried out by school-enterprise linkage, which is conducive to the overall integration of the curriculum system for logistics engineering specialty. The development of virtual simulation course projects, from topic selection, design, development to optimization, verification and application, pays equal attention to both virtual and real, and helps students further consolidate their professional skills, understand the professional frontier, and improve their interdisciplinary ability [17]. 3.1 Basic Ideas of Virtual Simulation Practice Course Reform The students’ comprehensive and systematic knowledge integration from language basis, mathematical logic, professional skills, off-campus practice and in-school experimental courses will help to achieve the objectives and requirements of training logistics engineering professionals in the context of new engineering [18]. Thus, the basic framework of virtual simulation practice course system for logistics engineering specialty based on OBE concept and school-enterprise linkage is constructed, as shown in Fig. 1.

930

Y. Pan et al.

Fig. 1. Virtual Simulation Practice Course System for Logistics Engineering

3.2 Specific Measures for the Reform of Virtual Simulation Practice Course The virtual simulation practice courses involved in the training program for logistics engineering professionals are mainly concentrated in professional core courses and centralized practice courses, such as Logistics System Simulation, Logistics System Planning and Design and other courses. These courses are organized by knowledge modules in traditional teaching and textbook design. The cases are independent, not coherent, and not easy to understand. Adopt the results-oriented approach, reconstruct the teaching content, excavate the teaching situation from the real enterprise operation status, extract the complex logistics engineering problems as the teaching items of the course, take students as the center, teachers as the guidance and assistance, and lead students to complete the analysis, design, innovation, application and problem solving in the form of project development and promotion, and put the learning into practice [19–21]. The course results are in the form of project reports and team results, giving full play to students’ subjective initiative and innovative spirit; The teaching results take solving the actual problems of enterprises as a reference, and carry out the scheme design and demonstration, which can achieve the goal of integration of specialty and innovation and integration of curriculum and innovation to a certain extent, and is conducive to the realization of the goal of virtual simulation course. The specific methods are as follows. (1) The integration of schools and enterprises, and the close connection and high cooperation between logistics engineering and logistics enterprises.

Reform and Practice of Virtual Simulation Practice Course

931

Deepening the cooperation between schools and enterprises, co-constructing the project development and course evaluation of virtual simulation courses, relying on the cooperation foundation established by enterprises and schools, can extract and design complex engineering problems of typical industrial applications, which plays an important role in improving the professional teachers’ teaching team and improving the professional theoretical level and teaching practical skills. School-enterprise linkage integrates industry and education into a theme, so that talent specifications comply with industrial needs, teaching resources match industrial needs, curriculum standards match professional standards, help students achieve “learning by doing” in the contact with logistics enterprises, comprehensively improve students’ comprehensive ability and professional quality, and achieve a high degree of cooperation between schools and enterprises [22]. (2) Result-oriented, combining curriculum with graduation design, innovation and entrepreneurship, and discipline competition, will help to achieve the integration of curriculum and innovation, and the integration of teaching and competition. Based on the results-oriented teaching project development and process organization, the curriculum consciously introduces the real questions of enterprise engineering projects and logistics simulation competitions, teachers and students learn and create together, and schools and enterprises develop together. At the same time, the curriculum evaluation mode is combined with the graduation design to solve the problems of insufficient class hours, single performance evaluation method, and outstanding results, which also improves students’ learning interest and innovation and entrepreneurship ability, and also cultivates students’ scientific expression Ability required by engineering certification, such as teamwork and self-study. (3) The joint construction of schools and enterprises and the penetration of curriculum ideology and politics will help students establish good professional ethics and professional quality According to the development requirements of the logistics industry, sort out the knowledge structure and application level of the virtual simulation course, effectively excavate and screen the ideological and political elements through the participation of enterprises and the combination of points, lines and areas, cultivate students to establish a correct professional view of logistics development, and do a good job in the responsibility of logistics personnel; Establish a correct concept of logistics development and stimulate students’ awareness of efficiency, service, cost, safety and low-carbon environmental protection. Through enterprise research and data collection, through modeling and optimization simulation, students are trained to develop a scientific spirit of scientific rigor, truth-seeking and pragmatic, and continuous exploration. (4) Multiple evaluation, establish the curriculum results of “model” + “scheme” preliminary evaluation and “course” + “final design” inspection Follow the principle of real combat, adhering to the scientific, rigorous, normative and effective ideological and political concept of curriculum, establish a dualdimensional and two-level curriculum achievement system of “model” + “program” preliminary evaluation and “curriculum” + “final design” test, and comprehensively evaluate students’ problem-solving ability, logical thinking ability, innovation ability and independent learning ability [23].

932

Y. Pan et al.

4 Construction of Evaluation Index System of Logistics Virtual Simulation Course Based on AHP Method The purpose of evaluating the virtual simulation course of logistics engineering specialty with the method of AHP is to find the key points of course reform [24, 25]. The evaluation needs to focus on the coherence and complementarity between “truth” and “virtual”. What called “truth” means that the teaching design of the virtual simulation course requires true data and true scene, and relies on the real enterprise background and enterprise environment, close to the reality of the enterprise; “Virtual” means to reproduce the scene in the school classroom through scientific teaching methods and means with the help of advanced computer networks and platforms. Students do not need to go to the enterprise site. Table 1. Evaluation Index and Description of Virtual Simulation Course of Logistics Engineering Items

Evaluating indicator

Teaching achievements

The assistance of the course to solve the key problems of the the project enterprises enterprise

Course organization

Originality of student work output

Different results from student-centered

Classroom teaching feedback

Supervision and student evaluation

ideological and moral education of the curriculum

Design embodiment in moral education

Construction of course teaching resource library

Established network teaching resource

Conforms to the actual production of the enterprise

Curriculum designed from real enterprises

Use of third-party teaching platform in teaching process

Use third-party platform for teaching

The difficulty degree of course The teaching situation is scenario design progressive

Teaching resources

Compliance with the requirements of course objectives

Meet the goal of certification education

Textbook selection

Choose planning textbook

Teaching team

Gradient of teaching team is reasonable

Practical class hours

Practice class hours meet teaching needs

According to the above analysis, fourteen evaluation indicators are summarized from three aspects of course scene mining, course organization and achievement evaluation

Reform and Practice of Virtual Simulation Practice Course

933

output to assess the virtual simulation course. See Table 1 for the assessment indicators of course evaluation. The evaluation indicators of logistics engineering virtual simulation course in Table 1 include both quantitative evaluation indicators and qualitative indicators, and the impact of these indicators on the logistics engineering virtual simulation course is different. In order to better measure the importance of each indicator for the training support of logistics professionals, the analytic hierarchy process (AHP) is used to set each indicator. 4.1 Establishment of Hierarchy Model Based on the evaluation index of virtual simulation course of logistics engineering specialty in Table 1, the hierarchy model of evaluation index of virtual simulation course of logistics engineering specialty is established, as shown in Fig. 2. 4.2 Evaluate and Calculate the AHP Model In view of the hierarchical structure model of the evaluation index of the virtual simulation course of logistics engineering specialty, the judgment matrix is established by comparing two different elements at the same level. Through the form of questionnaire and expert interview, the bank and enterprise expert consultation is carried out, the importance of each evaluation index is expressed in numbers, the qualitative index is converted into quantitative numbers, the relative weight is obtained, the consistency of the judgment matrix is checked, and then the total ranking weight of each level is calculated, and the ranking is carried out in order. Get the relative weight of each index. Table 2 shows the judgment matrix and relative weight of criterion layer A. Tables 3, 4 and 5 shows the judgment matrix and relative weights of teaching achievement B1 , curriculum organization B2 and teaching resources B3 . 4.3 Consistency Inspection Check the consistency of the judgment matrix of the criterion layer and the target layer. Maximum characteristic value: λ=

1 1 (Aω)i (n : Order) i=1 ωi n

(1)

λ−1 n−1

(2)

CI RI

(3)

Consistency indicators: CI = Consistency ratio: CR =

The consistency test and results obtained by calculation are shown in Table 6.

934

Y. Pan et al.

Fig. 2. Hierarchical structure model of evaluation indicators for virtual simulation courses of logistics engineering

Table 2. Criterion level judgment matrix and relative weight A

B1

B2

B3

Relative weight

B1

1

1/4

3

0.231

B2

4

1

5

0.665

B3

1/3

1/5

1

0.104

4.4 Summary and Correction From Tables 2, 3, 4 and 5, the relative weights of the indicators at the criterion level and the scheme level are obtained. From Table 6, the results of the consistency test are obtained. Then the relative weights of the scheme can be obtained by the weighted calculation method. The synthetic weight and ranking results of virtual simulation course evaluation are shown in Table 7.

Reform and Practice of Virtual Simulation Practice Course

935

Table 3. Judgment matrix and relative weight of teaching achievement evaluation B1

C11

C12

C13

Relative weight (B1 )

C11

1

1/2

3

0.334

C12

4

1

3

0.525

C13

1/3

1/3

1

0.142

Table 4. Course organization judgment matrix and relative weight B2

C21

C22

C23

C24

C25

C26

Relative weight (B2 )

C21

1

3

1/4

2

1/2

1/3

0.116

1

1/4

1/3

1/2

1/4

0.060

1

3

1

1

0.265

1

1/2

1/4

0.095

1

1

0.201

1

0.263

C22 C23 C24 C25 C26

Table 5. Judgment matrix and relative weight of teaching resources B3

C31

C32

C33

Relative weight (B3 )

C31

1

1/2

1

0.263

C32

1

1

1

0.413

C33

2

1

1

0.324

Table 6. Results and conclusions of consistency inspection Consistency inspection

Result

Conclusion

Criterion layer

λ max = 3.087; CR = 0.0836 < 0.1

Pass

Course scenario mining

λ max = 3.054; CR = 0.0517 < 0.1

Course organization

λ max = 6.311; CR = 0.0494 < 0.1

Evaluation and judgment

λ max = 3.054; CR = 0.0516 < 0.1

Due to the use of expert scoring, the analytic hierarchy process is greatly affected by subjective factors and has many qualitative components. Therefore, it can also be properly modified according to the actual situation of professional talent training, so that the curriculum evaluation is closer to the professional curriculum objectives and professional talent training objectives.

936

Y. Pan et al.

In the light of the results shown in the last column of Table 7, among the total 12 evaluation items, the first one is the situation that “the curriculum scenario conforms to the actual production of the enterprise”. This is in line with the concept of OBE. We should design the talent training system based on the actual situation of the enterprise. The second is that “the curriculum scenario meets the requirements of the degree of achievement of the curriculum objectives”, which further explains that the final requirements of talent cultivation play a decisive role in the design of the curriculum system. “The difficulty of curriculum scenario design” ranks the third, which tells us that we should carry out all-round balance and coordination in curriculum design and organization. Table 7. Evaluation weight and ranking of virtual simulation courses for logistics engineering Items

Evaluating indicator

Composite weight

Teaching achievements

The assistance of the course to the project enterprises

0.077

6

Originality of student work output

0.121

4

Classroom teaching feedback

0.033

11

ideological and moral education of the curriculum

0.077

5

Construction of course teaching resource library

0.040

9

Conforms to the actual production of the enterprise

0.176

1

Use of third-party teaching platform in teaching process

0.063

7

The difficulty degree of course 0.134 scenario design

3

Compliance with the requirements of course objectives

0.175

2

Textbook selection

0.027

12

Course organization

Teaching resources

Sorting

Teaching team

0.043

8

Practical class hours

0.034

10

The construction of virtual simulation course for logistics engineering specialty should focus on the course organization, especially in the case of platform coconstruction and school-enterprise cooperation, pay more attention to the relevant standards that conform to the actual production of enterprises and engineering education certification, and pay attention to the progressive teaching organization from easy to difficult.

Reform and Practice of Virtual Simulation Practice Course

937

5 Conclusion Applying the concept of OBE to the teaching and talent training of logistics engineering is to meet the requirements of social and economic development for talent team. The teaching of virtual simulation practice course needs to combine theoretical teaching with production practice from the actual situation of enterprises. The way of school-enterprise interaction is undoubtedly an effective way to promote the teaching design of schools to really focus on the reality of enterprises and society. School-enterprise interaction can effectively develop the teaching scene of logistics virtual simulation course, return “virtual” to “real”, and the course teaching is more practical and valuable. Achievement orientation can effectively carry out the teaching organization of logistics virtual simulation course, cultivate students’ practical and pragmatic scientific attitude, develop students’ rational thinking and innovative thinking, and improve students’ ability to use modern tools to solve practical problems. Only by integrating the two courses can we better achieve the goal of the virtual simulation course of logistics engineering specialty in the new engineering background, and the course construction and professional development can be more close to the development needs of industry enterprises. Acknowledgment. The research of this paper is supported by two projects: (1) The teaching reform project of Nanning University “Reform and Practice of Virtual Simulation Practice Course for Logistics Engineering Specialty Based on OBE Concept and School-Enterprise Linkage” (2022XJJG28); (2) The teaching reform project of the integration of specialty and innovation of Nanning University “The teaching reform and practice of the course” Modern Logistics Equipment “combining the systematization of work process under the mixed teaching mode and the integration of specialty and innovation” (2020XJZC05).

References 1. Zamir, M.Z., Abid, M.I., Fazal, M.R., et al.: Switching to outcome-based education (OBE) system, a paradigm shift in engineering education. IEEE Trans. Educ. 65, 1–8 (2022) 2. Zhiqin, L., Jianguo, F., Fang, W., Xin, D.: Study on higher education service quality based on student perception. Int. J. Educ. Manag. Eng. (IJEME) 2(4), 22–27 (2012) 3. Al-Hagery, M.A., Alzaid, M.A., Alharbi, T.S., Alhanaya, M.A.: Data mining methods for detecting the most significant factors affecting students’ performance. Int. J. Inf. Technol. Comput. Sci. (IJITCS) 12(5), 1–13 (2020) 4. Laghari, A.A., Jumani, A.K., Kumar, K., Chhajro, M.A.: Systematic analysis of virtual reality & augmented reality. Int. J. Inf. Eng. Electron. Bus. 13(1), 36–43 (2021) 5. Peng, Y., He, X., Huang, Y.: A Virtual simulation experiment system for requirement analysis. J. Phys.: Conf. Ser. 1757(1), P012194 (2021) 6. Lin, Z.: The construction of the virtual simulation experiment platform under the background of education informatization. In: Jansen, B.J., Liang, H., Ye, J. (eds.) International Conference on Cognitive based Information Processing and Applications (CIPA 2021). LNDECT, vol. 85, pp. 718–724. Springer, Singapore (2022). https://doi.org/10.1007/978-981-16-5854-9_92 7. Dubovi, I.: Learning with virtual reality simulations: direct versus vicarious instructional experience. Interact. Learn. Environ. 2022, 1–13 (2022)

938

Y. Pan et al.

8. Xie, X., Guo, X.: Influencing factors of virtual simulation experiment teaching effect based on SEM. Int. J. Emerg. Technol. Learn. 17(18), 89–102 (2022) ˙ 9. Bakhtadze, N., Zaikin, O., Zylawski, A.: Simulation experiment in a virtual laboratory environment as a ground for production competencies acquiring. In: Dolgui, A., Bernard, A., Lemoine, D., von Cieminski, G., Romero, D. (eds.) APMS 2021. IAICT, vol. 630, pp. 535–545. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-85874-2_57 10. Wang, X., Tan, Y.: Exploration of logistics virtual simulation experiment teaching in the training of innovative practical talents. Logist. Technol. 44(11), 172–174+178 (2021). (in Chinese) 11. Peng, C.: Construction of logistics management virtual simulation experiment teaching system based on post capacity. J. Taiyuan City Vocat. Tech. Coll. 06, 119–121 (2020). (in Chinese) 12. Hong, M.: Build virtual simulation logistics laboratory and cultivate innovative logistics talents. Electron. Technol. 39(01), 65–66 (2012) 13. Özgen, K.: The effect of project-based cooperative studio studies on the basic electronics skills of students’ cooperative learning and their attitudes. Int. J. Mod. Educ. Comput. Sci. (IJMECS) 2018(5), 1–8 (2018) 14. Wang, X.D., Wu, Y.: Exploration and practice for evaluation of teaching methods. Int. J. Educ. Manag. Eng. (IJEME) 3(29), 39–45 (2012) 15. Li, J., Zhou, Y.: The exploration and practice in innovative personnel training of computer science and technology. Int. J. Educ. Manag. Eng. (IJEME) 2(6), 47–51 (2012) 16. Khan, M.H.: A unified framework for systematic evaluation of ABET student outcomes and program educational objectives. Int. J. Mod. Educ. Comput. Sci. (IJMECS) 11(11), 1–16 (2019) 17. Wei, X.: Discovery and practice of EDA experimental teaching reform. Int. J. Educ. Manag. Eng. 1(4), 41–45 (2011) 18. Liu, G., Li, Q., Chen, H., et al.: Development and practice of cold chain logistics virtual simulation course. Packag. Eng. 42(S1), 191–195 (2021). (in Chinese) 19. Zhang, B., Yuan, Y., Huang, J., et al.: Virtual simulation of production logistics based on Plant Simulation. Sci. Technol. Innov. Appl. 11(21), 6–9+12 (2021). (in Chinese) 20. Huang, Y., Ji, L.: Exploration on the construction of virtual simulation experimental teaching center for logistics management based on VR technology. Exp. Technol. Manag. 37(08), 238–242 (2020). (in Chinese) 21. Ren, J.: Construction and practice of industrial fresh e-commerce and cold chain logistics virtual simulation experimental teaching center. Logist. Technol. 41(03), 125–129 (2022). (in Chinese) 22. Liu, Z., Li, Y., Chen, P., Zhao, L.: Research on the construction of virtual simulation logistics laboratory. Univ. Educ. 07, 54–56 (2020). (in Chinese) 23. Liu, H.: The application of virtual simulation technology in the practical teaching of logistics management specialty – taking the course of “warehouse and distribution management” as an example. J. Taiyuan City Vocat. Tech. Coll. 10, 160–161 (2017). (in Chinese) 24. Fu, J., Wu, Y., Yan, H.: The Application of database analytic hierarchy process (AHP) in teacher-to-teacher assessment. Int. J. Educ. Manag. Eng. (IJEME) 3(1), 27–33 (2013) 25. Kong, L.A., Wang, X.M., Yang, L.: The research of teaching quality appraisal model based on AHP. Int. J. Educ. Manag. Eng. 9(29), 29–34 (2012)

Application of Google Workspace in Mathematical Training of Future Specialists in the Field of Information Technology Olena Karupu1 , Tetiana Oleshko1 , Valeria Pakhnenko1 , and Anatolii Pashko2(B) 1 National Aviation University, Kyiv, Ukraine 2 Taras Shevchenko National University of Kyiv, Kyiv, Ukraine

[email protected]

Abstract. Education in the field of information technology in Ukraine should introduce a special innovative approach in the training of specialists in computer science and information technology. Problems of personality-oriented and competence-based models of education in the process of training future specialists in information specialties become topical. All professionals in the field of information technology must have specific personality traits and professional competencies formed in the studying different disciplines. Educational process has to result the formation of both hard and soft skills of students. Future IT specialists require a deep knowledge of the basic mathematical theoretical foundations and possession of the skills of applying mathematics. The study of the educational process in multinational academic groups seems to us quite interesting, since it has some peculiarities. The aim of this article is the consideration of new learning technologies and the features of their implementation in the context of distance education. We study methodological problems of collaborative approach in online teaching of mathematical disciplines applying Google Workspace for Education tools. Keywords: Teaching mathematics to information technology specialists · Teaching mathematics in multinational academic groups · Collaborative approach in education · Online and blended learning

1 On Mathematical Component in Professional Competence of IT Specialists In the professional development of future specialists of all technical specialties it is necessary to pay enough attention to forming of the basic mathematical knowledge and the skills of applying mathematical theory. One of the authors who most consistently apply this approach is N. O. Virchenko [1]. Education of future IT specialists should be problem-oriented. Broad and deep training of students should provide preparation of professionals able to improve their skills continually. This can be achieved by the fundamentalization of the educational process and the use of STEM technologies. Problematizing teaching and learning mathematics © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 939–949, 2023. https://doi.org/10.1007/978-3-031-36118-0_80

940

O. Karupu et al.

in STEM education were considered in [2–4]. Some problems of application of IT in teaching mathematical disciplines to future specialists in information technologies were considered in [5]. Research-oriented students of IT specialties need excellent mathematical knowledge and skills. Industry-oriented students of IT specialties need good mathematical knowledge and skills. On that account, the mathematical training of the first-year students of FCCSE of NAU is well above average level. The mathematical training of students of FCSC of KNU has very high level. It should be noted that in National Aviation University students of most specialties require fairly deep mathematical background. This is especially true for professionals in the field of information technology. The curricula of these specialties contain various mathematical disciplines. Problems encountered by students in the study of mathematical disciplines are often associated with their school background in mathematics and differ for students from different countries, although there are general characteristics of considered process. In addition, we consider it necessary to note the existence of certain difficulties in teaching students to solve applied problems having a technical orientation. In our opinion, this is due to the fact that in secondary schools in many countries more attention is paid to the consideration of applied problems having an economic nature. Insufficient level of previous school background in geometry, manifested when students study multivariable functions and multiple integrals, can be partially compensated by the use of CAS. Most students become aware of the existence of CAS and receive initial skills in using CAS while still studying in secondary school. We consider it desirable to provide guidance to students on the use of the various CAS, paying attention to the restrictions on their use.

2 Intensification of Integration Processes and Teaching to Mathematical Disciplines in English in Ukraine In recent years, we have seen an increase in the number of citizens of other countries who receive higher education in the universities of our country. Medical, engineering and IT professions are among the most popular for foreigners. Some foreigners are native English speakers or speak it at a fairly high level. A significant number of foreign students plan to work outside of Ukraine in the future. That is why these students choose education in English. In addition, the intensification of integration processes in the global educational and scientific space and the development of academic mobility significantly increases the interest of Ukrainian students in learning in foreign languages. Problems of implementation of English language in the educational process within the framework of internationalization of Ukrainian universities were analyzed in [6]. Trends and issues of provision of English-medium instruction were investigated in [7]. As a result, the current trend of modern national education in Ukraine is the continuous growth of the contingent receiving higher education in English. Some Ukrainian universities teach some students to some disciplines in English. Students of English-speaking groups may be divided into two categories: foreigners who receive higher education in Ukraine, and Ukrainian students who desire to study in English. And although students of both these categories have their own specific learning needs, the common denominator is

Application of Google Workspace in Mathematical Training

941

that the vast majority of participants in the considered educational process are not native English-speakers. Leading national universities of Ukraine are actively involved in the development of the education system, the introduction of new technologies, methods and models of organization of the educational process, educational and scientific work of students. Within the framework of the Higher Education Program of NAU in a foreign language, a team of experienced teachers of the Department of Higher and Computational Mathematics since 2006 has began conducting research on teaching mathematics in English to foreign and Ukrainian students. In particular, past few years, teachers of two Kyiv universities, NAU and KNU, have been implementing a project approach to the organization of educational and scientific student’s work. Also certain problems of teaching of mathematical disciplines to foreign students in English were investigated in [8]. Teaching Higher Mathematics in English to foreign students of construction specialties was considered in [9]. Some general issues of teaching to higher mathematics in English were considered in [10]. Methods of classifying foreign language communicative competences were investigated in [11]. Some features of teaching higher mathematics in English to students of computer specialties were investigated in [12]. When teaching mathematics to foreign and Ukrainian students teachers have to mind that English is not a native language for these students and that students had received secondary school education in their mother tongues. Separately, it should be noted that most foreign students come from countries using the traditional British system of notations for trigonometric functions, while Ukrainian students (and some foreign students) used a system of notations traditional for Ukraine during secondary school. Preparation of master’s theses in English at Taras Shevchenko National University of Kyiv and at National Aviation University allows students to demonstrate not only the level of proficiency in the specialty, but also their ability to join the international scientific community. Our experience of teaching in English of IT specialists in our universities confirms the need for the implementation of such projects. It should be noted that the engineering orientation and the scientific orientation of education at different universities are based on the same curricula, differing only in content. We also believe that it is interesting to assess the practical skills of professional English proficiency for graduates of both universities. We plan to consider these issues in our further research.

3 Application of Collaborative Approach in Multinational Academic Groups on Practical Training in Mathematical Disciplines In addition to scientific and practical competencies (hard skills), the professional competence of IT specialists also includes social, including communicative, competences (soft skills). Application of collaborative approach will help students develop both hard and soft skills. The general concept of collaborative approach and technologies in mathematics education was investigated in [13–15]. Preconditions for implementing of technique of CLIL (Content and Language Integrated Learning) in teaching mathematics were analyzed in [16, 17]. Approaches to modeling methods considered in [18–20] are used

942

O. Karupu et al.

in the guidance of term papers and scientific work of students. Features of the application of collaborative approach for mathematical education in international academic groups were investigated in [8, 21]. The investigation of the efficiency of certain methods of teaching educational material of mathematical disciplines and the organizing the educational process during lectures, practical training and individual work of students was carried out by traditional methods. That is, the current and the semester grades of different groups were compared; the subjective grades of students obtained through anonymous questionnaires and open discussions were analyzed. It should be noted that the availability of supporting materials is very important for most students. It should also be noted that supporting materials in the form of logical diagrams or flowcharts of the corresponding algorithms are better perceived by students studying in all specialties related to computer science and informatics. In addition, it should also be noted that to conduction lectures in multimedia classrooms is very useful for all students, since the use of various technical means provides excellent opportunities for visualizing educational material. Implementing the project approach in study of mathematical disciplines, we apply collective forms of work organization. For this purpose we practice dividing the staffs of academic groups into few teams for joint work with complex tasks, cross-checking the assimilation of the education material, creating presentations, etc., followed by discussion and comparison of results. In our opinion, especially interesting results were shown by the formation of teams consisting of students from different countries. Due to this organization of students’ work in multinational academic groups the level of academic success increases, English and interpersonal relationships improve. In the process of teaching to mathematical disciplines also it is necessary to give sufficient attention to the clarification of the peculiarities of the use of terminology in the solving of applied research problems. Also, teacher should give methods for the use of CAS (Maple, Mathematica, MATLAB, Scilab, MathCAD etc.).

4 Application of Google Workspace Tools in Mathematical Training of Future IT Specialists in Online and in Blended Learning Different problems of online and blended learning were investigated in [22–25]. The impact of COVID-19 on the academic performance of students was investigated by in [26]. Some problems of teaching in Computer Science were investigated in [27– 29]. Some problems of online and blended learning for mathematical education in multinational academic groups were investigated in [30, 31]. It should be noted that implementation of collective forms of training in practical classes in the context of the coronavirus epidemic has become very difficult. In recent years within the English-language project of NAU joint research to study the features of the project approach to the organization of educational and scientific work of students while distance learning has been conducted. In the 2019/2020 academic year, the conditions of study at many universities not only in Ukraine but also in many other countries have changed significantly: the first semester and beginning of the of the second semester

Application of Google Workspace in Mathematical Training

943

were held as usually under standard conditions, and then the second semester was continued in remote form. In the next 2020/2021 academic year, teachers and students of Kyiv universities faced new difficulties related to the introduction of a blended form of education at the beginning of the semester and the introduction of full distance learning (online) from October 19. During the second semester both lectures and practical classes were held entirely in remote form (online in Google Meet). This situation was extreme for everyone and had particularly catastrophic consequences for the first-year students. Unfortunately, in 2021/2022 academic year in Ukraine conditions for offline, online and blended learning worsened. Online learning in many universities is organized by means of cloud-based software platforms Zoom, Moodle and Meet. In NAU teachers realize distance learning in Google Workspace for Education. Tools of Google Workspace (formerly Google Apps and later G Suite) give teachers possibility to design and implement the educational process. We conduct classes online on Google Meet using the Jamboard. This requires special work of the teacher to identify the subjective experience of each student. Organization of the educational process in this way allows to more fully demonstrating skills and abilities of each student, to develop the creative potential of the student and his professional abilities, to gain experience in teamwork, which is important for information technology professionals. In particular, the use of the Google Drive cloud storage, which allows users to store, jointly edit their data on servers in the cloud and share it with other users on the Internet, is very useful for teachers. Application of Google Drive facilitates the effective interaction of participants in the educational process and teaches a digital style of work. Teacher can effectively plan time and optimize management processes, thus forming information and digital competence as a component of designing a new educational environment. Application of considered cloud technologies make possible to organize better joint work of teachers and students that will increase students’ achievement. The introduction of cloud technologies helps to effectively and quickly organize the methodical work of the teacher, improves the quality and efficiency of the educational process, as well as prepares students for life in the modern information society. From the very beginning of the 2021/2022 academic year at the university, teachers have created Google classes for all groups and streams, in which we place learning materials: theoretical statements and examples of problem solving. In addition to our developments, we offer students various online resources for additional use, such as Math24.net, Math is Fun and others. Distance learning (online and blended) in some universities is realized in the Google Workspace using Google Meet (Fig. 1). Effective organization of practicals, from our point of view, is especially difficult. Since in National Aviation University distance learning is realized in Google Workspace, then practical classes are conducted using Google Classroom and Google Meet. We also use Google Jamboard. Solving problems in online practical classes in mathematical disciplines, implemented with the application of Google Jamboard, in general was quite effective. However, it should be noted that it places quite high demands on both the computers on which students and teachers work, and the quality of the Internet.

944

O. Karupu et al.

Fig. 1. Practical Classes on Mathematical Analysis

In our opinion, the use of Google Jamboard to organize the work of student teams on practicals, workshops and consultations was especially interesting. During last two academic years this method to organize collective work on online practical classes in mathematical disciplines, in particular higher mathematics, was implemented for English-speaking groups of certain specialties of the Faculty of Cybersecurity, Computer and Software Engineering.

5 Experimental Results and Discussion We have analyzed the use of collaborative approach applying technologies of online learning based on Google Workspace Tools, classic online and blended learning by examining the results of studies of National aviation university students in the subjects “Higher Mathematics” and “Mathematical Analysis” in the first and second semesters of the 2021–2022 academic year. In the first semester we used collaborative approach applying technologies of online learning based on Google Workspace Tools, and in the second semester we used classic online and blended learning. The results of studies of students of English-speaking groups of “Software engineering” and “Computer Systems and Networks” specialties were investigated. Table 1 contains the analysis of the academic performance of English-speaking students majoring in “Computer Systems and Networks” in section “Mathematical Analysis” of subject “Higher Mathematics” and English-speaking students majoring in “Software engineering” in subject “Mathematical Analysis”.

Application of Google Workspace in Mathematical Training

945

Table 1. Final grades from subjects Semester

Final Grades (as a percentage) excellent

good

satisfactory

bad

“Higher Mathematics” First

21

21

58

0

Second

21

21

37

21

“Mathematical Analysis” First

25

54

21

0

Second

21

43

11

25

Figure 2 reflects the results presented in Table 1.

Fig. 2. Final Semester Grades (in percentages)

The samples under consideration are dependent. The same students study the same subject in different semesters and using different methods. This makes it possible to evaluate the effectiveness of different approaches to learning. The scheme of the algorithm for comparing two technologies is presented in Fig. 3.

Fig. 3. Algorithm for Comparing Two Technologies, Yes - we accept the null hypothesis, No - we reject the null hypothesis.

The results of the calculations. We consider ordered pairs (Xi , Yi ), where Xi are study results in the first semester, n di . Yi are study results in the second semester. Let’s calculate di = Yi − X i , ds = i=1 n

946

O. Karupu et al.

We will test the hypothesis H0 : ds = 0, about the equality of the mean zero. Alternative hypothesis with one-sided criterion H1 : ds > 0 or H1 : ds < 0, with a two-sided criterion H0 : ds = 0. Let’s calculate the statistics according to the formula  2 n  √ n ni=1 di2 − ds n i=1 di t= , Sd = Sd n(n − 1) The statistic has Student’s distribution with k = n − 1 degrees of freedom. Subject “Higher Mathematics” for students of specialty “Computer Systems and Networks”. Sample size n = 29. Estimated value of the criterion tr = 2.5359 for α = 0.05, critical value for the two-sided criterion tkp = 2.049. Since tr > tkp , the null hypothesis is rejected. Subject “Mathematical Analysis” for students of specialty “Software Engineering”. Sample size n = 28. Estimated value of the criterion tr = 2.8007 for α = 0.05, critical value for the two-sided criterion tkp = 2.052. Since tr > tkp , the null hypothesis is rejected. The test result shows that the approach proposed in the first semester is more effective.

6 Summary and Conclusion The need to combine distance and mixed forms of learning and to ensure the formation of the creative personality of the student and the development of student’s professional abilities in these conditions is the peculiarity of today educational process. Availability of many platforms for online classes requires finding new approaches to organize the educational process, both in its form and content. The research presented in the paper was based on a collaborative approach to organizing collective work. Conducting offline and online practical classes in mathematical disciplines was implemented in English-speaking multinational academic groups of certain specialties. Online classes and blended learning were conducted using Google Workspace tools. In particular, the implementation of a collaborative approach in online practicals with usage of Google Jamboard turned out to be very effective. This tool allows few teams to work simultaneously on different pages with further discussion and comparison of results. In general, results of this implementation in online practical classes were more successful for FC CSE than for other faculties. These results may be related to the specifics of students majoring in IT. Therefore, it is recommended to use this approach primarily in the training of students of IT specialties. The collaborative approach to organize the educational process for IT specialties allowed more fully identifying abilities and skills of each student, developing students’ creative and professional abilities. Future IT specialists also need skills of team work. Especially important is experience of team work in international teams. In our opinion, the obtained results allow to make some generalizations, confirm the successful choice of the direction of organizing the educational process and encourage further study of this approach. We have analyzed and compared the effectiveness of various approaches in the educational process when studying mathematical analysis. The testing result shows that

Application of Google Workspace in Mathematical Training

947

the collaborative approach with using Google Jamboard is more effective compared to other forms of organizing online learning based on Google Workspace Tools. In the future work, we plan to investigate the application of this approach in teaching other mathematical disciplines and different approaches to teams forming. Acknowledgment. This project is supported by Key Projects of China Southern Power Gird (GZ2014-2-0049).

References 1. Virchenko, N.O.: Selected Questions of the Methods of Higher Mathematics. Zadruga, Kyiv (2003). (in Ukrainian) 2. Li, Y., Schoenfeld, A.H.: Problematizing teaching and learning mathematics as “given” in STEM education. Int. J. STEM Educ. 6, Article no. 44 (2019). https://doi.org/10.1186/s40 594-019-0197-9 3. Abramovich, S., Grinshpan, A.Z., Milligan, D.L.: Teaching mathematics through concept motivation and action learning. Educ. Res. Int. 2019, Article ID 3745406 (2019). https://doi. org/10.1155/2019/3745406 4. Biziuk, A., Biziuk, V., Shakurova, T.: Analysis of teaching elements on technical and mathematical disciplines in modern distance education. Technol. Audit Prod. Reserves 4(2(60)), 28–32 (2021). https://doi.org/10.15587/2706-5448.2021.237455 5. Karupu, O.W., Oleshko, T.A., Pakhnenko, V.V., Pashko, A.O.: Applying information technologies to mathematical education of IT specialists in English-speaking academic groups. Bull. Taras Shevchenko Natl. Univ. Kyiv Ser. Phys. Math. 4, 70–75 (2019). https://doi.org/ 10.17721/1812-5409.2419/4.9 6. Bolitho, R., West, R.: The Internationalisation of Ukrainian Universities: The English Language Dimension. Stal, Kyiv (2017) 7. Kvasova, O., Westbrook, C., Westbrook, K.: Provision of English-medium instruction: trends and issues. Ars Linguodidacticae 5, 11–21 (2020). https://doi.org/10.17721/2663-0303.2020. 5.02 8. Karupu, O.W., Oleshko, T.A., Pakhnenko, V.V.: About teaching of mathematical disciplines in English to foreign students. East.-Eur. J. Enterp. Technol. 2(2(56)), 11–14 (2012). (in Ukrainian). https://doi.org/10.15587/1729-4061.2012.3657 9. Fedak, S.I., Romaniuk, L.A., Fedak, S.A.: Teaching the subject “Higher Mathematics” in English to foreign students of construction specialties. Acad. Commentaries. Ser.: Pedagogical Sci. 156, 106–111 (2017). (in Ukrainian) 10. Snizhko, N.V.: On the problems of teaching the course of higher mathematics in English. In: Science Week: Abstract Annual Science Practical Conference on Zaporizhzhia: ZNTU, pp. 292–293 (2018). (in Ukrainian) 11. Pashko, A., Pinchuk, I.: Methods of classifying foreign language communicative competence using the example of intending primary school teachers. In: Babichev, S., Lytvynenko, V., Wójcik, W., Vyshemyrskaya, S. (eds.) ISDMCI 2020. AISC, vol. 1246, pp. 98–113. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-54215-3_7 12. Rybalko, A.P., Stiepanova, K.V.: Features of teaching higher mathematics in English to students of computer specialties. Profession. Teach.: Theoret. Methodol. Aspects 12, 33–44 (2020). (in Ukrainian) 13. Clarke, P.A.J., Kinuthia, W.A.: Collaborative teaching approach: views of a cohort of preservice teachers in mathematics and technology courses. Int. J. Teach. Learn. High. Educ. 21(1), 1–12 (2009)

948

O. Karupu et al.

14. Seidouvy, A., Schindler, M.: An inferentialist account of students’ collaboration in mathematics education. Math. Educ. Res. J. 32, 411–431 (2020). https://doi.org/10.1007/s13394019-00267-0 15. Vlasenko, K., Chumak, O., Sitak, I., Chashechnikova, O., Lovianova, I.: Developing informatics competencies of computer sciences students while teaching differential equations. Revista Espacios 40(31), 11 (2019) 16. Tarasenkova, N., Akulenko, I., Kulish, I., Nekoz, I.: Preconditions and preparatory steps of implementing CLIL for future mathematics teachers. Univers. J. Educ. Res. 8(3), 971–982 (2020). https://doi.org/10.13189/ujer.2020.080332 17. Kyrpychenko, O., Pushchyna, I., Kichuk, Ya., Shevchenko, N., Luchaninova, O., Koval, V.: Communicative competence development in teaching professional discourse in educational establishments. Int. J. Mod. Educ. Comput. Sci. 13(4), 16–27 (2021). https://doi.org/10.5815/ ijmecs.2021.04.02 18. Pashko, A., Oleshko, T., Syniavska, O.: Simulation of fractional Brownian motion and estimation of Hurst parameter. In: Proceedings of the IEEE 15th International Conference on Advanced Trends in Radioelectronics, Telecommunications and Computer Engineering, pp. 632–637. IEEE (2020). https://doi.org/10.1109/TCSET49122.2020.235509 19. Pashko, A.O., Lukovych, O.V., Rozora, I.V., Oleshko, T.A., Vasylyk, O.I.: Analysis of simulation methods for fractional Brownian motion in the problems of intelligent systems design. In: Proceedings of the 2019 IEEE International Conference on Advanced Trends in Information Theory, pp. 373–378. IEEE (2019). https://doi.org/10.1109/ATIT49449.2019.9030478 20. Pakhnenko, V.V.: About some features of teaching ordinary differential equations to future specialists in the aviation industry. In: Proceedings of the XIII International Conference Modern Education – Accessibility, Quality, Recognition, pp. 136–128. DSMA, Kramatorsk (2021). (in Ukrainian) 21. Karupu, O.W., Oleshko, T.A., Pakhnenko, V.V.: Modeling future aviation and IT specialists’ professional skills development on mathematical practical training with application of information technologies. In: Proceedings of the 2021 IEEE 3rd International Conference on Advanced Trends in Information Theory (ATIT), pp. 215–220. IEEE (2021). https://doi.org/ 10.1109/ATIT54053.2021.9678904 22. Lim, D.H., Morris, M.L., Kupritz, V.W.: Online vs. blended learning: differences in instructional outcomes and learner satisfaction. J. Asynchronous Learn. Netw. 11(2), 27–42 (2007) 23. Borba, M.C., Askar, P., Engelbrecht, J., Gadanidis, G., Llinares, S., Aguilar, M.S.: Blended learning, e-learning and mobile learning in mathematics education. ZDM Math. Educ. 48(5), 589–610 (2016). https://doi.org/10.1007/s11858-016-0798-4 24. Means, B., Toyama, Y., Murphy, R., Baki, M.: The effectiveness of online and blended learning: a meta-analysis of the empirical literature. Teach. Coll. Rec. 115(3), 1–47 (2013) 25. Asarta, C.J., Schmidt, J.R.: The effects of online and blended experience on outcomes in a blended learning environment. Internet High. Educ. 44, Article no. 100708 (2020). https:// ejournal.undiksha.ac.id/index.php/JERE 26. Sakpere, A.B., Oluwadebi, A.G., Ajilore, O.H., Malaka, L.E.: The impact of COVID-19 on the academic performance of students: a psychosocial study using association and regression model. Int. J. Educ. Manag. Eng. 11(5), 32–45 (2021). https://doi.org/10.5815/ijeme.2021. 05.04 27. Kuzminykh, Ie., Yevdokymenko, M., Yeremenko, O., Lemeshko, O.: Increasing teacher competence in cybersecurity using the EU security frameworks. Int. J. Mod. Educ. Comput. Sci. 13(6), 60–68 (2021). https://doi.org/10.5815/ijmecs.2021.06.06 28. Nuga, O.A.: An application of the two-factor mixed model design in educational research. Int. J. Math. Sci. Comput. 5(4), 24–32 (2019). https://doi.org/10.5815/ijmsc.2019.04.03

Application of Google Workspace in Mathematical Training

949

29. Adebayo, E.O., Ayorinde, I.T.: Efficacy of assistive technology for improved teaching and learning in computer science. Int. J. Educ. Manag. Eng. 12(5), 9–17 (2022). https://doi.org/ 10.5815/ijeme.2022.05.02 30. Karupu, O.W., Oleshko, T.A., Pakhnenko, V.V.: On peculiarities of teaching linear algebra to future IT specialists within the program “Education in English” of the National Aviation University. Phys. Math. Educ. 4(26), 21–26 (2020). https://doi.org/10.31110/2413-1571-2020026-4-003 31. Karupu, O.W., Oleshko, T.A., Pakhnenko, V.V.: About teaching certain sections of higher mathematics to students of technical specialties in multinational academic groups at the National Aviation University. Sci. Educ. New Dimension. Pedagogy Psychol. IX(97)(246), 17–20 (2021). (in Ukrainian). https://doi.org/10.31174/SEND-PP2021-246IX97-04

The Current Situation and Development Trend of the Evaluation of Moral and Ideological Curriculum in China-Visual Analysis Based on CNKI Publications Yunyue Wu(B) , Yanzhi Pang, and Xiaoli Teng School of Transportation, Nanning University, Nanning 530200, China [email protected]

Abstract. The implementation of moral education in teaching is to adhere to the combination of knowledge impartation and value orientation, and cultivate students into talents with both morality and talent and all-round development. This is the focus of teaching reform in colleges and universities at this stage. This paper takes curriculum moral education as the research object. This paper uses Citespace software to visually analyze the knowledge map of 474 curriculum ideological and moral evaluation documents in CNKI database from 2017 to 2021 through the number of documents, co-occurrence of authors, co-occurrence of institutions, cooccurrence of keywords and clustering. The results show that the current research focuses on ideological and political courses, teaching reform, teaching design, moral and ideological education and professional courses. The future research should focus on the construction of curriculum moral and ideological indicators and evaluation system, integrate the spirit of ideals and beliefs into knowledge learning, and effectively improve the comprehensive quality and ability of college students. Keywords: Moral and ideological curriculum · Teaching evaluation · Citespace · Reform in education · Knowledge graph

1 Introduction At the National Ideological and Political Conference of Colleges and Universities in 2016, China’s leadership proposed that professional teachers should carry out Curriculum Ideological and Political Education, which means to promote education in ideology, morality and ethics in curriculum teaching. In December 2017, the Outline for the Implementation of the Quality Improvement Project of Ideological and Political Work in Colleges and Universities was issued, which clearly pointed out that the “ideological and political curriculum”, namely, moral and ideological education, should be included in teacher performance assessment. In June 2020, the Guiding Outline for the Construction of Moral and Ideological Education in Colleges and Universities issued by the Ministry of Education once again emphasized that its construction should be comprehensively promoted in all universities and disciplines. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 950–961, 2023. https://doi.org/10.1007/978-3-031-36118-0_81

The Current Situation and Development Trend of the Evaluation of Moral

951

The evaluation of teaching effect is not only a tool to check the effect of teaching reform, but also a reference for continuous improvement [1–4]. As a “baton” to test the implementation of “curriculum ethical and ideological education”, the evaluation of “curriculum ideological and moral education” is a problem that schools and teachers are generally confused and urgently need to solve in the process of comprehensive implementation of ideological and moral education. As an important tool for visualization research, Citespace is widely used in the analysis of research status and research trends. For example, Rawat K S and other scholars used the software to analyze the application of computers in the field of education [5]. For another example, scholars such as Zhang Sheng Liu sorted out the literature on curriculum ideological and political research [6]. Li Xiangxiang and others also used Citespace to draw a map of research hotspots and development trends of Moral and ethical education in university courses [7]. However, most of the existing studies have been carried out from the theme of curriculum Moral and ethical evaluation, and the research focus and trend of curriculum ideological and moral evaluation need to be further deepened. In this context, this paper uses CNKI database literature as the data source to visually analyze the current situation of curriculum Moral and ethical evaluation, and forecasts the research trend, hoping to provide reference for curriculum ideological and political evaluation.

2 Data Sources and Research Methods 2.1 Data Source The sample data of this paper are all from the CNKI database, with the theme of “Moral and ethical evaluation of courses”, the time range from 2017 to 2021, and the source category is the core of Peking University, CSSCI. A total of 474 journal papers were searched. After manual verification by the author and two other scholars, it was finally determined that 474 sample literature data were obtained, which are all valid and valuable references. 2.2 Research Methods In the work of scientific research, we need to find useful information in a large number of documents and read as comprehensively as possible. In the face of numerous vast literature databases, it is a hard job to select targeted and high-quality articles to school. It is necessary to excavate the frontiers of the discipline, find out the key documents worthy of intensive reading and research hotspots, which can not only broaden the horizon of authors, but also inspire them to generate new ideas and creativity. This is the first problem to be solved before conducting research. CiteSpace is such an excellent bibliometrics software. It can participate in advanced methods to visually display the relationship between documents in the form of scientific knowledge atlas in front of users, so that users can have a general understanding of future research prospects by combing the past research tracks. This paper uses Citespace software to visually analyze the sample literature. The Citespace software was developed and designed by Professor Chen Chaomei. Domestic

952

Y. Wu et al.

scholars have widely used it in information science, information and library science, computer science, business economics and other fields. Through the generation of maps, the structure, laws and distribution of scientific knowledge are presented intuitively and clearly.

3 Data Analysis For the data analysis of published literature, it mainly includes several aspects, namely: research space-time distribution, source journal distribution, author statistics and cooperation network analysis, as well as the analysis of the author’s affiliations and cooperation network. 3.1 Research on Space-Time Distribution The number of papers published can show the development trend of research topics. This paper makes statistics on the relevant literature published in China from 2017 to 2021. After searching and selecting, the final sample number of papers is 474, as shown in Fig. 1. It can be seen from the figure that in the past five years, the research on ideological and political evaluation of curriculum in China has shown a significant upward trend. In 2017, the number of papers issued was only 7, while in 2021, it has risen to 279, which also shows that the research on ideological and moral evaluation of curriculum has been paid more and more attention by the academic community.

Fig. 1. Distribution of documents published in 2017–2021 on moral and ideological courses evaluation

3.2 Distribution of Source Journals The source of journals reflects the quality of papers to a certain extent, and also reflects the spatial distribution characteristics of the research field. By analyzing the number of papers issued by journals, the focus and research groups in the research field can be clarified [8]. After analyzing the sample documents of ideological and political evaluation of courses (see Table 1), it is found that Chinese Foreign Languages, Sports Journal, China Audio Visual Education, Journal of Xinjiang Normal University (Philosophy and Social Sciences Edition), Journal of Beijing Sport University and other journals have the most papers on ideological and moral evaluation of courses.

The Current Situation and Development Trend of the Evaluation of Moral

953

Table 1. List of top 6 journals of ideological and moral evaluation of courses in 2017–2021 number

Journal

number of publications

1

Foreign Languages in China

7

2

Journal of Physical Education

7

3

Journal of Southeast University(Philosophy and Social Science)

4

4

China Educational Technology

3

5

Journal of Xinjiang Normal University(Edition of Philosophy and Social Sciences)

3

6

Journal of Beijing Sport University

3

3.3 Author Statistics and Cooperation Network Analysis The analysis of the number of papers and cooperation of the authors can clearly understand the key figures in this field and determine the research authority in this field [9]. In this paper, the author used Citespace software to measure the author and the author’s cooperation, and obtained the co-occurrence chart of the authors of the course ideology, morality and ethics evaluation study from 2017 to 2021 (see Fig. 2).

Fig. 2. Co-occurrence of author cooperation in 2017–2021 on the evaluation of moral and ideological curriculum

954

Y. Wu et al.

It can be seen from the figure that the cooperation of scholars is not obvious. Most scholars only have one article on moral and ethical evaluation of curriculum, and Shen Yang has the largest number (3) of articles. 3.4 Analysis of the Cooperation Network of the Sending Agency The co presentation chart of the publishing houses can show the academic status of the research institution in a certain field. This paper conducts measurement and statistics on the number of documents issued by the issuing institutions, and finally forms a cooccurrence chart of the research institutions (Fig. 3). It can be seen from the figure that the sending institutions are basically from universities, and the institutions with the largest number of documents are Northeast Normal University, Beijing Foreign Studies University, Hehai University, Wuhan University, Southeast University, etc. The above schools have made great contributions to the study of ideological and moral evaluation of courses.

Fig. 3. Co-occurrence of author organizations of 2017–2021 curriculum ideological and moral evaluation

4 Keyword Analysis 4.1 Keyword Co-occurrence Analysis “Co-occurrence”, as the name implies, refers to the phenomenon of the common occurrence of information described by the characteristics of literature. “Co-occurrence analysis” is a quantitative study of co-occurrence phenomenon, which aims to reveal the content relevance of information and the knowledge implied in the characteristic items. The

The Current Situation and Development Trend of the Evaluation of Moral

955

feature items extracted in this paper mainly focus on keywords, including co-occurrence analysis, clustering analysis and emergence analysis of keywords. Using the keywords in the bibliography, CiteSpace software is used to conduct keyword co-occurrence analysis on the sample literature of curriculum ideological and political evaluation. The author input the sample literature into CiteSpace 6.1 software and got the co-occurrence diagram of research keywords (Fig. 4). It can be seen from the figure that scholars mainly focus on moral and ethical courses, teaching reform, teaching design, ideological and moral education, professional courses, etc. How to carry out moral and ethical evaluation of courses is the focus of scholars’ research.

Fig. 4. Key words of 2017–2021 curriculum ideological and moral evaluation

4.2 Keyword Clustering Analysis The author conducted a cluster analysis of keywords and got Fig. 5. The smaller the number in the figure, the more keywords it contains, and the higher its compactness. According to the summary of the cluster diagram, the author divides it into three aspects. (1) The teaching reform of ideology, morality and ethics evaluation of professional courses. Curriculum is the foothold of curriculum moral and ethical education. Professional courses should carry out ideological and moral education together with moral and ethical courses, so ideological and political education in professional courses will become the focus of professional curriculum teaching reform. It is not only necessary to integrate moral and ethical elements into professional courses, but also to carry out curriculum ideological and moral evaluation reform. Some scholars have explored the morality, ideology and ethics evaluation of English curriculum, and

956

Y. Wu et al.

believe that the ideological and political evaluation of curriculum should include not only the evaluation of students’ acquisition, but also the evaluation of teaching process, as well as the evaluation of teachers [10]; Physical education teachers integrate the spirit of women’s volleyball into the volleyball general course, and design the evaluation requirements for ideology, morality and ethics content according to this, including five dimensions of curriculum moral and ideological education, including teacher level and development, curriculum design, teaching content, teaching implementation, student cognition and growth [11]; Teachers of traditional Chinese medicine have designed the evaluation index of “the same standard” for professional teaching evaluation and ideology, morality and ethics education evaluation, with teaching objectives, teaching design, teaching methods and teaching effects as the main evaluation elements [12]. (2) The evaluation principles of ideological and political evaluation of curriculum. The ultimate goal of implementing the curriculum ideology, morality and ethics reform in colleges and universities is to achieve the cultivation goal of “establishing morality and cultivating people”. When carrying out the curriculum ideological and moral evaluation, the fundamental requirement of evaluation should be “establishing morality and cultivating people” [13]. The ideology, morality and ethics evaluation of curriculum should follow the basic requirements of effectiveness orientation and student development, and form an evaluation system that avoids rigidity, advocates comprehensiveness, emphasizes perception and views effectiveness [14, 15]. (3) The practical path of ideological and moral evaluation of curriculum. An important system for the implementation of curriculum ideology, morality and ethics evaluation is to build a scientific evaluation system. At present, domestic research on the construction of curriculum ideological and ethical evaluation system mainly focuses on the design of evaluation indicators. The evaluation indicators are mostly based on the experience of experts and scholars, literature and national policy documents. On this basis, scholars have designed different evaluation indicator systems. For example, Tan Hongyan (2020), after a year of research and learning from the concepts and methods of professional evaluation and curriculum evaluation, believes that indicators at the level of schools and secondary units should include six dimensions: toplevel design, teacher team building, teaching system, textbook construction, quality assurance and output of results, while indicators at the level of teachers cover five parts: syllabus, teaching content, teaching methods, learning effectiveness and teaching reflection [16]; Sun Yuedong (2021) and other scholars used literature research, questionnaires and expert consultation methods to design an evaluation index system that combines peer evaluation, student evaluation and teacher self-evaluation [17]; The evaluation of students is mostly carried out by teachers according to the characteristics of the curriculum, focusing on the process evaluation and value-added evaluation of students [18, 19].

4.3 Keyword Emergence Analysis Emergent words are keywords with a sudden increase in citation frequency in a certain period of time, which can be used to reflect the research trend in a certain period of time;

The Current Situation and Development Trend of the Evaluation of Moral

957

Fig. 5. Keyword cluster diagram of 2017–2021 curriculum ideological and moral evaluation

Judge and identify frontier hotspots, and master the evolution of research topics in the field [20]. Generate the keyword emergence diagram of curriculum ideology, morality and ethics evaluation through citespace (as shown in Fig. 6). It can be seen from the figure that from the timeline, keywords can be divided into two periods. The first period is from 2017 to 2018, including key words such as ideological and moral construction, value guidance, educational function, and curriculum education. In this period, most scholars focus on the role of curriculum ethical and ideological, as well as the theoretical basis of curriculum ideological and moral evaluation [21]; After 2020, the emerging keywords include professional ideology, morality and ethics education, teaching methods, curriculum reform, teaching quality, etc. Scholars pay more attention to the reform and practice of curriculum ideological and political evaluation in the curriculum [22], and the construction of curriculum ideology, morality and ethics evaluation system [23, 24].

5 Results and Enlightenment of CiteSpace Analysis Ideological and moral curriculum is an educational and teaching concept, which plays a role in cultivating college students’ world outlook, outlook on life and values. It is also a way of thinking. The discipline thinking of moral education is used to refine the cultural genes and value paradigms contained in professional courses, transform them into concrete and vivid effective teaching carriers, and integrate the spiritual guidance of ideal and belief into students’ knowledge learning in its unique and silent way. In the ways of combination of infusion and infiltration, theory with practice, history and reality, explicit education and implicit education, commonness and individuality, as

958

Y. Wu et al.

Fig. 6. Key words of 2017–2021 curriculum ideological and moral evaluation

well as the combination of positive education and discipline, moral and ethical education in curriculum teaching is capable to be improved, so as to guide students’ moral development to the correct and healthy direction. In this paper, CiteSpace software is used to conduct a quantitative and qualitative analysis of the research literature on curriculum ideological and moral evaluation. It is found that although the research on curriculum ethical and ideological evaluation has achieved fruitful results, it still needs to be further deepened from the following aspects: (1) In terms of research methods. At present, most of the scholars’ research is based on personal experience summary, with more qualitative research and more practical research. However, theory is the basis for guiding practice. For how to evaluate the ideological and moral curriculum, we need to adopt a variety of research methods, and the combination of quantitative and qualitative needs to be further deepened; (2) As for the research content, the research focuses on evaluation subject, evaluation object, evaluation principle and evaluation index. Although some achievements have been made, a universally recognized evaluation system has not been formed; (3) From the perspective of system construction, although various schools around the country have carried out a lot of practice, some regions and universities have not yet formed a system and mechanism for how to carry out ideological and moral evaluation of courses, lacking a complete supervision and evaluation mechanism.

The Current Situation and Development Trend of the Evaluation of Moral

959

6 Conclusion and Outlook 6.1 Conclusion of the Research In this paper, Citespace software is used to analyze the current literature on ideological and moral evaluation of courses. Through word frequency co-occurrence, word frequency clustering and word frequency emergence, the following conclusions are drawn: (1) From the above analysis, it can be seen that the evaluation of curriculum ethical and ideological education is a hotspot in the current academic research. The exploration of the theoretical basis and practical logic of curriculum ideology, morality and ethics education will continue for a very long time, and the number of papers in the future will continue to rise. (2) Through keyword co-occurrence and cluster analysis, we can see that the domestic focus on the evaluation of moral and ideological teaching quality of the curriculum is mainly; The teaching reform of ideological and ethical evaluation of specialized courses, the evaluation principles of ideological and moral evaluation of courses and the practical path of ideological and political evaluation of courses. Through the keyword emergence analysis, we can see that the construction of the evaluation system of the ideological and moral teaching quality of the curriculum will be the focus of the study. 6.2 Outlook for Future Research Based on the above analysis, we can see that a lot of research has been done on the evaluation of the teaching quality of ideological and moral courses in China, with many achievements. However, the current research on the evaluation of ideological and moral education in curriculum teaching in colleges and universities is not deep enough, and there is still a certain gap with the actual needs of the reform of ideological and ethical courses. Future research should focus on the following two aspects. (1) The evaluation index of curriculum ideology and politics is the key to the evaluation of curriculum ideology and politics It is an important part of the construction of the curriculum ideological and moral evaluation system to establish an evaluation index for the teaching quality of ideological and moral education, which has both reliability and validity and is easy to operate. At present, no authoritative evaluation indicators have been formed. The evaluation indicators designed by scholars are only based on a single school, or a major, or a single course. To form high-quality evaluation indicators, a lot of research and data support are needed. (2) The evaluation system of ideological and moral education in curriculum teaching is the guarantee of the implementation of ideological and political evaluation of curriculum. The ideological and moral education of the curriculum cannot be achieved alone, but requires the cooperation of “five aspects of education”. Then we should do a good job in top-level design and formulate corresponding systems, such as making clear the status of curriculum ideological and political education, making good curriculum

960

Y. Wu et al.

ideology, morality and ethics planning, investing funds to ensure the implementation of curriculum ideological and moral education, improving teachers’ teaching ability of curriculum ideological and ethical education, and including curriculum ideology, morality and ethics performance into the annual assessment, etc. Acknowledgment. This paper is supported by three project funds: (1) The 2023 of the 14th Five-Year Plan for Educational Science in Guangxi “Research on the Quality Evaluation System of Ideological and Political Education in Colleges and Universities (2023C691)”; (2) The 2022 Guangxi Vocational Education Teaching Reform Research Project “Research and Practice on the Effective Connection of the Transportation Logistics Major Group Undergraduate Courses and the Construction of the Curriculum System” (GXGZJG2022B181); (3) The 2022 Guangxi Higher Education Undergraduate Teaching Reform Project “Research and Practice on the Evaluation System of the Teaching Quality of Applied Colleges and Universities in the OBE Perspective” (2022JGA390).

References 1. He, Y., Yang, M.: Study on quality evaluation system for university students. Int. J. Wirel. Microwave Technol. (IJWMT) 1(3), 29–34 (2011) 2. Nicolas Gravel, C.A., Levavasseur, E., Moyes, P.: Eval. Educ. Syst. 53(45), 5177–5207 (2021) 3. Zhong, L., Qi, C., Gao, Y.: Deep learning-assisted performance evaluation system for teaching SCM in the higher education system: performance evaluation of teaching management. Inf. Resour. Manag. J. 35(3), 1–22 (2022) 4. Amjad, M., Linda, N.J.: A web based automated tool for course teacher evaluation system (TTE). Int. J. Educ. Manag. Eng. (IJEME) 10(2), 11–19 (2020) 5. Rawat, K.S., Sood, S.K.: Knowledge mapping of computer applications in education using CiteSpace. Comput. Appl. Eng. Educ. 29(5), 1324–1339 (2021) 6. Liu, Z., Zhu, S., Liu, G.: Visualization analysis of curriculum moral and ideological research in China from the perspective of bibliometrics. Creat. Educ. 10(10), 2201–2218 (2019) 7. Li, X., Zhu, F., Sun, Z.: Research hotspot and development trend of ideological and political curriculum in colleges and universities - visual analysis based on Citespace knowledge map. J. Dali Univ. 5(01), 42–48 (2020). (in Chinese) 8. Du, L., Hao, Z.: The current situation, context and trend of family finance research - a visual analysis based on WoS and CNKI journal papers. J. Southwest Minzu Univ. (Human. Soc. Sci. Edn.) 40(02), 114–124 (2019). (in Chinese) 9. Zhang, H., Hou, Y.: Summary of research on innovation behavior of domestic employees based on knowledge map. Sci. Technol. Progr. Countermeas. 34(11), 153–160 (2017). (in Chinese) 10. Liu, B., Feng, L.: Construction of ideological and political system for English major courses: realistic difficulties and breakthrough paths. Foreign Lang. Audio Vis. Teach. 2022(04), 23– 28+112 (2022). (in Chinese) 11. Zhang, M., Yuan, F., Liang, Z.: Research on the ideological and political theory and practice of college volleyball curriculum in the context of the integration of sports and education – the design of the volleyball general course with the integration of women’s volleyball spirit. J. Beijing Sport Univ. 44(09), 156–165 (2021). (in Chinese) 12. Fang, W., Bao, Y.: Exploration of ideological and political education in traditional Chinese medicine. China Vocat. Tech. Educ. 35, 55–60 (2020). (in Chinese)

The Current Situation and Development Trend of the Evaluation of Moral

961

13. Ao, Z., Wang, Y.: Research on the value core of “ideological and political curriculum” in colleges and universities and the choice of its practice path. Heilongjiang High. Educ. Res. 37(03), 128–132 (2019). (in Chinese) 14. Lu, D.: Design and implementation of ideological and political evaluation of courses. Ideologic. Theoret. Educ. 03, 25–31 (2021). (in Chinese) 15. Du, Z., Zhang, M., Qiao, F.: Principles, standards and operating strategies for teaching evaluation of ideological and political education in science and engineering courses. Ideologic. Theoret. Educ. 2020(07), 70–74 (2020). (in Chinese) 16. Tan, H., Guo, Y., Wang, J.: Construction and improvement of ideological and political evaluation index system of university curriculum. Res. Teach. Educ. 32(05), 11–15 (2020). (in Chinese) 17. Sun, Y., Cao, H., Yuan, X.: Research on the construction of the evaluation index system of ideological and political teaching of science and engineering courses. J. Jiangsu Univ. (Soc. Sci. Edn.) 23(06), 77–88+112 (2021). (in Chinese) 18. Bi, J.: Building the “trinity” of “curriculum, thought and politics” – taking the economics course as an example. J. Shanxi Univ. Finan. Econ. 42(S2), 57–60+71 (2020). (in Chinese) 19. Xie, X., Wang, L.: Mining of ideological and political elements and teaching design of agricultural extension course based on MOOC. J. Southwest Normal Univ. (Nat. Sci. Edn.) 46(08), 132–139 (2021). (in Chinese) 20. Li, J., Chen, C.: SiteSpace: Scientific Text Mining and Visualization, 2nd edn. Capital University of Economics and Business Press, Beijing (2017). (in Chinese) 21. Qiu, W.: The value implication and generation path of ideological and political curriculum. Ideologic. Theoret. Educ. 07, 10–14 (2017). (in Chinese) 22. Dong, C., Fan, S., Gao, Y.: Theoretical basis and structural system construction for the establishment of ideological and political elements in physical education major courses. J. Phys. Educ. 28(01), 7–13 (2021). (in Chinese) 23. Jia, Z., Han, X.: Construction of evaluation system on multimedia educational software. Int. J. Educ. Manag. Eng. (IJEME) 3(1), 34–38 (2013) 24. Ning, H.: Analysis and design of university teaching evaluation system based on JSP platform. Int. J. Educ. Manag. Eng. (IJEME) 7(3), 43–50 (2017)

Target Performance of University Logistics Teachers Based on the Combination of AHP and Support Vector Machine in the Context of “Double First-Class” Project Construction Lei Zhang(B) School of Management, Guangzhou City University of Technology, Guangzhou 510800, China [email protected]

Abstract. In the future for a long period of time, “double first-class” has become the direction and subject of higher education reform and development. Based on the historical background of “double first-class” construction, the construction of teacher performance evaluation system in colleges and universities will change accordingly. The construction of teacher performance evaluation system is particularly important. As a means and tool of evaluation, it can be applied to the management of teacher performance evaluation in colleges and universities to verify the performance of teachers’ responsibilities, improve their initiative, enhance their sense of competition, and improve their teaching level. Thus, a good competition and incentive mechanism can be formed, which has gradually become an important part of university management. Under the background of “double first-class” construction, this paper selects the main indicators that affect the target performance of logistics teachers in universities and colleges, and constructs an evaluation system based on them. At the same time, the principle of analytic hierarchy process is used to build the hierarchy index system, and the support vector machine is used for comprehensive evaluation, in order to further improve the teaching level of logistics teachers in colleges and universities, scientific research ability, team assistance and other aspects of exploration, in order to do their utmost. Keywords: Logistics teacher · Target performance appraisal · AHP-SVM

1 Introduction The General Plan for Promoting the Construction of World-class Universities and Disciplines was issued by The State Council in October 2015, with the purpose of enhancing the international competitiveness and comprehensive strength of China’s higher education, accelerating the construction of world-class disciplines and a number of world-class universities. At the same time, with the advent of industry 4.0 era and the implementation of the strategy of “Made in China 2025”, higher requirements are put forward for higher vocational colleges to create first-class majors, strengthen connotation construction, improve the quality of talent training, strengthen the integration of production © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 962–971, 2023. https://doi.org/10.1007/978-3-031-36118-0_82

Target Performance of University Logistics Teachers

963

and education, and build first-class universities. Among them, teachers are the basic resources for personnel training and school teaching quality [1]. Only when the professional level and comprehensive quality of college teachers are improved, the connotation construction, first-class specialty and quality of personnel training of higher vocational colleges can be improved. At present, higher vocational colleges have entered a stage of rapid development, and performance appraisal is an effective tool for modern enterprises to improve their competitiveness. If it is scientifically applied to teachers’ management in colleges and universities, the performance of teachers’ responsibilities can be investigated and a good incentive mechanism can be formed [2]. However, there are many problems in the evaluation of teachers in higher vocational colleges. The evaluation system, methods and ways have not been formed and perfected, which is far from the requirements of the new situation. It’s very important. This paper puts forward the evaluation index and construction system of teachers’ goal performance in higher vocational colleges based on analytic hierarchy process under the new situation, and uses support vector machine to evaluate it, which is a beneficial exploration to give full play to teachers’ subjective initiative, enthusiasm and creativity and explore the establishment of an incentive mechanism [3]. Foreign scholars have made research on earlier performance evaluation, Chinese scholars in recent years have also made research in this field and made some achievements, in recent years, the performance evaluation of colleges and universities has been studied [4]. The research methods are mainly literature analysis, induction, investigation and comparative analysis. In terms of research content, there are many researches on the index system and evaluation methods of teacher performance evaluation [5]. In recent years, domestic scholars’ researches on how to carry out the reform of teacher performance evaluation under the background of “double first-class” construction are also increasing [6]. However, in general, the current research still has the following problems: First, the breadth and depth of the research is insufficient, there are more researches on general and universal problems, but less special researches [7]. Second, the research field of vision is insufficient. At present, most of the research in China is carried out for ordinary universities, and there are few researches on world-class universities [8]. Third, empirical research is insufficient. There are relatively many literatures based on the methods and index systems of teacher performance evaluation in colleges and universities, but most of them lack the test of empirical cases and lack of case support. Fourth, there are more qualitative studies, but not enough quantitative ones [9]. In the past, only one method was used to study the performance evaluation of teachers in colleges and universities. However, the performance appraisal of college teachers is characterized by complex assessment objectives, difficult to quantify assessment indicators and large subjectivity, so the method combining qualitative analysis and quantitative analysis is more suitable for high vocational school teacher performance assessment. Therefore, this paper puts forward an analytic hierarchy process-support vector machine (AHP-SVM) based performance evaluation method for college teachers.

964

L. Zhang

2 Principles Description 2.1 Principles of the Analytic Hierarchy Process T.L. Schatty, a well-known operations research scholar, proposed the Analytic Hierarchy Process (AHP) in the 1970s. The analytical hierarchy process (AHP) is a multi-objective decision-making method that combines qualitative and quantitative analysis. It can quantify decision-makers’ qualitative judgment. It can produce satisfactory results when the target structure is complex and data is scarce, particularly for complex problems that are difficult to fully quantify [10]. AHP divides complex problems into different elements and merges these elements into different levels by analyzing the factors involved and their interrelationships. This results in a multi-level structure. A judgment matrix can be established at each level by comparing the elements of the layer one by one according to a specified criterion. The weight of the layer elements to the problem can then be calculated by calculating the maximum eigenvalue of the judgment matrix and the orthodontic eigenvector. The combined weights of the elements of each level for the overall objective are calculated on this basis, and the weights of different schemes are obtained, which provides the basis for selecting the optimal scheme [11]. 2.2 Principles of Analytic Hierarchy Statistical learning theory was a machine learning method with a solid theoretical foundation that gradually developed on the basis of traditional statistics, gradually improved, and formed a relatively complete theoretical system in the 1990s. A new pattern recognition method, support vector machine, is proposed on this basis [12]. Based on Statistical Learning Theory (SLT), V. Vapnik proposed the Support Vector Machines (SVM) method [13]. It is a new machine learning method that has demonstrated many unique and exceptional advantages in solving nonlinear, small sample, and high dimensional model problems. It has produced good results in probability density estimation, pattern recognition, function approximation, and many other areas, and it can effectively realize the accurate fitting of high-dimensional nonlinear systems using small samples [14]. Its regression function is as follows: yi [(w · xi + b)] − 1 ≥ 0 i = 1, . . . , n

(1)

The optimal classification surface problem of the above formula is transformed into a dual problem by Lagrange optimization method, and the maximum value of the following function is obtained. Q(a) =

n i=1

ai −

  1 n ai aj yi yj k xi , xj i,j−1 2

(2)

The Lagrange multiplier corresponds to each sample, and the corresponding sample is the support vector, and the optimal classification function can be obtained [14].  n      f (x) = sgn W ∗ · x + b∗ = sgn (3) ai ∗ yi k xi , xj + b∗ i=1

Target Performance of University Logistics Teachers

965

Gaussian radial basis kernel function algorithm has a wide convergence range and only one parameter, which is easy to optimize, so it is the most widely used in support vector machines. In this paper, radial basis kernel function is also used for calculation [15–18, 19]. 2.3 Combination of Analytic Hierarchy Process and Support Vector Machine The AHP (Analytic Hierarchy Process) method decomposes the relevant elements of a problem into objective, standard, scheme and other levels, quantifies people’s subjective judgment objectively with a certain scale, and carries out qualitative and quantitative analysis on this basis. This method is suitable for human qualitative judgment, and the result of decision is difficult to be measured directly and accurately (Table 1). Table 1. Target performance appraisal system index set of vocational college teachers Target layer

Criteria layer

Target performance Basic teaching appraisal of teachers work(B1 ) in vocational colleges (A)

Indicator layer

Description

Number of class hours(C1 )

Complete the basic teaching workload

Graduation guidance (C2 )

Guide students in professional comprehensive practice

Graduation thesis (C3 )

The guidance content is detailed and the paper format is standard

Final examination paper (C4 )

The proposition is in line with the syllabus

Teaching related materials (C5 )

File format specification

Teaching plan (C6 )

The format of teaching plan is standard and the content is designed

Experimental teaching File format materials (C7 ) specification Number of course changes (C8 )

No semester transfer or temporary transfer

Invigilation times (C9 )

10 times per semester

Teaching effect (C10 )

The result of students’ evaluation of teaching was > 95 (continued)

966

L. Zhang Table 1. (continued)

Target layer

Criteria layer

Indicator layer

Discipline and Discipline specialty construction construction (C11 ) (B2 )

Description Participate in the discipline special construction plan

Teaching research and Establishment of reform work (C12 ) teaching research and reform projects Professional curriculum construction (C13 )

Participate in the construction of professional curriculum archives

Practice work (C14 )

Maintain and develop practice base

Teaching and research Carry out more than 5 office activities (C15 ) teaching and research activities per semester Instruct students to compete (C16 ) Scientific research work (B3 )

Three cup competitions

Scientific research School research achievements (papers, performance rating research, patents) standards (C17 ) Academic exchange (C18 )

Attend academic conference

Research institutions and platforms(C19 )

Provincial, departmental and university levels

Scientific research application (C20 )

Provincial, departmental and university levels

Master’s program construction (C21 )

Teacher’s report

Group activity (C22 )

Attendance and participation in group activities

SVM (Support Vector Machines) is a novel machine learning method based on statistical learning theory. It has demonstrated numerous distinct advantages in solving small sample, nonlinear, and high-dimensional pattern recognition problems, as well as good results in pattern recognition, function approximation, and probability density estimation.

Target Performance of University Logistics Teachers

967

The evaluation index system is divided into hierarchical structure by combining analytic Hierarchy Process (AHP) and support vector machine (SVM), and the weight of each index is determined by AHP, and the comprehensive evaluation results are obtained by SVM training.

3 Application and Test of SVM The objective performance indicators used to evaluate teachers in higher vocational colleges are all non-linear and interconnected. As a result, evaluation methods such as the analytic hierarchy process, the fuzzy comprehensive evaluation method, and the factor analysis method are all based on the independence and linear relationship of the indicators, which cannot fully and efficiently solve the evaluation of teachers’ target performance in higher vocational colleges. SVM must be used to train and self-learn the training samples in order to better describe the relationship between each evaluation index and the overall objective evaluation. Only through the statistical learning of the training samples can we constantly discover and explore the internal rules between input and output values. Support Vector Machine (SVM) is a data-driven “black box” modeling method with advantages such as self-learning habit, memorability, good “robustness”, simple operation and no more prior knowledge. Compared with the neural network method, it can effectively deal with the small sample problems that cannot be solved by it. Moreover, support vector machine has advantages in the evaluation of nonlinear problems. Using it for data processing, it can better overcome the nonlinear correlation between various indicators and make the evaluation results more accurate and objective. In the application process of SVM, the determination of kernel function is the most critical step, which needs to transform the input space into a high-dimensional space through the nonlinear transformation defined by kernel function. SVM is actually a linear classifier determined by support vector. 3.1 Selection of SVM Kernel Function Kernel functions most commonly used in support vector machines mainly fall into the following four types: (1) Gaussian radial basis kernel function

  K xi , xj = exp −γ xi − xj 2

(4)

(2) Polynomial kernel function     d K xi , xj = xi · xj + 1

(5)

    K xi , xj = xi · xj

(6)

(3) Linear kernel function

968

L. Zhang

(4) One-dimensional Fourier kernel function



    2 1 − 2q cos xi − xj + q2 K xi , xj = 1 − q2

(7)

Radial basis kernel function is the most widely used among the above four kinds of kernel functions. Because it has a wide convergence domain, contains only one parameter and is easy to optimize, it is a relatively ideal classification basis function. Therefore, this paper adopts radial basis kernel function for calculation. 3.2 Training of SVM Model The sample data of previous college teachers’ performance evaluation were collected and divided into two groups, each of which included those with good performance evaluation and those with poor performance evaluation. The data in this group were taken as training samples, Libsvm2.9 toolbox was adopted and gnuplot tool was used for parameter optimization to determine the final SVM. The training samples are input into SVM for training, and then the test samples are input into the model for testing. If the expected accuracy is not reached, the SVM is trained again, and satisfactory output results are obtained through repeated training. The 183 sets of performance evaluation data of teachers of logistics major in a vocational college in recent three years were used as sample data set, and the sample data were trained for the SVM model. After the training, 183 groups of samples were judged by the back estimation method. The discriminant results were consistent with the actual situation of gas risk occurrence, and the misjudgment rate of the back judgment was 0, indicating that the training results of the prediction model were highly correct and could be applied in practical practice.

4 The Empirical Evaluation of Teachers’ Performance in Vocational Colleges The complexity and multi-objectives of the qualitative evaluation of university teachers’ performance evaluation are difficult problems. This paper proposes a method combining analytic Hierarchy process (AHP) and support vector machine (SVM). On the basis of literature analysis and expert team, AHP is firstly used to hierarchize assessment indicators, decompose assessment objectives into different factors and form a multi-level assessment structure to build an indicator system. 22 evaluation indicators are made into an assessment scale, and the evaluator scores the performance of the evaluated on each item to form a sample data set. Secondly, SVM method is used to carry out hierarchical evaluation of evaluation indicators, and finally get the comprehensive evaluation of college teachers’ performance evaluation. This method combines qualitative and quantitative methods, is practical and easy to operate, and provides objective, accurate, comprehensive and feasible basis for teachers’ performance evaluation in vocational colleges.

Target Performance of University Logistics Teachers

969

In this paper, 183 sets of performance evaluation data of teachers of logistics major in a vocational university in recent three years are selected as sample data. (1) By using ROSETTA software, the continuous indicators among the 22 indicators of evaluation are discretized. Then, the 183 test samples are used to determine C 1 , C 3 , C 10 , C 11 , C 13 , C 16 , C 17 , C 20 as the main indicators. (2) Select the appropriate kernel function of support vector machine. At present, there are three types of inner product kernel functions that are widely used. Generally, radial basis function is considered first when selecting kernel function, mainly because of its relatively few parameters and small numerical difference. Gaussian function in support vector machines, has a wide convergence domain and only contains one parameter, which is easy to optimize. Therefore, radial basis kernel function is adopted in this study. In the MATLAB6.5 environment, LS-SVMlab1.5 calculation kit was used for calculation analysis. The radial basis function was used as the kernel function. The first 140 groups of 183 sample data were used as the training sample set, and the last 43 groups of samples were used as the test sample set to test the generalization ability of the model. (3) Sample training is an important task. 140 training samples were input into the SVMTRAIN tool for training, and the remaining 43 samples were input into the SVMPREDICT tool for testing. (4) Sample test is the last job. After the completion of sample training, through the accuracy of the actual output, it was found that the correct rate of discriminating 43 test samples reached 97.6%. Therefore, the model established in this paper has high accuracy. For the university teacher performance evaluation model constructed in this paper, it has the advantages of simple use, high efficiency, high precision, and not easy to be affected by subjective factors. Based on the accumulation of certain sample data, the model can be continuously trained, so as to further improve the accuracy of the model and provide scientific guarantee for teachers’ performance evaluation.

5 Conclusion The construction of the performance appraisal system for vocational and vocational teachers is a systematic project. The establishment of scientific performance indicators can maximize the work enthusiasm and initiative of the staff, and change the teacher management from form to connotation. It is necessary to bring the establishment of performance appraisal standards, appraisal procedures, and appraisal openness and transparency into the scope of system construction, and strengthen supervision. In order to build the first-class teaching team under the background of “double first-class” in higher vocational colleges, we need to pay more attention to the professional development of teachers and improve their teaching, scientific research and practical abilities through multiple channels. It is necessary for schools to formulate school development strategies from a long-term perspective according to different stages of development, so as to achieve a multi-dimensional and multi-level performance appraisal system for schools, so as to rapidly improve the comprehensive ability of teachers.

970

L. Zhang

To realize the strategic goal of “double first-class” construction, it is necessary to strengthen the construction of first-class teaching staff, which is also the urgent work of all colleges and universities at present. “Double first-class” development needs to be performance-centered and driven by innovation. It requires bold innovation, active exploration and quick implementation. At present, the research on teacher performance evaluation in higher vocational colleges is still in the initial stage of exploration, and there are still a lot of work to be done. Based on literature analysis and expert group discussion, this paper proposes an evaluation model based on AHP-SVM, and evaluates the established target performance evaluation index system of higher vocational teachers. The accuracy and rationality of this evaluation method are demonstrated by the comprehensive evaluation of the teachers’ goal performance in a vocational college. Through this method, it is helpful to explore the vocational teachers’ target performance evaluation and provide more extensive ideas.

References 1. Zhong, B.: Firmly promote the construction of world-class universities and disciplines. Educ. Res. 39(10), 12–19 (2018). (in Chinese) 2. Wang, H., Wu, F.: Research hotspots and development of teacher performance evaluation in Chinese universities. J. Natl. Inst. Educ. Adm. 11, 45–52 (2016) 3. Zhao, L., Huang, P.: Teacher performance evaluation in colleges and universities: pursuing the balance between quantity and quality. J. Xidian Univ. (Social Science Edition) 26(3), 91–95 (2016). (in Chinese) 4. Wang, B., He, M.: Salary system of university teachers under the background of “double first-class” construction. China High. Educ. 5, 14–17 (2017). (in Chinese) 5. Yao, X.: Discussion on university teacher performance management system to promote the development of “double first-class” strategy. Natl. Inst. Educ. Adm. 2, 57–62 (2017). (in Chinese) 6. Wang, B.: Salary reform in colleges and universities: improving quality and efficiency and stimulating teachers’ potential-salary system under the background of comprehensive reform in colleges and universities research. China High. Educ. 7, 9–13 (2016). (in Chinese) 7. Lu, D.: Research on teacher professional development of Stanford university and its enlightenment. China High. Educ. Res. 3, 48–54 (2014). (in Chinese) 8. Li, C., Zhang, l, Su, Y.: An empirical study on the relationship between salary structure, job satisfaction and job performance of university teachers. Fudan Educ. Forum 5, 89–95 (2016). (in Chinese) 9. Lu, W., Song, X.: A study on the performance evaluation of university administrators based on analytic hierarchy process – a case study of a university in Guangdong province. High. Educ. Explor. 10, 40–46 (2017). (in Chinese) 10. Wang, L., Xu, S.: Introduction to analytic hierarchy process, pp. 37–155. China Renmin University Press, Beijing (1990). (in Chinese) 11. Vapnik, V.: The nature of statistical learning theory, pp. 56–112. Tsinghua University Press, Beijing (2000). (in Chinese) 12. Zhang, X.: On Statistical learning theory and support vector machines. Acta Automatica Sinica 26(1), 32–34 (2000) 13. Zhang, L.: Evaluation of China’s petroleum security system: based on rough set and support vector machine. China Soft Science 11, 13–19 (2022). (in Chinese)

Target Performance of University Logistics Teachers

971

14. Zhang, Z.: Research on intrusion detection model of wireless sensor networks based on RS-SVM. Intell. Comput. Appl. 3, 319–320 (2019) 15. Júnior, P.R.M., Boult, T.E., Wainer, J., Rocha, A.: Open-set support vector machines. IEEE Trans. Syst. Man Cybern. Syst. 52(6), 3785–3798 (2022) 16. Piccialli, V., Sciandrone, M.: Nonlinear optimization and support vector machines. Ann. Oper. Res. 314(1), 15–47 (2022) 17. Yadav, O.P., Pahuja, G.L.: Bearing fault detection using logarithmic wavelet packet transform and support vector machine. Int. J. Image, Graph. Signal Process. (IJIGSP) 11(5), 21–33 (2019) 18. Zhou, L., Amoh, D.M., Boateng, L.K., Okine, A.A.: Combined appetency and upselling prediction scheme in telecommunication sector using support vector machines. Int. J. Mod. Educ. Comput. Sci. (IJMECS) 11(6), 1–7 (2019)

Research and Practice of Outcome-Oriented Innovative Talent Training System of Material Specialty Hu Yang and Qian Cao(B) School of Materials Science and Engineering, Wuhan University of Technology, Wuhan 430070, China [email protected]

Abstract. In the context of emerging engineering, the new education model based on outcome is the direction of teaching reform. Traditional training systems of material professionals can no longer meet the needs of the current society and industry. Thus, it is necessary to reform training systems to promote the balanced development of students’ professional competency, engineering literacy and development ability. Aiming at material specialty, this paper analyzes the problems existing in traditional talent training system, and studies the latest talent needs in society by tracking and investigating graduates and employers. Based on the above analysis and research, a new talent training course system, teaching system and evaluation system are constructed. After the implementation of this talent training system, students’ participation in innovation and entrepreneurship training, competitions and other extracurricular activities has increased significantly. Meanwhile, students’ non-technical ability has also been significantly improved. The acceptance rate of graduate students has reached more than 50%, and 73.97% of graduates are still engaged in material-related work. Keywords: Material specialty · Outcome-based education · Emerging engineering · Talent training system · Non-technical ability evaluation

1 Introduction Different themes and connotations of different industrial development era have put forward different needs for engineering and technical talents, and therefore promoted the continuous change and development of engineering education [1, 2]. Cultivating newtype engineering and technological talents to face the industry, the world, the future, and to support economic development of the innovation era, has become the fundamental purpose of Emerging Engineering Education [3–5]. The traditional talent training system of materials majors can no longer meet the development of Emerging Engineering Education. Thus, it is necessary to explore a new talent training system under the guidance of advanced teaching concepts or models. Outcome-based Education(OBE) is a novel outcome-oriented education model proposed by William G S. in 1994 [6, 7]. It takes learning outcomes that students achieve © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 972–983, 2023. https://doi.org/10.1007/978-3-031-36118-0_83

Research and Practice of Outcome-Oriented Innovative Talent

973

through the education process as the goal of instructional design and implementation. In recent years, many scholars have carried out teaching reforms around this new education model, and conducted several research and practice [8, 9]. Van M. E. et al. [10]. Conducted a study on the evaluation method of competency-based medical education, and proposed six criteria to evaluate the results of teaching reform. Sasipraba T. et al. [11]. Studied and applied the outcome-oriented education model assessment methods and evaluated student performance in capstone projects. The results showed that this assessment method contributes to the continuous improvement of teaching effectiveness. Shreeranga B. et al. [12]. Used a lean thinking approach to study the effectiveness of outcome-oriented teaching models, and showed that this method could not only enhance teaching and learning efficiency, but also improve student collaboration ability. However, at present, there are few studies on the outcome-based talent training system of material specialty. Materials science is a discipline with strong experimental and practical nature, and it is always developed under the traction of social needs [13, 14]. With the rapid development of social economy and related disciplines, advanced materials, technologies, and methods are continuously emerging [15–17]. Therefore, technical talents in material specialty should be equipped with strong engineering innovation capabilities to adapt to future development. For this reason, it is very important to study and build an experimental and practical training system that can comprehensively improve engineering innovation capabilities of talents. This article analyzes the problems of the traditional talent training system of material specialty. In response to the problems, a new training system of innovative talents is constructed based on the OBE concept. The evaluation and implementation methods of non-technical ability teaching is researched and established, and finally the practice has been carried out.

2 Problems Existing in the Traditional Training System of Material Specialty In recent years, with the rapid development of technology, material majors urgently need a group of outstanding talents that can support the transformation and upgrading of the engineering industry [18, 19]. However, the traditional training mode cannot meet the higher requirements. The existing problems could be summarized as follows. (1) Professional education and social needs are not highly fit. Under the background of social transformation, industrial transformation and technological innovation, the demand for talents in new fields and interdisciplinary disciplines is strong. The social needs are no longer a single professional and skilled talent, but an innovative person with a wealth of non-professional and technical capabilities. Traditional

974

H. Yang and Q. Cao

engineering students cannot meet the new social needs, and their career competitiveness is insufficient. (2) Inadequate support for the cultivation of engineering innovation ability of curriculum system. In the past, the curriculum system for material professionals was complicated. It rarely open innovative and entrepreneurial courses or set up internship and practical training credits. Students do not require to participate in competitions or innovation and entrepreneurial training programs before graduation. The training of engineering knowledge and engineering innovation ability is fragmented. The unreasonable curriculum system leads to insufficient training of college students’ ability to systematically design and effectively solve complex engineering problems, especially their engineering literacy and development ability are weak. (3) Insufficient conversion of engineering practice and competency training. Experimental and practical training are formalistic, the teaching mode and training content are single, effective resources are scarce, and practical training lacks innovation and challenges. Because students cannot contact actual production, they cannot understand the specific problems that arise in actual production, and cannot propose solutions to problems according to their professional knowledge, let alone make breakthroughs and innovations on the existing basis. In addition, due to the rapid development of technology, the latest technology and scientific research achievements cannot be converted into teaching content in time, thus traditional theoretical knowledge may deviate from actual production practice, which can easily cause students to be armchair strategists. In the long run, students’ practical ability gradually deteriorates and are unable to meet the needs of industry innovation. (4) Insufficient research of outcome evaluation. The teaching effect assessment is a very important link in the process of talent training [20–22]. Through the effectiveness evaluation of the outcome, continuous improvement of the teaching process can be achieved. However, in the actual process, evaluation is often the most overlooked link. At present, the evaluation system established by current training system of material specialty is not perfect, especially the implementation of non-technical ability training. The assessment and evaluation are vague, lacking effective assessment methods and evaluation standards.

3 Outcome-Oriented Innovative Talent Training System In view of the above problems in the training of material professionals, the researchers of this paper track and investigate graduates and employers, and detail the connotation of innovative talents in material engineering under the background of emerging engineering according to the adjustment of industrial structure and changes in social demand. Talents should not only possess professional ability and self-development ability of lifelong learning to carry out comprehensive research and development, technology creation

Research and Practice of Outcome-Oriented Innovative Talent

975

and engineering transformation during the whole process of materials from design, preparation to service, but also have correct moral and value orientation, as well as excellent engineering literacy. Through the analysis of the updated requirements for the training of material talents, a new curriculum system, teaching system and evaluation system are formulated. 3.1 Course System According to the needs of different levels and different ability aims, a curriculum system for the cultivation of innovative talents in materials majors are designed. The curriculum system consists of systematic theoretical teaching, modular experimental teaching, and advanced practical training (see Fig. 1).

Fig. 1. Outcome-based Innovative talent training system

In the constructed course system, theoretical teaching courses include the basic courses of engineering theory such as Fundamentals of Materials Engineering and Fundamentals of Materials Science. Meanwhile, a practical application courses of Thermal Equipment is also set up. The experimental teaching courses are mainly composed of basic experiments Research Methods and Testing Technology of materials, design experiments Materials Design, and professional comprehensive experiments Materials Preparation and Performance Experiments. According to the students’ professional base, the practical training courses apply design practice, production training and research practice, and offer courses such as Engineering Design Training, Job Internship, and Graduation Thesis, respectively. Through the construction of the above curriculum system, students will be instructed professional knowledge, technical ability, and at the same time their learning skills and non-technical ability could be improved. This curriculum system aims at the cultivation of innovative talents to adapts to the development

976

H. Yang and Q. Cao

of emerging engineering, to achieve effective support for different levels of ability goals and further promote the balanced development of students’ professional competence, engineering literacy and development ability. 3.2 Teaching System In terms of the new requirements of ability training, in this paper, traditional teaching methods are reformed and a multi-collaborative teaching chain of teaching-learningevaluation-research is established (see Fig. 2).

Fig. 2. The multi-collaborative teaching chain of teaching-learning-evaluation-research

Teaching: with the development of science and technology, more and more teaching methods appear [23–25]. This study integrates a variety of teaching methods such as hands-on teaching, digital teaching (digital design training, virtual factory roaming, etc.), and smart teaching (distance online teaching, virtual simulation, etc.), to promote students’ learning effect. Learning: through methods of offline hands-on training, virtual simulation humancomputer interaction experiment, lab-based micro-production line practice, production internship, etc., the teaching effect of experimental and practice courses could be improved. Evaluation: in the context of emerging engineering, the evaluation standards need to assess students’ professional ability, cooperation ability, innovation ability, communication ability and other comprehensive ability, so it is necessary to comprehensively evaluate the students’ pre-class performance, in-process behavior, and final examination. Research: transform the latest scientific research into teaching content and make it advanced. Improve teaching effectiveness by researching teaching methods and developing teaching resources, to promote the continuous progress of the teaching system.

Research and Practice of Outcome-Oriented Innovative Talent

977

3.3 Evaluation System This study mainly focuses on the reform of the experimental and practical teaching evaluation system, which not only examines students’ acquisition of technical skills, but also non-technical abilities such as and social adaptability, professional norms, team cooperation, project management, and lifelong learning. Incorporate experimental data recording, laboratory management system implementation, safe operation, waste liquid disposal, environmental hygiene, and other behaviors into the assessment system for evaluating professional norms, which may enable the cultivation of professional normative capabilities. In the whole process of experimental and practical teaching, the evaluation methods and results are closely related to the outcome objectives, which could improve the effectiveness of the non-technical ability evaluation results. Table 1 shows the evaluation strategy for achieving graduation requirements. To sum up, this new program is formulated for material specialty students of different majors and grades. Throughout the four academic years, it aims to improve students’ ability to design and solve complex engineering problems, cultivating innovative ability and project management capability. Table 1. Competition assessment standard Evaluation items Evaluation method

Engineering knowledge Analytical skills Design & development competency Research skills modern tools usage Social adaptability Professional norms Teamwork Communication skills Project management Lifelong Learning

Course scores

Questionnaires

Behavioral Evaluation

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

Student self-evaluation

Student peer evaluation

counselor evaluation

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

√

Extracurricular assessment

√

978

H. Yang and Q. Cao

4 The Practice and Effectiveness of the Training System The outcome-oriented material professional training system is practiced for students majoring in inorganic non-metallic materials at the School of Materials Science and Engineering of Wuhan University of Technology. 4.1 Analysis of Students’ Ability Attainment In order to ensure that students’ ability meet the training requirements, the courses involved in the newly established training system conduct an analysis of students’ competency attainment. The following uses an experimental course of material preparation and performance experiment as an example to analyze the degree of students’ competency attainment. (See Table 2). Table 2. Analysis of students’ competency attainment Course objectives

Evaluation basis

Evaluation value Full mark E1

Conversion ratio F1

Conversion points A1 = (E1 × F1 )

Sample average score G1

Sample conversion score B1 = (G1 × F1 )

Attainment value C   B1 / A1

1. Master the testing principles and methods of material physical and chemical properties, and design feasible experimental schemes

Prelab report

15

1

15

13.12

13.12

C 1 = 0.8747

2. Master the material performance test method, and utilize experimental equipment to obtain reasonable test data

Experimental process

25

1

25

22.20

22.20

C 2 = 0.8880

(continued)

Research and Practice of Outcome-Oriented Innovative Talent

979

Table 2. (continued) Course objectives

Evaluation basis

Evaluation value Full mark E1

Conversion ratio F1

Conversion points A1 = (E1 × F1 )

Sample average score G1

Sample conversion score B1 = (G1 × F1 )

Attainment value C   B1 / A1

3. Master experimental data processing and analysis methods, and be able to analyze and interpret experimental results

Experimental report of performance experiments (Data processing section)

45

0.5

22.5

36.91

18.46

C 3 = 0.8202

4. Understand and apply the correct method to analyze the relationship between material composition, structure, preparation and performance according to experimental data, and obtain reasonable and valid conclusions

Experimental reports of design experiments (Data processing section)

45

0.5

22.5

39.47

19.74

C 4 = 0.8771

(continued)

980

H. Yang and Q. Cao Table 2. (continued)

Course objectives

Evaluation basis

Evaluation value Full mark E1

Conversion ratio F1

Conversion points A1 = (E1 × F1 )

Sample average score G1

Sample conversion score B1 = (G1 × F1 )

Attainment value C   B1 / A1

5. Experimental Consciously process (data abide by real) engineering professional ethics and norms, and ensure the authenticity of experimental data;

5

1

5

4.95

4.95

C5 = 0.9900

6. Abide by laboratory rules and regulations, and be able to operate experimental instruments and equipment safely and standardly;

Experimental process (6s)

5

1

5

4.92

4.92

C6 = 0.9840

7. Complete Experimental experiments process in groups to (teamwork) build team spirit. The group members have a clear division of work and can complete their respective tasks independently

5

1

5

4.91

4.91

C7 = 0.9820

Sample information (Number of students)

Excellent(above 90 points): 10; good(80-89points): 21; average(70-79points): 5; pass(60–69 points): 0; fail(below 60 points): 0

Attainment

D=

C1 +C2 +C3 +···+Cn × 100 = 91.66 n

Research and Practice of Outcome-Oriented Innovative Talent

981

It can be seen from Table 2, after the reform of the curriculum, students’ professional abilities (engineering knowledge, problem analysis, research, etc.), engineering literacy (professional norms, teamwork, etc.) and development skills (communication, lifelong learning, etc.) have been cultivated, and the training aims have been achieved. This talent training system can not only promote students’ professional ability, but also effectively cultivate students’ engineering literacy and development ability. After the practice of this training system, students’ participation in innovation and entrepreneurship training, competitions and other extracurricular activities has increased significantly. The acceptance rate of graduate students reaches more than 50%, and 73.97% of graduates are still engaged in material-related jobs. 4.2 Questionnaire In the practice of the talent training system, the researchers in this article conducted a teaching effect questionnaire survey on relevant teachers and students. The results of the questionnaire survey are shown in Table 3. Table 3. Teaching Effect Questionnaire Items

Significantly improved

improved

General

Not improved

Engineering knowledge

48.87

10.69

20.36

20.08

Analytical skills

50.23

20.38

10.69

18.70

Design& development competency

62.35

10.36

15.38

11.91

Research skills

70.46

15.25

8.40

5.89

modern tools usage

60.45

20.19

9.65

9.71

Social adaptability

62.68

25.92

8.54

2.86

Professional norms

90.22

5.68

2.76

1.34

Teamwork

86.36

10.23

2.35

1.06

Communication skills

84.64

9.87

4.23

1.26

Project management

93.46

3.45

2.20

0.89

Lifelong Learning

92.53

4.58

1.89

1.00

It can be seen from Table 3 that after the construction of the new talent training system, more than 90% of teachers and students believe that students’ professional norms and lifelong learning ability have been significantly improved, more than 80% of teachers and students believe that team cooperation, and communication skills have been significantly improved, and about 70% of teachers and students believe that students’ research ability has been significantly improved. Evidently, the establishment of this talent training system has remarkably improved students’ non-technical ability.

982

H. Yang and Q. Cao

5 Conclusion In the stage of transforming material specialty from traditional engineering to emerging engineering, society has put forward higher requirements for the training system of material professionals: balanced development of professional competency, engineering literacy and development ability. In order to cultivate material talents that meet the demands of society and the industry, curriculum, teaching methods, evaluation methods need to be reformed to improve the training system for material professionals. Based on above facts, this paper formulates a new curriculum system, teaching system and evaluation system, respectively. The curriculum system consists of the systematic theoretical teaching, and advanced practical training. A diversified and collaborative teaching chain of teaching-learning-evaluation-research is established, and a methodology for evaluating non-technical competencies was implemented. After the practice of the training system, students’ participation in innovation and entrepreneurship training, competitions and other extracurricular activities increased significantly. The acceptance rate of graduate students reached more than 50%, and 73.97% of graduates were still engaged in material-related work. The students’ non-technical ability has improved significantly. This training system provides a reference for material majors in colleges and universities. Acknowledgments. This work has been supported by the teaching reform and research project of Wuhan University of Technology (Grant No. W2022061 and W2022064).

References 1. Jamilurahman, F., Mohammad, S.U.: A conceptual framework for software engineering education: project based learning approach integrated with industrial collaboration. Int. J. Educ. Manage. Eng. (IJEME) 5, 46–53 (2021) 2. Ozgen, K.: SASMEDU: security assessment method of software in engineering education. Int. J. Educ. Manage. Eng. (IJEME) 7, 1–12 (2018) 3. Gu, Y., Chen, J., Liu, Z.: Course resources and teaching methods of logistics management major under emerging engineering education. Lecture Notes on Data Eng. Commun. Technol. 83, 434–443 (2021) 4. Shen, J., Li, T., Wu, M.: The new engineering education in china. Procedia Comput. Sci. 172, 886–895 (2020) 5. Song, J., Shi, H., Liu, Y.: Exploration of core competence and teaching mode of new engineering. J. High. Educ. 8(23), 66–69 (2022). (in Chinese) 6. Liu, J., Feng, L., Wang, Z.: Assessment and implementation of college instruction based on OBE. J. North China Univ. Sci. Technol. 22(6), 79–84 (2022). (in Chinese) 7. William, G.S.: Outcome-Based Education Critical Issues and Answers. American Association of School Administrators, New York (1994) 8. Ines, G., Daniel, D., Belen, B.: Learning outcomes based assessment in distance higher education. A case study. Open Learn. J. Open Distance and e-Learning 37(2), 193–208 (2022) 9. Liu, L., Feng, W., Li, J., et al.: Application of BOPPPS model in the teaching design of the principle and method of micro-joining course. Adv. Comput. Sci. Eng. Educ. 134, 509–517 (2022) 10. Van, M.E., Hall, A.K., Schumacher, D.J., et al.: Capturing outcomes of competency-based medical education: the call and the challenge. Med. Teach. 43(7), 794–800 (2021)

Research and Practice of Outcome-Oriented Innovative Talent

983

11. Sasipraba, T., Kaja, R., Nandhitha, N.M., et al.: Assessment tools and rubrics for evaluating the capstone projects in outcome based education. Procedia Comput. Sci. 172, 296–301 (2020) 12. Shreeranga, B., Sathyendra, B., Ragesh, R., et al.: Collaborative learning for outcome based engineering education: a lean thinking approach. Procedia Comput. Sci. 172, 927–936 (2020) 13. Tian, J., Li, L., Zhou, C., et al.: Research on the construction of talent training Mechanism of materials subject in colleges and universities. Educ. Modernization 9(19), 60–63 (2022). (in Chinese) 14. Yuan, Y., Liang, S.: Discussion on the connection of undergraduate, postgraduate and doctoral courses in materials discipline. Education Modernization 6(51), 187–188,197 (2019). (in Chinese) 15. Qian, J., Nam, P.N., Yutaka, O., et al.: Introducing self-organized maps (SOM) as a visualization tool for materials research and education. Results in Materials 4, 100020 (2019) 16. Elistratkin, M.Y., Lesovik, V.S., Zagorodnjuk, L.H., et al.: New point of view on materials development. IOP Conf. Ser. Mater. Sci. Eng. 327(3), 032020 (2018) 17. Wibawa, B., Syakdiyah, H., Kardipah, S.: Technology development in thermochemistry materials. J. Phys: Conf. Ser. 1869(1), 012023 (2021) 18. Fan, Y., Dong, X., Guo, C., et al.: Teaching mode reform based on cultivation of innovative talents in materials science. China Mod. Educ. Equipment 1, 121–123 (2022). (in Chinese) 19. Wang, Z.: Computer Programming reform and practice for top talents in the material category oriented to engineering certification. The Science Education Article Collects 13, 85–87 (2022). (in Chinese) 20. Zhao, J., Xu, Y., Pan, H., et al.: Effectiveness of clinical classroom teaching evaluation system. Basic and Clinical Medicine 42(3), 529–532 (2022). (in Chinese) 21. Wang, S.: Evidence-based evaluation for instruction quality: taking the instruction quality assessment of mathematics in the USA as an example. Journal of China Examinations 2, 73–80 (2022). (in Chinese) 22. Soumi, M., Soumalya, C., Sayan, C.: Interactive web-interface for competency-based classroom assessment. Int. J. Educ. Manage. Eng. (IJEME) 1, 18–28 (2023) 23. Tang, H., Qian, Y., Wang, S.: Designing MOOCs with little. Cogent Education 9(1), 2064411 (2022) 24. Lin, Y., Feng, S., Lin, F., et al.: Adaptive course recommendation in MOOCs. Knowl.-Based Syst. 224, 107085 (2021) 25. Sun, J., Sun, C., Zhang, Y., et al.: Virtual simulation experiment of impulse voltage generator. Lab. Sci. 20(2), 64–67 (2022). (in Chinese)

Development of the Online Library of the University Department to Support Educational Activities: A Conceptual Model Nataliia Vovk(B) and Daryna Prychun Lviv Polytechnic National University, Stepana Bandery Street 12, Lviv 79013, Ukraine {nataliia.s.vovk,daryna.prychun.mdkib.2021}@lpnu.ua

Abstract. The research carried out an analytical review of the sources, which showed that today the issue of information support for the activities of university libraries has been studied at a sufficient scientific and educational-methodological level. Researchers consider this process both from the point of view of creating electronic libraries (electronic resources) in the Internet environment, and from the point of view of using social networks to establish communication and spread information about the activities of book collections. The basis of the research was carried out at the Department of Social Communications and Information Activities of the Lviv Polytechnic National University. Although library service to users at the university level is organized at a high level, library service at the department level is not sufficiently regulated. The proposed solution will allow not only to automate library service at the department level, but also to establish communications with readers - students of the department. The developed website of the Department of Social Communications and Information Activities library allows users to familiarize themselves with the document fund of the department, and registered users to access full-text documents Such e-library allows not only systematizing the works of university teachers (teaching manuals, textbooks, scientific articles, theses, etc.), but also to provide students with more opportunities for online learning, including the following: constant full-text access to such materials; independent preparation for classes; abstract analysis of such documents, etc.. In addition, the study presents general proposals for the development of information activities of the department’s library, the problems of implementing the online library and the possibility of avoiding and getting out of conflict situations. Keywords: university department · library · students · information support · website

1 Introduction The creation and provision of new library services is the result of transformations caused by technological progress, new document formats, innovative ways of circulating information, and user needs, as well as such social phenomena as the fragmentation of society, © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 984–999, 2023. https://doi.org/10.1007/978-3-031-36118-0_84

Development of the Online Library of the University Department

985

the aging of the population, and a very low level of reading information. The impetus in the process of providing library services, among which the book publisher is still the main one, was the wide use of information technologies. The majority of public, scientific and university libraries today already widely use all available methods and means of online reader service: electronic catalogs, access to Internet resources, online queue, etc. However, in the structure of library service for users in institutions of higher education, emphasis is placed only on the activities of the main university book collections. Thus, leaving out of consideration small educational department’s libraries, the fund of which, as a rule, includes the work of department employees: teaching aids, textbooks, lecture notes, other methodical publications, monographs, scientific articles, etc. Therefore, the relevance of the chosen topic is determined by the lack of work on the use of information technologies by the educational department’s libraries of higher education institutions. The purpose of the study is to develop a conceptual model of the online library of the university department. It is envisaged that such a library will allow not only systematizing the works of university teachers (teaching manuals, textbooks, scientific articles, theses, etc.), but also to provide students with more opportunities for online learning, including the following: constant full-text access to such materials; independent preparation for classes; abstract analysis of such documents, etc. According to the goal, there are the following research tasks: analysis of existing works on the research topic; development of a conceptual model of the electronic library of the university department; analysis of available software and its selection; development of the alpha version of the site, its analysis and further improvement. Today, the issue of the development of modern libraries has been sufficiently studied at the theoretical level, clear ways of their development have been determined and methods of promotion of modern libraries have been chosen. At the stage of implementation of these ideas, there is a kind of imbalance: on the one hand, quite high successes have been achieved (transformation of libraries into information centers and media libraries), the use of information technologies in the process of library service, on the other hand, insufficient attention has been paid to certain types of libraries, in particular to the educational department’s libraries of higher education institutions education.

2 Related Work The issue of information provision of library support for studies at universities is the subject of analysis of works by many researchers. This process is considered from different perspectives: both from the point of view of developing information services and systems, remote databases for library service to readers, and from the point of view of popularizing the activities of university libraries. Samah N. and others state that currently university students and teachers generally obtain reference materials from online databases and electronic resources accessed by university libraries. In this study, the authors identify the number and names of online databases most frequently searched, viewed, and downloaded by university students and staff for educational purposes. The data were obtained from the annual reports published by university libraries during 2014–2017. The most popular such databases, according

986

N. Vovk and D. Prychun

to research, are the following: IEEE Xplore Digital Library, ScienceDirect, EdITLib, Datastream and Scopus [16]. The most popular system for e-learning, in which it is possible to mark up electronic manuals and other electronic publications, is the system on the Moodle platfor. Bhupesh Rawat and Sanjay K. Dwivedi point out that Moodle (modular object oriented developmental learning environment) is one of the most widely used open source learning management systems. They can record any student activities including reading, writing, taking tests, performing various tasks, and even communicating with peers through collaborative forums such as chat, forum, glossary, wiki and workshop [3]. Ani O. and Ahiauzu B. investigated the levels of electronic information resources development in Nigerian university libraries. During the national seminar on electronic information for the library network (eIFL.net) in 2007, a survey was conducted to obtain information about libraries’ access to electronic information resources. The results of the study are presented in Fig. 1.

Fig. 1. Sources of electronic information resource development

The study findings show that the Internet is the main source of electronic information resources development in Nigerian university libraries. According to the survey results, there is a high level of electronic information resources development in Nigerian university libraries through direct subscription to electronic information (online databases, CD-ROM, etc.) than converting information into electronic form in the library through computerization and digitization [2]. Other studies show that the use of scientific databases is quite low. Such low indicators are caused by insufficient information about the existence of these library resources [6]. This shows that the participants of the educational process prefer those educational materials that are freely available on the Internet. Gulara A. Mammadova and other researchers define the concept “personalized learning”. Personalized learning is the adaptation of the pedagogy, curriculum and learning environments to meet the needs and styles of the individual students learning. The key elements that are customized in personalized e-learning are: the pace of learning,

Development of the Online Library of the University Department

987

the learning approach, learning activities, educational content [7]. Such content can be e-journals, textbooks, e-magazines, etc. Borrego A. and other researchers present the results of a survey on the use of ejournals by university teachers belonging to the Consortium of Academic Libraries of Catalonia. The results show that a significant part of teachers and researchers prefer electronic journals in contrast to printed formats [4]. McMartin F. in the study reports the results of a national survey of 4,678 respondents representing 119 higher education institutions in the United States regarding their use of digital resources for academic purposes. In particular, the following survey results are presented: demographic indicators commonly used in higher education to classify population groups cannot reliably predict the use of online digital resources; the assessment of online digital resources corresponds only to higher levels of use of certain types of digital resources; lack of time was a significant barrier to using the materials [11]. Librarians report an increase in journal reading among scholars and students due to the increase in the number of e-journals available and the improvement of tools for finding and accessing this information, especially access outside the university [14]. When using information technologies in library services for readers of university libraries, the competences of librarians and users of library funds play an important role. The question of the practice of managing digital resources in university libraries has also become one of the key areas of modern research [1, 10, 13, 15, 18]. Ocran T. provides clear recommendations for the successful implementation of library services in the mobile phone: “students should be educated to know the benefits that come with the use of mobile device to access library services while library personnel should be adequately trained for such services” [12]. Peculiarities of the library sites use in the information provision direction of library services [9] and other online resources are today the subject of study from the point of view of their use in branch university libraries [17]. In Fig. 2 presents the number of electronic documents in the electronic catalogs of the libraries of higher education institutions as of 2020 (according to the research of N. Levchenko). As of 2020, most libraries were at the same level according to the number of electronic documents in the library’s electronic catalogs. The scientific library of the Zaporizhia National University differs sharply from most. The current state of information technology development contributes to the rapid development of various online libraries. It is worth noting that most full-text databases (in the form of an online library) offer fiction. While electronic libraries of higher education institutions are mostly designed as electronic catalogs, where at best, the user can place an online order from a book publisher. To ensure the successful transition of Ukrainian libraries to the use of new national standards of Ukraine, harmonized with international normative documents, the National Library of Ukraine named after V. Vernadskyi initiated the creation of an interdepartmental working group, which included experts from leading libraries of various systems and departments. From September 1, 2016, new national standards of Ukraine entered into force, in particular ISO 16439:2016 (ISO 16439:2014) «Information and documentation – Methods and procedures for assessing the impact of libraries». This international

988

N. Vovk and D. Prychun

Fig. 2. Number of electronic documents in electronic catalogs of Ukrainian university libraries

standard was developed for the purpose of strategic planning and quality management of libraries; to facilitate comparison of library impact over time between libraries of a similar type and purpose; to increase the value of libraries for learning and scientific research, education and culture, social and economic life; to support policy decisions regarding the strategic goals of libraries.

3 Methodology It is advisable to use online libraries to develop a ADDIE instructional design model proposed by Peterson [5]. In order to create such an e-library, all design models needs the following stages: analysis, design, development, implementation, and evaluation; these stages are abbreviated as ADDIE. The efficacy of ADDIE is found in its next-phaseinforming nature, making it logical and viable for the development any kind of mobile education tool. Acclaimed for its resounding popularity in the development of educational instruction, it also presents an easy process, which even the amateur analyst, can follow. It also has the advantage of rapid prototyping and possesses the capacity for continual and formative feedback during instructional materials creation. Most importantly, the ADDIE model conserves the precious time of the developer due to the early fixes it allows at the initial phases. The description of the stages of the original model is described with the aid of Fig. 3. – the diagram depicting the ADDIE framework. In addition, the modeling method was used for the work, which made it possible to depict informational and functional models.

Development of the Online Library of the University Department

989

Fig. 3. The ADDIE Framework [5].

4 Development of an Online Library Model for the Department of Social Communications and Information Activities (Lviv Polytechnic National University) The scientific and technical library of Lviv Polytechnic National University is one of the largest book collections of higher educational institutions of Ukraine. The library is a structural unit of the university. It performs the functions of a library-bibliographic, scientific-informational, educational-supportive, cultural-educational institution, provides the educational process, scientific-pedagogical activity, scientific-research, educational, and cultural-educational work, with books and other information carriers that make up the fund libraries. The book and magazine fund of the library includes about 1 million 800 thousand copies, more than half of which is the fund of scientific literature. In addition to the library itself, individual copies of literature are located in the departments of the university. Such a department is the graduate department of social communications and information activities of the Institute of Humanities and Social Sciences. The Department of Social Communications and Information Activities (SCIA) was founded on August 23, 2011 with the aim of training bachelors and masters in the specialty “Information, library and archival affairs” in the field of knowledge “Culture”. In 2016–2022, the teachers of the SCIA department issued 35 study guides and 11 lecture notes (Fig. 4). In addition to the publication of educational and methodological literature, scientific and pedagogical workers are actively engaged in publishing activities in the direction of scientific publications: articles, monographs, abstracts of conferences, etc. So, for 2016–2022, 31 monographs and sections of monographs were published (Fig. 5). All scientific research is carried out by the teaching staff of the department on the registered topic “Management of social communication processes in the global information space”. As for the organization of the work of the library of the SCIA department, it is divided into two directions: • the placement of educational and methodological materials in electronic educational and methodological complexes in the virtual educational environment of the Lviv Polytechnic;

990

N. Vovk and D. Prychun

Fig. 4. Publication of educational literature by teachers of the SCIA department (2016–2022)

Fig. 5. Publication of monographs and sections of monographs by teachers of the SCIA department (2016–2022)

• the book exhibition of teachers’ manuals at the department. However, taking into account the fact that during the last three years the education has been conducted in a distance format, students’ access to such an exhibition, at least for familiarization is practically non-existent. In addition, users can familiarize themselves with the manuals of the department’s teachers, published by the Publishing House of the Lviv Polytechnic, and on the website of the publishing house itself. The manuals of the teachers of the SCIA department are posted on the website of the Publishing House in the section “IT, computers”. But there are two drawbacks: only those editions published by the Publishing House of the Lviv Polytechnic are posted on the website; on the site, users can only purchase a copy, but not view it online. In order to improve the work of the library of the department of social communications and information activities, it was decided to create a website, which will host an electronic catalog of available manuals, textbooks and other materials, links to Internet

Development of the Online Library of the University Department

991

resources where such works are already posted, and digitized educational and methodological materials scientifically - teaching staff of the department. For a better understanding of the site creation process, it is necessary to depict the information model of this task (Fig. 6).

Fig. 6. Information model of the problem

To reflect the implementation of the main process, it is advisable to develop a goal tree that allows you to understand the systematic solutions for solving the main goal (Fig. 7).

Fig. 7. Goal tree

The main goal “Development of the online library of the department of the higher education institution” is divided into 2 sub-goals: “Formation of the library fund” and “Development of the website”. Regarding the first sub-goal, part of the library fund of the SKID department is already available in electronic format (available from teachers or in the form of Internet resources), the other part needs to be digitized. The second sub-goal is divided into three: • “Development of the first and final versions of the site” - all versions developed using CMS WordPress must be approved by the head of the department; • “Navigation development” - it is planned to create a site menu and an electronic catalog on the site;

992

N. Vovk and D. Prychun

• “Feedback” - receiving feedback from users by the electronic library. Using the DFD-representation, in Fig. 8 presents the process of organizing activities for the development of the online library of the Department of Higher Education.

Fig. 8. DFD information support

The main process is the development of the online library of the department of higher education, there are 4 external entities: the head of the department, students, teachers of the department and the administrator of this library. Data flows include the following flows: • input (feedback, materials, online library project) • output (information about availability of materials, request, library content, reporting). The key purpose of creating such an online library is to place on it educational aids and textbooks issued by the teachers of the department. Then most students will be able to use them. For this, it is necessary, first, to receive such materials from the authors, to systematize them, if necessary, to digitize them and then place them on the website (library content). The head of the department must approve all versions of the site. In the case of receiving information about the absence of a certain book on the site, the administrator must, at the time determined by the head of the department, perform actions to download the manual / textbook to the department’s online library. The process of developing the online library of the Department of Higher Education can be divided into seven sub-processes (Fig. 9): 1. 2. 3. 4. 5. 6. 7.

Materials structuring Application processing Content systematization Information Systematization Request formation Website development Reporting Formation

Development of the Online Library of the University Department

993

Fig. 9. Decomposition of the 1st level

To implement the key task, it was decided to choose a Content Management System (CMS), which today significantly facilitates the work of web developers. Today, the most popular freely distributed CMS are the following: WordPress, Joomla, Drupal, MODX. Each of them has its own characteristics, advantages and disadvantages (Table 1). Table 1. Analysis of CMS Measurement time difference/s

WordPress

Joomla

Drupal

MODX

Keeping a record of user actions

−

−

+

+

Protection against automatic filling of forms

?

+

+

+

Caching pages

+

+

+

+

Distribution of user rights

-

+

+

+

SSL protocol support

+

+

+

+

Several interface languages

+

+

+

+

Support for multilingual sites

+

+

+

+

Web statistics

+

+

+

+

Chat

−

+

+

?

Forum

−

+

+

−

Photo gallery

+

+

+

+

Catalogue

−

+

+

+

Payment systems

+

+

+

+

To create an online library of the Department of Social Communications and Information Activities of the Lviv Polytechnic, choose the WordPress content management system.

994

N. Vovk and D. Prychun

One of the important issues in the development of the e-library of the university library is the determination of the reliability of the selected software and the subsequent protection of the created information resource. To ensure the reliability of the e-library, both at the first stage of its development and in its subsequent functioning, it is advisable to use various protection methods, in particular the following: an SSL certificate, which ensures a secure and confidential connection via the HTTPS protocol; ISO 27001 and 27018 certificates; additional layers of site protection, such as TLS 1.2; protection against DDoS attacks; monitoring the site for any suspicious activity and constantly checking the security of the software; application of two-stage verification of e-library visitors; application of GDPR and CCPA standards, etc. The developed website of the SCIA department library contains information about the library fund of the department. The main page contains information about the department’s new arrivals, which have been digitized and are available online (Fig. 10). The website menu contains quick links to the relevant sections of the library fund. If the appropriate item is selected, the user will go to the page with materials belonging to this category. The website menu contains quick links to the relevant sections of the library fund. If the appropriate item is selected, the user will go to the page with materials belonging to this category (Fig. 11).

Fig. 10. Main page of the site

Viewing library materials has differences for registered users. Anyone has access to the site and can be acquainted with the library fund of the department, but only reference and descriptive information is available for viewing. Registered users also have access to full-text documents (Fig. 12).

Development of the Online Library of the University Department

995

Fig. 11. Section “Education”

Fig. 12. Access to full-text documents

5 Proposals Regarding the Improvement of Document and Information Activities of the Institution New forms of organizing the functioning of universities stimulate the need to review the structure of libraries, their functional interaction with other units of the university. The strengthening of market trends in the field of education, the need to train competitive specialists make it necessary to provide a high level of information support for the educational process. With a noticeable increase in the total volume of the document and information fund of the libraries of higher education institutions, the index of renewal of library funds has a tendency to decrease. In the face of financial constraints, libraries need to

996

N. Vovk and D. Prychun

look for non-traditional methods of stocking them: book exchange, sponsorship, holding various promotions, etc. Considering the problem of insufficient amount of literature, especially educational, libraries should form full-text collections of educational and auxiliary materials. The use of modern information technologies has changed the standards of library work in general, reference, and bibliographic service to users in particular. The strategy of searching for information has changed, and access to information sources has expanded significantly. Higher education libraries today must develop - the use of digital technologies leads to new forms of working with users, especially in terms of the accumulation, formation and use of information resources. Libraries of higher education institutions should intensify their activities in the creation of electronic catalogs and thematic databases; provide users with access to numerous global information arrays via the Internet. Today, library users can search for information using both traditional reference and bibliographic equipment, as well as electronic catalogs, databases, etc. Libraries of higher education institutions implement a comprehensive approach to the formation of information culture of future specialists, using means and methods of correction and management of this process; provide an opportunity to combine cognitive and practical information activities of students. The management of the leading libraries of higher education institutions develops and implements the course “Fundamentals of information culture”, and uses other means of individual and group assistance to users. Improving the libraries activities in the formation of users the information culture requires, first of all, a state approach: a transition from episodic, non-systematic work to activities within the framework of a special state program, which involves training personnel, solving problems of coordinating the work of libraries and educational institutions, developing methodological support, creating necessary material base. The modern university library is not only an informational and educational center, but also a cultural and educational center. The libraries priority activity today is the expansion of cultural, recreational, and leisure-educational activities. One of the promising areas of activity of pedagogical and engineering-pedagogical university libraries is research and scientific-methodical work. Libraries study the information portfolios of users, conduct scientific work with old prints and rare editions, collections of domestic scientists-pedagogues, create a full-fledged system of secondary information documents, develop instructional and methodological materials, participate in various scientific projects, conduct scientific and practical conferences, etc. An integral component of modern library service in the libraries of higher education institutions is the expansion of the computer park, the use of multiple technologies, the presence of a local network and an automated library information system. This will allow automating all the main library processes and provides an opportunity for communicative information exchange. Today, university libraries have gained positive experience in working with library staff. However, the current system of professional development in the conditions of rapid technological changes in libraries cannot fully ensure the updating of theoretical and practical knowledge, mastery of modern information and library technologies, organizational and legal aspects of library management, etc. The rather high cost declared by

Development of the Online Library of the University Department

997

the organizers for the course participants does not allow this form of advanced training to be practiced widely. Libraries, especially their managers and administrative staff, need advanced training courses under special programs taking into account the specifics of their work.

6 Summary and Conclusion An analytical review of the sources showed that today the issue of information support for the activities of university libraries has been studied at a sufficient scientific and educational-methodical level. Researchers consider this process both from the point of view of creating electronic libraries (electronic resources) in the Internet environment, and from the point of view of using social networks to establish communication and spread information about the activities of book collections. Analysis of available Internet resources allows us to conclude that today there is a significant number of universally accessible online libraries. However, most of them offer users only fiction books. At that time, the university libraries of Ukraine mostly placed only electronic catalogs on their websites. In addition, it is worth noting that domestic libraries today are guided by various state standards, which are issued based on international ISO standards, which regulate processes of quality compliance in library services. The analysis of the websites of the leading universities of Ukraine showed that on all such information resources there is only a page (sometimes a separate site) of the library, which contains the following services: electronic catalog, resources for compiling indexes of the unified decimal classification, thematic selection of literature (in the form of a list of literature), access to Scopus and Web of Science databases. Instead, such an analysis showed the lack of full-text access to specialized literature for both students and other visitors to electronic libraries. It is advisable to start the development of full-text online library resources from the e-library of individual departments. They will become the alpha version of the e-library of the entire university. Library service at the Lviv Polytechnic National University is organized at a high level. The information provision of this activity is regulated by the relevant regulatory internal documents of the University. However, it is worth noting that library service at the level of departments is not sufficiently regulated. This will allow not only to automate library service at the level of the department, but also to establish communications with readers - students of the department. With the help of the development of an informational and functional model, the general concepts of the department’s library development were formed. Implementation of the online library requires the use of additional resources. One of the main ones is the availability of a domain and hosting. Lviv Polytechnic National University has its own servers for hosting the library website and its own domain. Such problems may arise in smaller universities or those with limited resources (personal servers, wide Internet channel). As of January 2023, students of the Department of Social Communications and Information Activities are testing the developed electronic library. Part of the students (registered users) use full-text access to the electronic library fund for preparing essays, term papers, graduation papers. The testing period will last until June 2023. Based on

998

N. Vovk and D. Prychun

user recommendations and errors detected during the website’s operation, the e-library will be modified and made available to the public. Results will be presented in further research.

References 1. Alphonce, S., Mwantimwa, K.: Students’ use of digital learning resources: diversity, motivations and challenges. Inf. Learn. Sci. 120(11–12), 758–772 (2019). https://doi.org/10.1108/ ILS-06-2019-0048 2. Ani, O.E., Ahiauzu, B.: Towards effective development of electronic information resources in Nigerian university libraries. Libr. Manag. 29(6–7), 504–514 (2008). https://doi.org/10. 1108/01435120810894527 3. Bhupesh, R., Sanjay, K.: An architecture for recommendation of courses in e-learning system. Int. J. Inf. Technol. Comput. Sci. 17, 39–47 (2017). https://doi.org/10.5815/ijitcs.2017.04.06 4. Borrego, À., Anglada, L., Barrios, M., Comellas, N.: Use and users of electronic journals at Catalan universities: the results of a survey. J. Acad. Librarianship 33(1), 67–75 (2007). https://doi.org/10.1016/j.acalib.2006.08.012 5. ChukwuNonso, N., Ikechukwu, U., Njideka, M.: GeoNaija: enhancing the teaching and learning of geography through mobile applications. Int. J. Educ. Manage. Eng. 6, 11–24 (2019). https://doi.org/10.5815/ijeme.2019.06.02 6. Dadzie, P.S.: Electronic resources: access and usage at Ashesi university college. CampusWide Inf. Syst. 22(5), 290–297 (2005). https://doi.org/10.1108/10650740510632208 7. Gulara, A., Firudin, T., Lala, A.: Use of social networks for personalization of electronic education. Int. J. Educ. Manage. Eng. 5, 25–33 (2019). https://doi.org/10.5815/ijeme.2019. 02.03 8. Hale, M.: Automated library acquisitions and the internet: a new model for business. New automation technology for acquisitions and collection development, pp. 65–82 (2019). https:// doi.org/10.4324/9780429329364-5. www.scopus.com 9. Martin, C.R., Rodriguez-Parada, C., Pacios, A.R., Osti, M.V., Bravo, B.R.: Transparency in the management of Spanish public university libraries through the internet. Libri 70(3), 239–250 (2020). https://doi.org/10.1515/libri-2019-0051 10. McCoy, M.: Why didn’t they teach us that? The credibility gap in library education. Conflicts in reference services, pp. 171–178 (2019). https://doi.org/10.4324/9780429354373-18. www. scopus.com 11. McMartin, F., Iverson, E., Wolf, A., Morrill, J., Morgan, G., Manduca, C.: The use of online digital resources and educational digital libraries in higher education. Int. J. Digit. Libr. 9(1), 65–79 (2008). https://doi.org/10.1007/s00799-008-0036-y 12. Ocran, T.K., Underwood, E.P.G., Arthur, P.A.: Strategies for successful implementation of mobile phone library services. J. Acad. Librarianship 46(5), 102174 (2020). https://doi.org/ 10.1016/j.acalib.2020.102174 13. Okeji, C.C., Mayowa-Adebara, O.: An evaluation of digital library education in library and information science curriculum in Nigerian universities. Digit. Libr. Perspect. 37(1), 91–107 (2020). https://doi.org/10.1108/DLP-04-2020-0017 14. Ollé, C., Borrego, A.: Librarians’ perceptions on the use of electronic resources at Catalan academic libraries: results of a focus group. New Libr. World 111(1–2), 46–54 (2010). https:// doi.org/10.1108/03074801011015685 15. Rahman, M.H.: Review of digital record management needs for academic libraries. Library Hi Tech News 37(3), 21–22 (2020). https://doi.org/10.1108/LHTN-11-2019-0083

Development of the Online Library of the University Department

999

16. Samah, N.A., Karno, M.R., Sharuddin, N., Tahir, L.M., Samah, N.A.: Examining the usage of and access to online databases for academic purposes: a study at an engineering- and technology-based university in Malaysia. Paper presented at the TALE 2019 - 2019 IEEE International Conference on Engineering, Technology and Education (2019). https://doi.org/ 10.1109/TALE48000.2019.9225991. www.scopus.com 17. Smith, D.: Re-visioning library support for undergraduate educational programmes in an academic health sciences library: a scoping review. J. Inf. Literacy 13(2), 136–162 (2019). https://doi.org/10.11645/13.2.2520 18. Zhao, X., Li, S., Xiao, Y., Huang, H.: Digital humanities scholarly commons at Beijing normal university library. Libr. Trends 69(1), 250–268 (2020). https://doi.org/10.1353/lib.2020.0031

Research on the Influencing Factors of Students’ General Competency Improvement in Application-Oriented Universities Jing Zuo1 and Tongyao Huang2(B) 1 Nanning University, Nanning 530022, Guangxi, China 2 Guangxi University of Foreign Languages, Nanning 530000, Guangxi, China

[email protected]

Abstract. Students’ competency cultivation is a very important measure of the effectiveness of talent development. In order to explore the influencing factor of student’s general competency in application-oriented Universities, this study take 9652 students from different majors in one application-oriented University as sample, regression analysis has been applied with a dependent variable (general competency improvement) and six independent variables (student’s learning behavior: in-class, after-class, interact with classmates; teacher’s teaching behavior; evaluation of teaching content; professional identity) to sort out the influencing factors. The results show that teacher’s teaching behavior, evaluation of teaching content, student’s learning behavior (in-class, after-class, and interact with student), and professional identity is significantly positively related to the improvement of students’ general competency; especially, the teacher’s teaching behavior, and evaluation of teaching content, these are the key factors of student’s general competency improvement. Keywords: General competency · The influencing factors · Regression analysis

1 Introduction Application-oriented university, is an important part in the development of higher education, which has the mission of cultivating high-quality and highly skilled talents in various fields, such as production, construction, service, and management. In the process of application-oriented talents development, students’ competency cultivation is a very important measure of the effectiveness of talent development, and also the main task of talents development in application-oriented universities [1]. In general, student competency includes general competency and professional competency, and cultivating students’ general competency can help students adapt to the changing social requirements and industry needs [2]. In order to help application-oriented universities improve the quality of talent development, this study applies the quantitative research methods to explore the factors that affect the cultivation of students’ general competency in application-oriented college. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 1000–1010, 2023. https://doi.org/10.1007/978-3-031-36118-0_85

Research on the Influencing Factors of Students’ General Competency

1001

2 Literature Review 2.1 Competency and General Competency In 1973, professor McClelland proposed the concept of competency in Testing for Competency rather than intelligence. Nadler believes that competency refers to a resource that can help improve competitive advantage [3]. Based on Françoise and Jonathan’ s definition using multidimensional framework, competency includes meta-competence, functional competence, social competence, and cognitive competence [4]; while some researchers propose that competence can be divided into Core Competency (key competences) and Specific Competency [5]. Most studies in the field of competency use general competences as synonyms of Core Competency and key competences, which refer to the competence applying to different work and life [6] without limiting specific majors or fields [7]. Many researches, at home and abroad, clarifies the definition of general competency (core competency). In Core competence of Chinese student, Core Competency contains six parts: cultural heritage, scientific spirit, learning to learn, healthy life, responsibility, practice and innovation [8]. A proposal on core competency for lifelong learning was adopted by the Council of the European Union in 2018, meanwhile core competency was been divided into eight parts: literacy competence, multilingual competence, cultural awareness, expression competence, and so on [9]. Secret’s Commission on Achieving Necessary Skills, established in the United States in 1990, they had determined the basic competency required to cope with the work through a lot of research [10], such as resources, interpersonal communication, information, system, technology [11]. Therefore, the definition of core competency in different countries has its own characteristics, which provides a useful reference for the development of the scale in this study. 2.2 The Influencing Factors of General Competency A large body of literature explores what factors could amplify or mitigate general competency, however, they mainly discuss the influencing factors of a specific part of competency, like cross-culture competency [12], scientific and technological innovation competency [13], critical thinking competency [14], autonomous learning competency [15]. In addition, some researchers focus on the factors that influence the mismatch of engineering students’ general competency [16], and the influence of institution and education environment on the formation of general competency [17]. The influencing factors of general competency are different as most of the previous researches were mainly about a specific part of competency in different scenarios, but it can mainly be divided into these three types: learner, influencer (teacher, tutor, classmate), and environment (school, major). Based on the previous researches, this study will also be conducted including these three types of influencing factors (learner’ learning behavior, teacher’s teaching behavior, teaching content and professional recognition).

1002

J. Zuo and T. Huang

3 Methodology 3.1 Questionnaire Design The questionnaire design of this study includes two parts: personal information (grade, gender, college, major) and five scales with fifty-three questions (general competency, student’s learning behavior, teacher’s teaching behavior, evaluation of teaching content, professional identity). The scale of general competency based on the SCANS standard of the United States [10], includes teamwork, information searching and processing, self-awareness, environmental adaptation, communication, design thinking, reading ability, critical thinking, organizational leadership, problem solving, autonomous learning, planning ability. The scales of student’s learning behavior, teacher’s teaching behavior, evaluation of teaching content, and professional identity are designed according to relevant literature. The scale (Table 1) has been modified according to the pilot test and the results of analysis on Reliability and Validity. 3.2 Data Collection and Sampling The researchers distributed 9652 questionnaires to freshmen, sophomore and junior in an application-oriented university, 8931 valid questionnaires (recovery rate is 92.5%) were collected (Table 2). Table 1. Scales and Samples Name of scale

Questions

Sample

Remark

General competency

12

Improvement of “teamwork” ability in this year

Likert scale (1–5: not at all improved, neutral, somewhat improved, very improved)

Teacher’s teaching behavior

11

Teachers add interaction in the classroom and pay attention to students’ participation

Likert scale (1–5: never, rarely, sometimes, often, always)

2

Finish reading and homework before class in this year

Student’s learning behavior

Before-class

(continued)

Research on the Influencing Factors of Students’ General Competency

1003

Table 1. (continued) Name of scale

Questions

Sample

Remark

In-class

3

Concentrate in class in this year

After-class

6

Review notes after class in this year

Interact with classmates

5

Work with classmate for course assignment or homework in this year

Evaluation of teaching content

7

Teaching content give emphasis to the combination of practice and theory

Professional identity

7

Pay attention to the hotspots or cutting-edge information of discipline development

Likert scale (1–5: strongly disagree, disagree, neutral, agree, strongly agree)

Table 2. Sample Descriptive Statistics Results Grade

Male

Female

Total

Students

Percent

Students

Percent

Students

Freshman

2346

26.3%

2378

26.6%

4724

52.9%

Sophomore

1061

11.9%

1006

11.3%

2067

23.1%

Junior

1004

11.2%

1136

12.7%

2140

24.0%

Total

4411

49.4%

4520

50.6%

8931

100.0%

Percent

4 Results 4.1 Analysis on Reliability Cronbach’s alpha coefficients of the scales can be seen in Table 3. The Cronbach’s alpha coefficients of general competency scale, teacher’s teaching behavior scale, student’s learning behavior scale, and teaching content scale are similar (α > 0.9, excellent reliability), while the one in professional identity scale is slightly lower than 0.9 (good reliability).

1004

J. Zuo and T. Huang Table 3. Reliability, Average and Standard Deviation

Scale

Number of students

Mean

Std. Deviation

Cronbach’s alpha

General competency

8931

3.89

0.73

0.965

Teacher’s teaching behavior

8931

4.01

0.71

0.966

Student’s learning behavior

8931

3.40

0.84

0.941

Evaluation of teaching content

8931

3.66

0.77

0.942

Professional identity

8931

3.58

0.75

0.886

4.2 One-Way Analysis of Variance The researcher conducted a one-way analysis of variance on the general competency of three groups of students from different grades (P < 0.05), differences between groups can be found. The f test statistic cannot check whether the variances of the samples are equal to the same value, so Tamhane has been utilized to check. The results show that there is no significant difference between sophomores and juniors, while a significant difference can be witnessed between freshmen and sophomores, juniors. 4.3 Regression Analysis In order to find out the factors that influence university students’ general competency and whether the factors vary by different grades, first, regression analysis of three different grades have been done, with another regression for all the samples. 4.3.1 Regression Analysis for Three Grades Linear regression analysis conducted between different grades with dependent variable (general competency improvement) and independent variables (student’s learning behavior (in-class) (abbr. in-class), student’s learning behavior (after-class) (abbr. After-class), student’s learning behavior (interact with classmates) (abbr. Interact); teacher’s teaching behavior (abbr. T’s behavior); evaluation of teaching content (abbr. Evaluation); professional identity (abbr. Identity)). According to the number of independent variables (Table 4), each regression analysis has seven models. Model 7 with a greater proportion of dependent variable changes (R2 = 0.46, 0.47,0.43) show that this is true for 40% of student’s general competency improvement. As shown in Table 5, for freshmen, student’s learning behavior (before-class) has no significant impact on the general competency improvement (P = 0.08 > 0.05); the coefficient of evaluation of teaching content and teachers’ teaching behavior is relatively the largest, with the largest impact on improving student’s general competency. For sophomore, student’s learning behavior (before-class) and student’s learning behavior (in-class) (P > 0.05) indicate no significant impact in general competency improvement,

Research on the Influencing Factors of Students’ General Competency

1005

Table 4. Three Regression Analysis Model Type

R

R2

Change R2

Change Statistics R2 Change F Change df1 df2

Freshman

Sig. F Change

0.68 0.46 0.46

0.00

42.652

1

4716 0.000

Sophomore 0.69 0.48 0.47

0.00

15.903

1

2059 0.000

Junior

0.00

10.844

1

2132 0.001

0.66 0.44 0.43

while the greatest impact is still the evaluation of teaching content and teachers’ teaching behavior. For junior, no significant impact can be seen in general competency improvement for the factors of student’s learning behavior (before-class), student’s learning behavior (in-class), and student’s learning behavior (interact with classmate) (P > 0.05), with the largest impact in evaluation of teaching content and teachers’ teaching behavior. By contrast, the influence of students’ learning behavior (in class), student’s learning behavior (interact with classmates), and professional identity on the improvement of students’ general competency in different grades gradually decreases, while the influence of students’ learning behavior (after-class) and teachers’ teaching behavior gradually increases. Table 5. Regression Analysis and the Influencing Factors (grades) Model 7

Freshman Regression Coefficient

Sophomore Std. Error

Regression Coefficient

Junior Std. Error

Regression Coefficient 0.84***

Std. Error

(constant)

0.51***

9.16

0.74***

9.52

Before-class

0.02

1.77

0.00

0.25

0.02

1.29

In-class

0.05***

3.77

0.03

1.81

−0.00

−0.08

After-class

0.05***

3.12

0.09***

3.89

0.11***

5.06

Interact

0.08***

5.93

0.07***

3.34

0.03

1.35

T’s behavior

0.28***

20.53

0.30***

14.61

0.36***

15.29

Evaluation

0.32***

18.91

0.26***

11.26

0.31***

9.40

Identity

0.09***

6.53

0.08***

3.99

0.06***

3.29

10.41

P.S.: *** shows that the correlation coefficient is significant at 0.01 level, ** shows that the correlation coefficient is significant at 0.05 level.

4.3.2 Regression Analysis for All Sample Because students’ before-class learning behavior is not significant in the regression analysis, this factor is excluded in the all-sample analysis. Linear regression analysis conducted in this study with a dependent variable (general competency improvement)

1006

J. Zuo and T. Huang

and six independent variables (student’s learning behavior: in-class, after-class, interact with classmates; teacher’s teaching behavior; evaluation of teaching content; professional identity). Model 6 with a greater proportion of dependent variable changes (R2 = 0.461) suggest that these factors is true for 46.1% of student’s general competency improvement. In Table 6, all factors have a significant impact on general competency, the impact is sorted from large to small as evaluation of teaching content, teacher’s teaching behavior, professional identity, student’s learning behavior (interact with classmates), student’s learning behavior (in-class), student’s learning behavior (after-class). Table 6. Coefficients Model

6

Unstandardized Coefficients (constant)

β

Std. Error

0.51

0.06

Standardized Coefficients

t

Sig.

9.191

***

Evaluation

0.32

0.02

0.27

19.025

***

T’s behavior

0.28

0.01

0.28

20.474

***

After-class

0.05

0.01

0.07

3.736

***

Identity

0.09

0.01

0.09

6.662

***

In-class

0.06

0.01

0.07

4.570

***

Interact

0.08

0.01

0.10

6.027

***

5 Discussion 5.1 The Influence of Evaluation of Teaching Content on the Improvement of Students’ General Competency According to Table 6, evaluation of teaching content is significantly positively related to the improvement of students’ general competency (β = 0.32, P < 0.01). In this study, students will evaluate the teaching content from the following seven aspects: “informative major course” and “integrate knowledge, ability and accomplishment” are the basic requirement of course; “give emphasis to the combination of practice and theory”, “pay attention to the intersection and integration of disciplines”, and “improve problem solving ability” are the features of application-oriented talents development; “expand student’s knowledge” and “catch the hotspots or cutting-edge information of discipline development” focus on depth and extension of teaching content. The findings show that evaluation of teaching content reflects the actual situation of curriculum construction, curriculum is a small step of talent development [18], classroom teaching is the main position of talent training, the construction of effective curriculum has a positive impact on students.

Research on the Influencing Factors of Students’ General Competency

1007

5.2 The Influence of Teacher’s Teaching Behavior on the Improvement of Students’ General Competency According to Table 6, a significantly relationship can be seen between teacher’s teaching behavior and the improvement of students’ general competency (β = 0.28, P < 0.01). Teacher’s teaching behavior refers to the actions taken by teachers to achieve certain educational and teaching purposes (group cooperative learning, utilize multimedia, instructional design). As teachers play a leading role in the classroom, the teaching methods adopted by teachers will influence students’ learning, and thus affect students’ learning effect and competency improvement [19]. 5.3 The Influence of Student’s Learning Behavior on the Improvement of Students’ General Competency In Table 6, the students’ learning behavior in-class, after-class and interaction with classmates is significantly related to the improvement of students’ general ability (β = 0.06, 0.05, 0.08, P < 0.01). Learning behavior is a series of behaviors adopted by students to achieve learning goals, including listen carefully in class, make study plan after class, join in competition with classmates. The learning behavior in-class, after-class and interaction with classmates are taken by students to improve academic achievement. Although academic achievement focuses more on achievement, competency improvement is also a kind of academic outcome [20]. In Table 6, there is no significant relationship between students’ before-class learning behavior and their general ability improvement (P > 0.05). Before-class learning behavior includes finishing homework or reading, and preview or organize knowledge points, which are also the behavior input taken by students to achieve their learning goals, but they are not significantly related to the improvement of general competency. The learning behavior focusing on rote memorization, short-term memory, and uncritical acceptance of knowledge [21], are more likely to be surface learning and it has nothing to do with competency improvement. 5.4 The Influence of Professional Identity on the Improvement of Students’ General Competency According to Table 6, there is a significant correlation between professional identity and student’s general competency (β = 0.09, P < 0.01). Professional identity refers to the degree of students’ satisfaction of their major, which indicates that students with a high satisfaction on their major are more likely to make improvement in their general competency. The basic findings are consistent with researches showing that high professional identity affects students’ learning, and will also have better learning initiative, for example, student with high professional identity may follow the hot spots or cuttingedge trends of their major and learn the professional knowledge more actively, which may help to improve their competency [22].

1008

J. Zuo and T. Huang

6 Conclusion 6.1 Teaching Content and Teachers’ Teaching Behavior Are the Key Factors on Improving Students’ General Ability Compared to the factor of students’ learning behavior and professional identity, teachers’ teaching behavior and evaluation of teaching content have a significant positive impact on the improvement of students’ general competency for freshmen, sophomores and juniors. In line with previous studies, the effective way for teachers to gain the recognition of students is to improve the teaching content and methods [23]. These data show that basic, practical, in-depth and extensive teaching content and diversified teaching methods, strategies and other teaching behaviors can help students improve their general competency, which requires universities to pay more attention to the development of curriculum construction and provide students with better learning resources. At the same time, we should give emphasis to the guidance of teachers’ teaching behavior by adopting diversified teaching methods and teaching strategies into classroom, such as seminar and group cooperative learning. 6.2 Teaching Content and Teachers’ Teaching Behavior Have the Greatest Impact on the Improvement of Students’ Competency, Which May Hinder the School’s Guidance on Students’ Learning Behavior As mentioned above, teachers’ teaching behavior and the evaluation of teaching content have a significant positive impact on the improvement of students’ general competency, and students’ learning behaviors also have impact. For example, for freshmen, students’ learning behaviors in class and after class, interaction with classmates can help students improve their general competency; for sophomores, students’ learning behavior after class and interaction with classmates can help students improve their general competency; for junior students, their learning behavior after class can help them improve their general competency. Students’ learning behavior has little impact regardless of their grades, which means students’ learning behavior does not play a key role in improving general competency. In this case, universities may neglect the guidance of students’ learning behavior. One possible reason is that students in application-oriented universities do not have strong learning ability and good learning habits. Most of them cannot learn independently and often rely on teachers’ teaching behavior or only pay attention to the school’s teaching contents. Therefore, their own learning behavior does not play a greater role than teachers’ teaching behavior and the evaluation factors of curriculum content, which may hinder colleges and universities from paying attention to the effective guidance of students’ learning behavior and the cultivation of learning ability. Findings from this research show that the role of students’ learning behavior is not as important as the factor of teachers’ teaching behavior and evaluation of teaching content, which may hinder universities from paying attention to the effective guidance of students’ learning behavior and the cultivation of learning ability. Acknowledgment. This paper is the research result of Guangxi’s 2021 private higher education research project “Research on Improving the Teaching Ability of Teachers in Private Applied Colleges and Universities” (No.2021ZJY667).

Research on the Influencing Factors of Students’ General Competency

1009

References 1. Soumi, M., Soumalya, C., Sayan, C.: Interactive web-interface for competency-based classroom assessment. Int. J. Educ. Manage. Eng. (IJEME) 13(1), 18–28 (2023) 2. Yuan, G.: Cultivation of application-oriented undergraduate general professional skills: value and approach. Mod. Educ. Manage. 377(08), 105–111 (2021). (in Chinese) 3. Nadler, D.A., Tushman, M.: The organization of the future: strategic imperatives and core competencies for the 21st century. Organ. Dyn. 27(1), 45 (1999) 4. Le Deist, F.D., Winterton, J.: What is competence? Hum. Resour. Dev. Int. 8(1), 27–46 (2005) 5. Zuo, J., Zhang, G.E., Huang, F.: Construction of ability and quality model of engineering talents based on analytic hierarchy process. In: Hu, Z., Zhang, Q., Petoukhov, S., He, M. (eds.) Advances in Artificial Systems for Logistics Engineering, pp. 356–366. Springer International Publishing, Cham (2022).https://doi.org/10.1007/978-3-031-04809-8_32 6. Tarimoradi, M., Zarandi, M.F., Türksen, I.B.: Hybrid intelligent agent-based internal analysis architecture for CRM strategy planning. Int. J. Intell. Syst. Appl. (IJISA) 6(6), 1–20 (2014) 7. Young, J., Chapman, E.: Generic competency frameworks: a brief historical overview. Educ. Res. Perspect. 37(1), 1–24 (2010) 8. Ministry of Education of the People’s Republic of China: Chinese Students Develop Core Literacy” was released [EB/OL] [2022–12–26]. http://www.scio.gov.cn/zhzc/8/4/Document/ 1491185/1491185.htm 9. EU publications: Key competences for lifelong learning [EB/OL] [2022–12–26]. https://op. europa.eu/en/publication-detail/-/publication/297a33c8-a1f3-11e9-9d01-01aa75ed71a1/lan guage-en 10. Secretary’s Commission on Achieving Necessary Skills. Background [EB/OL] [2022–12–26]. https://wdr.doleta.gov/SCANS/ 11. Wan, Z.: The classification, cultivation and inspiration of competence in American – taking SCANS as an example. Educ. Res. Exp. 160(05), 36–39 (2014). (in Chinese) 12. Wu, T., Lou, L., Zhou, H., et al.: The current status and influencing factors of intercultural communication competence of nursing students. Chin. J. Nurs. Educ. 19(07), 626–630 (2022). (in Chinese) 13. Zhuang, Y., Liu, Y.: The Influencing factors of lower grade undergraduates’ innovation capability in science and technology based on structural equation model. China High. Educ. Res. 4, 51–56 (2022). (in Chinese) 14. Zhang, Z., Zhao, M., Jie, W., et al.: Analysis of influencing factors of problem analysis ability of critical thinking among undergraduates in a medical college in Wuhan. Med. Soc. 35(04), 100–105 (2022). (in Chinese) 15. Xiao, X., Yuan, H., Liao, H., et al.: Study on influencing factors of autonomous learning ability of medical students in Guangzhou city during COVID-19 epidemic. Med. Soc. 35(02), 101–104+111 (2022). (in Chinese) 16. Yu, T., Gu, X.: The generic competency mismatch of engineering graduates and its influencing factors. Res. High. Educ. Eng. 05, 43–49 (2022). (in Chinese) 17. Rudenko, I.V., Fedorova, N.I., Muskhanova, I.V., et al.: Organizational and pedagogical conditions development of general competences formation in the educational activities of the university. J. Sustain. Dev. 8(6), 45 (2015) 18. Wang, Y.: Talent training and curriculum innovation of drama, film and television literature under the background of “New Liberal Arts.” Drama 198(04), 163–173 (2021). (in Chinese) 19. Wang, X., Wu, Y.: Exploration and practice for evaluation of teaching methods. Int. J. Educ. Manage. Eng. (IJEME) 2(3), 39–45 (2012) 20. Gull, F., Shehzad, S.: Effects of cooperative learning on students’ academic achievement. J. Educ. Learn. 9(3), 246–255 (2015)

1010

J. Zuo and T. Huang

21. Meeks, M.D., et al.: Deep vs. surface learning: an empirical test of generational differences. Int. J. Educ. Res. 1(8), 1–16 (2013) 22. Hou, Z., He, W., Wang, Z.: Research on the influence of tutor’s guidance style on graduate students’ knowledge sharing and innovation. Academic Degree & Graduate Education 279(02), 62–67 (2016). (in Chinese) 23. Zuo, J., Wu, X., Zhang, G., et al.: Statistical analysis based on students’ evaluation of teaching data in an application-oriented university. In: 2022 International Conference on Educational Innovation and Multimedia Technology (EIMT 2022), pp. 490–498. Atlantis Press (2022)

Dynamic Adjustment of Talent Training Scenario for Electrical Engineering and Automation Majors Based on Professional Investigation Haoliang Zhu(B) School of Intelligent Manufacturing, Nanning University, Nanning 530200, Guangxi, China [email protected]

Abstract. For the purpose of realizing the effective connection between the professional talent training program and the social needs, this paper has carried out a survey of previous graduates and related enterprises and industries for the electrical engineering and automation specialty of Nanning University. Starting from the needs of enterprises, this paper analyzes the scientificity and applicability of the development of talent training scenarios for electrical engineering and automation majors, invites enterprises to participate in the whole process of specialty setting, curriculum development and teaching reform, timely discovers the deficiencies of this major in school running ideas, specialty setting, curriculum system, student ability and quality training, student and other aspects, and dynamically adjusts the professional talent training plans and programs. While improving the talent training program, it is also necessary to constantly expand the channels of enrollment and employment. The research ideas and methods of this paper can provide some reference and reference for other colleges and universities. Keywords: Talent training · Major in electrical engineering and automation · Professional investigation · Talent training scheme · Dynamic adjustment

1 Introduction Electrical automation technology mainly refers to a multi-disciplinary comprehensive technology that uses electrical and electronic technology and devices to realize automatic control of production process [1]. Its emergence, development and increasingly extensive application have greatly improved productivity, and have penetrated into all aspects of industrial production, becoming an important basis and leading role in the development of modern industry [2–4]. The specialty of electrical engineering and automation is set up to train such specialized talents. Talent is the core competitiveness of enterprises and the cornerstone of industrial development and social progress [5, 6]. With the vigorous development of the equipment manufacturing industry and modern service industry in the Beibu Gulf urban agglomeration with Nanning as the core, relying on the developed industrial base of electrical © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 1011–1021, 2023. https://doi.org/10.1007/978-3-031-36118-0_86

1012

H. Zhu

and electronic industries in the Pearl River Delta, the demand for electrical automation technology professionals in regions including the Pearl River Delta has been strong in recent years, and the trend is increasing year by year, which requires colleges and universities to cultivate and cultivate a large number of professional theoretical knowledge that both adapt to the characteristics of the times, High-quality application-oriented talents with professional operating skills [7, 8]. In response to the new requirements of relevant industries for talents, colleges and universities should maintain real-time contact and effective communication with enterprises and industries at all times, and take the survey results and sampling data as feedback information in the formulation of talent training programs, and constantly make “dynamic” adjustments. Nanning University, as one of the first pilot universities for the reform of applied science and technology universities of the Ministry of Education, focuses on the development of professional clusters of applied disciplines, builds a disciplinary and professional system with engineering as the main body and multi-disciplinary and coordinated development of disciplines, and actively promotes the reform of the training mode of applied talents with the cultivation of applied talents as the main line [9]. In recent years, teaching reform has been continuously deepened around the cultivation of “applicationoriented talents”, and a series of effective measures have been formulated, especially in the formulation of talent training programs, and outstanding achievements have been achieved. In the construction of education and teaching system, professional research is an effective means, providing an important basis for the formulation and reform of talent training scheme [10]. The specialty of electrical engineering and automation is the dominant specialty of the school, serving the development needs of Guangxi’s “14 + 10” industrial cluster of 100 billion yuan [11]. At present, there are more than 700 students in the university, and there are more than 150 graduates of the first undergraduate course of this major in 2022. It is necessary to conduct comprehensive and in-depth research on graduates and conduct interviews with relevant enterprises to guide the revision of talent training plans.

2 Significance and Method of Investigation and Research Investigation and research is a scientific method for people to understand the real situation of things in a planned and purposeful way through various ways and methods, so as to understand and transform society. Through thinking and processing the data and materials obtained from the survey, we can gain the understanding of the nature and laws of objective things [12, 13]. 2.1 Significance of Professional Investigation and Research The professional research discussed in this paper includes both the research on graduates and the research on enterprises. The enterprise survey is mainly carried out for those enterprises that have a certain degree of relevance to the specialty. From the actual production of the enterprise, understand the job responsibilities and requirements for talent quality [14–16].

Dynamic Adjustment of Talent Training Scenario for Electrical Engineering

1013

The significance of conducting professional research can be summarized into four aspects. (1) (2) (3) (4)

Provide reference for the development of professional talent training programs; Provide reference for curriculum construction; Open up a path for school-enterprise cooperation; Provide learning opportunities for the construction of teaching staff.

Through in-depth communication, the school understands the characteristics, culture and talent needs of the enterprise. At the same time, it invites the technical personnel and managers of the enterprise to effectively participate in the school’s professional setting, curriculum development and teaching reform, which can improve the teaching quality of electrical engineering and its automation specialty, and timely find out the teaching ideas, professional setting, curriculum system, and student’s ability and quality training of the specialty Deficiencies in student management and other aspects and suggestions for modification [17–20]. 2.2 Research Methods Field survey, interview survey, conference survey, questionnaire survey, expert survey and statistical survey have all been proved to be effective methods of investigation and research [21, 22]. The two types of objects discussed in this paper - graduates and enterprises - can be investigated in different ways. (1) Graduate survey For graduates who have taken up their jobs, the most convenient way is to use questionnaire stars for questionnaire survey. Interviews, meetings and field surveys are also common methods [23, 24]. The main contents of the survey are: whether the graduates’ employment distribution, post treatment, professional knowledge, job distribution and interpersonal relationship meet the requirements of employers and posts; Check whether the students trained by the school are “satisfactory” to the enterprise; Analize whether the students are “satisfied” with the school’s professional setting, talent training objectives and procedures. (2) Research on enterprises The above-mentioned research methods are effective and often used for enterprise research. The most important purpose of enterprise research is to go deep into enterprises and industries and understand their requirements for talents [25]. In order to make the survey results broadly representative and scientific, the survey objects are divided into electronic and electrical enterprises in Nanning and its surrounding cities and the Pearl River Delta according to their geographical characteristics. According to the methods and steps of developing talent training programs, the contents of the survey are divided into two major aspects: first, the survey of the application of electrical automation technology in enterprises, the demand and source of talents, the law of professional growth of technicians, the professional positions and responsibilities of the corresponding students, the knowledge, ability and quality they have, and the key

1014

H. Zhu

courses they should study. The second is to investigate the actual employment positions and career growth process of graduates of electrical automation technology in recent years. The survey needs to be based on the growth and development of students [26, 27].

3 Questionnaire Survey on Graduates A questionnaire survey was conducted using a questionnaire star. The collected questionnaires were mainly aimed at graduates engaged in electrical and electronic industries. A total of 57 valid questionnaires were collected. 3.1 Employment Situation of Graduates Figure 1 shows that 61% of graduates are employed in Guangxi and 28% of graduates are employed in Guangdong, which shows that the talent training of this major is basically in line with the service orientation of the school, which is “based on the Beibu Gulf urban agglomeration with Nanning as the core, to cultivate high-quality application-oriented talents with” character, employment ability, entrepreneurship ability, foundation for further study, and development potential”, Therefore, the orientation of the talent training service of this specialty is basically accurate. In the future, the talent training orientation of this specialty should continue to adhere to “local” and “application-oriented”, and adhere to the implementation of talent training oriented to the training of talents serving local economic construction.

Fig. 1. Employment distribution of graduates

3.2 Salary of Graduates It can be seen from Table 1 that the average salary of graduates is 4850 yuan, slightly lower than the national statistics, but higher than the average level in Guangxi. From

Dynamic Adjustment of Talent Training Scenario for Electrical Engineering

1015

the student feedback, students are basically satisfied with talent training. Basically, it can also be reflected from the side that the quality of talent training in the electrical engineering and automation specialty of our university is affirmed, but the per-capita treatment of the graduates of this specialty is generally lower than the national level (5700 yuan per capita), which indicates that the lack of local industrial resources is also an important condition to constrain the professional development. Table 1. Wages and benefits of graduates Initial employment treatment Subtotal Proportion (RMB) > = 20000

1

1.75%

10000–20000

6

10.53%

5000–10000

17

29.82

0, i = 1, 2, . . . , m

(3)

(4)

j=1

(3) Establish the Weight Set The evaluation set is a set of various possible results of total evaluation. There are p possible results of total evaluation. The evaluation set can be expressed as V .   V = v1 , v2 , . . . , vp (5) (4) Single-factor Fuzzy Evaluation The evaluation is conducted from a factor alone to determine the membership degree of the evaluation object to the evaluation set elements, which is called singlefactor fuzzy evaluation. The evaluation object is evaluated according to the i factor ui , and the membership degree of the k element vk in the evaluation concentration is rik , then the evaluation result according to the i factor ui is Ri .   Ri = ri1 , ri2 , . . . , rip , i = 1, 2, . . . , n (6) The results of each single factor as a matrix of rows R. As the one-factor evaluation matrix. ⎡ ⎤ r11 r12 . . . r1p ⎢ r21 r22 . . . r2p ⎥ ⎢ ⎥ (7) R = (rik ) = ⎢ . . . ⎥ . . ... ⎦ ⎣ .. .. rn1 rn2 . . . rnp

The Effect of Integrating Higher Vocational Aesthetic and Ideological

1025

(5) Fuzzy Comprehensive Evaluation Fuzzy comprehensive evaluation is to consider the influence of all factors and get the correct evaluation results. If the one-factor evaluation matrix of the fuzzy comprehensive evaluation of the first-level indicators is set, the comprehensive evaluation set is B.  (8) B = H · R = b1 , b2 , . . . , bp The one-factor evaluation matrix of the fuzzy comprehensive evaluation of the second-level index is Ri , then the fuzzy comprehensive evaluation set of the secondlevel index is Bi . Bi = Hi · Ri , i = 1, 2, . . . , m

(9)

The fuzzy comprehensive evaluation results of second-level indicators are taken as the single factor evaluation of the fuzzy comprehensive evaluation of first-level indicators. ⎛ ⎞ B1 ⎜ B2 ⎟ ⎜ ⎟ (10) R=⎜ . ⎟ ⎝ .. ⎠ Bm mxp According to the maximum membership principle, if   bt = max bj , j = 1, 2, . . . , p

(11)

Then the evaluation results are subordinate to vi 2.2 Effect Evaluation The effect can be evaluated from the following five aspects. (1) Establish the Effectiveness Evaluation Index Set From 2.1, the evaluation index is divided into first-level indicators and second-level indicators. The primary indicators are U = {educatee, educators, educationalresources, integrationapproach}; the secondary indicators are U1 = {interest degree, enthusiasm for learning, participation, ability to participate}; U2 = {knowledge structure, teaching ability, scientific research ability}; U3 = {system construction, platform construction, capital investment, media publicity and guidance}; U4 = {design interesting ideological and political classroom to enrich campus aesthetic creation and experience, lead students’ better life and social practice, improve teachers’ aesthetic quality and cultivate aesthetic feelings}. The comprehensive evaluation index system and weight of the effect of the integration of higher vocational aesthetic education and ideological and political education are shown in Table 1. (2) Establish a Weight Set According to the index weights in Table 1, the weight set of each factor can be drawn as follows. Weight set of evaluation factors for first-level indicator H . H = (0.4096, 0.2951, 0.1056, 0.1897)

(12)

1026

L. Li and F. Hu

Table 1. Comprehensive evaluation index system and weight of the effect of the integration of higher vocational aesthetic education and ideological and political education Level 1 Indicators

Weight

Secondary Indicators

Weight

Total Weight

educatee (student)u1

0.4096

degree of interest u11

0.2911

0.0168

enthusiasm for learning u12

0.3252

0.0646

participation in activities u13 0.3409

0.1496

ability to participate u14

0.0428

0.0232

knowledge structure u21

0.4326

0.1305

teaching ability u22

0.4928

0.1568

ability to research u23

0.0746

0.0035

system construction u31

0.4547

0.1440

platform construction u32

0.3081

0.0761

funding u33

0.1012

0.0054

media publicity and guidance u34

0.1359

0.0084

design an interesting ideological and political class u41

0.5262

0.1874

enrich the campus aesthetic creation and experience u42

0.2079

0.0112

lead students to live a better life and social practice u43

0.1503

0.0178

improve teachers’ aesthetic quality and cultivate their aesthetic feelings u44

0.1256

0.0047

educators (teacher)u2

educational resources (educational institution)u3

Fusion way u4

0.2951

0.1056

0.1897

The weight set of the factors evaluating the secondary index H1 , H2 , H3 , H4 . H1 = (0.2911, 0.3252, 0.3409, 0.0428)

(13)

H2 = (0.4326, 0.4928, 0.0746)

(14)

H3 = (0.4547, 0.3081, 0.1012, 0.1359)

(15)

H4 = (0.5262, 0.2079, 0.1503, 0.1256)

(16)

(3) Establish the Evaluation Set The effect evaluation of the integration of higher vocational aesthetic education and ideological and political education is set as four levels, namely the evaluation set V = {Verygood , good , average, bad }. If the percentage-point system is used to represent the evaluation level, V = |90, 80, 70, 50|.

The Effect of Integrating Higher Vocational Aesthetic and Ideological

1027

Taking higher vocational colleges in Hubei Province as an example, the effect of the integration of higher vocational aesthetic education and ideological and political education is adjusted through the form of questionnaire adjustment. A total of 500 questionnaires were distributed and 469 questionnaires were recovered. After statistics, the evaluation of each factor in the index system is shown as in Table 2. Table 2. Summary table of the results of the secondary index item questionnaire Designation

Very good

good

average

bad

Designation

Very good

good

average

bad

u11

56

152

226

35

u31

100

340

27

2

u12

96

308

55

10

u32

99

344

23

3

u13

102

316

43

8

u33

45

353

58

13

u14

52

146

182

89

u34

65

363

37

4

u21

89

353

26

1

u41

132

331

5

1

u22

112

353

3

1

u42

86

350

26

7

u23

112

350

6

1

u43

83

337

34

15

u44

98

348

22

1

(4) Obtain the Fuzzy Relation Matrix Data processing in Table 2 obtained the evaluation proportion of each factor in the secondary index system, namely the one-factor evaluation matrix R1 , R2 , R3 , and R4 of the fuzzy comprehensive evaluation of the secondary index. ⎡

⎤ ⎡ ⎤ 0.1192 0.3241 0.4819 0.0763 0.1898 0.7527 0.0554 0.0021 ⎢ 0.2047 0.6567 0.1173 0.0213 ⎥ ⎥ ⎣ ⎦ R1 = ⎢ ⎣ 0.2175 0.6738 0.0917 0.0171 ⎦ R2 = 0.2388 0.7527 0.0064 0.0021 0.2388 0.7463 0.0128 0.0021 0.1109 0.3113 0.3881 0.1898 (17) ⎡ ⎡ ⎤ ⎤ 0.2132 0.7249 0.0576 0.0043 0.2814 0.7058 0.0107 0.0021 ⎢ 0.2111 0.7335 0.0490 0.0064 ⎥ ⎢ 0.1833 0.7463 0.0554 0.0149 ⎥ ⎢ ⎥ ⎥ R3 = ⎢ ⎣ 0.0959 0.7527 0.1237 0.0277 ⎦ R4 = ⎣ 0.1770 0.7186 0.0725 0.0320 ⎦ 0.1386 0.7740 0.0789 0.0085

0.2090 0.7420 0.0469 0.0021 (18)

1028

L. Li and F. Hu

⎡

⎤ 0.1192 0.3241 0.4819 0.0763 ⎢ 0.2047 0.6567 0.1173 0.0213 ⎥ ⎢ ⎥ ⎢ 0.2175 0.6738 0.0917 0.0171 ⎥ ⎢ ⎥ ⎢ 0.1109 0.3113 0.3881 0.1898 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 0.1898 0.7527 0.0554 0.0021 ⎥ ⎢ ⎥ ⎢ 0.2388 0.7527 0.0554 0.0021 ⎥ ⎢ ⎥ ⎢ 0.2388 0.7463 0.0128 0.0021 ⎥ ⎢ ⎥ ⎥ R =⎢ ⎢ 0.2132 0.7249 0.0576 0.0043 ⎥ ⎢ 0.2111 0.7335 0.0490 0.0064 ⎥ ⎢ ⎥ ⎢ 0.0959 0.7527 0.1237 0.0277 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 0.1386 0.7740 0.0789 0.0085 ⎥ ⎢ ⎥ ⎢ 0.2814 0.7085 0.0107 0.0021 ⎥ ⎢ ⎥ ⎢ 0.1833 0.7463 0.0554 0.0149 ⎥ ⎢ ⎥ ⎣ 0.1770 0.7186 0.0725 0.0320 ⎦ 0.2090 0.7420 0.0469 0.0021

(19)

(5) Calculate the Evaluation Results According to the above formula, the fuzzy comprehensive evaluation results of the secondary indicators can be obtained B1 , B2 , B3 , B4 . B1 = H1 · R1 = (0.1802, 0.5509, 0.2263, 0.0431)

(20)

B2 = H2 · R2 = (0.2176, 0.7522, 0.0281, 0.0021)

(21)

B3 = H3 · R3 = (0.1905, 0.7370, 0.0645, 0.0079)

(22)

B4 = H4 · R4 = (0.2389, 0.7277, 0.0338, 0.0092)

(23)

Taking the fuzzy comprehensive evaluation results of second-level index as single-factor evaluation of first-level index get R, and bring into formula above. B = H · R = (0.2035, 0.6635, 0.1142, 0.0209)

(24)

3 Evaluation Results From the above results, it can be seen that in the investigation of the effect of the integration of higher vocational aesthetic education and ideological and political education in Hubei Province, 20.35% of teachers and students think that the effect of the integration of higher vocational aesthetic education and ideological and political education is very good, 66.35% think it is good, 11.42% think it is general, and 2.09% of teachers and students think it is poor. According to the principle of maximum subordination, it can be concluded that the effect of integrating higher vocational aesthetic education and ideological and political education in Hubei Province is excellent. However, there are also

The Effect of Integrating Higher Vocational Aesthetic and Ideological

1029

some deep-rooted reasons worth analyzing, so as to further promote the integration of aesthetic education and ideological and political education [3–5], constantly realize the unity of “true, good and beautiful”, realize the high degree of ideological and political education and aesthetic experience, better realize the educational goal, and promote the cultivation of ideological and political quality of the whole people [6]. While in the study, it was found that students were still very enthusiastic about learning, that ideological and political teachers have good performance in knowledge structure, teaching ability and scientific research ability [7], but there is still a deficit of 2.1%, which needs to be further improved. The survey found that the system and platform construction of integrating aesthetic education and ideological and political education is quite perfect, but it needs to be strengthened in the aspects of capital investment, atmosphere construction and media publicity [8]. From the questionnaire survey, it is found that 23.89% of the students and teachers respond very well to the integration of higher vocational aesthetic education and ideological and political education in general, 72.77% of them think it is good, but 3.38% of them still think it is average and 0.92% of them think it is poor. Among them, “designing a tasty Civic Science classroom” and “improving teachers’ aesthetic quality and cultivating aesthetic feelings” are well developed [9], but “enriching aesthetic creation and experience on campus and leading students to social practice for a better life” are less satisfactory.

4 Evaluation Analysis 4.1 Improve Students’ Interest and Participation in Activities Except for art students, most students have no professional education or training in aesthetic education [10], and their ability to carry out activities is limited, which to some extent increases students’ sense of loss and reduces their interest in learning and enthusiasm for participation. Therefore, in the practice of integrating aesthetic education with ideological and political education, the strong motivation and desire for expression of young students can be explored and brought into play according to their characteristics and strengths, and their initiative and creativity can be actively mobilized, and different types of activities can be carried out to give students a greater sense of access. The “artistic”, “aesthetic”, “experiential”, and “emphatic” methods can be used to stimulate and train the educated to shape beautiful objects [11], actively explore existing cultural and artistic works, create situations with beautiful expressions, expand the emotional experience, enrich the aesthetic experience, and promote the depth of aesthetic understanding. In turn, they are able to receive theoretical inculcation, transform into practical ideological and political qualities, core literacy and key competencies, and internalize them into practical actions in a pleasant way. 4.2 Further Enhance Teacher’ Knowledge Reserve and Teaching and Scientific Research Ability The use of aesthetic education in ideological and political education is to unify and integrate the role of aesthetics in promoting the free and comprehensive development

1030

L. Li and F. Hu

of human beings with the goal of ideological and political education striving to achieve comprehensive human development. Through the emotional experience of beauty, the use of artistic conception, creation [12], expression and appreciation and other levels of thinking and practice of beauty, to mobilize the senses and thinking, the formation of aesthetic understanding, and get a sense of pleasure from the acquisition, and then the occurrence of changes in thought, prompting the adjustment of personal behavior, in order to achieve the unity of knowledge and action. Therefore, the knowledge reserve and teaching and scientific research ability of ideological and political teachers is to solve the problem of “explaining the deep, profound and vivid truth full of philosophical thinking, historical deposits and reality”. 4.3 School Funding Investment and Media Publicity Be Strengthened As the backbone of education, schools should fully play to the role of organization and leadership [13], enrich the construction of the system and platform, and appropriately increase some funds in the actual integration of aesthetic education and ideological and political education, increase the publicity and guidance, and strive to create a stronger atmosphere. 4.4 The Integration Way Be Further Broadened The national unified teaching materials for Civics reflect the highest level of current Marxist theory research, reflect the requirements of national will and ideology, and provide a strong theoretical foundation. And the rooting of the teaching material system requires the effective transformation of the teaching material system to the teaching system. The use of aesthetic education in ideological and political education means that the design of educational contents, educational methods and educational forms are set around the development of aesthetic activities, educational activities stimulate the leading role of the teacher and the enthusiasm [14], initiative and creativity of the taught in the educational activities with the help of the concept and methods of aesthetic education, the educator and the educated communicate with the intermediary of aesthetic objects, realize interaction through the production of beauty, promote the overall human development and improve the practical effect of thought transformation. Therefore, by expanding the ways of integration of higher vocational aesthetic education and ideological and political education, we can better realize the high degree of fit between ideological and political education and aesthetic experience, better achieve the education goal and promote the ideological and political quality of college students.

5 Conclusion The cultivation of aesthetic ability plays an important role in ideological and political education. Aesthetics has a unique advantage in the penetration of people’s deep consciousness. The emotional experience connects the past and future of the aesthetic subject, and the sensory pleasure as well as the imagination inspired in the process of transition to rationality are more deeply recorded in people’s physiological and psychological

The Effect of Integrating Higher Vocational Aesthetic and Ideological

1031

structures, reaching a depth that is difficult to achieve by purely rational thinking, thus making people’s understanding closely connected with people’s emotions, establishing emotional in people’s behavior system. The mechanism of stimulation, people’s judgment is not only utilitarian rational judgment, but also adds more actionable emotional judgment, which promotes the internalization of ideological and political education. Overall, this paper uses the fuzzy comprehensive evaluation method to assess the effectiveness of the current integration effect of higher vocational aesthetic education and ideological and political education, on the basis of which the following conclusions are drawn. 1) The current integration effect of higher vocational aesthetic education and ideological and political education is excellent, reaching 86.7%. However, there are some deepseated reasons worth analyzing. 2) Students’ enthusiasm for learning is still very high, but most of them have no professional education or training in aesthetic education, and their ability to carry out activities is limited except for art students, which to a certain extent breeds students’ sense of loss and reduces their interest in learning and enthusiasm for participation. Therefore, in the practice of integrating aesthetic education with ideological and political education, different types of activities can be carried out according to students’ characteristics and specialties, so that students can have more access to them. 3) Teachers of ideology and political science courses perform well in knowledge structure, teaching ability and scientific research ability, but there are still shortcomings that need further improvement. 4) The system and platform construction of the integration of aesthetic education and ideological and political education in higher vocational institutions is still quite perfect, but the financial investment, atmosphere creation and media publicity need to be strengthened. 5) The current ways of integrating aesthetic education with ideological and political education are generally well reflected. Among them, the educational content is designed to organically integrate the cultural vision, artistic literacy, aesthetic needs and state of mind of the educated, and deepen the understanding and experience of the beauty of ideological and political education through ideological leadership, political orientation and educational promotion. Acknowledgment. This project is supported by Hubei Province education planning 2021 annual key topics (2021GA092).

References 1. Li, L., Wang, Z., Tang, X.: Study on practical reform of ideological and political theory course in colleges and universities from the perspective of cultivating “core competency” and “key ability.” Educ. Circles 15, 66–68 (2019). (in Chinese) 2. Wang, Y.: Study on the effectiveness of ideological and political education in the major tasks of the armed police. J. Higher Educ. 15, 38–39 (2019). (in Chinese) 3. Sathe, M.T., Adamuthe, A.C.: Comparative study of supervised algorithms for prediction of students’ performance. Int. J. Modern Educ. Comput. Sci. (IJMECS) 2, 1–21 (2021)

1032

L. Li and F. Hu

4. Rafida, V., Widiyatni, W., Harpad, B., Yulsilviana, E.: Implementation of multi-attribute rating technique simple in selection of acceptance scholarship of PMDK (Case study: STMIK widya cipta dharma). Int. J. Modern Educ. Comput. Sci. (IJMECS) 2, 22–33 (2021) 5. Sriram, B.: Learner’s satisfactions on ICT innovations: omani learners viewpoints. Int. J. Modern Educ. Comput. Sci. (IJMECS) 12(5), 1–15 (2020) 6. Yunusovich, A.V., Ahmedov, F., Norboyev, K., Zakirov, F.: Analysis of experimental research results focused on improving student psychological health. International Journal of Modern Education and Computer Science (IJMECS) 14(2) (2022) 7. Xiao, W.: Integrity and innovation of ideological and political teaching in colleges and universities--interview with Lingjun Huang, dean of school of marxism in huazhong university of science and technology. People’s Daily, 2019–1–25(9) (in Chinese) 8. Noor, M.B., Ahmed, Z., Nandi, D., Rahman, M.:Investigation of facilities for an m-learning environment. International Journal of Modern Education and Computer Science (IJMECS) 1, 34-48 (2021) 9. Bouzid, T., Kaddari, F., Darhmaoui, H., Bouzid, E.G.: Enhancing math-class experience throughout digital game-based learning, the case of moroccan elementary public schools. International Journal of Modern Education and Computer Science (IJMECS) 13(5) (2021) 10. Divayana, D.G.H.: ANEKA-based asynchronous and synchronous learning design and its evaluation as efforts for improving cognitive ability and positive character of students. Int. J. Modern Educ. Comput. Science (IJMECS) 10, 14–22 (2021) 11. Kyrpychenko, O., Pushchyna, I., Kichuk, Y., Shevchenko, N., Luchaninova, O., Koval, V.: Communicative competence development in teaching professional discourse in educational establishments. Int. J. Modern Educ. Comput. Sci. (IJMECS) 8, 16–27 (2021) 12. Firdaus, H., Zakiah, A.: Implementation of usability testing methods to measure the usability aspect of management information system mobile application (Case study sukamiskin correctional institution). Int. J. Modern Educ. Comput. Sci. (IJMECS) 13(5), 58–67 (2021) 13. Perez, J.G., Perez, E.S.: Predicting student program completion using naïve bayes classification algorithm. International Journal of Modern Education and Computer Science (IJMECS), 13(3), 57–67 (2021) 14. Dolgikh, S.: Categorization in unsupervised generative self-learning systems. Int. J. Modern Educ. Comput. Sci. (IJMECS) 6, 68–78 (2021)

Teaching Reform and Practice of “Sensor and Detection Technology” Course Based on “Specialty and Innovation Integration” and Project-Based Education Deshen Lv(B) and Ying Zhang School of Intelligent Manufacturing, Nanning University, Nanning 530200, China [email protected]

Abstract. The course “Sensor and Detection Technology” is a required course for many colleges and universities in electronics, electrical, communication and automatic control and related majors. It is a course combining theory and practice and plays a very important role in the cultivation of innovative talents. Under the traditional teaching method, the course arrangement is mostly limited to the class teaching and corresponding in-school experiments. The course design is not well integrated with enterprise practice. Students are relatively passive in the course learning, and high-quality course resources are relatively scarce. In view of the above problems, the paper discusses the objectives and design ideas of curriculum teaching reform under the guidance of the principle of “integration of specialty and innovation”, the specific items of curriculum teaching and the evaluation methods, focuses on the deep integration of “professional education” and “innovation and entrepreneurship education”, optimizes the teaching content, aims to cultivate students’ innovative thinking, and lays a solid foundation for the society to cultivate innovative talents. Keywords: Sensor and detection technology · Integration of specialty and innovation · Innovation and entrepreneurship · Teaching reform

1 Introduction Sensor is the primary link to realize automatic detection and automatic control [1, 2]. As a device that converts external input signals into electrical signals, sensors have become an indispensable tool in today’s production and life [3–5]. Industrial production lines, security systems, car navigation systems, aerospace systems, household appliances, etc. are not difficult to see the traces of sensors [6–8]. The sensor and its detection technology have become a key field of research in the industry and academia [9]. Many colleges and universities have taken the course of “Sensor and Detection Technology” as a compulsory basic course for electronics, electrical, communication and automatic control majors [10, 11]. The purpose is to enable students to master sensor technology, understand relevant knowledge in measurement and control, computer © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 1033–1045, 2023. https://doi.org/10.1007/978-3-031-36118-0_88

1034

D. Lv and Y. Zhang

application, electronic information and other fields, and cultivate students’ ability of product design and innovative application [12–14]. Graduates of electronics, electrical, communication and automatic control majors have many jobs related to sensors, such as detection and installation of electronic products, intelligent instrument development, robot application development, etc. [15, 16]. It occupies a large proportion in the whole basic curriculum system. The teaching of this course attaches great importance to the combination of “specialty” and “innovation”. The former focuses on the vertical development of professional technology to understand the essence of knowledge and skills; The latter - innovation and entrepreneurship education needs to pay attention to the innovation and application of knowledge, which is conducive to the horizontal development of professional technology [17–19]. In recent years, the integration of specialty and innovation has been the focus of curriculum teaching reform in colleges and universities. It attaches importance to both professional education and innovation and entrepreneurship education [20, 21]. Traditional teaching only pays attention to professional education, which cannot effectively help students grasp and understand the content of “sensor and detection technology” and apply it to future work [22, 23]. In the context of “mass entrepreneurship and innovation”, it is required to closely link professional knowledge with innovation and entrepreneurship education to carry out curriculum reform [24, 25]. Therefore, it is urgent to link sensor technology with innovation and entrepreneurship education, and it is very urgent to develop a curriculum reform plan of “sensor and detection technology” based on the integration of specialty and innovation [26]. How to integrate innovation and entrepreneurship education into the teaching of “sensor and detection technology” is the key to reform [27, 28].

2 Teaching Status and Reform Content of the Course “Sensor and Detection Technology” 2.1 Current Situation of Course Teaching The teaching of specialized and creative integration course is a teaching reform method that has emerged in recent years. Many colleges and universities have not reformed the teaching of “sensor and detection technology” course, and still use the traditional teaching method. In traditional teaching methods, teachers regard themselves as the controller and dominator of the classroom, and students’ learning state is always particularly passive. Students need to listen to what the teacher says. Some teachers put too much emphasis on the discussion of principles, but did not guide students to carry out diversified practice, relied too much on the final written test results, and did not allow students to participate in the operation by themselves, resulting in very poor practical results of students. At the same time, the content of the textbook lags behind the development of the discipline and does not show the cutting-edge dynamics of sensor technology. In addition, most teachers use the method of classroom teaching to guide students to follow their own ideas and learn theoretical knowledge mechanically, without leaving enough room for students to think. When a question is thrown out, students will directly publish the correct answer before they have time to think. Over time, students’ initiative and enthusiasm will be greatly weakened.

Teaching Reform and Practice of “Sensor and Detection Technology” Course

1035

Some colleges and universities have carried out the curriculum reform of “sensor and detection technology” specialized innovation integration, but there are still some problems, such as the curriculum of specialized innovation integration is not reasonable enough, and the high-quality curriculum resources of specialized innovation integration are lacking. 2.2 The Curriculum is not Reasonable Enough In the process of “sensor and detection technology” innovation and entrepreneurship integration reform, many colleges and universities still separate sensor knowledge from innovation and entrepreneurship, and have not designed an organic curriculum integration mode, which makes learners mechanically contact with innovation and entrepreneurship content after the completion of professional learning in the course of curriculum learning, and does not have a connection between the two in their minds. Such curriculum settings have lost the original significance of innovation and entrepreneurship integration, It deviates from the original intention of innovation and entrepreneurship based on the advantages of sensor knowledge. The sensor curriculum ignored innovation and entrepreneurship practice at the beginning of the integration design, or the school lacked innovation and entrepreneurship practice resources, and the curriculum theory and practice system design was not perfect. 2.2.1 Lack of High-Quality Curriculum Resources In the specific practice of the integration of specialty and innovation, the construction of curriculum resources is particularly important, which is the implementation of details in practice. Most colleges and universities lack high-quality curriculum resources for “sensor and detection technology” courses. Professional education continues to be carried out in the traditional professional education mode, and innovation and entrepreneurship education is also carried out in the traditional way. The current development status of the industry is not included in the construction of curriculum resources, and the existing problems of the industry are analyzed. For the specific key and difficult points of the course, no in-depth analysis and sorting has been carried out, and the organic relationship between the key and difficult points and the innovation and entrepreneurship knowledge points has been sorted out, so as to achieve the one-to-one mapping between the course content and innovation. In the resource construction, real cases should be added, and the case rational analysis of sensor knowledge should be used while cultivating innovation literacy. The lack of strong participation of enterprises is also an important reason for the lack of resources in the construction of specialized creative courses. The enterprise has rich project cases, which can help the school promote the teaching of experiment and practice. At the same time, the rich technology and management advantages of enterprises can also inspire the teachers and students of the school to pursue innovation and carry out innovation and entrepreneurship. Increasing the participation of enterprises in the integration of specialty and innovation in colleges and universities can provide strong support in the construction of resources, help schools truly realize the integration of specialty and innovation, and promote talent cultivation.

1036

D. Lv and Y. Zhang

2.3 Ideas and Contents of Curriculum Reform Starting from the principle of “integration of specialty and innovation”, in order to make the curriculum design of “sensor and detection technology” more reasonable, provide high-quality and realistic projects, and promote the collaborative development of students’ professional technology and innovation and entrepreneurship. As shown in Fig. 1, the curriculum reform mainly includes three modules - the design of curriculum objectives, the implementation plan of project teaching method, and diversified assessment methods.

Fig. 1. Contents of the curriculum reform of “sensor and detection technology”

Module 1 “Design of teaching objectives” is to reset the old teaching objectives, adjust and deepen the knowledge objectives, ability objectives and quality objectives, and increase the goal of innovation and entrepreneurship; Module 2 “Project teaching method implementation plan” is to establish the idea and steps of project-based teaching method, deconstruct the old knowledge points, adopt project teaching method, reconstruct the items that meet the curriculum, and give the teaching items of “sensor and detection technology” curriculum; Module 3 “diversified assessment methods” is to establish a diversified process assessment and evaluation system, more objectively and scientifically reflect the learning effect of students, and help to cultivate diversified innovative talents.

3 Curriculum Reform of “Sensor and Detection Technology” Based on the Integration of Specialty and Innovation 3.1 Design of Teaching Objectives By learning “sensor and detection technology”, students can master the basic working principles of temperature sensor, humidity sensor, photoelectric sensor, mechanical sensor, etc., master the conversion circuit, error analysis and correction circuit of various sensors, and master the circuit analysis and detection circuit design methods of various sensors. Adhering to the comprehensive talent training objectives of specialty and innovation and entrepreneurship, the design of teaching objectives in the four directions of knowledge, ability, literacy and innovation and entrepreneurship is determined. The course objectives of “Sensor and Detection Technology” are shown in Table 1. The conceptual model of curriculum objectives can be expressed as: 4 4 3 3 Ai + Bj + Ck + Dm (1) Tkcqi = i=1

j=1

k=1

m=1

Teaching Reform and Practice of “Sensor and Detection Technology” Course

1037

Table 1. Course objectives of “sensor and detection technology” Teaching objectives

Concrete content

Knowledge objectives A

A1 -Master the basic method of measurement, the error representation method of sensor measurement data and the data processing method A2 -Familiar with the types, basic structures, working characteristics, working parameters and applications of various sensors A3 -Analyze the working process of various sensor control circuits, and select sensors reasonably according to the design needs A4 -Understand the application of new materials, new processes and new technologies in the field of sensors and detection technology

Capability objectives B

B1 -Recognize and identify various sensors commonly used B2 -Be able to use multi-meter, oscilloscope, signal generator and other instruments to detect the quality of various sensors B3 -Be able to select and design the detection circuit that actually meets the requirements according to the needs B4 - Be able to analyze the typical application circuits and working principles of various sensors

Quality objectives C

C1 -Cultivates students’ patriotic feelings and dialectical thinking ability, and has a rigorous and realistic attitude of seeking knowledge and a craftsman’s spirit of keeping improving C2 -Improves students’ ability to analyze and solve problems, as well as team cooperation C3 -Has the awareness of quality and serious work attitude, establishes a correct outlook on life, and strengthens the sense of social responsibility

Innovation and entrepreneurship objectives D1 -Professional knowledge and innovation D capabilities are integrated. Introduces the case of competitions into the learning of professional knowledge D2 -Has the ability of market research and scheme design, team cooperation and division of labor, unity and cooperation, as well as a certain degree of leadership D3 Cultivate the spirit of innovation and practice, stimulate students’ professional honor and industry pride, and lay a good foundation for work after graduation

1038

D. Lv and Y. Zhang

3.2 Course Design Ideas 3.2.1 The Course Adopts the Project-Based Teaching Method The project-based teaching method used in the teaching of “sensor and detection technology” can not only make students pay attention to the relationship between their living environment and professional knowledge, but also use the knowledge learned to find and solve practical problems, effectively improve their comprehensive qualities such as communication, cooperation and display, and make them become lifelong learners, which is a significant advantage of project-based teaching. The course teaching is guided by engineering practice problems, guided by the teacher’s explanation and analysis, and guided by the theory. The teaching with students’ hands as the main body is a modern and scientific teaching method that focuses on practical skills training and develops specific teaching contents. Under the guidance of the teacher, students can collect knowledge one by one according to the outline requirements designed for the content of each specific experimental project, consult various relevant books and materials, and then analyze and solve the problems existing in the actual problems through in-depth study and cooperation and discussion of research topics among the team members, until all the project tasks are completed, so as to fully meet the requirements of the actual teaching objectives of the course. 3.2.2 Design Idea of Project-Based Teaching In project-based teaching, teaching is generally carried out according to five steps: setting tasks, grouping, dividing tasks, assessing, reviewing and summarizing. The details are as follows: (1) Task setting According to the characteristics of the talent training teaching program and the content requirements of the syllabus, teachers closely combine the teaching content of the theoretical knowledge of the course with the tasks of professional practice to design the teaching tasks, intersperse innovation and entrepreneurship knowledge, and try to integrate, optimize and adjust multiple knowledge points, so as to achieve the dual purpose of strengthening the effective training of students’ skill quality and comprehensive practical application. In the process of independently completing the project tasks, Continuously gain a sense of experience and achievement, and stimulate their learning initiative and independent innovation consciousness. (2) Grouping Teachers can group students according to their wishes and the degree of mastery of course learning, and try to achieve the overall level balance of students in each group. A group leader is selected from each group to be responsible for the implementation process and reporting of the project, so as to ensure the smooth completion of the teaching work. (3) Tasks’ assigning Determine the specific projects of each group and their organization and distribution work, communicate and discuss within the group, propose project solutions based on division of labor and cooperation, and constantly adjust and improve them to obtain specific and effective solutions and complete the group tasks.

Teaching Reform and Practice of “Sensor and Detection Technology” Course

1039

(4) Communication and reporting The plans of each group should be communicated and reported in the whole class, so that they can inspire, learn from each other, comment on and discuss with each other, which is obviously beneficial to the common improvement of all students. (5) Review and summary When summarizing and commenting on the project results of each group of students, the teacher can clarify the relationship between the solution and the knowledge points, inspire students to make horizontal comparison of each group of projects and vertical comparison of the improvement effect of the plan, thus triggering in-depth thinking and training students’ thinking and expression abilities. 3.3 “Sensor and Detection Technology” Teaching Project The “sensor and detection technology” teaching project mainly includes the basic sensor detection technology, temperature detection, pressure detection, material level detection, wave detection, environmental detection and other projects, with a total of 64 class hours. Its main content and class hour distribution are shown in Table 2. The course project of “Sensor and Detection Technology” is based on the familiar projects of students, integrates the knowledge points and experimental contents of traditional theoretical teaching, combines theory with practice, implements students’ innovation and entrepreneurship in place, and trains students’ innovation ability to achieve twice the result with half the effort. 3.4 Assessment of Students’ Learning Effect Before the reform of “sensor and detection technology”, the total score = attendance score (5%) + classroom score (25%) + final exam score (70%). After the reform, the project is based on the project, and each project is designed based on the work process. The assessment method is shown in Table 3. In the previous teaching class, the subjective score of teachers is relatively high, and the proportion of final score is high, which is easy to lead to poor final exam scores of students who usually learn well, so the total score obtained deviates greatly from the actual learning effect of students. “The course of sensor and detection technology” is specially created for integrated teaching, with students occupying the dominant position and playing a leading role, focusing on the process assessment of project teaching, and viewing students’ learning achievements from multiple perspectives, which has great advantages.

1040

D. Lv and Y. Zhang

Table 2. “Sensor and detection technology” teaching items and class hour allocation table No

Project

Main contents

Hours

1

Item 1 Fundamentals of sensor detection technology

-Fundamentals of sensor and detection 8 technology -Measurement error and data processing -Data processing cases

2

Item 2 Temperature detection

-Thermocouple principle and detection circuit analysis -Thermal resistance principle and detection circuit analysis -Principle and detection of thermistor -Design of other temperature sensors and temperature detection circuits

12

3

Item 3 Pressure detection

-Principle and detection of strain pressure sensor -Principle and detection of piezoelectric pressure sensor -Principle and detection of capacitive pressure sensor -Principle and detection of inductive pressure sensor -Design of pressure detection circuit

12

4

Item 4 Speed detection

-Magnetoelectric induction sensor -Eddy current sensor -Hall sensor -Semiconductor magnetic sensor -Overview of photoelectric sensors -Photoelectric effect and typical devices -Analysis and Design of Speed Detection Circuit

12

5

Item 5 Detection of sound waves

-Overview of wave sensor -Ultrasonic testing application -Infrasound detection application -Design of ultrasonic testing circuit

10

6

Item 6 Environmental detection

-Gas sensor and its application -Humidity sensor and its application -Environment detection circuit design

10

Teaching Reform and Practice of “Sensor and Detection Technology” Course

1041

Table 3. Post-reform evaluation table of “sensor and detection technology” course Evaluation content

Evaluation items

Evaluation method

Proportion

Process assessment

Attendance and discipline

“Learning Pass” sign-in

5%

Team cooperation

Self and inter-group evaluation

5%

Project research and scheme design

Scheme evaluation

10%

Leadership and communication

Self and inter-group evaluation

10%

Basic professional knowledge

Quiz

10%

Knowledge application and Quiz creation

10%

Individual contribution

Self and inter-group evaluation

5%

Language expression and project analysis

PPT, speech, evaluated by group leaders

10%

Project results

Project results judged by team leaders

15%

Final exam

Teachers evaluate according to the test

20%

Result assessment

4 Achievements of the Current Stage of Innovation and Innovation Integration Reform 4.1 Significant Improvement of Students’ Course Performance Since the implementation of the innovation and innovation integration curriculum reform in the second semester of 2020, the curriculum content of “sensor and detection technology” has been restructured, the project-based teaching method has been adopted, the innovation and entrepreneurship content has been integrated, and the process assessment has been emphasized. The students’ learning achievements have been significantly improved. Figure 2 shows the total results of electrical engineering and its automation, mechanical design and manufacturing and its automation in the past three years. It can be seen that after the reform, the achievements of the two majors have generally improved by more than 5 points compared with the previous ones, with obvious progress. 4.2 Achievements in Innovation and Entrepreneurship Competitions have Improved Significantly Nanning University attaches great importance to the teaching reform of the integration of specialty and innovation. In the past three years, it has approved 60 projects of the

1042

D. Lv and Y. Zhang

Fig. 2. Three years’ scores in “sensor and detection technology” of two majors

integration of specialty and innovation, which has effectively improved the quality of talent training and achieved excellent results. In particular, outstanding achievements have been made in the establishment of Innovation and entrepreneurship training program for college students (abbreviated as IET) and the International “Internet + ” Undergraduate Innovation and Entrepreneurship Competition Guangxi Division (referred to as “INT + IEP” for short). The project approval and awards are shown in Table 4. Table 4. Awards of IET and “INT + IEP” Year

IET State level

IET Provincial level

“INT + IEP” Guangxi Provincial Competition Gold award

Silver award

Bronze award

2020

38

93

3

8

30

2021

39

103

14

26

31

2022

44

122

10

17

68

As can be seen from Table 4, the number of major founding projects and Internet plus awards are increasing year by year, as well as national level projects. Through the implementation of the integration of specialty and innovation teaching, the students’ professional level and innovation and entrepreneurship literacy have been continuously improved, and the new type of entrepreneurship and innovation talents have been growing. The course teaching team guided the students to achieve good results in the innovation and entrepreneurship competition, and the students obtained dozens of achievements and awards in the national and regional competitions of the above two major events. 4.3 Stimulate Graduates’ Entrepreneurial Enthusiasm Through the teacher’s classroom teaching and related content experiments, as well as the promotion of project teaching method, students have accumulated a lot of entrepreneurial experience while learning knowledge, which provides very effective help for their independent entrepreneurship. For example, through the environmental detection project,

Teaching Reform and Practice of “Sensor and Detection Technology” Course

1043

students not only acquired the relevant knowledge of environmental humidity and environmental gases, but also grasped the market demand of environmental detection instruments and the circuit design scheme of environmental measurement instruments. Among the graduates, 2 students were engaged in relevant entrepreneurial work, and achieved good results. The turnover of students’ entrepreneurial companies in the first year of selling environmental detection instruments alone exceeded 500,000 yuan. Students have accumulated a lot of entrepreneurial experience by using in-class projects for reference, which has also stimulated the entrepreneurial enthusiasm of graduates.

5 Conclusion The teaching reform of “sensor and detection technology” based on the integration of specialty and innovation, taking students’ project practice ability and innovation consciousness as the foothold of talent cultivation, has not only improved students’ theoretical knowledge and practical ability, but also improved students’ innovation ability. By adopting the project-based teaching method, the transformation of the main role of classroom teaching has been realized, so that each student has always played the main and leading role, and the rapid improvement of their comprehensive quality has been continuously realized, and the students’ internal enthusiasm for course learning, experiment and practice has been enhanced, and the expected teaching purpose has been achieved. At the same time, the overall teaching level of the school has been rapidly, effectively and comprehensively improved. The teaching practice has proved that the teaching reform of the “sensor and detection technology” course based on the integration of specialty and innovation has optimized the teaching content around the requirements of the deep integration of “professional education” and “innovation and entrepreneurship education”, which is more in line with the requirements of education and teaching for professional courses, at the same time, it is conducive to students mastering the application of the curriculum in the specialty and the frontier of professional development, and cultivating students’ innovative thinking. Acknowledgment. This project is supported by the third batch of curriculum reform project of the integration of specialty and innovation of Nanning University “Teaching reform and practice of ‘sensor and detection technology’ based on the integration of specialty and innovation” (2021XJZC04).

References 1. Anayat, S., Butt, S., Zulfiqar, I., Butt, S.: A deep analysis of applications and challenges of wireless sensor network. International Journal of Wireless and Microwave Technologies (IJWMT), 10(3), 32–44 (2020) 2. Kori, G.S., Kakkasageri, M.S.: Game theory based resource identification scheme for wireless sensor networks. International Journal of Intelligent Systems and Applications (IJISA) 14(2), 54–73 (2022) 3. Altay, A., Learney, R., Güder, F., et al.: Sensors in blockchain. Trends in Biotechnology 40(2), 141–144 (2022)

1044

D. Lv and Y. Zhang

4. Ashraf, S., Saleem, S., Ahmed, T.: Sagacious communication link selection mechanism for underwater wireless sensors network. International Journal of Wireless and Microwave Technologies (IJWMT) 10(4), 22–33 (2020) 5. Kawamoto, K., Shiomi, H., Ito, T., et al.: Vector sensor imaging. Optics and Lasers in Eng. 162, 107439 (2023) 6. Nair, T., Singh, A., Venkateswarlu, E., et al.: Generation of analysis ready data for indian resourcesat sensors and its implementation in cloud platform. International Journal of Image, Graphics and Signal Processing (IJIGSP), 11(6), 9–17 (2019) 7. Shia, H.H., Tawfeeq, M.A., Mahmoud, S.M.: High rate outlier detection in wireless sensor networks: a comparative study. International Journal of Modern Education and Computer Science (IJMECS), 11(4), 13–22 (2019) 8. Puranikmath, V.I., Harakannanavar, S.S., Kumar, S.: Dattaprasad Torse. Comprehensive study of data aggregation models, challenges and security issues in wireless sensor networks. International Journal of Computer Network and Information Security (IJCNIS), 11(3), 30–39 (2019) 9. Rambabu, C., Prasad, V.V.K.D.V., Prasad, K.S.: Multipath cluster-based hybrid mac protocol for wireless sensor networks. International Journal of Wireless and Microwave Technologies (IJWMT), 10(1), 1–16 (2020) 10. Chunjiao, Z.: Teaching design of sensor and detection technology course based on OBE concept. China Modern Education Equipment 11, 88–90 (2022). (in Chinese) 11. Chen, J.: Construction of practical training course assessment system for applied electronic technology specialty based on OBE concept -- taking the Comprehensive Practical Training Course for Sensors as an example. Computer Knowledge and Technology, 18(33): 115–118 (2022). (in Chinese) 12. Zebin, L.: Thinking on the teaching reform of sensor principle and application. Science and Technology Vision 34, 190–191 (2021). (in Chinese) 13. Siqi, W.: Teaching design of “sensor working principle and application” based on STEM theory. Physics Teaching Discussion 40(08), 30–35 (2022). (in Chinese) 14. Guo, H.: Research on the construction mode of open courses based on OBE concept -- take the sensor course of photovoltaic specialty as an example. Nanfang Agricultural Machinery, 53(16), 195–198 (2022). (in Chinese) 15. Cao, X., Chen, Y.: Practical research on the course “Fundamentals of Sensor and Engineering Test Technology” based on “project-based teaching”. Theoretical Research and Practice of Innovation and Entrepreneurship, 5(23), 162–165 (2022). (in Chinese) 16. Bai, Y., Liu, G., Chuai, Y., et al.: First-class course construction of detection technology and sensors under OBE concept. Metallurgical Manage. (19), 75–76 (2021). (in Chinese) 17. Sun, L., Wan, L., Wang, Y., Wang, X.: Research on BOPPPS teaching mode under the background of integration of specialty and innovation - taking the analogical innovation method in innovative thinking and innovation method as an example. Science and Education, (24), 79–83 (2022). (in Chinese) 18. Yang, K.: Exploration and practice of the integration of innovation and entrepreneurship education and professional education in the context of “special innovation”. Heilongjiang Education (Higher Education Research and Evaluation), (12), 42–44 (2022). (in Chinese) 19. Wu, J., Yang, B.: Analysis on the teaching reform of “product design” course based on the integration of specialty and innovation. Intelligence (35), 60–63 (2022). (in Chinese) 20. Geng, Y., Zhang, R.: Construction and implementation of specialized and creative chemistry curriculum system under the mechanism of “three complete education”. Science and Education Guide, (17), 33–35 (2022). (in Chinese)

Teaching Reform and Practice of “Sensor and Detection Technology” Course

1045

21. Liu, S.: Research on the teaching reform of Research Travel Course Design based on the integration of specialty and innovation. In: Proceedings of 2022 the 6th International Conference on Scientific and Technological Innovation and Educational Development, pp. 568–570 (2022) 22. Zhou, H., Zhang, W., Qin, H.: Research and practice on the “student-centered” teaching reform of specialized and creative integration in higher vocational colleges. Modern Vocational Education, (32), 121–123 (2022). (in Chinese) 23. Wu, C., Fei, F., Yao, E., et al.: Course reform practice of sensor principle for engineering education professional certification and training students’ ability to solve complex engineering problems. China Modern Education Equipment, (21), 70–72 (2021). (in Chinese) 24. Han, Y.: Reflection and reconstruction of the curriculum system of “integration of specialty and innovation” in higher vocational colleges. Jiangsu Higher Education, (12), 122–127 (2022). (in Chinese) 25. Hu, D.: Exploration of teaching reform in split class in the course of “Sensor Principle and Application”. Science and technology wind, (20), 100–102 (2022). (in Chinese) 26. Zhao, C., Wu, Y., Fan, K., et al.: Optimization of the education model of application-oriented universities from the perspective of “integration of specialty and innovation”. Journal of Texas University, 38(06), 107–110 (2022). (in Chinese) 27. Li, X., Zhang, J., Zhang, H., et al.: Teaching reform of “sensor” course under the new situation. Industry and Information Education, (12), 69–72 (2022). (in Chinese) 28. Wang, H.: Survey and thinking on entrepreneurship education of local college students. International Journal of Education and Management Engineering (IJEME), 2(7), 25–30 (2012)

Improved Method for Grading Bilingual Mathematics Exams Based on Computing with Words Danylo Tavrov(B) and Liudmyla Kovalchuk-Khymiuk National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, 37 Peremohy Avenue, Kyiv 03056, Ukraine [email protected]

Abstract. As the Content and Language Integrated Learning approach to studying subjects such as mathematics through foreign languages gains more recognition, the problem of assessing students’ performance in bilingual settings remains largely understudied. In this work, we propose an improved method for grading mathematics exams using words that takes into account correctness, optimality, thoroughness, and tidiness of solutions written by students. We outline the general multilayer framework for grading bilingual mathematics exams, where assessment at each layer is computed as a fuzzy set. Our approach is an improvement upon previously published methods and considers four main competences that a student should obtain. Applicability of our method is illustrated using model data. Keywords: bilingual education · CLIL · computing with words · perceptual computing · type-2 fuzzy set

1 Introduction In a modern globalized world, the idea that students should learn at least one foreign language as early as possible has gained wide popularity. One of the forms of such learning is known by the name of Content and Language Integrated Learning (CLIL), broadly defined as [1] the educational approach, whereby learners study content subjects (computer science [2], mathematics, history, biology, and so on) in a foreign language. The idea is that students acquire knowledge of a content subject and at the same time master language competences. This idea has proven to be very effective, which is why the European Council recommended [3] that, among other provisions, CLIL practices be further implemented. Some studies indicate [4, 5] that learners in bilingual classes were more motivated than those who studied through their primary language, and that CLIL has a positive impact on mathematical performance. In general, learning in a foreign language is considered an advantage [6] to students. In this work, we discuss the problem of assessing students’ performance in bilingual setting. CLIL is not equivalent to learning a foreign language. Subjects such as mathematics are taught not in, but rather with the help of a foreign language. This means that, © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 1046–1056, 2023. https://doi.org/10.1007/978-3-031-36118-0_89

Improved Method for Grading Bilingual Mathematics Exams

1047

in general, such assessment should [1] be carried out in a foreign language, but with the intent of testing knowledge gained in the subject being studied. At the same time, importance of the foreign language should not be downgraded. Both the content subject and the foreign language are involved equally [7] in defining learning goals. Proper assessment in such classes should thus take into account linguistic competences, subject competences, sociolinguistic competences, and what can be called soft skills competences (work in a team, debates, and so on). In this work, given expertise of the second author, we focus on teaching mathematics in French for students whose primary language is Ukrainian. Therefore, emphasis is put on linguistic and subject (mathematics) competences. In [8], the authors discuss a framework for assessing both subject and language dimensions of students’ work. The authors split subject questions into recall, application, and analysis tiers depending on their cognitive requirements. Linguistic requirements of a given task are treated at three levels—vocabulary, sentence, and text. However, teaching mathematics in a bilingual setting is complicated by the fact that [9] students perform code-switching, i.e. they switch languages during arithmetic computation and conversation. As a result, there do not currently exist assessments for language proficiency in the context of mathematical competence, for different age groups and topics. Further complications arise from the fact that the overall grade cannot be computed as an arithmetic sum of mathematical and linguistic parts. Even though some authors claim otherwise [10], it is our conviction that every problem on the exam touches upon several competences at once. Moreover, assigning specific numeric grades to problems on an exam is highly subjective, not in the sense that the teacher is not impartial, but in the sense that she is typically assigning grades based on her experience. Even when rough guidelines exist, the final call is upon the teacher who needs to consider not only correctness of a solution but also how optimal it is, how thorough the explanation is, how neat and tidy a student’s work is, and so on. In this paper, we propose to grade exams using fuzzy sets to capture uncertainty and subjectivity involved. Fuzzy sets have been applied in other works [11, 12], but we propose an improvement of our previous assessment framework [13] tailored to grading mathematical exams for Ukrainian students studying mathematics in French. We use a computing with words (CWW) [14] approach that enables the teacher to use words (expressed as fuzzy sets), making her work user-friendly and allowing to minimize subjectiveness by treating imprecision related to linguistic concepts in a formal way. The paper is structured as follows. In Sect. 2, we describe our framework for grading bilingual mathematics tests and discuss modelling linguistic concepts using fuzzy sets. Section 3 gives a version of an exam that can be graded using our approach, and results for synthetic data are provided to illustrate its validity. Section 4 concludes.

2 CWW Based Method for Grading Mathematics Exams 2.1 Grading Process Outline Let us consider a mathematics exam of N problems that need to be graded (a concrete example will be shown in Sect. 3). In accordance with the above discussion, we propose to obtain an overall grade according to a decision-making process shown in Fig. 1.

1048

D. Tavrov and L. Kovalchuk-Khymiuk

Such decision making is called hierarchical and distributed [15], because it involves aggregating independent grades for individual problems in an hierarchical fashion.

Fig. 1. Hierarchical and distributed grading of mathematical exams

A teacher evaluates each problem according to four criteria: – correctness (C1). In some cases, a solution can be unequivocally correct, and in others, it can be “mostly” correct (e.g., when all the steps of the solution are correct save for some minor arithmetic mistakes); – optimality (C2). For problems, where it is appropriate, a chosen method is typically either optimal or not at all. We suggest grading this criterion in a binary way; – thoroughness (C3). To assess linguistic competences of a student, it is important to consider explanations they provide, which can only be assessed subjectively; – tidiness (C4) of a solution, including its legibility and structure. In general, these criteria are evaluated using words, as they are mostly subjective and hard to formalize. In some cases, it is acceptable to use numbers 0 (if an answer is absent or it is not optimal) or 1 (if an answer is absolutely correct). At each node in Fig. 1, inputs are incoming from the lower-level nodes connected to it, and the output is computed as a linguistic weighted average (LWA) discussed below. Improving upon our previous findings in [11], we consider the following competences: solving a problem (competence 1), understanding a problem (competence 2), explaining a chosen solution (competence 3), and ability to integrate language and mathematics (competence 4). The last competence is evaluated directly by the teacher based on the overall impression from the exam being graded. Arcs connecting nodes are weighted, also with words (or in some cases with numbers 0 or 1), to allow for different characteristics to have different effects on the overall grade. Weights can be assigned by independent experts, which makes the process of grading more objective. The teacher does not know which criteria and competences are more important, and thus cannot influence the overall result, even subconsciously.

Improved Method for Grading Bilingual Mathematics Exams

1049

2.2 Aggregating Grades Using Linguistic Weighted Average To calculate a weighted average of several words, we use perceptual computing described in [15] as the most detailed and developed approach to date. To average words, we need to represent them as mathematical objects. Following [16], we will use (interval) type-2 fuzzy sets (T2FS) as scientifically correct from the point of view of modeling first-order uncertainty of words. Type-2 fuzzy sets were originally introduced [17] and further developed by J. M. Mendel (see, e.g., [18] for an extensive treatment of this subject). All the words that can be used by a teacher in the system, together with their T2FS models, collectively make up a codebook. A T2FS A is a set, to which any element can belong with some membership degree, expressed as an interval. For instance, numeric grade 0.9 can belong to a T2FS “very good” with membership [0.85; 0.95]. Unlike type-1 fuzzy sets, for which only one membership value is required for each number, using intervals enables the researcher to capture the fact that different people understand the same linguistic term differently. We can represent A using two membership functions: a lower membership function (LMF), μA˜ : X → [0; 1], and an upper membership function (UMF), μA˜ (x) : X → [0; 1], where X is some universe of discourse. We take X = [0; 1]. Then we can say that any value x ∈ X belongs to A with an interval degree expressed as [μA˜ (x); μA˜ (x)]. In applications, it is customary to use trapezoidal membership functions: ⎧ x−a ⎪ ,a≤x≤b ⎪ ⎪ ⎨ b−a 1, b≤x≤c μ(x; a, b, c, d ) = d −x (1) ⎪ ,c≤x≤d ⎪ ⎪ d −c ⎩ 0 , otherwise If LMF and UMF are trapezoidal, we obtain a T2FS from Fig. 2. Numbers 0 (or 1) are represented as T2FSs by setting a = b = c = d = 0 (or 1) for both LMF and UMF. To compute weighted average of several real numbers x 1 , …, x n , with associated real positive weights w1 , …, wn , one would compute n 

y=

xi wi

i=1 n 

(2) wi

i=1

When values X 1 , …, X n and their associated weights W 1 , …, W n are expressed as T2FSs, (2) can be technically rewritten as n 

Y =

Xi Wi

i=1 n 

(3) Wi

i=1

This expression is computed as a series of interval weighted averages [15].

1050

D. Tavrov and L. Kovalchuk-Khymiuk

2.3 Assigning a Final Grade After the words are assigned by a teacher for each criterion of each problem, and they are averaged according to the connections depicted in Fig. 1, one obtains a T2FS G at the “Final grade” node. There are several options that the teacher has at this point: – she can select a word from the codebook that has the most similar T2FS; – she can rank the students according to their T2FSs; – she can assign G to one of the predefined classes corresponding to different grades.

Fig. 2. Trapezoidal interval type-2 fuzzy set

Following [11], we proceed according to the third option and assign one of the four grades expressed as intervals: A = [0.75; 1], B = [0.5; 0.75], C = [0.25; 0.5], or D = [0; 0.25]. We do it using a well-known Vlachos and Sergiadis’s T2FS subsethood measure [19] adapted to our case as follows: N 

ssVS (G, I ) =

i=1

min(μG (xi ), 1I (xi )) + N  i=1

μG (xi ) +

N  i=1 N 

min(μG (xi ), 1I (xi )) (4) μG (xi )

i=1

where x i ∈ X, i = 1, …, N, are equally spaced points of the support of G, I ∈ {A, B, C, D}, and 1I is the indicator function. We select the class that yields the highest value of (4). Classes can be defined differently depending on a specific application.

Improved Method for Grading Bilingual Mathematics Exams

1051

3 Application of CWW to Grading Mathematical Exam 3.1 Structure of the Exam In this section, we present an example of a mathematical exam that enables us to illustrate the strongest features of the CWW applied to grading. This exam was created by the second author of this paper as a practicing educator in the field of bilingual mathematical education in French and Ukrainian. 1. En utilisant le tableau des dérivées des fonctions usuelles, calculer [Using the table of derivatives of the common functions,√ calculate]: (a) y = sin x – cos x, x 0 = π (b) y = –8 x – x –1 , x 0 = 1 2. Soient C et C’ les courbes d’équations respectives, y = x 3 – 2x + 3 et y = 2x 2 – 3x + 3 [Let C and C’ be the curves described by two equations]: (a) Déterminer les coordonnées des points communs à C et C’ [determine the coordinates of their common points] (b) Déterminer les équations des tangentes à C et C’ en chacun de leurs points communs [derive tangent lines to C and C’ at their common points] 3. Calculer la dérivée des fonctions définies par les expressions ci-dessous en utilisant les règles de dérivations et en simplifiant le résultat [Calculate the derivative of the functions defined by the expressions below using the rules of derivation and simplifying the result]: (a) f (x) = ln((3x 2 – 4)3 ) (b) f (x) = (x 2 + 3)e2x – 5 4. Mettre les signes convenables [Insert correct sign]: Soit une fonction f définie et dérivable sur un intervalle I [Let f be a function defined and differentiable on an interval I]. Si [If] f’ (x) ? 0, alors f est décroissante sur I. [ f is decreasing on I] Si f’ (x) ? 0, alors f est croissante sur I. [The same for “increasing”] Si f’ (x) ? 0, alors f est constante sur I. [The same for “constant”] 5. Étudier complètement la function f (x) = (–x 2 + 3) / (x + 2). Pour cela parcourir les étapes ci-dessous [Completely analyze the function f . Follow the steps]: (a) Déterminer le domaine de définition de f [Determine its domain] (b) Déterminer les points d’intersections de f avec les axes [Determine points of intersection with the axes] (c) Déterminer les équations de ses asymptotes [Derive equations of its asymptotes] (d) Déterminer la dérivée de f . Puis déterminer les coordonnées des extremums de f . [Find its derivative and points of extrema] 6. Esquisser soigneusement le graphique de la fonction f (x) = (–x 2 + 3) / (x + 2) [Sketch the graph of f ] 7. Déterminer une primitive de la fonction f lorsque celle-ci est définie par [Find antiderivative]: √ (a) f (x) = (x – 1) / (x 2 – 2x)3 (b) f (x) = x2 2x3 − 3 (c) f (x) = 6 cos2 (4x) sin x 8. Dire si les affirmations suivantes sont vraies ou fausses. Justifier chaque réponse [True of False, justify]: (a) Toute fonction intégrable possède une unique primitive [Antiderivatives are unique]

1052

9.

10.

11.

12.

13.

D. Tavrov and L. Kovalchuk-Khymiuk

(b) Le résultat du calcul d’une intégrale définie est toujours positif [Definite integral is always positive] Calculer les intégrales suivantes [Compute the following integrals]: 2  π/2 3 2 xdx dx (b) −1 5xex −1 dx (c) π/6 3 cos (a) 1 3x23x+1 sin x +2x+2  On considère une fonction f affine par morceaux. Calculer f (t) dt en utilisant deux méthodes [Compute an integral of a given piecewise affine function using two methods] Complete le Théorème (Cauchy): Si la fonction f est—sur [a; b] alors est—sur [a; b] [Complete the Cauchy theorem: of a function f is [blank] on [a; b], it is [blank] on [a; b]] Dans un ZOO, on veut construire 3 enclos rectangulaires de même taille délimités par les clôtures dégrillages, selon du plan ci-dessous. L’aire totale des enclos doit être de 288 m2 . Pour minimiser les frais de clôture, quelles doivent être les dimensions de chaque enclos pour la longueur totale des clôtures utilisées soit minimale ? [In a zoo, we want to build 3 rectangular enclosures of the same size delimited by the fences, according to a given plan. The total area should be 288 m2 . Find the dimensions of each enclosure, so that the total length of fencing is minimal.] Soit f et g deux fonctions définies respectivement par f (x) = x 3 – x + 3 et g(x) = 2x + 1 [Consider two functions]: (a) Calculer les coordonnées des points d’intersection du graphe de f avec celui de g [Find points of intersection] (b) Calculer l’aire de la surface plane hachurée délimitée par les graphes de f et g [Compute the area between two graphs]

3.2 Construction of Codebook and Assignment of Weights For the method discussed here, the codebook contains words of two types: – words used to evaluate problems. Following [11, 13], we use eight words: excellent (EX) very good (VG), good (G), sufficient (S), satisfactory (SA); – words used as weights on connecting arcs in Fig. 1. Following [11, 13], we use five words: highly influential (HI), influential (I), moderately influential (MI), weakly influential (WI), not influential (NI). All the words are represented as T2FSs with support in [0; 1]. Each T2FS was created using a person FOU approach [20] that requires knowledge elicitation from one expert. The second author of this paper provided appropriate data for this method based on her extensive expertise in this domain. Resulting T2FSs are given in Table 1 (rounded to two significant digits) in reference to (1). All weights on the arcs from Fig. 1 were also assigned by the second author of this paper. Weights on the arcs connecting criteria, problems, and competences are given in Table 2. Competences are connected with the final grade node as follows: competence 1 has weight I, and other competences have weight HI.

Improved Method for Grading Bilingual Mathematics Exams

1053

3.3 Discussion of the Results for Synthetic Students Validity of the proposed approach can be illustrated using model data corresponding to three synthetic students. Grades for each problem are given in Table 3. Grades for competence 4 were assigned as follows: student 1 got S, and students 2 and 3 got 1. Final grades for each student are given in Table 4 as T2FSs. Also, we show subsethood (4) for each of the four classes (A, B, C, D). Based on these results, a teacher would give a B to the first student, and A to students 2 and 3. Table 1. Parameters of T2FSs used to represent words from the codebook Word

LMF

UMF

a

b

c

d

a

b

c

d

EX

0.96

0.97

0.99

1.00

0.96

0.97

0.99

1.00

VG

0.84

0.85

0.91

0.93

0.83

0.85

0.91

0.94

G

0.76

0.77

0.80

0.81

0.75

0.77

0.80

0.81

S

0.66

0.67

0.71

0.72

0.65

0.67

0.71

0.73

SA

0.56

0.57

0.61

0.62

0.55

0.57

0.61

0.63

BA

0.47

0.48

0.52

0.53

0.46

0.48

0.52

0.53

U

0.37

0.38

0.42

0.43

0.37

0.38

0.42

0.44

VB

0.12

0.23

0.32

0.33

0.05

0.23

0.32

0.34

HI

0.94

0.95

0.99

0.99

0.94

0.95

0.99

1.00

I

0.83

0.85

0.90

0.91

0.82

0.85

0.90

0.92

MI

0.71

0.73

0.78

0.79

0.70

0.73

0.78

0.80

WI

0.59

0.61

0.66

0.67

0.58

0.61

0.66

0.69

NI

0.26

0.41

0.54

0.55

0.21

0.41

0.54

0.57

As can be seen from these model results, a teacher is able to arrive at final grades using words and binary numbers to grade criteria of each problem on the exam. The results adhere to general expectations, in the sense that the students who get higher grades for individual problems get higher overall grades as well. The case of Student 1 is important, as it illustrates that even getting very substandard grades for tidiness (criterion 4) does not lower the overall grade very much. This is explained by the fact that criterion 4 is considered not very influential according to Table 2 for most problems. Weights of each criterion can be assigned not by a teacher, but by her supervisor, which makes the process of grading more objective, as the teacher does not have access to the inner workings of the system.

1054

D. Tavrov and L. Kovalchuk-Khymiuk Table 2. Weights on arcs connecting nodes in the framework

Problem

C1

C2

C3

C4

Comp. 1

Comp. 2

Comp. 3

Problem 1

HI

MI

MI

WI

HI

I

NI

Problem 2

HI

I

I

NI

HI

HI

I

Problem 3

HI

MI

MI

WI

HI

I

I

Problem 4

HI

0

0

0

NI

HI

NI

Problem 5

HI

I

HI

I

HI

HI

I

Problem 6

HI

I

I

HI

HI

HI

I

Problem 7

HI

MI

MI

WI

HI

I

NI

Problem 8

HI

I

HI

I

I

HI

HI

Problem 9

HI

MI

MI

WI

HI

I

NI

Problem 10

HI

MI

I

MI

HI

I

WI

Problem 11

HI

0

0

0

NI

HI

NI

Problem 12

HI

HI

HI

WI

HI

HI

HI

Problem 13

HI

I

I

I

HI

HI

MI

Table 3. Grades for synthetic students Problem

Student 1

Student 2

Student 3

C1

C2

C3

C4

C1

C2

C3

C4

C1

C2

C3

C4

1

EX

1

G

U

EX

1

VG

G

G

1

S

S

2

VG

1

G

U

EX

1

G

G

G

1

G

G

3

G

0

VB

U

G

0

G

G

EX

1

EX

G

4

0

0

VB

U

1

0

G

G

1

0

G

G

5

G

1

VB

VB

VG

1

VG

VG

G

1

VG

VG

6

S

1

VB

VB

EX

1

EX

VG

EX

1

EX

VG

7

G

1

G

U

G

1

G

SA

G

1

G

G

8

G

1

VB

VB

EX

1

VG

VG

VG

1

VG

VG

9

0

0

BA

U

1

0

G

G

1

0

G

G

10

VG

1

G

U

VG

1

G

G

EX

1

G

G

11

0

0

G

U

1

0

G

G

1

0

G

G

12

S

0

G

U

EX

1

VG

G

EX

1

VG

G

13

G

0

VB

VB

G

0

G

G

EX

1

EX

G

Improved Method for Grading Bilingual Mathematics Exams

1055

Table 4. Parameters of T2FSs for final grades for synthetic students and degrees of subsethood Student

Grade LMF

Grade UMF

Subsethood

a

b

c

d

a

b

c

d

A

B

C

D

1

0.49

0.52

0.58

0.61

0,47

0.52

0.58

0.62

0

0.95

0.05

0

2

0.86

0.87

0.90

0.91

0.85

0.87

0.90

0.92

1

0

0

0

3

0.90

0.91

0.93

0.94

0.89

0.91

0.93

0.94

1

0

0

0

4 Conclusions and Further Research In this paper, we presented an improved method for grading bilingual mathematics exams using computing with words. The field of assessing students’ performance in bilingual settings remains largely understudied, and we aimed to fill the gap by proposing a method that would allow teachers to grade bilingual mathematics exams in a friendly and coherent way. Our method differs from existing ones in two respects. First, it enables a teacher to use words in place of arbitrary and hard-to-determine numeric grades. Grading is very subjective, therefore computing with words helps qualify subjectiveness by formalizing uncertainty using type-2 fuzzy sets. Second, grading is organized in several layers. The teacher is responsible for evaluating correctness, optimality, thoroughness, and tidiness of solutions. These evaluations are aggregated using relative influence of each problem on each of the four core competences, and of each competence on the overall grade. By specifying relative weights of each problem and competence, it is possible to separate the grading process from the determination of the outcome, which makes assessment more objective. Our approach can be developed by adding more words to the codebook to make it more nuanced. Field experiments involving teachers and students should be conducted to investigate the benefits and improvements in grading that our approach can offer.

References 1. Eurydice Report. Content and Language Integrated Learning (CLIL) at School in Europe. Brussels (2006) 2. Xingle, F., Zhaoyun, S., Yan, C., Yupu, B.: The exploration and research on bilingual education of computer discipline. IJEME 2(10), 52–58 (2012) 3. Council of the European Union: Council Recommendation of 22 May 2019 on a comprehensive approach to the teaching and learning of languages. Off. J. Eur. Union 62, 15–22 (2019) 4. Mearns, T., de Graaff, R., Coyle, D.: Motivation for or from bilingual education? A comparative study of learner views in the Netherlands. Int. J. Biling. Educ. Biling. 23(6), 724–737 (2020) 5. Surmont, J., Struys, E., Van Den Noort, M., Van de Craen, P.: The effects of CLIL on mathematical content learning: a longitudinal study. Stud. Second Lang. Learn. Teach. 6(2), 319–337 (2016) 6. Wang, L., Han, X., Li, M.: Research on bilingual teaching of graduates for computer specialty in financial and economical colleges. IJMECS 1(1), 53–59 (2009)

1056

D. Tavrov and L. Kovalchuk-Khymiuk

7. Tardieu, C., Dolitsky, M.: Impact of the CEFT on CLIL. Integrating the task-based approach to CLIL teaching. In: Agudo, J., de, D.M. (ed.) Teaching and Learning English Through Bilingual Education. Cambridge Scholars (2012) 8. Lo, Y.Y., Fung, D.: Assessments in CLIL: the interplay between cognitive and linguistic demands and their progression in secondary education. Int. J. Biling. Educ. Biling. 23(10), 1192–1210 (2020) 9. Moschkovich, J.N.: Bilingual/multilingual issues in learning mathematics. In: Lerman, S. (ed.) Encyclopedia of Mathematics Education, pp. 75–79. Springer, Dordrecht (2020) 10. Massler, U., Stotz, D., Queisser, C.: Assessment instruments for primary CLIL: the conceptualisation and evaluation of test tasks. LLJ 42(2), 137–150 (2014) 11. Tavrov, D., Kovalchuk-Khymiuk, L., Temnikova, O., Kaminskyi, N.-M.: Perceptual computer for grading mathematics tests within bilingual education program. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2018. AISC, vol. 754, pp. 724–734. Springer, Cham (2019). https://doi.org/10.1007/978-3-319-91008-6_71 12. Liu, Y., Zhang, X.: Evaluating the undergraduate course based on a fuzzy AHP-FIS model. IJMECS 12(6), 55–66 (2020) 13. Tavrov, D., Kovalchuk-Khymiuk, L., Temnikova, O., Kaminskyi, N.-M.: Evolutionary algorithm for fine-tuning perceptual computer for grading mathematics tests within bilingual education program. In: Shahbazova, S.N., Kacprzyk, J., Balas, V.E., Kreinovich, V. (eds.) Recent Developments and the New Direction in Soft-Computing Foundations and Applications. SFSC, vol. 393, pp. 173–187. Springer, Cham (2021). https://doi.org/10.1007/978-3030-47124-8_15 14. Zadeh, L.A.: Fuzzy logic = computing with words. IEEE Trans. Fuzzy Syst. 4(2), 103–111 (1996) 15. Mendel, J.M., Wu, D.: Perceptual Computing: Aiding People in Making Subjective Judgments. John Wiley & Sons, Inc., Hoboken, New Jersey (2010) 16. Mendel, J.M.: Computing with words: zadeh, turing, popper and occam. IEEE Comput. Intell. Mag. 2(4), 10–17 (2007) 17. Zadeh, L.A.: The concept of a linguistic variable and its application to approximate reasoning1. Inf. Sci. 8, 199–249 (1975) 18. Mendel, J.M.: Type-2 fuzzy sets and systems: an overview. IEEE Comput. Intell. Mag. 2(1), 20–29 (2007) 19. Vlachos, I., Sergiadis, G.: Subsethood, entropy, and cardinality for interval-valued fuzzy sets—an algebraic derivation. Fuzzy Sets Syst. 158, 1384–1396 (2007) 20. Mendel, J.M., Wu, D.: Determining interval type-2 fuzzy set models for words using data collected from one subject: person FOUs. In: 2014 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), pp. 768–775 (2014)

Education System Transition to Fully Online Mode: Possibilities and Opportunities Md. Asraful Haque1(B) , Tauseef Ahmad2 , and Shoaib Mohd1 1 Department of Computer Engineering, Aligarh Muslim University, Aligarh 202002, India

[email protected] 2 Department of Information Technology, Rajkiya Engineering College, Azamgarh 276201,

India

Abstract. Digital transformation is one of the main objectives of educational institutions nowadays. Institutions may gain from the digital revolution by enhancing teaching and learning and their ability to run their operations more effectively. Both of these factors are crucial for better serving students. Online education became necessary due to a number of factors, including the increased accessibility and affordability of technology, the need for more flexible and convenient learning options, and the COVID-19 pandemic which forced schools and universities to close their physical classrooms and move instruction online. People who may not be able to attend traditional in-person classes owing to job, family, or other commitments will find that online education is an excellent alternative. Technical difficulties, such as slow internet connectivity or software glitches, can interfere with online learning and make it challenging for students to access instructional resources. Addressing these challenges and improving the online education experience requires a systematic and ongoing review of the online education system, with a focus on ensuring access to quality education for all students, addressing equity and inclusiveness concerns, and fostering innovation and improvement in the delivery of online education. The online education system has many benefits, as well as some drawbacks, and is likely to continue growing in popularity in the future. This article provides a general overview of online education systems, the technological requirements for their implementation, some potential future issues, and some approaches to make online education effective. Keywords: Online education · e-Learning · Information technology · IPR

1 Introduction Academic institutions, such as schools, colleges, and universities, have long served as the hubs of knowledge creation and exchange. Advanced digital technologies transformed and enriched so many other industries but education sector was still in the early stages of adoption until a few years ago. The health emergency due to COVID-19 pandemic forced the educational institutes towards the use of online mode or digital platforms [1]. Knowledge and information are no longer tied to the physical locations of educational institutions. Instead, a variety of platforms, software, encyclopedias, open-source © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 1057–1069, 2023. https://doi.org/10.1007/978-3-031-36118-0_90

1058

Md. A. Haque et al.

databases, and web browsers may be used to collect information and expertise on a wide range of topics, allowing users to supplement their learning. Online learning can be less expensive than traditional brick-and-mortar institutions, as it eliminates the need for physical classrooms, campus infrastructure, and other expenses [2]. Online learning can give students more control over their learning experience, which can boost motivation and engagement. It allows students to learn on their own schedule and can reduce the stress of commuting or trying to fit classes into a busy schedule. Educators are adopting technology in classrooms today, from massive open online courses (MOOCs) to flipped classrooms, there are new approaches and strategies to improve student learning. The drastic changes in education are generating news because they are having a significant influence on student learning. The ability to access educational resources from any location with an internet connection has led to an increase in the popularity of online education in recent years. According to the Impact Report of the renowned online learning platform Coursera, which was published in 2021, the overall number of enrollment more than twice in 2020 and jumped by 32% the following year, reaching at 189 million (Fig. 1) [3].

Fig. 1. Increasing trend of enrolled students in Coursera

Although digital education offers many advantages and gives unique access to highquality education, it also has several drawbacks that might make it difficult to succeed [4–8]. For example, the lack of digital infrastructure disrupts online education. Not all students have equal access to technology and the internet, which can create disparities in educational opportunities and outcomes. The students should have a minimum degree of computer literacy to learn effectively in an online context. The faculty members to be equipped with digital skills and knowledge to effectively design, deliver, and evaluate online programs, which can be challenging for some institutions and educators.

Education System Transition to Fully Online Mode: Possibilities

1059

Certain fields such as Science, Technology, Engineering and Medical require hands-on experiences which are not always possible through online education. The effectiveness of online education can depend on factors such as the specific course or program, the teaching methods used, the technology infrastructure and support provided, and the individual student’s learning style and needs [9]. Despite the challenges this new trend brings, it has many opportunities as well. It is necessary to review and analysis the different facets of online learning techniques for the following reasons: • • • • • • • • • •

Rapid increase in online education due to COVID-19 pandemic. Need to assess effectiveness and efficiency of online education. Address challenges and improve student experience. Ensure access to quality education for all students. Keep up with advancements in technology and education. Address equity and inclusiveness concerns in online education. Ensure compliance with regulations and accreditation standards. Evaluate the impact on employment and career opportunities. Prepare for potential future disruptions and emergencies. Foster continuous improvement and innovation in online education.

2 Online Learning Methods and Evaluation Process E-learning is a method of delivering education and training through digital means, using technologies such as online platforms, mobile apps, and multimedia resources. The methodology of e-learning involves the design and development of educational content, the selection of appropriate technologies, and the delivery and management of the learning experience. There are several different learning methods used in online education (Fig. 2), including: 1) Asynchronous Learning: This method allows students to access course materials and complete assignments at their own pace, without the need for real-time interaction with the instructor or other students. This method is often used in self-paced online courses. 2) Synchronous Learning: This method involves real-time interaction between students and instructors, such as through live online lectures, webinars, and virtual office hours. This method is often used in online classes that are delivered in real-time and that have a set schedule. 3) Blended Learning: This method combines the best of both asynchronous and synchronous learning, by incorporating both self-paced and real-time activities. This method is often used in online classes that have some scheduled meetings and some self-paced activities. 4) Self-directed Learning: This method allows students to take responsibility for their own learning by giving them the autonomy to choose their learning activities and assessments. This method is often used in MOOCs (Massive Open Online Courses) and other self-paced online programs. 5) Gamified Learning: This method uses game elements such as points, badges, and leaderboards to make the learning experience more engaging and motivating. This method is often used in online courses that are targeted to children and younger students.

1060

Md. A. Haque et al.

6) Project-Based Learning: This method allows students to apply their knowledge and skills to real-world problems or projects. This method is often used in online courses that focus on applied skills, such as computer programming or design.

Fig. 2. Online Learning Methods

Each approach has benefits and drawbacks of its own, and the best approach to adopt will depend on the course-objectives, the students’ background, and the available resources. In general, a combination of different methods is often used to create a more engaging and effective learning experience for students. The evaluation process in e-learning assesses the effectiveness and impact of the learning program. It involves collecting data on various aspects of the learning experience, such as student engagement, content effectiveness, and overall satisfaction. Some of the most common evaluation methods include: 1) Online quizzes and exams: These can be used to assess students’ knowledge and understanding of course content, and can include multiple-choice, short-answer, and essay questions. 2) Discussions and forums: Online discussions and forums can be used to assess students’ critical thinking, communication, and collaboration skills, as well as their understanding of course content. 3) Projects and presentations: Online students can submit projects, research papers, and presentations that demonstrate their understanding of course content and their ability to apply what they have learned.

Education System Transition to Fully Online Mode: Possibilities

1061

4) Self-reflection and peer evaluations: Students can be asked to reflect on their own learning and to evaluate the work of their peers. This can provide valuable feedback to both the students and the instructor. 5) Analytics: Data analytics can be used to track students’ engagement with course content and activities, such as how much time they spend on the course website, how many times they access the course materials, and how many times they participate in online discussions. 6) Online proctoring: Remote proctoring tools can help to ensure academic integrity of online exams and assessments, by monitoring students during the test, flagging suspicious behavior and capturing video and audio of the student during the test. An effective evaluation of students’ performance in online courses should take into account the unique characteristics of online learning and the different ways in which students engage with course content [10–12].

3 Tools and Technology for Online Education The use of modern technologies in online education can enhance the interactive and engaging nature of online learning, making it more effective and efficient. However, they require adequate infrastructure, resources and training for both teachers and students [13]. A wide range of digital tools and technology have been developed to meet the students’ evolving educational needs [14, 15]. The tools and technologies needed for an educational system to function virtually are listed below: 1) Reliable internet access: Students and educators need a stable internet connection to access online resources and participate in virtual classes. 2) Hardware and software: Students and educators will need devices such as smartphones, laptops or tablets, as well as appropriate software for video conferencing, online learning management systems, and other tools. 3) Virtual Classrooms: These are interactive online environments where students and teachers can meet, collaborate, and engage in real-time learning. Examples include Zoom, Google Meet, and Webex. 4) Digital content and resources: Educators need access to digital resources such as e-books, PPTs, videos, and other materials to support online instruction. 5) Professional development: Educators will need training and support in using digital tools and best practices for online instruction. 6) Learning Management System (LMS): A LMS is a software application for the administration, documentation, tracking, reporting, and delivery of educational courses, training programs, or learning and development programs. Examples include Blackboard, Canvas, and Moodle. 7) Artificial Intelligence (AI) and Gamification: AI and ML are increasingly being used to create personalized and interactive learning experiences, as well as to assist with assessment and feedback. The gamification approach uses game design elements, such as points, badges, and leaderboards, to make learning more engaging and interactive.

1062

Md. A. Haque et al.

8) Social Learning Platforms: These are online platforms that allow students to collaborate, share resources, and build networks with their peers, such as Edmodo, Schoology, and Classcraft. 9) Technical support and data security: A dedicated technical support team is required to troubleshoot and solve any technical issues that may arise with the infrastructure. Implementing security measures and protocols to protect student and faculty data is crucial. Implementing digital infrastructure for e-learning system requires a comprehensive approach, involving not just technology but also the development of policies, training, and support to ensure a successful transition to online education [16].

4 Possible Consequences in Online Education System Modern technologies, such as artificial intelligence, virtual and augmented reality, and online learning platforms, are likely to have a significant impact on the way teaching is conducted in the future. These technologies have the potential to personalize learning, provide more interactive and immersive experiences, and make education more accessible to people all over the world. Additionally, the use of data analytics and machine learning can help teachers to better understand how their students are learning, and make adjustments to their teaching methods accordingly. However, online learning can lack the face-to-face interaction and engagement that is present in traditional classroom settings, which can make it more difficult for some students to stay motivated and engaged. Ensuring the quality of online education can be a challenge. The future of online education is likely to be shaped by a number of factors, including advances in technology, changing student demographics, and shifting societal and economic trends. We conducted a survey to determine the direction and future of online education. We have gathered feedback from 200 teachers and 500 students in the 18 to 25 year old age range. All are from India and are proficient with computers and the Internet. Six questions made up the survey’s foundation. The participants were asked to respond with (Yes/No) when appropriate and to offer any comments they might have (optional). The survey findings are presented in Table 1. 69.14% of participants think that offering education entirely online will lower academic standards. However, the majority of them support blended learning. Blended learning, which combines online learning with face-to-face instruction, will become more prevalent, providing students with the best of both worlds. According to 84.86% of respondents, online learning has a negative impact on students’ mental health. 75.86% believe that e-learning will increase socioeconomic inequality. 63.71% of participants believe that cyber-attacks and hacking make online education susceptible to loss or theft of important data and intellectual property, including educational assets like video lectures, eBooks, and other interactive content. However, the majority of participants disregard the problems with discipline and gender inequality. We prefer to talk about four potential issues with a future online education system based on the survey findings.

Education System Transition to Fully Online Mode: Possibilities

1063

Table 1. A survey on future of online education Significant issues with online education

Students opinion (Out of 500)

Teachers opinion (Out of 200)

Yes

No

Yes

No

Will e-learning lead to a drop in educational quality?

329

171

155

45

69.14%

Will students face a lot of psychological problems?

412

88

182

18

84.86%

Will e-learning increase socioeconomic divide?

367

133

164

36

75.86%

44

456

28

172

10.29

118

382

69

131

26.71%

Will the violations of IPR increase? 305

195

141

59

63.71%

Will e-learning lead to further gender inequality? Will the flexibility of e-learning lead to a lack of discipline among students?

Results (agreed-upon percentage)

4.1 Lack of Quality Education The shift to online education has led to concerns about the quality of education being delivered. One factor affecting the quality of online education is the lack of in-person interaction between students and teachers. Online education relies heavily on technology and asynchronous communication, which can make it difficult for teachers to engage students and provide immediate feedback. The absence of in-person interaction also means that students may miss out on the social and emotional benefits of traditional in-person education, such as building relationships with classmates and teachers. In the future, the education system is likely to become less human-driven and more technologydriven as advancements in technology continue to shape the way we learn. The potential over-reliance on modern tools like ChatGPT may lead to a decrease in critical thinking skills. Moreover, the content generated by these AI-systems like ChatGPT may not be accurate or relevant [17]. This could result in students receiving incomplete or incorrect information, leading to misunderstandings or misinformation [18]. 4.2 Psychological Issues Online education has become a prevalent form of learning in recent years; however it is not without its psychological challenges. The lack of social interaction and face-to-face interaction can lead to feelings of isolation and disconnection, which can negatively impact a student’s mental health and motivation [19]. Technical difficulties and frustration can also disrupt the learning experience and cause added stress and anxiety. Online education requires a high level of self-discipline and motivation to stay on task and complete assignments, which can be difficult for some students. The virtual learning

1064

Md. A. Haque et al.

environment can reduce the sense of community and belonging for online students, leading to feelings of loneliness and a lack of connection. Additionally, online education can limit the amount of feedback and support students receive from instructors and peers, leading to feelings of uncertainty and a lack of confidence in their ability to succeed. 4.3 Socioeconomic Divide The increased reliance on online education as a result of the COVID-19 pandemic has raised concerns that it may exacerbate socioeconomic divides. On one hand, online education offers more flexible and accessible options for students, especially for those who live in rural or remote areas or have disabilities. This can level the playing field for students who previously faced barriers to accessing traditional in-person education. However, on the other hand, online education also requires access to technology and a stable internet connection, which many students from low-income families may not have. These students may therefore fall behind their peers and face further disadvantage. Additionally, the lack of in-person interaction and support can also negatively impact students who come from disadvantaged backgrounds, as they may not have access to the same resources and support systems as their more privileged peers. Overall, while online education has the potential to bridge some of the gaps in access to education, it can also deepen existing inequalities if steps are not taken to address the technology and resource disparities among students. 4.4 Increased Violations of Intellectual Property Rights (IPR) The breach of intellectual property rights for e-learning materials is one of the major concerns with online education systems [20]. Content utilized in online learning environments, such as instructional software, e-books assessments, videos, audio recordings, photographs, and other multimedia components, is considered intellectual property. Depending on how they are used and their unique nature, these resources may be covered by intellectual property regulations i.e. copyright, trademark, or patent laws. The sharing of educational resources is actually a good thing. Students, scholars and even other teachers get exposure to you. But when they go too far, stealing, plagiarism and piracy are all unethical practices that can be a risk, as it can lead to loss of revenue and reputation. Intellectual property rights in e-learning can be threatened by various factors, including: copyright infringement, plagiarism of course content, unauthorized distribution of course materials, misuse of trademarks and branding, theft of trade secrets and proprietary information, piracy of online courses and exams, unauthorized use of patented technology, and more [21]. E-learning is becoming more susceptible to hacking and cyber-attacks, which can lead to the loss or theft of valuable data and intellectual property. E-learning materials must be managed in compliance with the applicable privacy laws and regulations.

Education System Transition to Fully Online Mode: Possibilities

1065

5 Suggestions for Effective Online Education With the increasing use of digital tools and resources, students will have access to a wealth of information and resources at their fingertips, enabling them to explore and learn in new and innovative ways. While online education has made education more accessible for many students, it has also created new challenges and limitations that can affect the quality of education being delivered. Effective online learning requires careful planning and strategies to ensure that the learning experience is engaging, interactive, and tailored to the needs of the students [22]. The following are some planning strategies for effective online learning: 5.1 Defining Learning Objectives Clearly defining the learning objectives and outcomes for the course will help to ensure that the content and activities are aligned with the goals of the students. Tailoring the course to the needs of the students, such as providing flexible scheduling, accommodating the difficulty level of the content, and accommodating the teaching methodologies, will help to accommodate different learning styles and preferences. To ensure the quality of online programs, accreditation is necessary. Typically, accreditation is carried out by independent, nationally recognized accrediting bodies. These bodies have specific criteria and standards that institutions must meet in order to be accredited. The question of accreditation, which is a process of evaluating institutions and programs to ensure that they meet certain standards, is still a concern in online higher education. 5.2 Developing Proper Counseling System Counseling is important in an online education system as it provides critical support and guidance to students. Online education can pose unique challenges such as isolation, technology difficulties, and time management issues. Counselors can help students navigate these challenges and succeed in their studies. Encouraging self-paced learning and providing flexibility in terms of when and how students complete assignments will help to accommodate different schedules and learning styles. 5.3 Facilitating Communication and Collaboration Providing opportunities for communication and collaboration, such as virtual office hours, discussion forums, and group projects, will help to create a sense of community and support among students. Creating a sense of community and connection, such as through virtual meetings, student clubs, and group projects, will help students to feel connected to the class and to the instructor, even though they are learning remotely. 5.4 Designing Interactive and Engaging Content Designing interactive and engaging content that makes use of multimedia, simulations and other interactive elements will help to keep students engaged and motivated. Incorporating technology, such as learning management systems, virtual reality, and artificial intelligence, to enhance the learning experience can increase engagement and improve the students’ learning outcomes.

1066

Md. A. Haque et al.

5.5 Developing Effective Feedback Mechanism and Support System Providing clear and timely feedback on assignments and assessments will help students to understand what is expected of them and to make progress in their learning. Regularly review and assess the online learning system, and make changes as necessary to improve the experience of students and instructors. Providing support and resources, such as tutoring, office hours, or academic advising, will help students to stay on track and to manage their time more effectively. 5.6 Implementing an IPR Management Policy It is important for online education institutions and educators to have a clear understanding of IPR laws and to have policies in place to prevent and address these threats. Figure 3 shows some standard practices that combine both technical and legal safeguards for Intellectual Property Rights (IPR). • Copyright registration: registering the e-learning materials with the appropriate government copyright office provides legal protection against unauthorized reproduction and distribution. • Watermarking: adding a visible or invisible mark to the e-learning materials that identifies the owner can help deter unauthorized use and aid in enforcement of IPR. • Digital rights management (DRM) technology: this technology is used to control access to and usage of the e-learning materials, such as by requiring a login or preventing the materials from being downloaded or printed. • Licensing: using a license agreement to clearly set out the terms and conditions under which the e-learning materials can be used, such as specifying that the materials can only be used for personal or non-commercial use. • Legal action: taking legal action against individuals or organizations that are found to be infringing on the IPR of the e-learning materials. The architecture of IPR protection will vary depending on the nature of the e-learning materials, the intended audience, and the laws of the country where the materials are being used. It is expected that online education will become more widely recognized and accepted, and will be seen as a viable alternative to traditional, on-campus education. Online education should become more affordable and accessible, as more institutions and organizations invest in the infrastructure and resources needed to deliver high-quality online education. It’s also important to provide feedback and support to the students throughout the course to help them achieve their goals.

Education System Transition to Fully Online Mode: Possibilities

1067

Fig. 3. IPR Protection Methods

6 Summary and Conclusion The trend of online education has been on the rise in recent years, and it has accelerated significantly due to the COVID-19 pandemic. The future of education in an online environment is likely to be characterized by a continued growth in the number of online courses and degree programs offered, as well as an increase in the use of technology to enhance the learning experience. Online education has the potential to make education more accessible to a wider range of students, including those who live in remote or underserved areas, those who have work or family obligations that make it difficult to attend traditional on-campus classes, and those who are seeking to further their education while balancing other responsibilities. The use of virtual and augmented reality technologies, artificial intelligence, and other advanced technologies is also likely to become more prevalent in online education in the future. These technologies can enhance the interactive and immersive nature of online learning, making it more engaging and effective for students. Additionally, digital transformation offers institutions exciting opportunities to enhance their teaching and learning and the ability to effectively manage their operations – all of which are key to better serving students. However, according to the survey results, many people believe that maintaining the quality of education and ensuring that online courses are rigorous and aligned with standards. An education system that is overly reliant on technology will not help students develop ethical values. We believe that while technology will play a central role in shaping the future of education, the human touch will still be critical. Teachers and instructors will continue to play a crucial role in guiding students, providing emotional support, and fostering critical thinking skills. The future of education will be a balance between technology and human interaction, with each playing a complementary role in promoting learning and growth. Cyber security to protect intellectual property and student data in a completely online environment is also a major concern. It is important to note that the survey was conducted on a small sample data collected from a specific geographical region. In the future, we will address

1068

Md. A. Haque et al.

these issues through a large-scale survey because it is critical to ensuring a successful and high-quality online education experience for students.

References 1. Ahmed, N., Nandi, D., Zaman, A.G.M.: Analyzing student evaluations of teaching in a completely online environment. Int. J. Mod. Educ. Comput. Sci. 14(6), 13–24 (2022) 2. Nguyen, T.: The effectiveness of online learning: beyond no significant difference and future horizons 11(2), 309–319 (2015) 3. 2021 Impact Report, Serving the world through learning, Coursera (2021). https://about.cou rsera.org/press/wp-content/uploads/2021/11/2021-Coursera-Impact-Report.pdf 4. Dalgaly, T.: Benefits and drawbacks of online education. In: International Scientific Conference on Innovative Approaches to the Application of Digital Technologies in Education, Stavropol, Russia (2020) 5. Basar, Z.M., et al.: The effectiveness and challenges of online learning for secondary school students – a case study. Asian J. Univ. Educ. (AJUE) 17(3), 119–129 (2021) 6. Hassan, M.M., Mirza, T., Hussain, M.W.: A critical review by teachers on the online teachinglearning during the COVID-19. Int. J. Educ. Manage. Eng. 10(6), 17–27 (2020). https://doi. org/10.5815/ijeme.2020.05.03 7. Hassan, M.M., Mirza, T.: The digital literacy in teachers of the schools of Rajouri (J&K)-India: teachers perspective. Int. J. Educ. Manage. Eng. 11(1), 28–40 (2021) 8. Arkorful, V., Abaidoo, N.: The role of e-learning, the advantages and disadvantages of its adoption in higher education. Int. J. Educ. Res. 2(12) (2014) 9. Nawaz, A., Khan, M.Z.: Issues of technical support for e-learning systems in higher education institutions. Int. J. Mod. Educ. Comput. Sci. 4(2), 38–44 (2012) 10. Dixson, M.D.: Measuring student engagement in the online course: the online student engagement scale (OSE). Online Learn. 19(4) (2015) 11. Abbasi, S., Ayoob, T., Malik, A., Memon, S.I.: Perceptions of students regarding e-learning during Covid-19 at a private medical college. Pak. J. Med. Sci. 36, 57–61 (2020) 12. Zheng, M., Bender, D., Lyon, C.: Online learning during COVID-19 produced equivalent or better student course performance as compared with pre-pandemic: empirical evidence from a school-wide comparative study. BMC Med. Educ. 21, 495 (2021). https://doi.org/10.1186/ s12909-021-02909-z 13. Andrews, R.: Does e-learning require a new theory of learning? Some initial thoughts. J. Educ. Res. Online 3(1), 104–121 (2011) 14. Nermend, M., Singh, S., Singh, U.S.: An evaluation of decision on paradigm shift in higher education by digital transformation. Procedia Comput. Sci. 207, 1959–1969 (2022) 15. Saykili, A.: Higher education in the digital age: The impact of digital connective technologies. J. Educ. Technol. Online Learn. 2(1), 1–15 (2019) 16. Huang, R.H., Liu, D.J., Tlili, A., Yang, J.F., Wang, H.H.: Handbook on facilitating flexible learning during educational disruption: the Chinese experience in maintaining undisrupted learning in COVID-19 outbreak. Smart Learn. Inst. Beijing Norm. Univ. UNESCO, 1–54 (2020) 17. Haque, M.A.: A brief analysis of ‘ChatGPT’ – a revolutionary tool designed by OpenAI. EAI Endorsed Trans. AI Robotics 1(1), e15 (2023) 18. Mintz, S.: ChatGPT: threat or menace? Are fears about generative AI warranted? Inside Higher Ed. (2023). https://www.insidehighered.com/blogs/higher-ed-gamma/chatgpt-threator-menace

Education System Transition to Fully Online Mode: Possibilities

1069

19. Sakpere, A.B., Oluwadebi, A.G., Ajilore, O.H., Malaka, L.E.: The impact of COVID-19 on the academic performance of students: a psychosocial study using association and regression model. Int. J. Educ. Manage. Eng. 11(5), 32–45 (2021) 20. Lemoine, A.P.: The impact of online learning on global intellectual property issues. In: Munigal, A. (ed.) Scholarly Communication and the Publish or Perish Pressures of Academia, pp. 279–311. IGI Global (2017) 21. Casey, J., Proven, J., Dripps, D.: Managing intellectual property rights in digital learning materials. TrustDR (2007) 22. Alonso, F., López, G., Manrique, D., Viñes, J.M.: An instructional model for web-based e-learning education with a blended learning process approach. Br. J. Edu. Technol. 36(2), 217–235 (2005)

Collaborative Education Mechanism of In-Depth Integration Between Schools and Enterprises Under the Background of the Construction of EEE -- Taking the Major of Data Science and Big Data Technology of Nanning University as an Example Zhixiang Lu(B) and Suzhen Qiu College of Artificial Intelligence, Nanning University, Guangxi 530200, China [email protected]

Abstract. School-enterprise cooperative education is an important starting point for the construction of new engineering. How to cultivate “Emerging Engineering Education (EEE)” talents in line with the requirements of the times has become the focus of the construction of application-oriented universities. Taking the construction of the information technology specialty of Nanning University as an example, this paper analyzes the background of the construction of emerging engineering education, and forms a new talent cultivation mode of “dual mainline of integration of industry and education, and dual mainline of integration of professional education and vocational education”, which opens up a new path for the cultivation of applied talents, including the joint development of industry-education cooperation projects, joint design of courses, joint preparation of textbooks, joint training of teachers, and sharing of resources Jointly determine the project, jointly teach, jointly manage, jointly evaluate the quality and jointly assist in employment. The practice shows that the deep integration and collaborative education mechanism between Nanning University and enterprises is conducive to improving the quality of applied and innovative talents. Keywords: Emerging Engineering Education (EEE) · School-enterprise integration · Collaborative education

1 Introduction The role of higher education in talent training and social progress is huge and cannot be ignored [1–3]. The fourth industrial revolution and industrial transformation are emerging, affecting the form and form of higher education. From “Reindustrialization” in the United States to “Industry 4.0” in Germany, all countries around the world seize the opportunity of the times, vigorously develop engineering education, and have introduced relevant reform policies and measures [4–6]. Under such a background, China’s © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 1070–1084, 2023. https://doi.org/10.1007/978-3-031-36118-0_91

Collaborative Education Mechanism of In-Depth Integration

1071

engineering education has entered a new stage of innovation and reform, and higher education has turned into a connotative development. The “Made in China 2025” strategy emphasizes that as China’s economic development evolves to a new stage with more advanced form, more complex division of labor and more reasonable structure, engineering talents should not only have solid technical capabilities, but also have comprehensive qualities such as globalization perspective, cultural awareness, innovation and entrepreneurship [7]. National leaders have repeatedly stressed that “we should promote collaborative innovation between industry, university and research, actively participate in the implementation of innovation-driven development strategy, and focus on cultivating innovative, compound and application-oriented talents”. In October 2017, the 19th National Congress of the Communist Party of China clearly put forward the major reform task of deepening the integration of industry and education, and promotes the coconstruction and sharing of resources in the collaborative interaction between industry and education [8]. In the same year, the Ministry of Education launched the construction of “Emerging Engineering Education”, also abbreviated as EEE, and accelerated the training of high-level engineering and technological innovation talents in emerging fields [9]. The construction of EEE science is proposed in response to the needs of industrial transformation and upgrading, national strategy and the transformation of engineering education paradigm. It emphasizes the orientation of industrial demand and requires the integration of multiple resources and advantages [10]. How to coordinate the interests and responsibilities of both schools and enterprises and achieve benign interaction is the core issue of building a 3-E and middle-class education integration and collaborative education mechanism.

2 The Necessity and Problems of School-Enterprise Collaborative Education 2.1 The Necessity of College-Enterprise Collaborative Education Under the current situation, it is very necessary for colleges and universities to vigorously promote the school-enterprise cooperation in educating people. 2.1.1 Integration of Production and Education is an Important Way to Cultivate High-Quality Application-Oriented Talents In October 2019, the National Pilot Implementation Plan for the Integration of Industry and Education emphasized that the integration of industry and education is an important basis for the reform of human resources in China, and it is to solve the mismatch between talent cultivation and industrial demand in China’s economic and social development. It is necessary to give full play to the promotion role of enterprises and industries in talent cultivation [11]. In engineering practice, talent training can further strengthen the understanding of professional knowledge and improve professional quality by combining practical production with theoretical knowledge [6, 12]. As an important mode of collaborative education in the context of industrial transformation and upgrading, industryeducation integration can effectively promote the integrated development of education and industry, and is an important way to cultivate high-quality application-oriented talents [13].

1072

Z. Lu and S. Qiu

2.1.2 The Needs of Local Regional Economic Development In order to implement the spirit of the Notice of the State Council on Printing and Distributing the Action Plan for Promoting the Development of Big Data, promote industrial innovation and development, and speed up the construction of the China-ASEAN Information Port, the People’s Government of Guangxi Zhuang Autonomous Region issued a series of documents, including the Development Plan for the Information and Innovation Industry of Guangxi Zhuang Autonomous Region (2021–2025), the Development Plan for the Digital Economy of Guangxi Zhuang Autonomous Region (2018–2025), and clearly proposed the construction of big data infrastructure, The construction of digital Guangxi was fully launched [14–16]. Nanning University adheres to the localization and application-oriented orientation of running a school, serving the local regional economy and implementing the integration of production and education is the fundamental to the development of the school. The development of regional economy and society requires schools to cultivate highquality applied talents, and local applied schools should determine the goal of talent cultivation by serving the regional economy. Therefore, only through the integration of industry and education and school-enterprise cooperation to improve the ability to serve local industry enterprises can we better serve the local regional economic development. Guangxi’s information technology application and innovation industry has just started. The professional construction should be closely combined with the information and innovation industry, and the orientation of professional talent training should adapt to local industries. Through in-depth cooperation between schools and enterprises, integration of industry and education, and joint training of application-oriented talents, promote regional economic development. 2.1.3 The Needs of Professional Construction The construction of the EEE, specialty should be in line with the market demand from the aspects of specialty orientation, talent training mode, teaching resource construction, teaching process, teaching assessment, etc. [17–19]. Nanning University is a resource-constrained local newly-built application-oriented university, with relatively weak resources such as teaching staff, teaching software and hardware, which is difficult to meet the training needs of information technology professionals. However, information technology-related enterprises’ advantageous resources in technology development, application, management, and software and hardware environment are exactly the resources needed for professional construction. The specialty should actively connect with the market, carry out in-depth cooperation with the industry, complement each other’s advantages, jointly establish a market-oriented new data science and big data technology talent training mode, and cultivate high-quality application-oriented talents serving regional economic development [20, 21]. Through the in-depth cooperation between schools and enterprises, the educational logic and industrial logic are connected to realize the connection between professional education and industrial production, and the connection between curriculum teaching and professional positions. At the same time, the problem of insufficient teachers and teaching resources in the school is solved,

Collaborative Education Mechanism of In-Depth Integration

1073

and the problem of the disconnection between theoretical teaching and actual production is solved. 2.2 Main Problems Faced by School-Enterprise Collaborative Education 2.2.1 There is no Mature and Stable Education Mode for the Construction of EEE China’s Emerging Engineering Education construction is still in its infancy, and there are many deficiencies in the training mode of new engineering talents. In the context of the construction of 3-E, the training objectives of new engineering talents have gradually changed. The industry-university-research partners have not formed a mature collaborative education model in this field. The existing education model is relatively simple, mostly from the perspective of colleges and universities to talk about cooperation, ignoring the need for high-quality skilled talents in the construction of new engineering talents. In the construction of the EEE talents training model, the synergy of the various subjects of industry-university-research is insufficient, As a result, colleges and universities, enterprises, research institutes and other resource groups have not formed a mature collaborative education model. 2.2.2 Incomplete Multi-party Responsibility and Authority Mechanism The imperfection of the responsibility and right mechanism of the integration of industry and education in China’s engineering education is mainly due to the lack of incentive and restraint mechanism, which leads to the low participation enthusiasm of all subjects; On the other hand, the division of responsibilities and rights of each subject is not clear, which leads to hindered coordination and low efficiency. The specific performance is as follows: First, there are few incentive and restraint mechanisms in the collaborative education mechanism of industrial and educational integration of engineering education in China, and few policies mention incentive and restraint mechanisms. The collaborative education of industrial and educational integration mainly depends on the cooperation oriented by the industrial demand of universities and enterprises. Second, the division of responsibilities and rights of each subject is not clear, which leads to many obstacles to synergy and poor synergy effect. For universities, the integration of production and education can promote students to combine theoretical knowledge and practice, which is conducive to the cultivation of talents required by the industry. For enterprise employers, participating in collaborative education can make it easier to obtain talents, enhance the publicity effect of enterprises, and reduce the cost of employment. In the actual operation process, there are still some problems, such as the university’s education responsibility and the low participation of enterprises. 2.2.3 Lack of Scientific Evaluation System In the operation process of industry-education integration and collaborative education mechanism, both schools and enterprises assume the responsibility of education, and the evaluation subject should be carried out jointly by multiple parties. The research history of the collaborative education evaluation of the integration of industry and education in engineering education is relatively short, and the relevant rules and regulations are not

1074

Z. Lu and S. Qiu

perfect, showing the current situation that the evaluation subject is lack of diversification, the evaluation content is not systematic and comprehensive, and the evaluation method is not scientific enough.

3 Construction of a Collaborative Education Model with Deep Integration of Schools and Enterprises 3.1 College and Enterprise In-Depth Integration and Joint Construction of Professional Ideas With the reform concept of “professional construction and industrial development in the same direction and in the same direction as the same industry”, the demand side promoting the structural reform and construction of the professional supply side as the reform direction, and comprehensively implementing the reform task of “cultivating morality and educating people”, adhere to the “open” path of professional construction and development, take the integration of industry and education as the starting point, with the help of external forces of emerging industries, and through school-enterprise cooperation to build, change, improve and strengthen majors, Constantly strengthen the construction of professional connotation, promote the application-oriented transformation of professional, improve the level of professional construction, and cultivate professional characteristics. 3.2 Deepen the Reform of the Cooperative Education Mechanism Between Schools and Enterprises, and Continue the Reform and Innovation of the Cultivation Mode of Application-Oriented Talents In the training and model reform of applied talents, school-enterprise cooperation is a very effective way. (1) Professional ecological construction. With “resource integration, team integration, management integration, teaching integration, and cultural integration” as the starting point of reform, deepen the reform of industry-education integration mechanism based on the community of interests, deepen cooperative development, create a new education community that integrates talent training, scientific research, technological innovation, enterprise services, student entrepreneurship and other functions, and vigorously promote the sound development of professional ecology. The construction of the new mechanism for the construction of the “five integration” industry-education in-depth integration specialty is shown in Fig. 1. (2) The continuous improvement and innovation of the talent training mode of the cooperation between industry and education. 1) Based on the path of “reverse design and positive implementation”, according to the professional certification standards of engineering education and the needs of the school’s orientation, industry chain and innovation chain, put forward the objectives and requirements of talent cultivation, reconstruct the curriculum system according to the post ability and career development, and develop a new plan for the cultivation of applied talents. 2) Through the “ten joint” construction of the industrial college’s industry-education cooperation

Collaborative Education Mechanism of In-Depth Integration

1075

plan, curriculum, teaching materials, teacher training, resource sharing, project setting, teaching, management, quality evaluation and employment assistance, the new mode and path of industry-education cooperation and application-oriented talent cultivation in line with application-oriented universities will be reconstructed, and the role of the two education subjects in the process and each link of application-oriented talent cultivation will be given full play.

Fig. 1. New mechanism for the construction of the “five integration” industry and education deep integration specialty

Through the implementation of the “ten joint” of industry and education, explore the talent cultivation path of “professional and industrial docking, curriculum and real job docking, classroom and work situation docking”, and realize the docking of talents’ professional skills and work fields. 3) Innovate the training process of “double mainline education”, implement new collaborative education measures such as “production and education, double teachers and double abilities” teaching, “production and education, double bases” practice, “production and education, double cooperation” training, and “production and education, double standards” evaluation, so that the educational integration of production and education synergy runs through the whole process of talent training, and achieve a good connection between the talent supply side and the industry demand side. Figure 2 shows the framework of the talent cultivation model of industry-education cooperation.

1076

Z. Lu and S. Qiu

Fig. 2. Talents cultivation mode of industry-education cooperation

3.3 Establish a Teacher Development and Training Mechanism, Focus on the Training Goal of “Double Teachers and Double Abilities”, and Build an Excellent Team The building of an excellent teaching team should focus on the goal of “double teachers and double abilities”. (1) Innovate the training mechanism and promote the transformation and construction of teachers. Establish a teacher development center, focus on the goal of team construction of “structural transformation, capacity transformation and role transformation”, implement a series of measures such as the “dual ability promotion drive” mechanism of teachers, “innovative teaching team cultivation”, strengthen the reform of the construction of teachers and grass-roots teaching organizations, optimize the team structure, improve the team level, and promote the overall transformation of teachers’ role to application-oriented undergraduate education, Cultivate a team of young teachers who grow together with the school. (2) Based on the integration mechanism of industry and education, promote the structural reform of the team. Focusing on the structural objectives of the teaching staff under the mode of industrial college, giving full play to the institutional advantages of the cooperation with iFLYTEK to jointly build majors, promoting the structural reform of the team based on the concept of “team integration”, establishing high-level talent introduction and training measures, and building a high-level mixed teaching staff with “school-enterprise connectivity, full-time and part-time combination, and complementary advantages”. (3) Implement the “double leadership” system of schools and enterprises, and innovate the construction mode of teaching grass-roots organizations. Establish a professional teaching and research office and an enterprise teaching department to implement the grass-roots teaching organization construction of “double leadership system” of schools and enterprises; Establish the curriculum responsibility system, promote the construction of “one course, two teachers” embedded teaching organization, and form a working environment, teaching and research pattern and management mode

Collaborative Education Mechanism of In-Depth Integration

1077

of “resource sharing and complementary advantages”; We will establish a college student employment and entrepreneurship service guidance center, an innovation and entrepreneurship guidance team, and a school-enterprise mixed innovation guidance team to integrate ideological and political education and industry education into professional education, so as to jointly promote the quality of students’ employment and entrepreneurship. (4) Focus on the training goal of “double teachers and double abilities” to build an excellent team. Through the project establishment and construction of applicationoriented scientific research team, innovative teaching team, curriculum ideological and political teaching team, core technology training and academic exchange of leading enterprises, teacher ability training workshops and other training modes and measures, improve the “double teachers and double abilities” quality of teachers, improve the overall level of the team, promote the effectiveness of the construction of first-class professional, first-class team, first-class curriculum, and promote the ability of professional leaders and the comprehensive development of teachers. 3.4 Promote Curriculum Construction and Reform Around Application-Oriented Characteristics Curriculum construction and reform should highlight the characteristics of applicationoriented schools and specialties. (1) Curriculum system reform. Based on the needs of the application “type”, application “positioning” and application “ability” of the curriculum, establish the curriculum logic system and curriculum construction objectives, deconstruct and reconstruct the traditional curriculum system according to the technology application logic and work action logic, and form a new application-oriented curriculum system. (2) Curriculum content reform. Based on the needs of “change” in technology, “crossborder” in production and teaching, “reality” in scenes, and “effectiveness” in application of the curriculum implementation, and based on the reform strategy of “circularization of work tasks, work-orientation of teaching tasks, and systematization of work processes”, the curriculum content is reconstructed, the curriculum implementation carrier is constructed, the curriculum implementation situation is optimized, and the curriculum quality evaluation is reformed. (3) First-class curriculum building. Take the first-class curriculum standard as the benchmark, build the golden curriculum, promote the full coverage of the ideological and political construction of the professional curriculum, and improve the overall level of the professional core curriculum group; Clearly standardize the process and objectives of each course construction, and implement the reform to achieve results through the establishment of course archives. 3.5 Promote the Construction and Reform of Teaching Mode Around Application-Oriented Characteristics The reform of teaching mode should be carried out around the characteristics of application type. (1) Classroom teaching reform. Based on the ability required by a position working in the field of big data, we will position the training objectives, design the teaching

1078

Z. Lu and S. Qiu

content, determine the teaching methods, manage the teaching process, evaluate the teaching effect, and form a new teaching model of output-oriented ability. We launched the “project-based task-driven” teaching reform demonstration, promoted the innovation of teaching methods and technical means such as flipped teaching, design experiment and innovative experiment, and formed a new model and method suitable for classroom teaching in applied technology universities. (2) Practice teaching reform. 1) Correctly understand the goal and important position of practical teaching, innovate the professional practical teaching system, form the curriculum system structure of “double system” of theory and practice, and reflect the applicability, professionalism, technology and innovation of practical teaching; 2) Give full play to the advantages of the construction mechanism of the Industrial College, build and develop the experimental training center, build and continuously optimize the online experimental platform, standardize and strengthen the construction of the joint off-campus training base with strong complementarity and relatively stable with the campus, and improve the utilization rate and effect of the in-campus and off-campus experimental training teaching base; 3) In the reform of practical teaching mode, we should strive to achieve “four changes”, that is, from demonstration to actual combat, from verification to design, from restriction to openness, and from singleness to comprehensiveness, and focus on cultivating students’ engineering practice ability and innovation ability; In the implementation of practical teaching innovation, actively expand the practical teaching path of school-enterprise cooperation, industry-teaching integration, and work-study alternation, so that students can truly experience practical training; In the reform of practical teaching methods, give full play to the main role of students, and strive to promote “teaching as one” and “cooperation between production, teaching and research”; In the construction of practical teaching carriers, we should enrich the teaching carriers, widely carry out innovation and entrepreneurship practice, skill training and certification, discipline competition, campus cultural activities, and various social practice activities to guide students’ personality development. 3.6 Cooperation Between Production and Teaching, Technology Guidance, and Strengthening the Construction of Application-Oriented Teaching Resources The construction of teaching resources can be started from the two aspects mainly. (1) Construction of application-oriented teaching materials. The joint venture develops professional core courses, selects and compiles applied course textbooks, and focuses on highlighting the application and progressiveness of textbooks. (2) Construction of digital technology teaching resources. 1) Relying on new technology, develop and enrich technical engineering training programs and work scenario experience; Guided by the basic concepts and standards of engineering education professional certification, establish a professional knowledge map and skill map based on the big data talent training process, and help form a new intelligent learning model of “precision and intelligence”; 2) Industry-teaching collaboration continuously improves and optimizes the teaching platform, continuously improves the level of closed-loop management of “teaching, learning, training, management and

Collaborative Education Mechanism of In-Depth Integration

1079

testing” of the platform in the talent training process, improves the scientificity and accuracy of teaching management, and improves the organizational efficiency and teaching quality of teaching. 3.7 Promote the Comprehensive Reform of Professional Education and Teaching with Advanced Quality Assurance Concept Adhering to the quality concept can guarantee the comprehensive reform of professional education and teaching. (1) Adhere to the student-centered concept and promote the reform of the main function of quality. 1) Reform the system and mechanism, and strengthen the main function of school quality: adhere to the “people-oriented”, and strengthen the central position of undergraduate teaching; 2) Reform the teaching mode and strengthen the quality subject function of teachers and students: first, change the education and teaching concept and promote the reform of student-centered teaching mode; The second is to use modern information technology to transform traditional teaching and drive the modernization of education. (2) Adhere to the output-oriented concept and promote the reform of quality factor allocation. 1) Reform the allocation of process elements to produce excellent quality through excellent education and teaching process: actively respond to the requirements of the times, and gradually promote the implementation of the strategy of cultivating outstanding talents through the selection and training of discipline and professional competitions, the further study and the cultivation of innovative achievements; The second is to vigorously promote innovation and entrepreneurship education and cultivate innovative, application-oriented and compound talents by using the “Create the Future” college students’ innovation and entrepreneurship practice education platform, the brand of college students’ career planning competition, the construction of specialized and innovative integrated courses, and the innovation and entrepreneurship competition; 2) Reform the allocation of resource elements and support the excellent education and teaching process with high-quality education and teaching resources: deepen the construction of “resource integration and team integration” of school-enterprise cooperation. First, build a mixed teaching team with noble teachers’ ethics, strong teaching ability and willingness to invest, and be a good guide for students in the workplace; The second is to rely on the introduction of advanced technology to improve the modernization level of teaching resources and improve teaching quality. (3) Adhere to the concept of continuous improvement and promote the reform of quality assurance methods. Establish advanced quality assurance concepts to provide effective guidance for quality assurance work; Improve quality standards and play an important role in talent training; Improve the quality assurance mechanism and enhance the quality assurance effect; Create a strong quality culture and improve the quality assurance level. Take the effectiveness of professional service for economic and social development, the combination of professional and industrial enterprises, the connection of training process and production practice, and the matching of training quality and industrial demand

1080

Z. Lu and S. Qiu

as the observation points of quality evaluation, and establish a new multiple evaluation mechanism through the integration of “school and enterprise double standards”.

4 Achievements of School-Enterprise Integration and Collaborative Education Mechanism Construction Significant results have been achieved through the construction of the collaborative education mechanism for the in-depth integration of schools and enterprises. 4.1 Course Effectiveness The evaluation of the quality of talent training mode can establish the evaluation indicators of talent training quality from the aspects of curriculum effectiveness, employment quality, basic work ability, etc. Statistical analysis is made on the survey data of graduates of Data Science and Big Data Technology in our university after graduation in recent two years through questionnaires. The employment of students half a year after graduation is relatively stable, and they can correctly evaluate the importance of their ability requirements in work. Conduct data analysis on the importance and satisfaction of the curriculum surveyed by the graduating students to analyze the effectiveness of the curriculum. The impact of curriculum importance and satisfaction on curriculum effectiveness is shown in Table 1. Table 1. Relationship between satisfaction of curriculum importance and curriculum effectiveness Course importance

Course satisfaction

Course effectiveness

high

high

high

high

low

low

low

/

general

According to the survey and analysis of 45 courses in the talent training plan, the number and proportion of various courses in 2021 and 2022 are shown in Table 2. Table 2. Number and percentage of professional courses in 2021 and 2022 Class label

2021

Important but not satisfied

24

53.33%

13

28.89%

6

13.33%

24

53.33%

15

33.33%

8

17.78%

Important and satisfied unimportance

2022

Collaborative Education Mechanism of In-Depth Integration

1081

Assume that the coefficient of important and satisfying class courses is 1, the quantity is n1 , the coefficient of important but not satisfying class courses is 0.5, the quantity is n2 , and the coefficient of unimportant class courses is 0, the quantity is n3 . Then the course effectiveness index is: C=

n1 × 1 + n2 × 0.5 + n3 × 0 n1 + n2 + n3

(1)

The value of C is between [0, 1]. The lower the value, the lower the importance and satisfaction. The higher the value, the higher the importance and satisfaction. Table 2 shows that the proportion of important and satisfying courses will increase in 2022. According to the above formula, the curriculum effectiveness index of this major in 2021 and 2022 is 0.40 and 0.68, respectively. It can be seen that the school-enterprise in-depth industry-university integration collaborative education model can dynamically adjust the curriculum according to the market demand to meet the actual work needs of talents. 4.2 Employment Quality The quality of employment is an important indicator to evaluate the quality of talent training. The factors that affect the quality of employment include salary level, work status, resignation factors, etc. The evaluation of salary level mainly depends on whether the actual salary is higher than the bottom line of salary expectation, the working status mainly depends on whether it meets the value of career expectation, and the reason for resignation depends on the number of employers who have graduated to the current position. The quality of employment is divided into three categories, good, average and poor, and the comprehensive judgment is made based on the factors such as graduation direction, income, work place, number of employers, and reasons for leaving. We collected 443 survey data of graduates of data science and big data technology in 2021 and 304 in 2022. The statistical analysis is shown in Table 3. Table 3. Classified statistics of employment quality Class label

2021

Good employment quality

211

47.63%

197

64.80%

Average employment quality

147

33.18%

58

19.08%

85

19.19%

49

16.12%

Poor employment quality

2022

Suppose that the coefficient of “good” employment quality is 1, the records is J1 , the coefficient of “average” employment quality is 0.5, the number of records is J2 , the coefficient of “poor” employment quality is 0, and the number of records is J3 , the employment quality index is as shown in formula (2): J=

j1 × 1 + j2 × 0.5 + j3 × 0 j1 + j2 + j3

(2)

1082

Z. Lu and S. Qiu

The value of J is between [0,1]. The lower the value, the worse the quality of employment. The higher the value, the better the quality of employment. It can be seen from Table 3 that the proportion of good employment quality will increase in 2022. According to the above formula, the employment quality index of this major in 2021 and 2022 is 0.66 and 0.74, respectively. It can be seen that the school-enterprise deep industry-university integration collaborative education model flexibly adjusts the talent training program, and the employment quality is getting better and better. 4.3 Application and Promotion Adhering to the “open” path of professional construction and development, taking the integration of industry and education as the starting point, with the help of external forces of emerging industries, and through school-enterprise cooperation in building, changing, optimizing, and strengthening majors, we have built an industry-education integration mechanism of “resource integration, team integration, management integration, teaching integration, and cultural integration”, and formed a co-ordering of industry-education cooperation programs, co-setting of courses, co-compiling of textbooks, co-training of teachers, resource sharing, and co-setting of projects The “ten joint”, also called “ten common” specialty construction path of teaching, management, quality evaluation and employment assistance, the innovation of the “double main line education” training process, the implementation of the “production and education, double teachers, double abilities” teaching, “production and education, double bases” practice, “production and education, double cooperation” training, “production and education, double standards” evaluation and other collaborative education measures, so that the education integration of production and education cooperation runs through the entire process of talent training, Realize a good connection between the talent supply side and the industry demand side. Over the past five years of construction, more than 20 colleges and universities from inside and outside the province have come to our university every year to investigate the construction of the collaborative education mode of industry-teaching integration, and the reform results have been used for reference and application by 15 similar colleges and universities. He has been invited to share experiences at the National Computer Education Conference, the International Forum on the Integration of Industry and Education and other conferences and universities for more than 30 times.

5 Conclusion Nanning University and enterprises have deeply integrated production and education, and jointly built a EEE specialty under the mode of artificial intelligence industrial college, aiming at the cultivation of application-oriented talents, seeking development based on the integration of production and education, adhering to the theoretical guidance of OBE professional certification of “student center, ability output, and continuous improvement”, and exploring the effective path of production and education integration based on the “five integration” and “ten-joint” mechanism innovation of schools and enterprises The construction of the new model of application-oriented talent cultivation of “production and education cooperation with two main bodies and two main lines” has achieved remarkable results, showing strong development vitality.

Collaborative Education Mechanism of In-Depth Integration

1083

The 3-E construction puts forward new requirements for the training specifications of talents, and also brings new challenges to colleges and universities and enterprises. It is the need of the new engineering construction, the need of industrial transformation, and the need of national development to jointly build the education model that meets the needs of economic development. Due to the relatively new theme of “new engineering” construction, the systematic analysis of Emerging Engineering Education and “industry-university-research collaborative education” is not thorough enough, the school-enterprise collaboration lacks the participation of the government and industry, and there is no quality monitoring system that restricts all parties, and the cultivation of high-quality applied talents is not guaranteed. Therefore, it is necessary to increase the strength of the government and industry and establish a quality monitoring system for collaborative education of schools, governments, banks and enterprises. Through the multi-party collaborative quality monitoring system, improve the enthusiasm of enterprises to participate and cultivate high-quality application-oriented talents that meet the needs of regional economy. Acknowledgment. This research is supported by the Collaborative Education Project of the Ministry of Education in 2020: Research on The Curriculum Reform of The School-Enterprise Collaborative Education Model (202002273034).

References 1. Waha, L.W., Yeeb, W.S., Soona, W.K., et al.: Collaborative learning experience in higher education: students’ perspective. Malaysian Constr. Res. J. 15(Special 1), 273–286 (2022) 2. Skedsmo, G., Huber, S.G.: Measuring teaching quality, designing tests, and transforming feedback targeting various education actors. Educ. Assess. Eval. Accountability 32(3), 271– 273 (2020). https://doi.org/10.1007/s11092-020-09333-9 3. Hamada, H.: Action research to enhance quality teaching. Arab World English Journal, 2019, Conference Proceedings (1), 4–12 (2019) 4. Nawaz, R., Sun, Q., Shardlow, M., et al.: Leveraging AI and machine learning for national student survey: actionable insights from textual feedback to enhance quality of teaching and learning in UK’s higher education. Appl. Sci. 12(1), 514 (2022) 5. Mantle, M.J.H.: How different reflective learning activities introduced into a postgraduate teacher training programme in England promote reflection and increase the capacity to learn. Res. Educ. 105(1), 60–73 (2019) 6. Alkhathlan, A.A., Al-Daraiseh, A.: An analytical study of the use of social networks for collaborative learning in higher education. Int. J. Mod. Educ. Comput. Sci. (IJMECS) 9(2), 1–13 (2019) 7. Li, T., Li, F., Lu, G.: Analysis of the way to improve the quality of engineering talent training for “Made in China 2025.” High. Eng. Educ. Res. 06, 17–23 (2015). (in Chinese) 8. Chen, J.: Put the deep integration of industry and education in an important position. People’s Daily/2018–04–03 (7). (in Chinese) 9. Zhong, D.: The connotation and action of new engineering construction. High. Eng. Educ. Res. 3, 1–6 (2017). (in Chinese) 10. Li, Z.: Key points for design and implementation of the new round of audit and evaluation scheme. High. Eng. Educ. Res. (03), 9–15 (2021). (in Chinese)

1084

Z. Lu and S. Qiu

11. Central People’s Government of the People’s Republic of China. National Pilot Implementation Plan for the Integration of Industry and Education [EB/OL] 2. http://www.Gov.Cn/xin wen/2019-10/11/content_5438226.htm. 2019-10-11. (in Chinese) 12. Inayat, I., Amin, R., Inayat, Z., Badshah, K.: A collaborative framework for web based vocational education and training (VET); findings from a case study. Int. J. Mod. Educ. Comput. Sci. (IJMECS) 5(12), 54–60 (2013) 13. Ma, E.: Transforming education requires collaboration at every level. Child. Educ. 98(4), 36–39 (2022) 14. The State Council: Notice of the state council on printing and distributing the action outline for promoting the development of big data. Chinese Government Procurement (9), 15–21 (2015). (in Chinese) 15. Peng, X.: Implementation of the big data strategy “digital construction” in Hezhou is accelerated. Guangxi Economy 10, 23–24 (2018). (in Chinese) 16. Lu, D., Yang, C.: Guangxi power grid corporation issued the “fourteenth five-year plan” development plan. Guangxi Electric Power 5, 7–8 (2022). (in Chinese) 17. Meng, Q., Li, W.: Consideration and mechanism innovation of the cultivation of legislative talents in China under the mode of cooperation between government, industry, university and research. Hebei Law 40(10), 76–96 (2022). (in Chinese) 18. Liu, X.: Research on college students’ ideological and political education in the process of industry-education integration. Northeast Forestry University, Harbin (2021). (in Chinese) 19. Wei, W.: Research on the cultivation of application-oriented undergraduate talents from the perspective of capability-based education. Jiangsu Higher Education (02), 44–48 (2017). (in Chinese) 20. Li, K., Liang, S., Zhao, Y.: Exploration and practice of the construction of new engineering courses in cooperation between application-oriented undergraduate universities and enterprises – taking Shenyang Institute of Technology as an example. J. Southwest Jiaotong Univ. (Social Science Edition) 22(01), 134–141 (2021). (in Chinese) 21. Salas-Tarin, C., Gunderman, R.B.: Educating learners in academic-industrial collaboration. Acad. Radiol. 28(9), 1321–1322 (2021)

Formal and Non-formal Education of Ukraine: Analysis of the Current State and the Role of Digitalization Maryna Nehrey(B) , Nataliia Klymenko, and Inna Kostenko National University of Life and Environmental Sciences of Ukraine, 15 Heroyiv Oborony Street, Kyiv 03041, Ukraine [email protected]

Abstract. The paper offers an analysis of formal and non-formal components of education in the context of digitalization processes, focusing on indicators in the wartime period. An analysis of formal and non-formal learning in Ukraine under martial law was carried out. The connection between the education of the population and the indicators of the country’s innovative development is proven. The influence of the educational level of the population on the level of development of the digital economy was studied. The factors of demand for formal education through the dynamics of educational institutions are analyzed. It has been established that the spread of non-formal education in Ukraine is accelerating due to the introduction of open educational platforms and projects, and education is acquiring signs of digitization. Keywords: education system · digitalization · formal learning · non-formal learning · regression

1 Introduction Digitalization is one of the up-to-date trends in education and science, as well as in the economy and society. The Cabinet of Ministers of Ukraine has approved the Concept for the Development of the Digital Economy and Society of Ukraine for 2018–2020 and the Concept for the Development of Digital Competencies until 2025 [1]. This gave a significant boost to the development of the digitalization of education. It has been proven that the level of digitalization and the level of education are highly correlated [2, 3]. The new ranking goals of the Concept for the Development of the Digital Economy and Society of Ukraine are to increase positions in the Networked Readiness Index, Global Innovation Index, ICT Development Index, and Global Competitiveness Index. In general, Ukraine has an unstable position in this indicator and is constantly in the third quantile [4]. The continuous development of information and communication technologies, as well as changes in the structure of in-demand professions during the pandemic and wartime, are leading to a shift in the balance between formal and non-formal education. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 1085–1098, 2023. https://doi.org/10.1007/978-3-031-36118-0_92

1086

M. Nehrey et al.

In the late 1960s, the concept of the global education crisis was introduced [5]. Its definition was based on the idea that the formal education system was adapting rather slowly to socioeconomic changes in society. The importance and potential of education that can be obtained outside of traditional educational institutions, which is called formal education, gained the scientists’ focus. In our view, the problem of education’s inconsistency with modern requirements is in the balance between formal and non-formal components, and the underestimation of the role of non-formal education. Therefore, the paper proposes an analysis of the impact of the level of education on the innovative development of the state and investigates the transformational processes of the structure of formal and non-formal education under the influence of digitalization and wartime challenges.

2 Literature Review Scientific and technological progress, economic and political dynamics in the world in recent years, and now the challenges of full-scale military actions, have led to the necessity of finding new approaches and models of education that create conditions for a continuous learning process and understanding the world. The lifelong learning model and non-formal education have become significant components of the modern educational environment. Non-formal education of youth and adults is becoming increasingly important in the context of social adaptation, ensuring sustainable and balanced development of society. Adult education is considered a social indicator of the human dimension of the state policy, one of the ways to achieve socio-economic well-being, and a tool for promoting the ideas of the information society. Therefore,the demand for non-formal education is growing [6]. Scholars and politicians are discussing lifelong learning. On the one hand, the authors declare that it can be a combination of formal and non-formal education, on the other hand, this term is used to explain any form of systematic learning that takes place outside a formal educational institution [7–12]. Non-formal education programs are growing in many countries [13]. Non-formal learning is broadly aligned with organized institutional learning models (such as schooling, universities), while informal learning describes everyday learning that people do throughout their lives. Informal learning is less well understood, despite its specific use in different contexts of international policy. Formal learning is relatively well defined in the research literature, which makes it easier to study. It fits well into narrow curriculum models (i.e., models that focus particularly on organized learning). Informal and non-formal learning do not fit well within the narrow confines of a curriculum model and require us to use a broader conceptualization of curriculum. Informal and non-formal learning are complex but powerful concepts, and they pose challenges to curriculum design. Non-formal learning is a hybrid of other forms of learning, which means that it occurs when formal and non-formal elements are combined. Learning in formal systems is simple in terms of all participants in the educational process [7, 14]. Formal learning often ends with certification and recognition of diplomas or qualifications. Formal learning can also be part of the assessment function, as this is required for formal systems [8, 15].

Formal and Non-formal Education of Ukraine

1087

Performance-based budgeting and funding in higher education is an offshoot of the public administration and service reform often associated with the term new public administration. This reform began to spread in the 1980s. Today, many European countries have introduced some form of performance-based funding in higher education [16–19]. It is obvious that the educational infrastructure of Ukraine has become a strategic target for the Russian invaders, as more than 2,600 educational institutions have been damaged by bombing and shelling, more than 330 of them have been completely destroyed. Educational institutions located on captured Ukrainian lands are unable to fulfill their primary function of providing a favorable and safe educational environment and providing support to participants in the educational process [20].

3 Methodology The paper aims to study the potential of educational services in the context of formal and non-formal education in Ukraine during the martial law period. Dynamic changes in the digitalization of the country’s economy tend to affect structural indicators in education. The paper highlights the development of formal education in terms of ownership, focuses on the increasing demand for non-formal education, and presents audience metrics for the recommended platforms by the Ministry of Education and Science. The following hypotheses were proposed to study the potential of formal and nonformal education in Ukraine in the context of economic digitalization and war: Hypothesis 1. There is a relationship between the Global Innovation Index and educational level. The level of development of the digital economy (innovative development) depends on the educational level of the population. To confirm this, we used econometric modeling and linear regression. To test the significance of linear regression models, the F-criterion was used, and then the calculated value was compared with the critical value of the corresponding Fisher distribution at a given level of significance. Hypothesis 2. There are structural changes in the supply of formal and non-formal education in Ukraine under the influence of institutional changes. To confirm Hypothesis 2, a comparative analysis was used, and the analytical indicator used was the growth (decline) coefficient. This is a relative indicator that characterizes the ratio of a given level to the level taken as a basis for comparison. Absolute indicators focus on the closure of institutions by form of ownership and time periods, the introduction of external independent evaluation, the establishment of an admission system based on open lists by educational levels and specialty through the electronic office of the applicant, reforms to eliminate low-competitive HEIs (2014– 2019), the impact of the pandemic and the digitalization of the educational process.

4 Data For the empirical study, we chose indicators of supply and demand for formal and nonformal education in Ukraine for the period of 10–12 years. The study uses data from the World Bank and the State Statistics Service of Ukraine.

1088

M. Nehrey et al.

We examine the relationship between educational level and economic development indicators to substantiate the main principles of the Concept of Development of the Digital Economy and Society of Ukraine. The analysis is based on data from the World Bank and the State Statistics Service of Ukraine. The Global Innovation Index was analyzed to investigate the correlation between educational level and the digitalization of the economy. The sample volume is 114 values. The share of the population with a complete higher education in the total population aged 25 years and older (educational level) is used as an indicator of educational attainment. To represent the innovative development of countries, a weighted average indicator is taken - the country’s score on the Global Innovation Index, which includes 81 variables that characterize in detail the innovative development of countries at different levels of economic development (hereinafter referred to as the GII). The sample includes the years 2017–2018, which were chosen to substantiate the main provisions of the Concept of Development of the Digital Economy and Society of Ukraine, the identified factors of influence, and assessments of the results. The study of formal education in Ukraine used data on the number of applicants by educational level over the past 10 years from the Unified State Database on Education, as well as the dynamics of changes in the number of educational institutions, considering the form of ownership. To study non-formal education in Ukraine, we used the recommended list of educational platforms from the Ministry of Education and Science of Ukraine, and the Similarweb service for competitive analysis and classroom metrics. The data sample includes data for September-December 2022.

5 Empirical Results 5.1 The Relationship Between Education and Indicators of the Country’s Competitiveness Education and human capital development are of course crucial for rapid, sustainable economic growth. Scientists, politicians, and experts are actively discussing the development of formal and non-formal education using various technologies [21–31]. However, the real share of human capital in national wealth remains small. In support of this, we examine gross value added by economic activity (GVA), which is calculated as the difference between output and intermediate consumption of each economic activity, reduced by the amount of financial intermediary fees. It includes the primary income generated by production participants. The share of GVA of educational services in Ukraine in 2000 amounted to 4.35%, then in 2010 - 5.6%; since 2013, it began to decline to the level of 4.4% in 2016, in 2020 it amounted to 5.0% [32]. In the OECD countries, according to Eurostat, the gross value added of education is accounted for in a separate group along with public administration, defense, health, and social services, but this indicator is not available for Ukraine. Traditionally, the UK has one of the highest indicators, its figure in 2020 was 20.94%, Germany - 19.43%, countries bordering Ukraine - Poland - 15.28%, and Hungary - 17.47%. [33]. Previous studies have determined the impact of education on basic indicators of economic development. According to the regression model, with the current level of education, Ukraine should have 95% confidence intervals of GDP per capita ranging from 36 to 48 thousand dollars, with actual GDP per capita not exceeding

Formal and Non-formal Education of Ukraine

1089

3 thousand dollars. According to the regression model, the expected increase in GDP per capita with a 1% increase in educational attainment is approximately 1 thousand dollars. Based on these data, it can also be concluded that Ukrainian education is not sufficiently effective in terms of economic development, despite having a high potential that needs to be realized. Education has a significant impact on the competitiveness and digitalization of the country’s economy. According to the approved Concept for the Development of the Digital Economy and Society of Ukraine, the key goals of the implementation of the Concept for the Development of the Digital Economy and Society of Ukraine are to increase positions in the Networked Readiness Index, Global Innovation Index, ICT Development Index, and Global Competitiveness Index [4]. Economically developed countries, such as the United States, Canada, and Singapore, have the highest positions in terms of the educational level of the population in the world. In 2019, Ukraine ranked 85th out of 141 in The Global Competitiveness Index. It shows the best results in the Skills indicator, which reflects the overall level of qualification of the workforce, as well as the quantitative and qualitative indicators of secondary and higher education (44th position). Ukraine has an unstable ranking in this indicator and is constantly in the third quantile. Ukraine has the best performance in the structure of the critical thinking of the future workforce (26th position). This leads to the conclusion that the level of human capital potential can be increased with appropriate government regulation in the further future. The study used data on the share of the country’s population with higher education (%), the country’s Global Innovation Index (score) and the country’s GDP (current US$). To test the hypothesis that there is a relationship between the educational level and the digitalization of the economy, it was hypothesized that there is a natural relationship between the global innovation index and the educational level. Highly developed countries are characterized by the relationship between education and the technological process: investments in education contribute to the development of technological innovations, thereby making human capital and labor more productive and generating income growth. As an indicator of educational attainment, we used the share of the population with a complete higher education in the total population aged 25 and older (hereinafter referred to as educational attainment). To represent the innovative development of countries, a weighted average indicator is taken - the country’s score on the Global Innovation Index, which includes 81 variables that characterize in detail the innovative development of countries at different levels of economic development (Fig. 1). The dependence of a country’s Global Innovation Index on its educational level is increasing. The functional dependence equation indicates that the expected initial level of the Global Innovation Development Index is 26.5 points. A 1% increase in the proportion of the country’s population aged 25 and older with a complete higher education will lead to an increase in the Global Innovation Development Index by an average of 1.06 points. The modeling results show that 54.19% of the variance in the Global Innovation Index is explained by changes in educational attainment. Y = 1.057x + 26.53

(1)

The regression standard error is 8.26. If the educational level indicator includes statistical information on the share of the country’s population aged 25 and older with

1090

M. Nehrey et al.

Fig. 1. Global Innovation Index and population level scatter diagram, Source: Global Innovation Index and the World Bank

incomplete higher education, Ukraine will have a value of 39.92%. In this case, the determination index will increase by 0.02%, and the standard error for the value of the Global Innovation Index will decrease by 0.02%. We tested the hypothesis that there is a correlation between the education level, the level of economic digitalization, and GDP per capita as a baseline indicator of economic development. It is assumed that GDP per capita and educational level influence the development of a country’s digital economy (Table 1.). The analysis showed that 76.84% of the variance of GDP per capita is explained by the variance of the population’s educational level and the Global Innovation Index, with their statistical significance equal to 0.05. 5.2 Analysis of the Formal Education Market of Ukraine in the Conditions of Martial Law Formal education in Ukraine is well developed. It is largely public, both in terms of demand and supply. This trend is typical for both secondary and higher education. Complete secondary education is compulsory in Ukraine, so the largest share of pupils, students, and trainees in Ukrainian educational institutions is made up of students of basic and complete secondary education (62% in the 2020–2021 academic year) [34]. The general demographic situation in the country, the number of places funded by the budget, and the cost of education are the main factors influencing the formation

Formal and Non-formal Education of Ukraine

1091

Table 1. Parameters of regression dependence between the investigated indicators Regression

Determination coefficient

Standard error

Significance level

Model “GII 1.057x + 26.53 educational level”: y is the country’s GII (points), x is the population with higher education aged 25 and over (%)

54%

8.26

0.05

Model “GII 26.37x + educational level 0.34x1 + 0.34x2 and GDP per capita”: y is the country’s GII (points), x1 is GDP per capita (thousand USD) x2 is the population with higher education aged 25 and over (%)

76.84%

5.83

0.05

Source: Global Innovation Index and the World Bank

of demand for education. The demographic situation in the country caused a certain decline in the total number of people, which occurred from 2007 to 2017. In the period from 2016 to 2021, there were trends toward a decrease in the number of secondary school graduates by an average of 4–7% annually, while the percentage of enrolled university students as a share of the total number of secondary school graduates is growing. According to the Unified Database on Education, the demand for bachelor’s degrees in 2019 increased by 6.86% compared to 2018. The consequences of military operations in the country - migration processes, mortality, occupation of the territory, destruction of educational institutions and other infrastructure related to the educational process - will have a significant and unpredictable impact. An analysis of the current structure of Ukrainian higher education institutions by the form of ownership has shown that there have been no significant changes in the education reform since 2014. A slight decrease occurred in 2018–2019 by 5.4% compared to 2015– 2016, mainly due to private HEIs. According to the register of educational entities, as of January 01, 2020, there are 151 private higher education institutions in Ukraine, and the number of higher education students enrolled in them is only 10% (72525 people) of the total number of students. At the same time, more than half of these private HEIs (86 HEIs) have less than 200 students, and only 18 private HEIs have a contingent of more than 1000 students. This can be explained by the significant influence of the ranking,

1092

M. Nehrey et al.

popularity, and elite status of the HEIs, and the quality of their educational services, which are generally considered higher in Ukraine for public HEIs than for private ones. We have analyzed data on the dynamics of the number of vocational pre-university, vocational, and higher education institutions in Ukraine by type of ownership and changes that have occurred in the time periods since the introduction of external independent evaluation (EIE), the establishment of an admission system based on open lists by educational levels and specialty to all HEIs, including the electronic office of the applicant and targeted allocation of budget places based on the specified priorities and EIT, as well as reforms to eliminate low-competitive HEIs (2014–2019), the impact of the pandemic and the digitalization of the educational process (2020–2021), the outbreak of hostilities on the territory as a result of Russia’s attack (2022–2023) (Table 2.). Table 2. Dynamics of the number of institutions of professional pre-higher, professional and higher education in Ukraine The year of blocking the subject of educational activity in the unified state electronic database on education Institutions of higher education, amount 2022-2023 сhange (blocking), total, %

In total

Form of ownership of educational institutions State Private Communal

600 17.92%

350 19.35%

210 15.32%

40 18.37%

сhange (open), total, % 21.48% 14.06% 37.10% 8.16% Institutions of higher education: blocking the subject of educational activity in the unified state electronic database on education 2010-2014

1

1

0

0

2014-2019

51

31

17

3

2020-2021

35

22

11

2

2022-2023 44 30 10 4 Institution of professional (vocational and technical) education and Institutions of vocational pre-higher education, amount 2022-2023 сhange (block), total, %

1506 23.75%

1029 19.86%

231 39.21%

246 20.90%

сhange (open), total, % 8.76% 6.70% 19.74% 3.86% Institution of professional (vocational and technical) education and Institutions of vocational pre-higher education: blocking the subject of educational activity in the unified state electronic database on education 2010-2014 0 0 0 0 2014-2019

144

79

43

22

2020-2021 2022-2023

132

67

43

22

193

109

63

21

Source: Unified State Electronic Database on Education

Formal and Non-formal Education of Ukraine

1093

Since the beginning of Russia’s full-scale invasion of Ukraine, a significant number of higher education institutions have been under threat of forced relocation or evacuation. In the areas of active hostilities (Kharkiv, Chernihiv, Sumy, Donetsk, Kherson, Zaporizhzhia, Mykolaiv regions and Kyiv), educational buildings have been damaged. Ukraine already has experience in evacuating universities. Some universities in Donetsk, Luhansk, and Crimea were moved to safer cities within these regions or to other regions (20 universities are targeted) after 2014. Based on the comparative analysis, we can conclude that the number of Institutions of professional (vocational and technical) education and Institutions of vocational pre-higher education decreased the most between 2010 and 2023. A significant number of higher education institutions were closed in the period of 2022–2023. There is currently no data on the number of affected students and teachers who were in the area of active hostilities. There are examples of targeted evacuations of students and teachers from the area of active hostilities (486 people from Kharkiv, 1700 foreign students from Sumy). According to the Ministry of Education and Science, by August 2022, more than 2,000 secondary and higher education institutions were damaged and more than 220 were destroyed, including 131 vocational pre-university and higher education institutions. For various reasons, about 2,000 academic staff was not able to continue teaching and research after February 24, 2022, and 63 educational institutions reported a shortage of teachers. The biggest obstacles to continuing education are the lack of Internet connection (79%), the deteriorating security situation in places of study (46%), and the lack of technical means for online teaching and learning (39%). [35]. 5.3 Development of Non-formal Education in Ukraine: Study of Typical Platforms and Projects Compared to formal education, which has a clear geographical link, in the case of non-formal education, it is very difficult to draw a physical border between markets. The Ministry of Education and Science of Ukraine has recommended a list of online platforms for non-formal education [36]. A comparative analysis of the platforms was conducted based on data on the average monthly number of visitors, their structure, time of visit, and the number of pages viewed after viewing the initial login page. Data received from open web analytics services (Table 3.). The most popular platforms in terms of audience metrics are www.coursera.org and www.udemy.com, which have a significant number of visitors and page views. The most dynamic platform in terms of the number of users is openuped.eu. Ukrainians prefer international platforms for learning foreign languages. Domestic online platforms were used to study narrowly specialized areas. This can be explained by the monetization of the international non-formal education market. Typical monetization options include selling a program of a set of courses, selling a course, selling a subscription for a period, selling certificates when the course is provided free of charge, selling additional services such as consultations, assignment checking, etc., selling visitor data to advertisers for targeted advertising, processing analytics on the uploaded content and compiling the necessary research for a fee, providing additional paid services such as material storage and organizing discussion platforms.

1094

M. Nehrey et al.

Table 3. Comparative analysis of educational platforms for non-formal education (audience metrics during October-December 2022) №

Name of the project Type (site)

Monthly visits

The structure Visit of visitors by Duration geographical feature (most)

Bounce Rate

Pages / Visit

The average changes in visitors

Country of origin: USA www.coursera.org

Fixed time

52.83M

USA -21%, 00:11:06 India -12%, others – 67%

37.17%

8.16

1.61%

www.khanacade my.org

Open schedule

43.29M

USA – 47%, 00:08:02 India – 6%, others – 47%

49.42%

5.29

−14.95%

www.edx.org

Fixed time

13.17M

USA – 18%, 00:06:45 India –10%, others – 72%

42.36%

6.30

−1.62%

www.canvas.net

Open schedule

975,017

USA – 51%, 00:01:10 others – 49%

58.96%

2.46

−4.71%

www.udemy.com

Fixed time

101.3M

USA – 12%, 00:10:09 India –18%, others – 70%

38%

5.73

1.96%

online.stanford.edu

Mix

855,722

USA – 16%, 00:01:27 India –9%, others – 75%

51.33%

3.66

0.77%

www.codecademy. com

Open schedule

8.542M

USA – 28%, UK, India –6%, others – 60%

00:09:07

44.18%

5.62

−1.24%

www.ted.com

Open schedule

10.57M

USA – 18%, 00:03:32 India –5%, others – 77%

57.22%

2.80

−11.63%

Country of origin: United Kingdom www.futurelearn. com

Open schedule

3.750M

UK – 20%, 00:04:53 India –9%, USA – 8%, others – 63%

58.56%

4.87

−11.08%

uk.duolingo.com

Fixed time

168,694

Ukraine 00:00:43 – 33%, Canada – 15%, others – 52%

41.88%

3.11

4.51%

(continued)

Formal and Non-formal Education of Ukraine

1095

Table 3. (continued) №

Name of the project Type (site)

Monthly visits

The structure Visit of visitors by Duration geographical feature (most)

Bounce Rate

Pages / Visit

The average changes in visitors

Open schedule

45,356

Russia, 00:03:58 Germany – 18%„ India – 14%, others – 50%

50.76%

3.58

−40.4%

Fixed time

4.454M

Egypt – 21%, 00:09:51 USA – 13%, India -8%, others – 58%

36.53%

7.90

−1.9%

Mix

8,214

Brazil - 75%, 00:01:10 Italy – 6%, others –19 %

66.87%

1.60

75.89%

prometheus.org.ua

Mix

782,576

Ukraine 00:06:45 –88%, Poland – 3%, others –9 %

49.48%

5.15

−10.86%

www.ed-era.com

Open schedule

456,178

Ukraine 00:05:43 –92%, Poland – 3%, others –5 %

53.72%

4.66

−18.85%

vumonline.ua

Open schedule

28,469

Ukraine 00:07:30 –97%, others –3 %

31.92%

6.92

−9.85%

https://osvita.diia. gov.ua/

Open schedule

302,623

Ukraine –87%, UK, Germany, Netherlands – 2%, others –7 %

53.65%

4.23

−3.33%

Country of origin: Germany iversity.org

Country of origin: Egypt www.udacity.com

Country of origin: Brazil openuped.eu

Country of origin: Ukraine

00:02:21

Source: Ministry of Education and Culture of Ukraine and the Similarweb service

One of the initiatives is the Diia. Digital Education platform supported by the Ministry and the Committee for Digital Transformation of Ukraine, which offers free education in an innovative format of educational series. In total, the portal already has more than 75 educational series on digital literacy for lawyers, teachers, doctors, journalists, civil servants, schoolchildren, etc., which increases the level of competitiveness in the labor market.

1096

M. Nehrey et al.

Since the beginning of 2022, there have been more than 10 million internally displaced persons in Ukraine. Despite the full-scale war, Ukrainian IT companies are maintaining their capacities and developing, so they need qualified employees. At the beginning of the war, there was a shortage of about 30,000 IT specialists to meet the market demand. In 2022, the IT Generation project was launched. Almost 48 thousand Ukrainians took part in the program, submitting more than 211 thousand applications. The high demand for educational retraining programs in Ukraine signals the relevance of non-formal education in the current Ukrainian context. On the other hand, the demand also indicates trends in the labor market, which will necessarily be accompanied by an increase in the level of human potential realization and digitalization of economic processes, and thus the country’s economic development.

6 Summary and Conclusion The study accepts the hypothesis that the digitalization of the economy depends on the level of education. 54% of the variance of the Global Innovation Index is determined by the variance of educational attainment. At first sight, this may not seem like a large enough value. However, Global Innovation Index includes 81 indicators (7 sub-indices) in its calculation, so explaining about 54% of the variation in countries can be considered quite significant. The study of the impact of educational level and GDP per capita on the Global Innovation Index showed that 76.84% of the variance in GDP per capita is explained by the variance in educational level and the Global Innovation Index. We conclude that Ukraine has the potential to have a Global Innovation Index about 30% higher than the actual level, and a GDP per capita 86.21% higher. The analysis of Ukraine’s position in the global rankings of competitiveness and efficiency of education systems showed the need to change the views of the country’s population on the effectiveness of knowledge as a source of human capital. The trends and features of formal education in Ukraine are analyzed in detail. By all quantitative indicators, public higher education institutions significantly outperform private higher education institutions. It is important to note that despite the relatively stable number of HEIs over the past 5 years (until 2020), only 12% of all HEIs are in sufficient demand (70% of students study there). Since the start of the full-scale invasion of Ukraine in 2022, more than 2,000 secondary and higher education institutions have been damaged and more than 220 destroyed, including 131 vocational pre-university and higher education institutions. The biggest obstacles to continuing education are the lack of internet connection (79%), the deteriorating security situation in places of study (46%), and the lack of technical means for online teaching and learning (39%). Despite the full-scale war, the spread of non-formal education in Ukraine is accelerating through the introduction of open educational platforms and projects, and education is becoming more digitalized. Demand for foreign language learning and skills development for IT professions is growing. Current research focuses on the potential of non-formal education in Ukraine, which is growing under the influence of the economy’s digitalization and the effects of the war. Furthermore, the emphasis is on the demand structure of formal education, which is mainly state-funded and thus subject to strict legislative budget constraints. Further

Formal and Non-formal Education of Ukraine

1097

analysis will consider the impact of individual factors on the structural indicators of the education market and the simulation of the main scenarios for the Ukrainian education system, considering the current challenges.

References 1. On the approval of the Concept for the Development of the Digital Economy and Society of Ukraine for 2018–2020 and the Approval of the Plan of Measures for its Implementation. https://zakon.rada.gov.ua/laws/show/67-2018-p 2. Volkova, N.P., Rizun, N.O., Nehrey, MV.: Data science: opportunities to transform education. In: CEUR Workshop Proceedings, vol. 2433, pp. 48–73 (2019). http://ceur-ws.org/ 3. Skrypnyk, A., Klimenko, N., Kostenko, I.: The formation of digital competence for the population as a way to economic growth. Inf. Technol. Learn. Tools 78(4), 278–297 (2020) 4. Global Innovation Index (GII). World Intellectual Property Organization. https://www.wipo. int/global_innovation_index/en/ 5. Coombs, P.H.: World Educational Crisis: a Systems Approach. Oxford University Press, New York (1968) 6. Hoppers, W.: Non-formal Education and Basic Education Reform: a Conceptual Review. UNESCO IIEP (2006) 7. Pienimäki, M., Kinnula, M., Iivari, N.: Finding fun in non-formal technology education. Int. J. Child-Comput. Interact. 29, 100283 (2021) 8. Johnson, M., Majewska, D.: Formal, Non-formal, and Informal Learning: What are they, and how can we Research them?. Cambridge University Press & Assessment Research Report (2022) 9. Romi, S., Schmida, M.: Non-formal education: a major educational force in postmodern era. Cambridge J. Educ. 39(2), 257–273 (2009) 10. Kucherova, H., Honcharenko, Y., Ocheretin, D., Bilska, O.: Fuzzy logic model of usability of websites of higher education institutions in the context of digitalization of educational services. Neuro-Fuzzy Model. Tech. Econ. 10, 119–135 (2021). https://doi.org/10.33111/ nfmte.2021.119 11. Davydenko, N., Buriak, A., Titenko, Z.: Financial support for the development of innovation activities. Intell. Econ. 13(2), 144–151 (2019). https://doi.org/10.13165/IE-19-13-2-06 12. Glazunova, O., Morze, N., Golub, B., Burov, O., Voloshyna, T., Parhom, O.: Learning style identification system: design and data analysis. ICT Educ. Res. Indus. Appl. Integr. Harmonization Knowl. Trans. 2732(2732), 793–807 (2020) 13. Rogers, A.: Non-Formal Education. Springer, New York (2005) 14. Yeasmin, N., Uusiautti, S., Määttä, K.: Is non-formal learning a solution to enhance immigrant children’s empowerment in Northern Finnish communities? Migr. Dev. 1–19 (2020) 15. Alnajjar, E.A.M.: The impact of a proposed science informal curriculum on students’ achievement and attitudes during the covid-19. Int. J. Early Childhood Spec. Educ. (INT-JECSE), 13(2), 882–896 (2021) 16. Herbst, M.: Performance-based funding, higher education in Europe. In: Teixeira, P.N., Shin, J.C. (eds.) The International Encyclopedia of Higher Education Systems and Institutions. Springer, Dordrecht (2020) 17. Skrypnyk, A., Klimenko, N., Kostenko, I.: Indicative cost of educational services as a way to optimize Ukrainian higher education. Euro. Cooper. 4(48) (2020) 18. Rigopoulos, G.: Assessment and feedback as predictors for student satisfaction in UK higher education. Int. J. Mod. Educ. Comput. Sci. (IJMECS) 14(5), 1–9 (2022)

1098

M. Nehrey et al.

19. Trianggoro, W.: Tony a the role of curriculum and incubator towards new venture creation in information technology international. J. Educ. Manage. Eng. (IJEME) 9(5), 39–49 (2019) 20. Education and war in Ukraine Cedos. https://cedos.org.ua/researches/osvita-i-vijna-v-ukr ayini-24-lyutogo-1-kvitnya-2022/#visa_osvita 21. Volkova, N.P., Rizun, N.O., Nehrey, M.V.: Data science: opportunities to transform education. In: Proceedings of the 6th Workshop on Cloud Technologies in Education (CTE 2018), Kryvyi Rih, Ukraine, 21 December 2018, CEURWS.org, pp. 48–73 (2018) 22. Shakhovska, N., Montenegro, S., Kryvenchuk, Y., Zakharchuk, M.: The neurocontroller for satellite rotation. Int. J. Intell. Syst. Appl. 11(3), 1–10 (2019) 23. Izonin, I., Tkachenko, R., Shakhovska, N., Lotoshynska, N.: The additive input-doubling method based on the SVR with nonlinear kernels: small data approach. Symmetry 13(4), 612 (2021). https://doi.org/10.3390/sym13040612 24. Mochurad, L., Kotsiumbas, O., Protsyk, I.: A model for weather forecasting based on parallel calculations. In: Advances in Artificial Systems for Medicine and Education VI 21 Jan 2023, pp. 35–46. Cham, Springer Nature Switzerland (2023). https://doi.org/10.1007/978-3-03124468-1_4 25. Granaturov, V., Kaptur, V., Politova, I.: Determination of tariffs on telecommunication services based on modeling the cost of their providing: methodological and practical aspects of application. Econ. Ann. XXI. 156(1–2), 83–87 (2016) 26. Matviychuk, A., Lukianenko, O., Miroshnychenko, I.: Neuro-fuzzy model of country’s investment potential assessment. Fuzzy Econ. Rev. 24(2), 65–88 (2019). https://doi.org/10.25102/ fer.2019.02.04 27. Kobets, V., Yatsenko, V.: Influence of the fourth industrial revolution on divergence and convergence of economic inequality for various countries. Neuro-Fuzzy Model. Tech. Econ. 8, 124–146 (2019). https://doi.org/10.33111/nfmte.2019.124 28. Babenko, V., Zomchak, L., Nehrey, M., Salem, A.-B., Nakisko, O.: Agritech startup ecosystem in ukraine: ideas and realization. In: Magdi, D.A., Helmy, Y.K., Mamdouh, M., Joshi, A. (eds.) Digital Transformation Technology. LNNS, vol. 224, pp. 311–322. Springer, Singapore (2022). https://doi.org/10.1007/978-981-16-2275-5_19 29. Munawar, H.S., Khalid, U., Jilani, R., Maqsood, A.: Version management by time based approach in modern era. Int. J. Educ. Manag. Eng. (IJEME), 7(4), 13–20 (2017) 30. Platforms for improving skills and self-development. https://mon.gov.ua/ua/news/platformidlya-vdoskonalennya-navichok-i-samorozvitku 31. Imanov, G., Hajizadeh, M.: The quality assessment of the development of information economy in the republic of Azerbaijan. Neuro-Fuzzy Model. Tech. Econ. 7, 111–126 (2018). https://doi.org/10.33111/nfmte.2018.111 32. State Statistics Service of Ukraine. http://www.ukrstat.gov.ua/ 33. World Bank. http://www.worldbank.org/ 34. World Economic Forum. https://www.weforum.org 35. Klymenko, N., Nosovets, O., Sokolenko, L., Hryshchenko, O., Pisochenko, T.: Off-balance accounting in the modern information system of an enterprise. Acad. Acc. Financ. Stud. J. 23 (2019) 36. Karunaratne, N.D.: The globalization-deglobalization policy conundrum. Mod. Econ. 3(4), 373–383 (2012)

Impact of YouTube Videos in Promoting Learning of Kinematics Graphs in Tanzanian Teachers’ Training Colleges Beni Mbwile1,2(B) and Celestin Ntivuguruzwa1 1 African Centre of Excellence for Innovative Teaching and Learning Mathematics and Science

(ACEITLMS), University of Rwanda-College of Education (UR-CE), Kayonza, Rwanda [email protected] 2 University of Dar es Salaam-Mkwawa University College of Education, Iringa, Tanzania

Abstract. Tanzanian students have been performing very low on the topic of motion in a straight line which consists of kinematics graphs and formulas for several years. Their low performance is attributed to the dominance of teachercentred teaching methods such as the lecture method which is contrary to what is being advocated in the current competence-based curriculum. This study investigated the impact of YouTube videos as one of the learner-centred strategies that can promote students learning of kinematics graphs in Tanzanian teachers’ training colleges. 150 pre-service Physics teachers from two colleges were distributed into two groups and taught through YouTube videos and lecture methods. Developed and validated physics teachers’ kinematics graphs conceptual questions were used to measure pre-service teachers’ achievement by administering tests before and after interventions. Simple descriptive and inferential statistics were analysed by using Statistical Package for Social Sciences version 22 and Microsoft Excel. The group of pre-service teachers who were taught kinematics graphs by using YouTube videos performed better than those taught through lectures by achieving average normalised learning gains of 22.87% compared to 1.54% for lecture class on the post-test. Also, there was a statistically significant difference (p-value < 0.001) in performance during the post-test between the lecture method and YouTube video classes while there was no significant difference during the pretest. The study recommends that Physics educators should incorporate YouTube videos in teaching for the adequate performance of students in kinematics graphs. Keywords: Kinematics graphs · pre-service teachers · teachers’ training colleges · YouTube videos

1 Introduction Improving the quality of education through the use of technology is becoming increasingly important in educational curricula [1]. Technology always has been part of the teaching and learning process since its inception, whether it was hard copies, writing

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 1099–1109, 2023. https://doi.org/10.1007/978-3-031-36118-0_93

1100

B. Mbwile et al.

instruments, or audio-visual media [2]. To cope with the growing technological needs of societies, most countries have employed efforts through curricula reforms [3]. The efforts Tanzania underwent to improve the teaching of science began during the Arusha Declaration in 1967 when the education system had to adopt the philosophy of Education for Self-Reliance [4]. The philosophy emphasized a need for curriculum reform that will enable learners to become creative thinkers, independent, and able to apply theory to real-life [3]. There are four major reforms of curricula from the year 1961 to 2010 which took place in Tanzania [5]. The reforms aimed at improving the teaching and learning of all subjects with more weight being given to Science, Technology, Engineering, and Mathematics (STEM). Despite all efforts that the government of Tanzania is putting in to improve the way STEM is being taught particularly teaching physics, students’ performance in the subject is the lowest compared to other science subjects like chemistry and biology [6, 7]. There are several topics in which students perform very low and hence contribute to the overall poor performance of students in physics, one of them is motion in a Straight line [8]. Concepts of motion in a straight line are presented as kinematics formulas and displayed as kinematics graphs [9]. Kinematics graphs consist of position, velocity, and acceleration as the ordinate and time as the abscissa. There are numerals fundamental errors found when students are plotting, interpreting, and analysing kinematics graphs. For example, in illustrating the position-time graph students often make mistakes in plotting graphs of rest objects [10], difficulty in interpreting area under kinematics graphs [11], and confusion between slope and height [12]. In Tanzania, the topic of motion in a straight line is taught to form two students. Results from the National Examination Council of Tanzania indicate that students are not performing well on the topic in both Form Four and Form Two national examinations [13]. For example, according to [14–18], students had a low performance for consecutively five years in the Form Two nation assessment as shown in Fig. 1.

Fig. 1. Students’ performance on motion in a straight-line topic

Impact of YouTube Videos in Promoting Learning of Kinematics

1101

Figure 1 above entails students’ highest average performance for consecutively five years was 20.9% and the lowest was 11.7% in 2018 and 2020 respectively. The performance of students is considered to be good, average or weak if their percentage lies in the interval of 65–100, 30–64, and 0–29 respectively [15]. Therefore, for five years students’ achievements on the topic were weak. Several factors are associated with students’ poor performance in kinematics such as students’ insufficient knowledge of reading graphs, low ability to solve tests in graphical form, and low understanding of the formula used to describe graphs [10]. Other factors not fixed to kinematics include students’ negative attitudes toward physics [19] and the dominance of traditional methods of teaching like lecturing and taking notes [20, 21]. Traditional methods make teaching less interactive and more difficult for learners to understand. Thus, deliberate efforts are required to increase students’ performance in the topic and Physics in general. One of the efforts is to use YouTube Videos to teach motion in a straight line. YouTube’s website was founded in February 2005 and it enables viewers to upload anything from video clips to full-length videos [22]. Sometimes teachers face the challenge of a shortage of teaching and learning resources while thousands of useful videos are uploaded daily on YouTube [23]. Most of these videos have free access, what is required is an internet connection. The benefit of YouTube videos is that they can be used as a teaching tool (students watch YouTube videos related to the subject) and as a teacher resource (teachers access materials for teaching) [22]. Moreover, YouTube and other social networking sites are being used by schools and university faculties to disseminate information [24]. The recent impact of the COVID-19 outbreak which imposed a lockdown on the entire world forced some education institutions to use online teaching websites, YouTube included [25, 26]. Online mode of teaching and learning was somehow available already but attracted intense attention during the period of COVID-19 [27]. Different ways such as video conferencing and audio were used by educators during online learning. Sometimes educators used to upload videos on learning websites like YouTube for students to use [28]. Little is known in Tanzania about the usefulness of YouTube videos in facilitating the teaching and learning of kinematics graphs. Therefore, this study investigated the impact of YouTube videos in promoting learning kinematics graphs in Tanzanian teachers’ training colleges. 1.1 Research Objectives • Assess significant difference in performance between pre-service teachers in YouTube videos and lecture method classes during pre-test • Find out significant difference in performance between pre-service teachers in YouTube videos and lecture method classes during post-test 1.2 Research Hypothesis There is no statistically significant difference in performance between pre-service teachers in YouTube videos and lecture classes

1102

B. Mbwile et al.

1.3 Significance of the Study The findings of this study will help tutors in colleges, pre-service teachers and teachers in general to widen the opportunity of using YouTube videos in teaching kinematics graphs as well as other topics in physics subject. In addition, findings will contribute to the existing body of knowledge through publication whereby future researchers can use them as a reference.

2 Literature Review 2.1 YouTube Videos as a Teaching Method The study by [23] revealed that YouTube videos can play a part in classroom instruction. Through YouTube videos, new concepts can be presented and information can be disseminated. In addition, it was revealed that University faculties use YouTube as a medium for delivering lectures to learners [24]. Also, [29] found that teachers use YouTube to display students’ designed art crafts. Furthermore, [30] discovered that teachers use YouTube to keep recorded videos during students’ science experiments. Moreover, [31] indicated that students who are far away from normal classrooms would still continue their studies by watching videos uploaded on YouTube. Meanwhile, [32] mentioned that video lectures tend to reduce loneliness by increasing students’ attention towards the subject matter being presented. Similarly, [33] revealed the inclusion of videos in teaching and learning is of paramount benefit in tests and examinations. Likewise, it was argued that investment in YouTube technology can be used for training teachers and saving costs related to face-to-face training [21]. YouTube videos are treasures of resources of materials which can stimulate multiple senses of students and enhance learning [23]. Therefore, learning through YouTube videos can aid the understanding of the subject matter to students. 2.2 Challenges of Kinematics Graphs Numerous findings have shown students’ difficulties when dealing with kinematics graphs. For example, students often make mistakes in plotting graphs of rest objects [10]. The study by [34] revealed two kinds of difficulties including difficulty in connecting graphs to physical concepts and misconceptions in connecting graphs to the real world. [9] revealed students’ difficulties with kinematics graphs such as variable confusion, graph transformation, and forming graphs from given kinematics equations. The study further noticed that some students were able to calculate slopes of straight-line graphs which start from the origin but fail when it does not pass through zero. Meanwhile, [12, 35], found that very few students could understand what the slope of a line graph represents. Findings by [11] reported four difficulties facing students when dealing with kinematics graphs; difficulties in calculating slopes from curved graphs, challenges in describing shapes of kinematics graphs, misconceptions in converting between kinematics graphs, and difficulties in interpreting areas under kinematics graphs. Therefore, finding the means of addressing students’ challenges on kinematics graphs is very important.

Impact of YouTube Videos in Promoting Learning of Kinematics

1103

3 Research Methodology 3.1 Approach and Design This study adopted the quantitative research approach with an experimental research design. In this design, one group of participants (experimental) receives an intervention which is the focus of the researcher and the other (control) receives interventions in normal practices or does not at all [36]. Therefore, the experimental design having control and experimental groups was used in this study. 3.2 Research Participants A total of 150 respondents comprising 75 pre-service Physics teachers from each of the 2 selected teachers’ training colleges in Tanzania participated in the study. Half of the respondents were in year 1 and the other half were in year 2 perusing diploma teachers’ programmes by majoring in Physics subject. Males were 118 and females were 32 with ages ranging from 21 to 23. Pre-service Physics teachers were targeted because they are future teachers and each year take part in teaching practice in ordinary-level secondary schools where kinematics graphs are taught. All respondents were sampled randomly from their respective classes and two teachers’ training colleges. 3.3 Instruments and Validation Pre-service teachers were assessed by using the developed and validated kinematics graphs conceptual questions. Some questions were adapted from previous researchers [9] and others were constructed. To ensure validity and reliability, multiple-choice questions were examined by a panel of 11 Physics experts from Iringa, Tanzania. In addition, test items were piloted followed by pre-service Physics teachers’ focus group discussion and then re-piloted after 3 months. Also, test re-test reliability revealed a strong correlation with correlation coefficients of .783 and .878 for single and average measures respectively. 3.4 Data Collection Procedures Two teachers’ training colleges (each with 75 participants) were involved, one being the control group and the other as the experimental group. Pre-service Physics teachers in experimental and control groups were taught using YouTube videos and lecture methods respectively. The process used for this study is shown in Fig. 2 below. In Fig. 2 above, the pre-test was administered to the control and experimental groups. It was given to assess whether groups had a similar level of understanding and misconceptions about kinematics graphs. For 10 weeks, pre-service teachers in the experimental group received interventions through a series of YouTube Videos accessed at www.You Tube.com while in the control group were taught through a series of lectures. A total of 34 carefully selected YouTube Videos were used to teach pre-service teachers kinematics graphs. Four Physics tutors (2 per group) were involved in the teaching process. In the experimental group, pre-service teachers were divided into 5 groups, each with 15 respondents.

1104

B. Mbwile et al.

Fig. 2. The process of data collection for control and experimental groups

3.5 Data Analysis Procedures Simple descriptive and inferential statistics were analysed by using Statistical Package for Social Sciences version 22 and Microsoft Excel. The study also compared the average normalised learning gain scores according to interventions that were given for each group. Average normalised learning gain is the percentage ratio of the mean difference between the post- and pre-test over the mean difference between the highest expected score (100%) and pre-test score [37].

4 Results and Discussions 4.1 Results The overall pre-service teachers’ performance for two interventions indicates that YouTube video interventions improved pre-service teachers’ performance more than the lecture method. Table 1 below summarises the mean, standard deviation, minimum, and maximum performance of pre-service teachers for pre-and post-test in all groups. Table 1. The overall pre-service teachers’ achievements from two interventions Pre-service Teachers’ Performance S/N

Interventions

N

Test

Mean

STD

Minimum

Maximum

1

YouTube Videos

75

Pre-test

37.12

10.3

12

44

64

Post-test

51.50

9.9

32

72

75

Pre-test

36.69

7.9

24

48

63

Post-test

37.65

6.95

28

52

2

Lecture Method

Table 1 above indicates that all interventions improved pre-service teachers’ performance, but performance in the lecture class was not significant i.e., ranging from a mean score of 36.69 to 37.65. In contrast, YouTube Videos interventions provided a considerable improvement in pre-service teachers’ achievement which ranged from a mean score

Impact of YouTube Videos in Promoting Learning of Kinematics

1105

of 37.12 to 51.5. The standard deviation for the lecture method group was smaller than the YouTube Videos group in pre-and post-test. While the maximum standard deviation was 10.3 for the pre-test of the YouTube group, the minimum was 6.9 for the post-test in the lecture Method. The highest pre-service teachers’ score for post-test in the lecture group was 52 out of 100 while in the YouTube video class was 72 out of 100. Pre-service teachers in the YouTube video class had a mean difference of 14.38 and those in the lecture class had a mean difference of 0.96. The mean difference between the YouTube class and the lecture, class equate to an average normalized learning gain of 22.87% and 1.54% respectively. Figure 3. Shows average normalized learning gain and mean scores for pre-and post-test according to the interventions given per group.

Fig. 3. Teaching interventions and outcome results

The reports from Fig. 3 indicate that there was a significant gain in the YouTube videos class compared to the lecture class. The findings suggest that pre-service teachers who were taught by using YouTube videos significantly outperformed those who were taught by using the lecture method. Meanwhile, Pre-service teachers’ performance across groups (independent samples test) indicates no statistical significance difference in performance between pre-service teachers who were taught by using YouTube videos and those taught through the lecture method during the pre-test. However, findings indicate a statistically significant difference in performance between pre-service teachers in the YouTube videos class and lecture class for the post-test as shown in Table 2 below. Table 2. Pre-service teachers’ performance in pre-and post-test across groups Test

Groups

N

Mean

Pre-test

Control

75

36.69

Experimental

75

37.12

Control

63

37.65

6.95

Experimental

64

51.50

9.91

Post-test

STD 7.90

T

Df

Sig. (2-tailed)

Effect Size

−0.285

148

0.776

0.047

−9.102

113

0.05 and p-value < 0.001) implies no statistically significant difference between pre-service teachers in the experimental group and those in the control group during the pre-test while there was a statistically significant difference for the post-test. These findings signify that before interventions pre-service teachers in both the control and experimental groups had a similar understanding of kinematics graphs. Similarly, the effect size of 0.047 and 1.62 for the pre-and post-test indicates low and high effect respectively. According to [38], the size effect below 0.2 is regarded as a small effect size, above 0.5 is medium, and above 0.8 is a large effect size. Therefore, YouTube videos produced a higher effect among pre-service teachers than the effect produced by lectures. 4.2 Discussions YouTube videos as a learner-centred method of teaching and learning were shown to enhance pre-service teachers’ conceptual understanding of kinematics graphs more than the lecture method. Therefore, the null hypothesis, which states to be no statistically significant difference between pre-service Physics teachers who were taught through YouTube videos and those through the lecture method was rejected. YouTube videos instructional strategy received a larger effect size, higher mean performance and statistically significant increase (p . < faculty >< year of admission>@ < high school domain>.edu.ua. The next step was to find the last names of the students, their faculties, and the year of entry to generate a pool of addresses. This information could be obtained from admission orders, which are also contained on the website of the educational institution, as well as from publicly available national databases of entrants of all higher education institutions [30, 31]. The correctness of the generated e-mail addresses from the source data and the existence of the corresponding addresses in the e-mail domain was checked using the Burpsuite package. It was found that the feature of interaction between the client and the server part of the Gmail application allows you to enter the address field and check its existence before entering the password. Therefore, the server responses allow you to check the existence of a pool of generated addresses. It was planned to send letters to as many targets as possible at the same time. This was intended to preclude the possibility of communication between targets if someone discovered the malicious nature of the letter and wished to inform others about it. Also, it allowed testing the reaction of system administrators to the attack on a large number of targets. According to the content, the phishing letter was supposed to come from the educational department of the university and contained a request to get acquainted with the rules of academic integrity when taking exams. The mailing was carried out only to those students who enrolled in the current year and did not yet have an examination session. This should have justified the sending of a letter of acquaintance with the rules of academic integrity. The letter contained a link to the Google Form you had to fill out. Given the above ethical considerations, several markers were left in the letters sent, which should help the attentive user to doubt the authenticity of the letter. The user who noticed one or more markers, according to the plan, should not follow the link from the letter, and after going—should not enter and send information about themselves. Researchers note several signs by which recipients of letters can assume that the letter is phishing [33]. We have selected features from those that affect the success of phishing [34] but can be noticed by the average user without the use of special knowledge given the ethical limitations of the experiment (see Table 1). Our task was to test the vulnerability of targets to a real phishing attack, so the number of markers that can be contained in one letter corresponded to our idea of what a real phishing letter might look like.

Social Engineering Penetration Testing

1139

Table 1. Category and description of the phishing markers implemented into the email Marker

Generalized description

Familiarity

Sending a letter, not from a corporate email address. In this institution, the teaching and support staff has e-mail addresses provided in the domain of the institution of the type < name > @ < high school domain >.edu.ua. The sender’s address was different

Email address policy

To send letters, the address of the type of science. @gmail.com was registered, i. e. with a hint of the scientific department. At the same time, in an educational institution, science and education are separated and have different meanings, as students know. The letter was not about science, but about education. The signature in the letter is formulated as “educational department of the university,” not scientific

Visual perception

The letter did not mention the university’s name, did not use its logo, and did not provide the name or contact details of the sender, which is not typical for this type of letter

Inconsistency of parts

The title of the letter was formulated to attract the attention of the student, so it contained an instruction to control learning. At the same time, the letter contained a request for information on academic integrity

Atypical structure

The need to send two Google Forms in a row separately is an excessive complication and not justified by the declared purpose of the survey

Attentiveness to details The text of Google Forms contained a large text of the document, which students usually do not read and only mark the checkbox to read it. There was a disclaimer between the text and the phishing field for the student to enter their data stating that the form was part of a phishing vulnerability study and that the sender of the letter did not collect any personal data. The disclaimer was presented in the same font as the text of the main document, which was required of the student

Indicators of successful phishing of a specific target (recipient of the letter) were selected as follows: • Go to an external link from a phishing letter (link click). In this case, the letter was sent by an unknown sender and in a real situation, when you follow the link, for example, malware could be downloaded [32]. • Google Forms contained two fields for entering your email addresses and one field for entering a user’s password. Correct completion of such forms by the recipient

1140

R. Marusenko et al.

of the phishing letter and sending of data to an unknown person also testified to the success of phishing (data sent). An integral indicator of the success of the attack was the time of its detection and the effectiveness of response measures by the school, as well as the cessation of users’ sending of their data. The appearance of the sheet is shown in Fig. 2.

Fig. 2. Sample of the phishing letter with a link to Google Form (in Ukrainian)

Please note that by providing students with additional phishing markers, it was decided to register an address for sending emails to science < high school domain> @gmail.com. At the same time, we tested the possibility of allowing such an address. Thus, it was suggested that despite the vulnerability fix discovered by T. Cotten [35], today similar actions with modifications of the list header by separate HTML tags allow the hiding of the sender’s address. Modification of the letter changes the result of its processing by the recipient’s browser. While the sender’s address still contains the letter in the body, the default browsers do not display it in either the mobile or desktop versions of Gmail. Such a letter is received with a successful check of SPF, DKIM, and DMARC filters and without prior notice to recipients or senders. A list of mystical links to the Google Form with the first document to learn about the rules of academic volunteering. After filling it out and sending it after switching to other forms. This structure allowed us to test two signs of phishing success. Students who followed the link from the list initially provided only the e-mail address. Based on this fact, we could estimate that (a) the recipient of the list followed the external links from the list from a stranger; (b) the student failed to read the document and sent an unknown personal data file. Another form contained more obvious signs of phishing— the password field, so we predicted a large number of failures to enter it. Indentation,

Social Engineering Penetration Testing

1141

even when sending messages of another form of experimenters, is a small result of filling in the first and can find out that the student goes to a third-party address from the given list. The design of the two Google forms, which contain a link to the letter, was as follows (Fig. 3). The second form contained an optional field for entering the password. It was not specified what password was meant. First, the need to pass the student’s password to an outsider could not be motivated. Secondly, the educational part of the institution has the necessary (increased compared to the student) access rights to the e-learning environment, the password of which could be transmitted by the student. Therefore, this field should have become an additional indicator of phishing for the student.

Fig. 3. Structure of the first (a) and the second (b) phishing Google Forms

5 Research Results For phishing, it has been assumed that the target email domain has basic restrictions on sending emails, which for Gmail is 500 emails per day for a free account [36]. We assume that exceeding the limit could lead to the blocking of the mailing list, the sender’s account, and the notification of the domain administrator about unusual activity on the server. The distribution of letters was planned in a continuous block to a separate faculty. On the one hand, this would allow obtaining a homogeneous sample of responses of targets— by age and type of professional interests. On the other hand, it allowed simulation mailing by the training department, which is usually carried out simultaneously for all users. In general, it was planned to start with the faculties of the humanities and end with the faculty of computer science. Such an order was intended to prevent the rapid detection of the attack, as students of technical specialties have a broader knowledge of the field

1142

R. Marusenko et al.

of IT. University staff did not report the experiment, as it could lead to information leaks about the nature of the experiment and “socially desirable” behavior. The latter could be manifested in the desire to demonstrate better results of student attack recognition or response by IS administrators, and the experiment itself to level. 261 letters were sent. The mailing was carried out by one letter with an attached list of recipients. An unknown factor was the presence or absence of limits in the corporate mail network for the mass distribution of identical letters from one address, which also does not belong to the pool of corporate addresses. It was assumed that the second group of letters to another faculty would be sent the next day. There was a significant delay in sending the first response to the Google Form, which is not typical of this type of interaction. In a previous study, we found that students respond quickly to a letter they intend to respond to if such a letter does not require additional complex actions. The mode of time of the answers lay within the first hour after the mailing [25]. The experimenters sent a letter to their address confirming the delay, which was probably due to mass mailing from the newly registered address. We received a total of 99 responses to the first Google Form and 5 responses to the second. The distribution of response time to the first form is interesting (Fig. 4). In the first hour, 59 responses were received. The other 38 responses were received for the next 17 h (including night and morning) until the experiment was stopped. Two more responses arrived two days later. This depends on the number of answers depending on the time is well approximated by the function: N (t) = e

  π 2−t/ 100

,

(1)

where t is the response time, min. From this, we can conclude that a potential attacker gets most of the desired response from the targets in just one hour. On the other hand, IS protection personnel has significantly less time to protect at least half of the users who could potentially be attacked and compromise the school’s IS and themselves. It took only 40 min for the attacking party to get answers from 50% of users (50 people) who were exposed to this phishing. Mass distribution of phishing emails to users of one IS resembles a reverse DDoS attack. An attacker attacks many targets located outside the physical perimeter of the IS, which are associated with weak and slow information interaction. This poses difficult challenges, as university staff has very little time to detect the full range of targets and warn them of the true nature of the mailing when they detect an attack. Instant messaging tools should be used to prevent this. Because clicking on the link in the letter takes the target outside the conditionally controlled perimeter of the IS, even a complete physical blocking of the network operation of the IS is not able to save the target from the harmful effects of the attacker and the transmission of sensitive data to the attacker. Analysis of responses to Google Form 1 showed that of the 99 responses received (see Table 2). Thus, 36.8% of targets responded positively to the phishing letter and fell victim to an experimental attack. One can conclude about the success of the attack design in the form that was proposed. It is significant that among the addresses there is the address of the university lecturer, which is obvious given the different structure of the teaching addresses of this institution.

Social Engineering Penetration Testing

1143

Fig. 4. Response time for the phishing email answer

Table 2. Distribution of active response of addressees for a phishing attack Answers

Generalized description

Amount

Part, %

88

33.7

7

2.7

Contained non-corporate e-mail addresses, i. e. students provided their second address. The check showed the authenticity of all seven addresses

2

0.8

Contained a random set of letters. Probably these students recognized phishing

1

0.4

Corresponded to the postal address but is invalid

Contained addresses from the mailing pool, i. e. students provided real data

Note that the e-mail address field did not check the format of the e-mail address, which allowed you to enter any sequence of characters if desired. This was intended to record the responses of those respondents who, when filling out the form, found phishing and wanted to demonstrate it

It was introduced at the stage when the fact of mailing became public and the content of the phishing letter was analyzed by the teaching staff. Another interesting fact is the receipt of two responses two days after the mailing when its phishing nature was publicly known, and the authors of the experiment have already openly communicated this with stakeholders. This once again confirms our conclusion that a simultaneous attack on distributed targets is a threat given the difficulty of preventing all targets, even if it is detected. The second form was filled in by five respondents, four of whom entered passwords that looked like real ones, which is 1.5% of respondents. This result demonstrates a high level of document recognition with obvious signs of phishing.

1144

R. Marusenko et al.

We should also note the passionate position of one respondent who, in response to a phishing letter, sent a message that he did not immediately understand the phishing purpose of the mailing, but now realized it. There was also a question about who is the organizer of the mailing. Since the experiment had not been completed at that time, the experimenters sent a standard response thanking them for participating in the experiment and asking them to wait for it to complete. Later, the employees of the institution became aware of the mailing. The discovery of the fact of sending phishing letters made it necessary to inform the organizers of its true purpose. Further sending of letters to other faculties was not carried out, as it was no longer possible to ensure non-dissemination of information about the true purpose of the letters. The response to the experimental attack turned out to be comprehensive, involving employees from various departments and trying to find a mailing list organizer. It should be noted that despite the measures taken, two responses to the letter were received two days after the end of the experiment. Also, the peculiarity of the simultaneous attack on a large number of targets did not allow software and hardware measures to promptly warn all victims of the attack and prevent their compromise. The results can be summarized as follows. The response and mitigation of an attack require simultaneous and fairly rapid corrective action on the target of the attack. In this case, it could be alerting targets to attack. At the same time, phishing in an institution of this scale could be carried out from different addresses. In turn, this fact, as well as finding the targets of the attack in different geographical locations would not allow us to quickly identify all the victims. As shown in the experiment, the time required to “compromise” the target is short enough to manually notify the target of the attack and prevent the consequences. The software protects the perimeter before the letter hits the target, i. e. before the impact on the latter. If this influence has already occurred, then the software and hardware are powerless. This study differs from previous studies [20] in the combination of social engineering and man-made attacks. The experiment was preceded by the analysis of information messages that are sent within the corporate network of the university. The masking of the link to a Google form for collecting sensitive information was done taking into account similar emails that are sent to students regularly. This approach gives an order of magnitude better results than mailing to open sources, even taking into account the context [21]. From the experiment, it can be seen that the choice of time and subject of a phishing email stimulates recipients to actively respond. In this case, half of the responses were received in less than an hour, so the speed of response to such incidents should be no more than a quarter of an hour.

6 Conclusions and Future Work Conducting social engineering penetration testing requires a careful attitude to the ethical side of the design of the experiment, so as not to level its results on the one hand, and the other—not to cause a negative impact and protect the privacy of the targets of the experiment. Finding such a balance will always require consideration of the conditions

Social Engineering Penetration Testing

1145

of a particular IS, the specifics of users of such IS, the incident response system, and the type of information processed in the IS. Phishing is essentially a distributed attack that develops quite rapidly. An effective and timely response is a big challenge. Social engineering, like any attack, exploits the simplest and most effective possible ways to achieve a goal. In this case, it is the sending of letters, which is difficult or impossible to prohibit or restrict the filter of harmful letters. The primary measure of protection against this type of attack is currently training and raising respondents’ awareness. Developing effective and efficient programs in the conditions of the constant evolution of attacks by methods of social engineering requires constant search and improvement of methodology. In this paper, we have demonstrated a basic version of social engineering penetration testing, which should be noticed by the attentive user. Further development of staff competencies to respond to more complex types of attacks by methods of social engineering variation of testing may be the use of sender addresses with the so-called. Cousin domain, typos in the address, vulnerabilities that allow hiding it, and other techniques. Also, the text of the letter may contain more compelling instructions and may not contain markers that allow you to recognize phishing. However, in this case, researchers will again face the ethics of penetration testing. Therefore, we believe that the study of this area is far from complete. Acknowledgment. The authors are grateful to Borys Grinchenko Kyiv University administration for assistance in conducting experiments, as well as personally to the deputy head of IT in the Education Laboratory Oksana Buinytska for successfully detecting attack simulation and prompt response.

References 1. Singh, A., Kumar, A., Bharti, A.K., Singh, V.: An e-mail spam detection using stacking and voting classification methodologies. Int. J. Inf. Eng. Electr. Bus. (IJIEEB) 14(6), 27–36 (2022). https://doi.org/10.5815/ijieeb.2022.06.03 2. Ahraminezhad, A., Mojarad, M., Arfaeinia, H.: An intelligent ensemble classification method for spam diagnosis in social networks. Int. J. Intell. Syst. Appl. (IJISA), 14(1), 24–31 (2022). https://doi.org/10.5815/ijisa.2022.01.02 3. Fan, W., Kevin, L., Rong, R.: Social engineering: i-e based model of human weakness for attack and defense investigations. Int. J. Comput. Netw. Inf. Secur. (IJCNIS) 9(1), 1–11 (2017). https://doi.org/10.5815/ijcnis.2017.01.01 4. Smith, J.K., Shorter, J.D.: Penetration testing: a vital component of an information security strategy. Issues Inf. Syst. XI.1, 358–363 (2010). https://doi.org/10.48009/1_iis_2010_3 58-363 5. Jazzar, M., Yousef, R.F., Eleyan, D.: Evaluation of machine learning techniques for email spam classification. Int. J. Educ. Manage. Eng. (IJEME) 11(4), 35–42 (2021). https://doi.org/ 10.5815/ijeme.2021.04.04 6. Positive Technologies: Cybersecurity threatscape: Q3 2022. https://www.ptsecurity.com/wwen/analytics/cybersecurity-threatscape-2022-q3/. Accessed 16 Feb 2023 7. The history of penetration testing. https://alpinesecurity.com/blog/history-of-penetration-tes ting. Accessed 16 Feb 2023

1146

R. Marusenko et al.

8. Scarfone, K., et al.: Technical guide to information security testing and assessment. Recommendations of the National Institute of Standards and Technology. NIST SP800–115. https://nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication800-115.pdf. Accessed 16 Feb 2023 9. European Parliament, Regulation 2016/679 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data. https://eur-lex.eur opa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32016R0679. Accessed 16 Feb 2023 10. Payment Card Industry Data Security Standard. Requirements and security assessment procedures, version 3.0. https://www.pcisecuritystandards.org/minisite/en/docs/PCI_DSS_v3.pdf. Accessed 16 Feb 2023 11. Campbell, N., Lautenbach, B.: Telstra security report. https://www.telstra.com.au/content/ dam/shared-component-assets/tecom/campaigns/security-report/Summary-Report-2019LR.pdf. Accessed 16 Feb 2023 12. Kessel, P.: EY global information security survey. https://assets.ey.com/content/dam/eysites/ey-com/en_ca/topics/advisory/ey-global-information-security-survey-2018-19.pdf. Accessed 16 Feb 2023 13. Pescatore, J.: SANS top new attacks and threat report. https://www.sans.org/reading-room/ whitepapers/threats/top-attacks-threat-report-39520. Accessed 16 Feb 2023 14. Dimkov, T., et al.: Two methodologies for physical penetration testing using social engineering. In: 26th Annual Computer Security Applications Conference, pp. 399–408 (2010). https://doi.org/10.1145/1920261.1920319 15. Barrett, N.: Penetration testing and social engineering. Inf. Sec. Tech. Rep. 8(4), 56–64 (2003). https://doi.org/10.1016/s1363-4127(03)00007-4 16. Nguyen, T.H., Bhatia, S.: Higher education social engineering attack scenario, awareness & training model. J. Colloquium Inf. Syst. Secur. Educ. 8(1), 8 (2020) 17. Indrajit, R.E.: Social engineering framework: understanding the deception approach to human element of security. Int. J. Comput. Sci. Iss. 14(2), 8–16 (2017). https://doi.org/10.20943/012 01702.816 18. Kelm, D., Volkamer, M.: Towards a social engineering test framework. In: 11th International Workshop on Security in Information Systems, pp. 38–48 (2010). https://doi.org/10.5220/ 0004980000380048 19. Buriachok, V., et al.: Technology for information and cyber security in higher education institutions of Ukraine. Inf. Technol. Learn. Tools 77(3), 337–354 (2020). https://doi.org/10. 33407/itlt.v77i3.3424 20. Marusenko, R., Sokolov, V., Buriachok, V.: Experimental evaluation of phishing attack on high school students. In: Hu, Z., Petoukhov, S., Dychka, I., He, M. (eds.) ICCSEEA 2020. AISC, vol. 1247, pp. 668–680. Springer, Cham (2021). https://doi.org/10.1007/978-3-03055506-1_59 21. Marusenko, R., Sokolov, V., Bogachuk, I.: Method of obtaining data from open scientific sources and social engineering attack simulation. Adv. Artif. Syst. Logist. Eng. 135, 583–594 (2022). https://doi.org/10.1007/978-3-031-04809-8_53 22. Hu, Z., Buriachok, V., Sokolov, V.: Implementation of social engineering attack at institution of higher education. In: International Workshop on Cyber Hygiene, pp. 155–164 (2019) 23. Finn, P., Jakobsson, M.: Designing and conducting phishing experiments, 1–21 (2006) 24. Mouton, F., et al.: Necessity for ethics in social engineering research. Comput. Sec. 55, 114–127 (2015). https://doi.org/10.1016/j.cose.2015.09.001 25. Resnik, D.B., Finn, P.R.: Ethics and phishing experiments. Sci. Eng. Ethics 24(4), 1241–1252 (2017). https://doi.org/10.1007/s11948-017-9952-9 26. Faily, S., McAlaney, J., Iacob, C.: Ethical dilemmas and dimensions in penetration testing. In: 9th International Symposium on Human Aspects of Information Security & Assurance, pp. 233-242 (2015). https://doi.org/10.13140/rg.2.1.3897.1360

Social Engineering Penetration Testing

1147

27. Pierce, J., Jones, A., Warren, M.: Penetration testing professional ethics: a conceptual model and taxonomy. Aust. J. Inf. Syst. 13(2), 193–200 (2006). https://doi.org/10.3127/ajis.v13i2.52 28. Creasey, J., Glover, I.: A guide for running an effective penetration testing programme. https://www.crest-approved.org/wp-content/uploads/CREST-Penetration-TestingGuide.pdf. Accessed 16 Feb 2023 29. Hadnagy, C.: Social engineering: The science of human hacking (2018) 30. Introduction.EDUCATION.UA. https://vstup.osvita.ua. Accessed 16 Feb 2023 31. Applicant search service. http://abit-poisk.org.ua. Accessed 16 Feb 2023 32. Kotov, V., Massacci, F.: Anatomy of exploit kits. In: Jürjens, J., Livshits, B., Scandariato, R. (eds.) ESSoS 2013. LNCS, vol. 7781, pp. 181–196. Springer, Heidelberg (2013). https://doi. org/10.1007/978-3-642-36563-8_13 33. Parsons, K., et al.: Do users focus on the correct cues to differentiate between phishing and genuine emails? arxiv:1605.04717 34. Jampen, D., Gür, G., Sutter, T., Tellenbach, B.: Don’t click: towards an effective anti-phishing training. a comparative literature review. HCIS 10(1), 1–41 (2020). https://doi.org/10.1186/ s13673-020-00237-7 35. Cotten, T.: Ghost emails: hacking Gmail’s UX to hide the sender. https://blog.cotten.io/ghostemails-hacking-gmails-ux-to-hide-the-sender-46ef66a61eff. Accessed 16 Feb 2023 36. Google: Gmail sending limits in Google Workspace. https://support.google.com/a/answer/ 166852. Accessed 16 Feb 2023

Author Index

A Abdullayeva, Fargana J. 289 Adelodun, Abdulwasiu Bolakale Adetunji, Charles O. 721 Agrawal, Niharika 269 Ahmad, Tauseef 1057 Ahmed, Yusuf Kola 748 Ajami, Sima 299 Akhmetov, Bakhytzhan 595 Akhmetov, Berik 595 Akinyemi, Joseph D. 440 Akyol, Kemal 401 Alidadi, Tahmineh 299 Alifov, Alishir A. 682 Ananthajothy, Gabaalini 450 Anna, Synko 328 Atighehchian, Arezoo 299 B Bacherikov, Dmytro 636 Bao, Chun 1121 Bao, Qianxian 75 Barde, Snehlata 759 Bazilo, Constantine 527, 636 Boyko, Taras 788 Bubela, Tetiana 499 Bulgakova, Oleksandra 137 Burchak, Pavlo 59 C Cai, Changyu 341 Cai, Hao 913 Cao, Qian 972 Chebanyuk, Olena 514 Chen, Jianqiu 901, 1121 Chen, Yi 875 Cheng, Ni 37 D Dan, Qiong 13 Danmusa, Nasir Ayuba

748

748

David, Acheme I. 721 Deibuk, Vitaly 95 Demchuk, Olena 605 Dissanayake, Maheshi B. 450 Dolgikh, Serge 690 Dong, Lingsan 860 Dovhaniuk, Oleksii 95 Dragan, Olena 106 Drozd, Valeriia 207 Drozdenko, Lyubov 700 Dvulit, Zoriana 106 Dzhumelia, Elvira 788 E Eleftherios, Zacharioudakis

236

F Farzaliev, M. G. 682 Fedyshyn, Tetiana 499 Filimonov, Sergey 527, 636 Filimonova, Nadiia 527, 636 Filimonova, Olena 527 G Gao, Ying 875 Ghita, Bogdan 429 Gigashvili, Ketevan 260 Gnatyuk, Sergiy 371, 690 Gobov, Denys 574 Goncharenko, Tetiana 246 Gowda, Mamatha 269 Guan, Dichen 85 H Haidai, Serhii 560 Halytskyi, Daniil 236 Haque, Md. Asraful 1057 Harasymchuk, Oleh 486 Hegde, Rajalaxmi 709 Hegde, Sandeep Kumar 709

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 Z. Hu et al. (Eds.): ICCSEEA 2023, LNDECT 181, pp. 1149–1152, 2023. https://doi.org/10.1007/978-3-031-36118-0

1150

Hentosh, Lesia 645 Holovchenko, Maxim 207 Honcharenko, Oleksandr 472 Hots, Nataliya 803 Hu, Feifei 1022 Hu, Zhengbing 371 Huang, Caigui 539 Huang, Chuan 827, 845 Huang, Fang 85 Huang, Qi 419 Huang, Tongyao 1000 I Iavich, Maksim 260, 585 Ibitoye, Ayodeji O. J. 440 Ishchenko, Vitalii 815 Ivakh, Roman 815 J Janisz, Karina 371 Jiang, Lijun 1110 K Kang, Jing 913 Karacı, Abdulkadir 401 Karanwal, Shekhar 391 Karupu, Olena 939 Khajuria, Himanshu 318 Khan, Faheem Ahmed 269 Klymenko, Nataliia 1085 Kochan, Nataliia 803 Kochan, Orest 788, 815 Kochan, Roman 803 Kochan, Volodymyr 803 Kohut, Uliana 803 Kondratenko, Galyna 626 Kondratenko, Yuriy 126, 626 Kopp, Andrii 246 Kornienko, Yaroslav 560 Korniyenko, Bogdan 560 Kosovan, Ivan 163 Kostenko, Inna 1085 Kovalchuk, Pavlo 605 Kovalchuk, Volodymyr 605 Kovalchuk-Khymiuk, Liudmyla 1046 Kravchenko, Oleksandr 645 Kravets, Peter 47 Kreinovich, Vladik 126

Author Index

Kuts, Viktor 486 Kuzminykh, Ievgeniia 429 Kyrianov, Artemii 472 Kyryliuk, Taras 95

L Lebedenko, Yurii 47 Legeza, Viktor 670, 700 Lei, Meng 37 Lei, Xiang 13 Lemeshko, Oleksandr 223 Li, Guangping 3 Li, Lili 1022 Li, Liwei 885, 927 Liang, Chengquan 26 Lipska, Vlada 774 Lishchuk, Kateryna 207 Liu, Changjian 827, 845, 860 Liu, Dalong 885, 927 Liu, Yilei 13 Liu, Yuhong 550 Liubeka, Andrii 560 Lotfi, Farhad 299 Loutskii, Heorhii 472 Lu, Zhixiang 1070 Lupei, Maksym 173 Lupei, Nitsa 173 Lv, Deshen 1033

M Makinde, Samuel 721 Maksymovych, Volodymyr 486 Makulov, Kayirbek 595 Mania, Esma 260 Marcovets, Oleksandr 193 Martyniuk, Hanna 595 Marusenko, Roman 1132 Mathur, Saransh 429 Maznyk, Liana 106 Mbwile, Beni 1099 Mitsa, Oleksandr 173 Mo, Wenhao 341 Mohammadi, Reyhaneh Rasekh 299 Mohd, Shoaib 1057 Morozov, Kostiantyn 236 Moshenskyi, Andrii 657 Mukha, Iryna 207

Author Index

1151

N Navaratnam, Rudsika 450 Nayak, Biswa Prakash 318 Nehrey, Maryna 1085 Neshchadym, Oleksandr 700 Novak, Dmitriy 657 Novatskyi, Anatolii 47 Ntivuguruzwa, Celestin 1099 Nwankwo, Chukwuemeka P. 721 Nwankwo, Wilson 721 O Ojagverdiyeva, Sabira S. 289 Olawale, Adeolu Johnson 349 Oleshchenko, Liubov 59, 147, 615, 657, 670 Oleshko, Tetiana 939 Onifade, Olufade F. W. 440 Onyshchenko, Viktoriia 183 Orekhov, Sergey 246 Orlovskyi, Dmytro 246 Oyewale, Christianah T. 440 P Pakhnenko, Valeria 939 Pan, Yanfang 885, 927 Pang, Yanzhi 901, 950 Pashko, Anatolii 939 Pavlov, Alexander 207 Pazderska, Ruslana 193 Perelygin, Sergey 410 Petoukhov, Sergey V. 309 Petrovska, Iryna 499 Petryk, Mariana 193 Proskurin, Dmytro 371 Prychun, Daryna 984 Prystavka, Pylyp 371, 690 Przystupa, Krzysztof 486, 499 Pustovit, Oleksandr 472 Q Qiu, Suzhen

1070

R Rasim, Alguliyev M. 289 Romankevich, Vitaliy 236 Romankevitch, Alexei 236 Ruda, Mariia 788 Rudakova, Antonina 47 Rudakova, Hanna 47

Rusinov, Volodymyr 472 Ryabyy, Myroslav 371 S Salahudeen, Kafilat Atinuke 748 Salamon, Ivan 788 Saminu, Sani 748 Savchyn, Vitalii 645 Seliuchenko, Marian 106 Seliuchenko, Nadiia 106 Shabatura, Mariia 486 Sharkan, Vasyl 173 Shybytska, Nataliia 116 Shymkovych, Volodymyr 47 Sidenko, Ievgen 626 Singh, Chintan 318 Skladannyi, Pavlo 1132 Sleiman, Batoul 223 Sokolov, Volodymyr 1132 Solovei, Olga 574 Song, Wenyu 419 Stakhiv, Roman 486 Su, Huasheng 1110 Su, Lianju 85 Suleiman, Taofik Ahmed 748 Sun, Hanbing 827 Sun, Yun 459 T Tang, Yuexia 3 Tavrov, Danylo 1046 Teng, Xiaoli 950 The, Quan Trong 410 Thevarasa, Niluksha 450 Tian, Jia 827 Timchenko, Victor 126 Tolokonnikov, G. K. 382 Trukhov, Artem 626 Tvalavadze, Tea 260 U Ukhurebor, Kingsley E. 721 Umezuruike, Chinecherem 721 V Vargha, Sabolch 173 Vasylkivskyi, Igor 815 Vdovyn, Mariana 361 Verma, Apurva 759

1152

Verma, Vijayant 759 Vernik, Mykhailo 147 Vlasiuk, Yevhenii 183 Voitovich, Oleksandr 605 Volokyta, Artem 472, 774 Vovk, Nataliia 984 W Wang, Haiyou 75 Wang, Liyuan 419 Wang, Ruoxi 459 Wang, Shurui 419 Wang, Yue 913 Wei, Mingqi 85 Wei, Wanyi 1110 Wu, Hongyu 845, 860 Wu, Yunyue 950 X Xu, Min 13 Y Yagaliyeva, Bagdat 595 Yan, Xueqin 13 Yang, Hu 972 Yang, Xiaozhe 75

Author Index

Yang, Zheng 13 Yao, Yuntao 341 Yeremenko, Oleksandra 223 Yevdokymenko, Maryna 223 Yu, Shaohuai 419 Yu, Xunran 845, 860 Yurchuk, Iryna 163 Z Zhan, Jun 459 Zhang, Lanfang 1121 Zhang, Lei 962 Zhang, Qiang 913 Zhang, Xiaoqing 827 Zhang, Ying 26, 1033 Zhao, Lei 1110 Zhi, Yanli 341 Zhou, Qiang 550 Zhou, Qiqi 845 Zhou, Yu 341 Zhu, Haoliang 1011 Zhukov, Yuriy 626 Zomchak, Larysa 361 Zosimov, Viacheslav 137 Zubair, Abdul Rasak 349, 748 Zuo, Jing 1000