Dynamics in Logistics: Proceedings of the 7th International Conference LDIC 2020, Bremen, Germany (Lecture Notes in Logistics) 3030447820, 9783030447823
Since 2007, the biennial International Conferences on Dynamics in Logistics (LDIC) offers researchers and practitioners
130 58 45MB
English Pages 582 [575] Year 2020
Preface
Contents
Part I: Maritime Logistics and Multi-Modal Transport
Recommendations for Human Resources Development in Danube Inland Ports
Abstract
1 Introduction
2 Methodology
3 Literature Review
4 Findings
4.1 Current Human Resource Structure in Danube Ports
4.2 Current and Future Training Requirements
4.3 Personas – Description of Future Employees Types
5 Discussion and Recommendations for Human Resources Development
6 Conclusion, Limitations and Research Outlook
Acknowledgment
References
System Dynamics Modeling of Logistics Hub Capacity: The Dubai Logistics Corridor Case Study
Abstract
1 Introduction
2 Literature Analysis: System Dynamics Approach in Logistics
3 The Dubai Logistics Corridor
4 Proposed Approach
4.1 Causal Loop Diagram Representation of the Model
5 The SD Model and Case-Scenario Analyses
6 Validation
7 Implications, Limitations and Conclusion
References
Towards Intelligent Waterway Lock Control for Port Facility Optimisation
1 Introduction
2 Use Case
3 Waterway Lock Operations
4 Architecture for Intelligent Lock Control
4.1 Data Acquisition and Data Integration
4.2 Prediction Models
4.3 Multi-Agent System
5 Risk Management and Evaluation
6 Summary and Next Steps
References
On the Influence of Structural Complexity on Autonomously Controlled Automobile Terminal Processes
Abstract
1 Introduction
2 Automobile Terminal Planning Processes
3 Autonomous Control of Logistics Processes
4 Automobile Terminal Scenario
4.1 Structural Configuration of the Scenario
4.2 Modelling Incoming and Outgoing Volumes
4.3 Conventional Control Method
4.4 Autonomous Control Method
5 Simulation Results
6 Summary and Outlook
Acknowledgements
References
Literature Classification on Container Transport Systems for Inter-terminal Transportation
Abstract
1 Motivation
2 Introduction to Inter-terminal Transportation
3 Literature Review and Classification
3.1 Classification Scheme
3.2 Classification Tables
4 Research Perspectives
4.1 General
4.2 Process Integration at the Nodes
4.3 Process Integration at the Edges
5 Conclusions
References
A Simulation Study of a Storage Policy for a Container Terminal
Abstract
1 Introduction
2 Literature Review
3 Storage Policy for a Container Terminal
4 Simulation Design
5 Experiment Design and Discussion
6 Conclusion
Acknowledgment
References
Investigation of Vessel Waiting Times Using AIS Data
Abstract
1 Introduction
2 Previous Work
3 Data and Data Pre-processing
4 Results
5 Discussion
References
Resource Sharing as a Management Concept for Digital Logistics Terminals
Abstract
1 Introduction
2 Resource Sharing as a Management Concept?
2.1 Definition of a Shared Resource
2.2 Use and Provision of a Shared Resource
2.3 Conceptual Framework
2.4 Overview of Potential Applications
3 Use Case: Digital Logistics Terminal
3.1 Intermodal Transport and Digital Logistics Terminals?
3.2 Application of the Shared Resources Concept
4 Conclusion
References
Coordinating the Seaport-Hinterland Interface: Theoretical and Methodological Insights from Scientific Literature
Abstract
1 Introduction
2 Need for Coordination at the Seaport-Hinterland Interface
3 Methodology
3.1 Background on Systematic Literature Reviews
3.2 Data Collection and Analysis
4 Results
4.1 Characteristics of Literature Sample
4.2 Analysis of Central Problems in the Research Area
4.3 Analysis of Theoretical Perspectives and Methodological Approaches
5 Conclusion, Critical Reflection and Outlook
Appendix A – Search and Evaluation Process
Appendix B – References for Systematic Literature Review
References
Sensor Simulation and Evaluation for Infrastructure-Free Mobile Sensor Carrier Platforms
Abstract
1 Variety of Sensor Platform Types
2 Benefits and Applications of Mobile Infrastructure-Free Sensor Platforms
3 Requirements in Human-Machine and Machine-Machine Collaboration
4 Sensor Analysis
4.1 Time-of-Flight Sensor Errors
5 Simulation of Time-of-Flight Sensors
6 Evaluation of Simulation Results
7 Conclusion
Acknowledgements
References
Expansion Planning at Container Terminals
Abstract
1 Introduction
2 Methodology
3 Results and Discussion
3.1 Industry: Project Reflections
3.2 Case Study: Specific Scientific Examples
3.3 Method: General Discussions
4 Conclusions and Future Research Directions
Appendix: Reviewed Literature
References
A Hybrid Robust-Stochastic Optimization Approach for the Noise Pollution Routing Problem with a Heterogeneous Vehicle Fleet
Abstract
1 Introduction
2 Literature Review
3 Model Formulation
3.1 Hybrid Robust-Stochastic Optimization Methodology
3.2 Formulation of the Base Deterministic Formulation
3.3 Formulation of the Stochastic NPRP
3.4 Formulation of the Hybrid Robust-Stochastic NPRP (HRNPRP)
4 Computational Experiments
5 Conclusions
References
Part II: Supply Chain Management and Coordination
The Influence of Cognitive Biases on Supply Chain Risk Management in the Context of Digitalization Projects
Abstract
1 Introduction
2 Literature Review
2.1 Supply Chain Risk Management at Times of Digitalization
2.2 The Decision-Making Process and Cognitive Biases
3 Findings – Cognitive Biases’ Influence on SCRM
3.1 Phase 1 – Identifying
3.2 Phase 2 – Assessing and Evaluating
3.3 Phase 3 – Mitigating and Controlling
4 Handling the Human Factor in SCRM Through Debiasing
5 Conclusion
References
Flow Management Tools and Techniques for Logistics Performance: An Application to the Logistics Service Sector in Cameroon
Abstract
1 Introduction
1.1 The Concept of Logistics Performance
1.2 The Theory of Resources
1.3 The Theory of Dynamic Capabilities
1.4 Research Hypotheses
1.4.1 Application of New Solutions
1.4.2 Investment in Recruitment and Training of Qualified Personnel
2 Research Methodology
2.1 Sample and Data Collection
2.2 Measuring Variables
2.3 Factor Analyses
3 Findings
3.1 Descriptive Analysis of the Sample
3.2 Hypothesis Testing and Discussion
3.2.1 Relationship Between the Application of New Solutions and the Quality of the Logistics Service
3.2.2 Relationship Between Investment in Recruitment and Training of Qualified Personnel and Cost Reduction
3.3 Discussion and Conclusion
References
Blockchain and Risk in Supply Chain Management
Abstract
1 Introduction
2 Theoretical Background
3 Methodology
4 Conceptual Model and Discussion
5 Concluding Remarks
References
Prioritizing Supply Chain Agility Factors Using Fuzzy Analytic Network Process (FANP)
1 Introduction
2 Literature Review
3 Research Methodology
3.1 Fuzzy Analytic Network Process (FANP)
4 Conclusions
References
Influence of Supply Chain Management & Logistics in the Wake of China Pakistan Economic Corridor (CPEC) on Domestic Industry in Pakistan
Abstract
1 Introduction
2 Model Development
3 Research Methodology
3.1 Data and Sampling Technique
3.2 Reliability of Model
3.3 Confirmatory Factor Analysis (CFA)
3.4 Estimation of SEM
3.5 Model Fit
4 Results and Discussion
5 Conclusion
References
A Disruption Management Model for a Production-Inventory System Considering Green Logistics
Abstract
1 Introduction
2 Model Development
2.1 Model Formulation
2.1.1 Objective Function
3 Results and Discussion
3.1 Numerical Analysis
4 Conclusion
Acknowledgment
References
A Concept for a Consumer-Centered Sustainable Last Mile Logistics
Abstract
1 Introduction
2 Current Stage of Consumer-Centered Sustainable Logistics
2.1 Green/Sustainable City Logistics
2.2 Food Logistics from the Consumer’s Perspective
2.3 Critical Reflection on Current State of Research
3 Method Development
3.1 Objective and Boundary Conditions
3.2 Suggested Approach
4 Conclusion and Outlook
References
The Omnichannel Retailing Capabilities Wheel: Findings of the Literature
Abstract
1 Introduction
2 Literature Review
3 Descriptive Synthesis of Results
3.1 Ordinary Capabilities for Omnichannel Retailing
3.2 Dynamic Capabilities for Omnichannel Retailing
4 Conclusion
References
Sustainable Retail Supply Chain Management – A Bibliometric Viewpoint
Abstract
1 Introduction
2 Methodology
3 Results and Discussion
4 Conclusions
Appendix 1
References
Part III: Distributed and Collaborative Planning and Control
Autonomous Production Control Methods - Job Shop Simulations
1 Introduction
2 State of the Art
2.1 Autonomous Production Control
2.2 Existing Autonomous Production Control Methods
2.3 Research Gap
3 Methodology
3.1 Simulation Setup
3.2 Selected Autonomous Production Control Methods
4 Results
5 Conclusion
References
Individual Predictive Maintenance Approach for Diesel Engines in Rail Vehicles
Abstract
1 Introduction
1.1 Motivation
1.2 Objective
2 State of the Art
2.1 Maintenance Strategies for Diesel Engines in Rail Freight
2.2 Methods for Predicting the Remaining Useful Life for Preventive Maintenance
2.3 Meta-learning for Selecting and Parameterizing Appropriate Forecasting Models
3 Solution Approach
4 Economic Potential Due to Predictive Maintenance of Diesel Engines in Rail Freight
5 Conclusion
Acknowledgements
References
Modelling Autonomous Production Control: A Guide to Select the Most Suitable Modelling Approach
1 Introduction
2 Modelling of Production Planning and Control
3 Minimal Models
4 Linear Programming
5 Discrete-Event Simulation
6 Conclusion
References
Predictive Manufacturing Tardiness Inference in OEM Milk-Run Operations
Abstract
1 Introduction
2 Order Release Tardiness Effects on OEM Milk-Run Operations
2.1 Order Delivery Tardiness Forecast
2.2 An Approximate Tardiness Forecast Example
2.3 Milk-Run Shifts Without Machine Breakdown
2.4 Milk-Run Shifts with Machine Breakdown
3 Conclusion and Research Prospects
Acknowledgements
References
Implementation of a Total Cost of Ownership Model for Last-Mile Logistics as a Constraint Satisfaction Problem
1 Introduction
2 TCO Model
2.1 Production Factors
2.2 Cost Functions
2.3 Feasible Production Combinations
3 Constraint Based TCO Implementation
3.1 TCO MODEL as a PROLOG CSP
3.2 CSP-TCO Application
3.3 Usecases
4 Simulation
5 Conclusions
References
A Theoretical Framework Assessment Proposal for a Complexity Degree Measurement on a Supply Chain Network
Abstract
1 Introduction
2 Literature Review Background
2.1 Supply Chain Complexity
2.2 Drivers in Supply Chain Complexity
2.3 SC Risk Management
3 Methodology
4 Supply Chain Complexity Theoretical Assessment Framework
5 Test Case and Preliminary Results
6 Conclusions
References
Supply Chain Integration: A Bibliometric Analysis
Abstract
1 Introduction
1.1 Starting Points of Consideration
1.2 Research Objectives and Methodology
2 What is Supply Chain Integration?
3 Results
4 Discussion and Conclusion
Appendix 1
Appendix 2
References
Development of a Decision Support Model for Managing Supply Chain Design Problems in Global Service Supply Chains
Abstract
1 Introduction
2 Literature Review
3 Model Development
3.1 Introduction
3.2 Basic Setup
3.3 The Mixed-Integer Linear Program (MILP)
3.4 The Analytical Hierarchy Process (AHP)
3.5 Integrating AHP and MILP to Calculate the Pareto Frontier
4 Conclusion
References
Shelter Site Selection and Allocation Model for Efficient Response to Humanitarian Relief Logistics
1 Introduction
2 Related Work
3 Methodology
4 Case Study
5 Results and Discussion
5.1 Computational Results
5.2 Sensitivity Analysis
6 Conclusion and Future Research
References
Part IV: Modeling, Simulation, and Optimization
A Hypercube Queuing Model Approach for the Location Optimization Problem of Emergency Vehicles for Large-Scale Study Areas
Abstract
1 Introduction
2 Literature Review
2.1 Coverage and P-Median Models
2.2 Hypercube Queuing Model
3 Large-Scale SQM
3.1 Model Formulation
3.2 Aggregation Algorithm
4 Solution Technique, Study Area and Results
4.1 Study Area
4.2 Results
5 Conclusion
References
Dynamic Optimization Model for Planning of Multi-echelon Logistic System Activity
Abstract
1 Introduction
2 Problem Statement
3 Optimization Model for Random Demand Over Planning Horizon
4 Conclusion
References
Simulation-Based Sensitivity Analysis of Dynamic Contract Extension Elements in Supplier Development
1 Introduction
2 Related Literature
3 Model Description
4 Experimental Setup
5 Numerical Results
6 Conclusion
References
Searching for Production Robustness Through Simulation-Based Scheduling Optimization
Abstract
1 Introduction
2 Some Works Using DES on Production Scheduling
3 Main Steps Behind the Proposed Method
4 Applying the Proposed Method to a Three Machine Schedule Example
4.1 Input Data: The Production Schedule
4.2 Input Data: The Production Schedule in Excel
4.3 The ARENA Simulation Model
4.4 Simulation Results and Analysis
5 Final Considerations
Acknowledgment
References
A Multiagent System for Truck Dispatching in Open-pit Mines
1 Introduction
2 Problem Definition
3 Formalization
4 An Alternative Solution Approach: Multiagent System
4.1 Scheduling MAS Architecture
4.2 Interaction in the Scheduling MAS
4.3 Decision Making
5 Results and Discussion
6 Conclusions
References
Drone Delivery Using Public Transport: An Agent-Based Modelling and Simulation Approach
Abstract
1 Introduction
2 Drone Delivery Using Public Transport (DDPT) Concept
3 Methodology
3.1 Simulation Setup
3.2 DDPT Models
3.3 Truck Delivery Models
3.4 Key Performance Indicators (KPIs)
4 Results and Discussion
4.1 Experiment Generation
4.2 Experiment Results
5 Conclusion
References
Part V: Intelligent Production and Logistics Systems
Perspectives on the Application of Internet of Things in Logistics
Abstract
1 Introduction
2 Methodology
3 Results
3.1 Bibliometric Analysis
3.2 Content Analysis
4 Conclusion
Acknowledgements
References
AIDA 4.0: Architectures of Industry 4.0 Demonstrated Through Application Scenarios in Business Game
Abstract
1 Introduction
2 Theoretical Background
2.1 Impact of Business Games on Learning
2.2 Struggles to Find an Optimal Supply Chain Strategy in Times of Digitalization
2.3 Related Studies and Research Question
3 A Fictional Company “Bike Manufacturer” and AIDA 4.0 Game
4 Strategies for Digitalization and Operations Scheduling
4.1 Conventional Centralized Supply Chain with Predictive Scheduling
4.2 The Architecture of Partially Digitalized Supply Chain
4.3 The Architecture of Fully Digitalized Supply Chain – Reactive Scheduling
5 Conclusions
References
Machine Learning in Production Scheduling: An Overview of the Academic Literature
Abstract
1 Introduction
2 Research Methodology
3 Bibliometric Analysis of the Portfolio
4 Main Machine Learning Techniques Applied in Production Scheduling
5 Conclusion and Further Research
Acknowledgements
References
Software-Defined Mobile Supply Chains
Abstract
1 Introduction
2 Literature-Based Positioning of the Idea of SD-MSCs
3 Mobile Production Units
3.1 Oil Industry
3.2 High Sea Fishery
3.3 Perishable Goods
3.4 Blood Supply Chains
3.5 Advantages of Mobile Production Assets
4 From Mobile Assets to Software-Defined SCMs
4.1 Impacts on Supply Chain Planning and Decision Making
4.2 The Meaning of “Software Definition”
5 Outline of a Research Agenda for SD-MSCs
6 Conclusions and Future Work
Acknowledgement
References
Clustering for Monitoring Logistical Processes in General Cargo Warehouses
1 Introduction
2 Monitoring of Logistical Workflows in General Cargo Warehouses
2.1 Monitoring Processes in General Cargo Warehouses
3 Method
3.1 Monitoring of Logistical Processes
3.2 Database for Storing Time Series
3.3 Clustering
3.4 Experimental Structure and Procedure
4 Results
4.1 Experiment 1: DBSCAN, raw_data_2, 00:00–01:00 a.m.
4.2 Experiment 2: DBSCAN, raw_data_2, 01:00–02:00 a.m.
5 Discussion
6 Conclusions
References
Using RFID to Monitor the Curing of Aramid Fiber Reinforced Polymers
Abstract
1 Introduction
1.1 Motivation
1.2 State of the Art
1.3 Research Approach
2 Measurement Background
2.1 Dielectric Analysis (DEA)
2.2 Radio Frequency Identification (RFID)
3 Material
3.1 Integration of RFID Transponders Within Composites
3.2 Plasma Treatment of Polymer Films
4 Results of Fiber Composites Cure Monitoring
5 Further Research for Surface Treatment of Polymer Films
6 Conclusion
Acknowledgment
References
Handover Abilities in Reconfigurable Material Flow Systems for Topology Computing
1 Introduction
2 State of the Art Regarding Reconfiguration of MFSs
3 Introduction to an Agent-Based Control Concept and Its Drawbacks
3.1 Agent-Based Reconfiguration of Automated MFSs
3.2 Drawbacks of the Agent-Based Approach Regarding MFM Interfaces and Topology Detection
4 Agent-Based Control Enlarged with Handover Abilities
4.1 Neighborhood Topology Detection
4.2 Agent-Based Control Concept Including Handover Abilities
4.3 Evaluation of Enlarged Concept with Lab-Sized Demonstrator
5 Conclusion and Outlook
References
PalletAssist - Concept for a Multisensor-Based Assistance System for Safe Handling of Palletized Goods with Forklift Trucks
Abstract
1 Motivation
2 State of the Art
3 Concept of the Assistance System
3.1 Outline of the Initial Situation and Problem Situation
3.2 Requirements for the Assistance System
3.3 System Design
3.4 Software Modules
4 Conclusion and Outlook
Appendix A
References
An Edge Neural Cyber-Physical Production System: Products Monitoring Their Production Recommend Adaptations of Their Schedule
1 Introduction
2 An Edge Nerual Production Monitoring System
2.1 The Products' Digital Twins
2.2 The Neural Positioning Network
3 A Test in the GME's Production Process
4 Conclusion
References
Part VI: Human-Machine Interaction
Augmented Reality in the Packing Process: A Model for Analyzing Economic Efficiency
Abstract
1 Introduction
2 Methodology
3 Conclusion and Future Research
Appendix 1
References
Assessment of Cognitive Strain in Digital Logistics Work: Background, Analysis and Implications
Abstract
1 Introduction
2 Conceptual Framework Cognitive Ergonomics
3 Analytical Framework
4 Implications for Digital Logistics Work
5 Conclusion and Outlook
References
Human Factor in Forecasting and Behavioral Inventory Decisions: A System Dynamics Perspective
Abstract
1 Introduction
2 Literature Review
2.1 Demand Forecasting
2.2 Judgmental Forecasting
2.3 Inventory
2.4 Inventory Decisions
2.5 Effects on Forecasting Adjustment to Inventory Decisions
2.6 System Dynamics
3 Methodology
3.1 Research Question and Model
4 Simulation and Analysis of the Model
5 Evaluation
6 Conclusion
References
Improving Human-Machine Interaction with a Digital Twin
1 Introduction
2 Fundamentals
2.1 Levels of Autonomy
2.2 Adaptive Automation
2.3 Digital Twin
3 Container Unloading Systems – State of the Art
4 Adaptive Automation in Container-Unloading
5 Application
6 Results and Outlook
References
Requirements for an Incentive-Based Assistance System for Manual Assembly
Abstract
1 Introduction
2 Related Work on Assistance Systems and Incentive Systems
3 Assistance Systems Requirements in Manual Assembly
4 Requirements Impact Analysis Based on Assembly Process Modelling
5 Conclusion and Outlook
Acknowledgment
Appendix
References
Evaluation of Human-Computer-Interaction Design in Production and Logistics by Using Experimental Investigations
1 Introduction
2 User-Centered Evaluation of HCI Systems
3 Exemplary Use Cases of Human-Machine-Interaction Design and First Results of Evaluation Studies
3.1 Design of the User Interface in Cyber-Physical Production Systems
3.2 Design of AR Systems for Production Environments
3.3 Design of a Training Device for Collaboration Improvements
4 Conclusion
References
Author Index
Recommend Papers
![Dynamics in Logistics: Second International Conference, LDIC 2009, Bremen, Germany, August 2009, Proceedings [1 ed.]
3642119956, 9783642119958](https://ebin.pub/img/200x200/dynamics-in-logistics-second-international-conference-ldic-2009-bremen-germany-august-2009-proceedings-1nbsped-3642119956-9783642119958.jpg)
![Dynamics in Logistics: First International Conference, LDIC 2007, Bremen, Germany, August 2007, Proceedings [1 ed.]
9783540768616, 3540768610](https://ebin.pub/img/200x200/dynamics-in-logistics-first-international-conference-ldic-2007-bremen-germany-august-2007-proceedings-1nbsped-9783540768616-3540768610.jpg)
![Dynamics in Logistics : Twenty-Five Years of Interdisciplinary Logistics Research in Bremen, Germany [1 ed.]
9783030886622, 9783030886615](https://ebin.pub/img/200x200/dynamics-in-logistics-twenty-five-years-of-interdisciplinary-logistics-research-in-bremen-germany-1nbsped-9783030886622-9783030886615.jpg)
File loading please wait...
Citation preview
Lecture Notes in Logistics Series Editors: Uwe Clausen · Michael ten Hompel · Robert de Souza
Michael Freitag Hans-Dietrich Haasis Herbert Kotzab Jürgen Pannek Editors
Dynamics in Logistics Proceedings of the 7th International Conference LDIC 2020, Bremen, Germany
Lecture Notes in Logistics Series Editors Uwe Clausen, Fraunhofer Institute for Material Flow and Logistics IML, Dortmund, Germany Michael ten Hompel, Fraunhofer Institute for Material Flow and Logistics IML, Dortmund, Germany Robert de Souza, The Logistics Institute - Asia Pacific, National University of Singapore, Singapore
Lecture Notes in Logistics (LNL) is a book series that reports the latest research and developments in Logistics, comprising: • • • • • • • • • • • • • • • • • • •
supply chain management transportation logistics intralogistics production logistics distribution systems inventory management operations management logistics network design factory planning material flow systems physical internet warehouse management systems maritime logistics aviation logistics multimodal transport reverse logistics waste disposal logistics storage systems logistics IT
LNL publishes authored conference proceedings, contributed volumes and authored monographs that present cutting-edge research information as well as new perspectives on classical fields, while maintaining Springer’s high standards of excellence. Also considered for publication are lecture notes and other related material of exceptionally high quality and interest. The subject matter should be original and timely, reporting the latest research and developments in all areas of logistics. The target audience of LNL consists of advanced level students, researchers, as well as industry professionals working at the forefront of their fields. Much like Springer’s other Lecture Notes series, LNL will be distributed through Springer’s print and electronic publishing channels. More information about this series at http://www.springer.com/series/11220
Michael Freitag Hans-Dietrich Haasis Herbert Kotzab Jürgen Pannek •
•
•
Editors
Dynamics in Logistics Proceedings of the 7th International Conference LDIC 2020, Bremen, Germany
123
Editors Michael Freitag BIBA - Bremer Institut für Produktion und Logistik GmbH Bremen, Germany Herbert Kotzab Logistics Management University of Bremen Bremen, Germany
Hans-Dietrich Haasis Maritime Business and Logistics University of Bremen Bremen, Germany Jürgen Pannek Gifhorn, Germany
ISSN 2194-8917 ISSN 2194-8925 (electronic) Lecture Notes in Logistics ISBN 978-3-030-44782-3 ISBN 978-3-030-44783-0 (eBook) https://doi.org/10.1007/978-3-030-44783-0 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 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
Since 2007, the biennial International Conference on Dynamics in Logistics (LDIC) offers researchers and practitioners from logistics, operations research, and engineering as well as from computer science an opportunity to meet and to discuss the latest developments in this particular research domain. From February 12 to 14, 2020, LDIC 2020 was held in Bremen (Germany) for the seventh time. This time, the conference featured the doctoral workshop of the International Graduate School for Dynamics in Logistics with the theme “Ensuring Practical Relevance and Scientific Contribution for Good Logistics Research” as a satellite event as well as tours through the Robotics Labs of the Institute for Artificial Intelligence and through the LogDynamics Lab of the Bremen Research Cluster for Dynamics in Logistics (LogDynamics). Similar to its six predecessors, the LogDynamics Research Cluster organized this conference in cooperation with the Bremer Institut für Produktion und Logistik (BIBA), which is a scientific research institute affiliated to the University of Bremen. The spectrum of topics of the LDIC 2020 reached from the dynamic modeling, planning and control of processes over supply chain management and maritime logistics to intelligent technologies and robotic applications for cyber-physical production and logistics systems. LDIC 2020 provided a forum for the discussion of advances in that matter. The conference program considered four invited keynote speeches and 51 scientific papers selected by a double-blind reviewing process. All selected scientific papers are arranged within these LDIC 2020 proceedings. By this, the proceedings give an interdisciplinary outline on the state of the art of research in dynamics in logistics as well as identify challenges and solutions for logistics today and tomorrow. The volume is organized into the following main areas: • • • •
Maritime Logistics and Multi-Modal Transport, Supply Chain Management and Coordination, Distributed and Collaborative Planning and Control, Modeling, Simulation, and Optimization,
v
vi
Preface
• Intelligent Production and Logistics Systems, and • Human—Machine Interaction. There are many people whom we have to thank for their help in one or the other way. For pleasant and fruitful collaboration, we are grateful to the members of the international program committee: Julia Arlinghaus, Magdeburg (Germany) Hyerim Bae, Pusan (Korea) Till Becker, Emden/Leer (Germany) Tobias Buer, Maskat (Oman) Sergey Dashkovsky, Würzburg (Germany) Jianhui Du, Zhengzhou (China) Enzo M. Frazzon, Florianópolis (Brazil) Axel Hahn, Oldenburg (Germany) Michael Henke, Dortmund (Germany) Soondo Hong, Pusan (Korea) Johann Hurink, Twente (Netherlands) Alexander Hübner, München (Germany) Dmitry Ivanov, Berlin (Germany) Hamid Reza Karimi, Milano (Italy) Aseem Kinra, Bremen (Germany) Matthias Klumpp, Essen (Germany) Antônio G. N. Novaes, Campinas (Brazil) Kulwant S. Pawar, Nottingham (UK) Erwin Pesch, Siegen (Germany) Jörn Schönberger, Dresden (Germany) Victor Tsapi, Ngaoundéré (Cameroon) Thorsten Wuest, Morgantown, WV (USA) Carrying the burden of countless reviewing hours, we wish to thank our LDIC colleagues Frank Arendt, Michael Beetz, Anna Förster, Otthein Herzog, Aseem Kinra, Hans-Jörg Kreowski, Walter Lang, Michael Lawo, Burkhard Lemper, Rainer Malaka, Nicole Megow, Klaus-Dieter Thoben, Yilmaz Uygun, Hendro Wicaksono as well as Ilja Bäumler, M. Khurrum Bhutta, Xavier Brusset, Matthias Burwinkel, Katharina Daschkovska, Alberto de Marco, Hendrik Engbers, Nadège Ingrid Gouanlong, Aleksandra Himstedt, Hawa Hishamuddir, Orlando Fontes Lima Junior, Nils Meyer-Larsen, Hettige Niles Perrera, Sarah Pfoser, Moritz Quandt, Magnus Redekker, Ingrid Rügge, Tobias Sprodowski, Marius Veigt, Guilherme Ernani Viera, and Jasper Wilhelm for their help in the selection process. We are also very grateful to Aleksandra Himstedt, Ingrid Rügge, Matthias Burwinkel, and countless other colleagues and students for their support in the local organization and the technical assistance during the conference. Moreover, we would like to acknowledge the financial support by the BIBA, the cluster
Preface
vii
LogDynamics, and the University of Bremen. Finally, we appreciate the excellent cooperation with Springer, which continuously supported us regarding the proceedings of all LDIC conferences. May 2020
Michael Freitag Hans-Dietrich Haasis Herbert Kotzab Jürgen Pannek
Contents
Part I: Maritime Logistics and Multi-Modal Transport Recommendations for Human Resources Development in Danube Inland Ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sarah Pfoser, Lisa-Maria Putz, and Eva Jung System Dynamics Modeling of Logistics Hub Capacity: The Dubai Logistics Corridor Case Study . . . . . . . . . . . . . . . . . . . . . . . Alberto De Marco, Hussein Fakhry, Marco Postorino, Zakaria Mammar, and Hakim Hacid Towards Intelligent Waterway Lock Control for Port Facility Optimisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Thimo Schindler, Christoph Greulich, Dennis Bode, Arne Schuldt, André Decker, and Klaus-Dieter Thoben On the Influence of Structural Complexity on Autonomously Controlled Automobile Terminal Processes . . . . . . . . . . . . . . . . . . . . . . Michael Görges and Michael Freitag Literature Classification on Container Transport Systems for Inter-terminal Transportation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nicole Nellen, Michaela Grafelmann, Justin Ziegenbein, Ann-Kathrin Lange, Jochen Kreutzfeldt, and Carlos Jahn
3
21
32
42
52
A Simulation Study of a Storage Policy for a Container Terminal . . . . . Henokh Yernias Fibrianto, Bonggwon Kang, Bosung Kim, Annika Marbach, Tobias Buer, Hans-Dietrich Haasis, Soondo Hong, and Kap Hwan Kim
62
Investigation of Vessel Waiting Times Using AIS Data . . . . . . . . . . . . . Janna Franzkeit, Hannah Pache, and Carlos Jahn
70
ix
x
Contents
Resource Sharing as a Management Concept for Digital Logistics Terminals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Paul Gerken, Herbert Kotzab, and Hans G. Unseld
79
Coordinating the Seaport-Hinterland Interface: Theoretical and Methodological Insights from Scientific Literature . . . . . . . . . . . . . Patrick Specht and Herbert Kotzab
89
Sensor Simulation and Evaluation for Infrastructure-Free Mobile Sensor Carrier Platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Artur Schütz and Maik Groneberg Expansion Planning at Container Terminals . . . . . . . . . . . . . . . . . . . . . 114 Marvin Kastner, Ann-Kathrin Lange, and Carlos Jahn A Hybrid Robust-Stochastic Optimization Approach for the Noise Pollution Routing Problem with a Heterogeneous Vehicle Fleet . . . . . . . 124 Hani Shahmoradi-Moghadam, Omid Samani, and Jörn Schönberger Part II: Supply Chain Management and Coordination The Influence of Cognitive Biases on Supply Chain Risk Management in the Context of Digitalization Projects . . . . . . . . . . . . . . 137 Julia C. Arlinghaus, Manuel Zimmermann, and Melanie Zahner Flow Management Tools and Techniques for Logistics Performance: An Application to the Logistics Service Sector in Cameroon . . . . . . . . . 148 Nadege Ingrid Gouanlong Kamgang, Adama Bidisse, and Victor Tsapi Blockchain and Risk in Supply Chain Management . . . . . . . . . . . . . . . 159 Rami Alkhudary, Xavier Brusset, and Pierre Fenies Prioritizing Supply Chain Agility Factors Using Fuzzy Analytic Network Process (FANP) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 Haniyeh Dastyar, Ali Mohammadi, and Moslem Ali Mohamadlou Influence of Supply Chain Management & Logistics in the Wake of China Pakistan Economic Corridor (CPEC) on Domestic Industry in Pakistan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 Ayesha Khan, Sayed Mehdi Shah, Hans-Dietrich Haasis, and Michael Freitag A Disruption Management Model for a Production-Inventory System Considering Green Logistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 Hawa Hishamuddin, Mohd Azizi Abd Aziz, Noraida Azura Md Darom, Mohd Nizam Ab Rahman, and Dzuraidah Abd Wahab A Concept for a Consumer-Centered Sustainable Last Mile Logistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 Michael Freitag and Herbert Kotzab
Contents
xi
The Omnichannel Retailing Capabilities Wheel: Findings of the Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 Bastian Mrutzek, Herbert Kotzab, and Erdem Galipoglu Sustainable Retail Supply Chain Management – A Bibliometric Viewpoint . . . . . . . . . . . . . . . . . . . . . . . . 215 Kristina Petljak and Herbert Kotzab Part III: Distributed and Collaborative Planning and Control Autonomous Production Control Methods - Job Shop Simulations . . . . 227 Ziqi Zhao, Oliver Antons, and Julia C. Arlinghaus Individual Predictive Maintenance Approach for Diesel Engines in Rail Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236 Hendrik Engbers, Simon Leohold, and Michael Freitag Modelling Autonomous Production Control: A Guide to Select the Most Suitable Modelling Approach . . . . . . . . . . . . . . . . . . . . . . . . . 245 Oliver Antons and Julia C. Arlinghaus Predictive Manufacturing Tardiness Inference in OEM Milk-Run Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254 Antônio G. N. Novaes, Orlando F. Lima Jr., José Eduardo Souza De Cursi, Jaime Andres Cardona Arias, and José Benedito Silva Santos Jr. Implementation of a Total Cost of Ownership Model for Last-Mile Logistics as a Constraint Satisfaction Problem . . . . . . . . . . . . . . . . . . . . 263 Bernd Nieberding and Johannes Kretzschmar A Theoretical Framework Assessment Proposal for a Complexity Degree Measurement on a Supply Chain Network . . . . . . . . . . . . . . . . . 274 Silvio Luiz Alvim, Jose Benedito Silva Santos Jr., and Carlos Manuel Taboada Rodriguez Supply Chain Integration: A Bibliometric Analysis . . . . . . . . . . . . . . . . 286 Herbert Kotzab, Ilja Bäumler, and Paul Gerken Development of a Decision Support Model for Managing Supply Chain Design Problems in Global Service Supply Chains . . . . . . . . . . . 299 Juri Reich, Tina Wakolbinger, and Aseem Kinra Shelter Site Selection and Allocation Model for Efficient Response to Humanitarian Relief Logistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309 Panchalee Praneetpholkrang and Van-Nam Huynh
xii
Contents
Part IV: Modeling, Simulation, and Optimization A Hypercube Queuing Model Approach for the Location Optimization Problem of Emergency Vehicles for Large-Scale Study Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321 Felix Blank Dynamic Optimization Model for Planning of Multi-echelon Logistic System Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331 Mykhaylo Ya. Postan, Sergey Dashkovskiy, and Kateryna Daschkovska Simulation-Based Sensitivity Analysis of Dynamic Contract Extension Elements in Supplier Development . . . . . . . . . . . . . . . . . . . . . 341 Haniyeh Dastyar and Jürgen Pannek Searching for Production Robustness Through Simulation-Based Scheduling Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351 Guilherme Ernani Vieira and Enzo Morosini Frazzon A Multiagent System for Truck Dispatching in Open-pit Mines . . . . . . 363 Gabriel Icarte, Paulina Berrios, Raúl Castillo, and Otthein Herzog Drone Delivery Using Public Transport: An Agent-Based Modelling and Simulation Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374 Raheen Khalid and Stanislav M. Chankov Part V: Intelligent Production and Logistics Systems Perspectives on the Application of Internet of Things in Logistics . . . . . 387 Ícaro Romolo Sousa Agostino, Charles Ristow, Enzo Morosini Frazzon, and Carlos Manuel Taboada Rodriguez AIDA 4.0: Architectures of Industry 4.0 Demonstrated Through Application Scenarios in Business Game . . . . . . . . . . . . . . . . . . . . . . . . 398 Julia Feldt (geb. Wagner) and Henning Kontny Machine Learning in Production Scheduling: An Overview of the Academic Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409 Satie L. Takeda-Berger, Enzo Morosini Frazzon, Eike Broda, and Michael Freitag Software-Defined Mobile Supply Chains . . . . . . . . . . . . . . . . . . . . . . . . 420 Uwe Aßmann, Udo Buscher, Sven Engesser, Jörn Schönberger, and Leon Urbas Clustering for Monitoring Logistical Processes in General Cargo Warehouses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431 Andreas Neubert
Contents
xiii
Using RFID to Monitor the Curing of Aramid Fiber Reinforced Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441 Marius Veigt, Marco Cen, Elisabeth Hardi, Walter Lang, and Michael Freitag Handover Abilities in Reconfigurable Material Flow Systems for Topology Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451 Juliane Fischer, Haitham Elfaham, Ulrich Epple, and Birgit Vogel-Heuser PalletAssist - Concept for a Multisensor-Based Assistance System for Safe Handling of Palletized Goods with Forklift Trucks . . . . . . . . . . 462 Thomas Depner, Liu Cao, and Hagen Borstell An Edge Neural Cyber-Physical Production System: Products Monitoring Their Production Recommend Adaptations of Their Schedule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477 Magnus Redeker and Axel Wagenitz Part VI: Human-Machine Interaction Augmented Reality in the Packing Process: A Model for Analyzing Economic Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493 Tim Woltering, Andre Sardoux Klasen, and Carsten Feldmann Assessment of Cognitive Strain in Digital Logistics Work: Background, Analysis and Implications . . . . . . . . . . . . . . . . . . . . . . . . . 504 Matthias Klumpp, Vera Hagemann, and Martina Schaper Human Factor in Forecasting and Behavioral Inventory Decisions: A System Dynamics Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516 Kavith Balachandra, H. Niles Perera, and Amila Thibbotuwawa Improving Human-Machine Interaction with a Digital Twin . . . . . . . . . 527 Jasper Wilhelm, Thies Beinke, and Michael Freitag Requirements for an Incentive-Based Assistance System for Manual Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 541 Christoph Petzoldt, Dennis Keiser, Thies Beinke, and Michael Freitag Evaluation of Human-Computer-Interaction Design in Production and Logistics by Using Experimental Investigations . . . . . . . . . . . . . . . . 554 Moritz Quandt, Hendrik Stern, Supara Grudpan, Thies Beinke, Michael Freitag, and Rainer Malaka Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 567
Part I: Maritime Logistics and MultiModal Transport
Recommendations for Human Resources Development in Danube Inland Ports Sarah Pfoser(&), Lisa-Maria Putz, and Eva Jung University of Applied Sciences Upper Austria, Roseggerstraße 15, 4600 Wels, Austria {sarah.pfoser,lisa-maria.putz,eva.jung}@fh-steyr.at
Abstract. Triggered by the growing and fluctuating freight volumes arriving in European seaports, peaks and bottlenecks are caused in hinterland terminals such as inland ports. Measures to increase the performance of inland ports previously focused on technological advancements such as infrastructure and ICT. However, recent developments show an increased awareness of human resources. Inland ports constitute a ‘socio-technical’ system since port performance is equally dependent on technological equipment as on skilled employees. However, the importance of human resources is not yet fully recognized in the inland port sector. The aim of this paper is to present recommendations for human resources development in inland ports. The Danube region is the geographic scope of the study including Austria, Hungary, Croatia, Bulgaria and Romania. A survey completed by eleven inland port authorities (employing 1,487 staff members in total) is presented. Additionally, three transnational expert rounds are held to develop so-called “personas”. These personas describe requirements and challenges for different types of future port employees. Results of the survey and the transnational expert rounds are used to better assess the current and future needs concerning human resources development in inland ports and to derive recommendations. Results suggest that the main challenge for Danube inland ports is the risk of staff shortages due to the high number of old employees. In addition, the inland port sector lacks training offers for port employees. Keywords: Human resources development Sustainability
Inland ports Danube region
1 Introduction Inland ports represent logistics hubs which facilitate transhipment between transport modes in a transport system (Dooms and Macharis 2003) and act as regulators of freight flows in supply chains (Rodrigue et al. 2010). In previous years, inland ports have hardly been recognized as a major factor that supports the improvement of transport flows in the whole supply chain since earlier research focused only on the efficiency of sea ports (Wiegmans et al. 2015). Ongoing measures to improve efficiency of inland ports mainly focus on infrastructural and technological development and hardly consider human resources (European Commission 2013; Witte et al. 2014). Political measures, companies and research focused on the improvement of infrastructure and often neglected the human factor (Fawcett et al. 2008; Rožić et al. 2016). © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 3–20, 2020. https://doi.org/10.1007/978-3-030-44783-0_1
4
S. Pfoser et al.
In fact, research frequently lacks analysing the role of human resources in supply chain management and rather focuses on operational issues such as transhipment infrastructure (Ellinger and Ellinger 2014; Gowen and Tallon 2003). In particular, little attention has been paid to the development of human resources in inland ports (Meletiou 2006). Thus, the goal of this paper is to provide recommendations for human resources development in inland ports. Human resources development can be defined as “a combination of training, career development, and organizational development [which] offers the theoretical integration needed to envision a learning organization, but it must also be positioned to act strategically throughout the organization” (Marsick and Watkins 1994, p. 355). In order to identify the recommendations for human resource development in inland ports, a study including a survey and three transnational expert rounds was conducted to derive the recommendations for human resources development in inland ports. The geographical scope of our study is on the Danube riparian countries. The Danube connects ten countries from the Black Forest region in Germany to the Black Sea in Romania on a distance of about 2,800 km and is hence Europe’s second largest river (Dolinsek et al. 2013). The Danube area constitutes one of the most important economic regions in Europe (via donau 2018). We received responses from Austria, Slovakia, Hungary, Croatia, Bulgaria and Romania. According to CCNR, these six EU Danube countries account for 15% of total goods transport performance on European inland waterways (CCNR 2017). Table 1 illustrates the cargo volumes that are handled in the ports of these six EU Danube countries. In the past three years, the cargo volumes were on a stable level with a tendency of increased volumes in 2017 (Danube Commission 2018). Table 1. Overview of cargo handling in EU Danube ports (data from Danube Commission 2018) Country
Austria Slovakia Hungary Croatia Bulgaria Romania
Cargo volume handled in 2015 (figures in thsd. tons) 7,449 2,009 5,978 566 4,547 24,462
Cargo volume handled in 2016 (figures in thsd. tons) 7,493 1,969 5,439 677 n.a. 25,096
Cargo volume handled in 2017 (figures in thsd. tons) 7,981 2,127 5,799 631.6 5,570 23,785
To cope with the increasing transport volumes that are shipped on the Danube and to meet the growing importance of Danube ports as multimodal hubs, highly qualified personnel will be needed. This is also reflected in the EUSDR, the EU Strategy for the Danube Region, where human resources represent a key issue. “People and skills” is one out of eleven priority areas that are focused by the EUSDR. This priority area aims to tackle unemployment in the Danube region, improve education and training, ensure
Recommendations for Human Resources Development in Danube Inland Ports
5
equal learning and labour opportunities for all persons and contribute to a closer cooperation between educational, training, labour market and research institutions (EUSDR 2018). The present paper addresses these ambitions as we aim to derive recommendations for human resources development in European inland ports.
2 Methodology Data collection for this study took place in three steps. The first step of our study was to conduct a literature review to identify the current status of human resources development in inland ports in Europe. Based on the results of the literature review, an online survey was developed to identify the current status of human resources management and the demand for human resource development measures in the identified countries. The questionnaire for this survey was structured in four sections. The first section included general questions such as general information about the respondent and the port. The second section included questions concerning the human resource structures of the ports (i.e. number of employees, age distribution, level of education, hierarchical level). The third section dealt with the current training and educational offers which are implemented in the ports. The fourth section of the questionnaire identified the future need for training and development measures of Danube inland ports. The questionnaires were distributed to port authorities in all ten Danube riparian countries in the period from February to March 2018. In total, eleven ports from five Danube riparian countries (Austria, Hungary, Croatia, Bulgaria and Romania, see Fig. 1) completed the survey.
Fig. 1. Geographical scope of the study
As a third step, we organized three transnational expert rounds with international experts from the inland port sector to evaluate current and future needs of the inland port sector concerning measures for human resources development. These expert
6
S. Pfoser et al.
rounds took place in May 2018 in Vienna (Austria), Budapest (Hungary) and Constanta (Romania). Each expert round was held in the national language (German, Hungarian and Romanian) to reduce misunderstandings due to foreign language barriers. In total, 135 experts from research and the inland port sector (such as representatives from port authorities, shippers or logistics service providers) attended the transnational expert rounds. In addition, potential future inland port employees such as logistics students were invited to attend the expert rounds. During the expert rounds, the participants developed so-called ‘personas’ which support the definition of requirements of future inland port employees. Personas represent a method to develop an abstract description of real people on an individual basis (Miaskiewicz and Kozar 2011). Using personas, demographic characteristics such as job tasks and requirements (e.g. tasks for future port employees) can be described (LeRouge et al. 2013). Based on the results from the survey and the transnational expert rounds, recommendations for human resources development were derived for internal stakeholders (port authority) and external stakeholders (economic stakeholders – e.g. port companies, public policy stakeholders – e.g. government bodies, community stakeholders). The differentiation between internal and external stakeholders is necessary since both stakeholder groups require different types of measures for human resources development (Notteboom and Winkelmans 2002).
3 Literature Review Recent trends in the transport industry such as an increasing transport of containers have a significant impact on port operations and management. These changes also bare challenges for human resources in ports. Challenges arise due to changing logistical activities, which require new qualifications for port employees in terms of education and skills. In addition, applying the principles of modern management in inland ports may be difficult since the port industry is a traditional sector (Muntean et al. 2010). The increasing number of innovations in technology such as advances in infrastructure or communication between transport modes have a great influence on jobs in the port industry. In fact, current jobs may be changed or completely new jobs or disciplines will emerge, which are not yet present in the port industry (Meletiou 2006). The hypothesis that a growing abundance of new technologies causes a rise in the employment of high-skilled staff as well as an increase in income inequality, has become known as the ‘Skill-Biased Technological Change’ thesis (Violante 2012). This hypothesis deals with the proven statement that job requirements and therefore qualifications, in terms of relevant skills required regarding the progressive development, caused by the development process will change. Based on these statements, better skilled workers are preferred over less-skilled workers (Trkulja 2008). It is evident that the transport industry as well as the port sector face gradual changes in terms of skill requirements caused by the rising blur of the technology. In the last decades, technological advances in the port sector focused on reducing the dependency on human effort, knowledge and skills. Only in recent years developments show a recurring importance of human resources (Meletiou 2006). Meletiou (2006) argue that a port should be seen as a ‘socio-technical’ system, because port performance is equally
Recommendations for Human Resources Development in Danube Inland Ports
7
dependent on technical applications and skilled employees. In order to see this correlation work successfully emphasis has to be given to human resource development in terms of training (Meletiou 2006). Human resources development can be defined as educational measures such as training to develop work-based knowledge or expertise aiming to foster the performance of organizations, companies or communities such as ports (McLean and McLean 2001). The aim of human resources development is to improve performance of individuals and groups in an organizational setting. Training and education can be seen as crucial parts of human resources development (Umesh 2014). Taking into account that a majority of employees in the transport sector are older than 36 years (Ecorys et al. 2015), lifelong learning measures are crucial to provide up-to-date training to current employees in the transport sector (Meletiou 2006). Due to constant changes in technology and the emerging trend of digitalization in the transport sector lifelong learning plays a crucial role (Meletiou 2006). Given the demographic change and the potential skill shortages in freight transport a central demand is to increase the employment rate of older workers aged 50 to 64. As the hiring rates of older workers are still very low compared to other age groups, a critical point is to implement measures to employ people on the long term. Although firms have promoted early retirement in the past, awareness has increased that firms are dependent on their existing workforce. As a result, firms are developing strategies to hold on to the potential of older employees (Boockmann et al. 2018). Even though the importance of human resources is recognized, there are no standardized rules for professional training of inland port employees on European level. Considering the heterogeneous standards, a regulatory option might be considered, including EU rules for port employees exceeding health and safety issues (Muntean et al. 2010).
4 Findings In this section, we summarize the main results from the survey conducted in eleven Danube inland ports located in Austria, Hungary, Croatia, Bulgaria and Romania. It will be shown that the human resource structure of these ports varies remarkably. Especially the port of Constanta plays a crucial role as it is relatively large compared to the other ports and engages more than 60% of the employees under investigation in this study. In addition, we present and analyze the personas, which have been developed during the three transnational expert rounds. 4.1
Current Human Resource Structure in Danube Ports
To be able to develop human resource measures, it is important to first understand human resource structures in inland ports. We analyzed the age structure of the employees in inland ports since age is considered to have an impact on human resource development (Ishizaki et al. 1998; Nitrini et al. 2008). In total, the eleven participating port authorities employ 1,487 employees. The majority of employees (86%, 1,276 employees) is older than 36 years. In fact, 45% are between 36 and 50 years (669 employees) and 41% are older than 50 years (607 employees). Figure 2 summarizes the age distribution of the sample. As seen in Fig. 2, in all observed countries the majority
8
S. Pfoser et al.
of the employees is older than 36 years, except for Hungary. In Hungary, the majority of people employed by port authorities are between 21 and 35 years old (53%). In particular, in the port of Adony (Hungary) the employees are relatively young since 32 port employees (68%) are between 21 and 35 years old.
older than 50 years
9%
3%
11%
16%
31%
39%
40%
50%
50%
40% 31%
27%
51%
between 36 and 50 years 50%
between 21 and 35 years
53%
up to 20 years
AUSTRIA
HUNGARY
CROATIA
BULGARIA
ROMANIA
Fig. 2. Age distribution of port employees
Second, in Table 3 we analyzed the level of education of the employees. The respondents were clustered in five levels of education (1) academic degree (i.e. university degree), (2) higher education (i.e. vocational secondary school), (3) schoolleaving qualification (i.e. middle school), (4) no occupational training (i.e. early school leavers), (5) other (i.e. people rejoining the labor market). The most common reported level of education by respondents is a school-leaving qualification (58%). This result corresponds with the results of an analysis conducted by the European Commission in 2009 for the transport and logistics sector in Europe. More than half of the employees (58%) in the European transport and logistics sector have medium qualification referring to school-leaving qualification. In Central and Eastern Europe, more than three quarters of employees (81%) in the transport and logistics sector completed compulsory schooling and are medium qualified (European Commission 2009). We found that high qualified employees which refers to higher or academic education represent the second largest group of educational level. In fact, 10% of all employees at port authorities have an academic degree and 27% completed higher education. This finding is contrary to the study conducted by the European Commission in 2009, which suggested that low qualified employees represent the second largest group (28%). According to the European Commission (2009), high-qualified employees represent the smallest share in the European transport and logistics sector (European Commission 2009). However, only 3% of people employed by port authorities considered in this study have no occupational training. As shown in
Recommendations for Human Resources Development in Danube Inland Ports
9
Table 2. Educational level of port employees (1) Academic degree (2) Higher education (3) School-leaving qualification (4) No occupational training (5) Other Total
Austria 18 22 129 25 5 199
Hungary 7 12 23 5 17 64
Croatia 5 3 – – – 8
Bulgaria – 59 246 – – 305
Romania 120 310 466 15 911
Table 2, in particular in Croatia (port of Vukovar) a high share of employees at port authorities have an academic degree or completed higher education. Table 4 categorizes the port employees in five management level (1) top management (e.g. CEO), (2) middle management (e.g. head of department), (3) lower management (e.g. administration or accountant), (4) operational level (e.g. crane driver, warehouse worker) or (5) other. The majority of employees (52%) are working on operational level followed by lower management (e.g. administration or accountant) (34%). Few employees can be found in the top management such as the board of directors or CEO and middle management such as head of department or logistics manager. Table 3. Management level of port employees (1) Top management (2) Middle management (3) Lower management (4) Operational (5) Other Total
Austria 4 12 49 34 – 99
Hungary 4 9 16 34 – 63
Croatia 2 – 6 – – 8
Bulgaria 6 19 42 151 89 307
Romania 9 45 361 496 – 911
Figure 3 shows the distribution of male and female employees for the responding countries. The majority of people employed by port authorities included in the survey were male (77%). This is also the case on national level (see Fig. 3). Worth mentioning, is that on port level two inland ports show different results concerning gender distribution. In the port of Vienna (Austria), more women (78 women) than men (38 men) are employed. This result is similar to the port of Baja (Hungary), with three female and one male employee. The gender distribution on management level is summarized in Table 4. Results show that the majority of female employees (62%) is working in lower management positions (e.g. administration) followed by operational positions such as crane drivers (25%). The majority of male port employees can be found in operational (61%) or lower management positions (25%). It is evident that all management levels are
10
S. Pfoser et al. Austria
Hungary
male
44% 56%
female
male
31% 69%
Croatia
50% 50%
female
73%
female
male
23%
female
male
27%
Bulgaria
male
Romania
77%
female
Fig. 3. Gender distribution of port employees on national levels
Table 4. Gender distribution categorized in management level of port employees Top management
Male Female Middle management Male Female Lower management Male Female Operational Male Female Other Male Female Total Male Female
Austria 3 1 9 3 35 14 34 0 –
Hungary 3 1 7 3 1 13 31 5 –
Croatia 1 1 –
81 18
42 22
4 4
3 3 – –
Bulgaria 6 0 12 7 29 13 106 43 82 7 225 70
Romania 6 3 24 21 183 178 455 41 – 668 243
Total 19 6 52 34 251 221 626 89 82 7 1020 357
dominated by male employees. The share of male employees is the highest in the top management level (76%) and the operational level (88%), which represent both a male domain in inland ports. Concerning the management level, results demonstrate that the majority of port employees are working on operational level. This result is in line with the statistics presented in a report elaborated by Christidis et al. (2014). The authors concluded that in the transport industry a small share of employees works in top and middle management positions and a higher share of employees in the lower management or operational level (Christidis et al. 2014).
Recommendations for Human Resources Development in Danube Inland Ports
4.2
11
Current and Future Training Requirements
The survey included questions about trainings that are currently offered by port authorities. In Austria, operational training (e.g. operation of logistical equipment), administrative training (e.g. IT training, accounting), social skills training (e.g. communication skills) and safety training are provided by port authorities. In general, these trainings are carried out by experienced employees for new employees on an individual basis. Trainings on port level (e.g. for all companies located in the port area and port authorities) are only organized in the case of safety trainings. The trainings offered by port authorities in Hungary focus on logistics, administration, safety and sustainability. The ports in Hungary do not provide trainings concerning social skills or law. Trainings at ports are organised annually or if required by employees. New employees may demand training by port authorities if required. Currently, there are no funding sources in Hungary, which would support training in the inland port sector. Substantial financial support would be needed for trainings in port operation, port management (incl. port strategy, lean management), as well as trainings for operators of cranes, forklifts, front loaders or truck drivers. Funding for training on the topic of IT and language skills are also considered relevant by Hungarian port authorities. In Croatia, the port authority Vukovar as a public institution is not in charge of offering any trainings for employees in the port. Training measures of Bulgarian ports focus on operational, administrative and safety training. In contrast, the port of Constanta offers a wide range of trainings for port employees. Training measures include operational, administrative, social skills and safety training as well as trainings on the topic of sustainability and law. Trainings are in most cases organized if demanded by employees (e.g. new employees are hired). Port authorities were also asked whether they are planning to adapt their current training programs for employees in the near future. Only two port authorities from Austria and Hungary mentioned that they are planning to adapt their current training programs in the future. Respondents mentioned that further training is required in the fields of administration, operation, social skills, law and safety. Port authorities indicated that they prefer online learning materials/online courses, expert rounds moderated by experts from the industry, field-trips to other ports or companies located in port areas and specified courses organized by educational institutions such as universities for training purposes in the future. The survey also included a question whether port authorities think that standardized training in Danube inland ports may increase the competitiveness of inland ports and inland waterway transport in the Danube region in the future. Results show a contrary result: respondents indicated that on the one hand standardized training may facilitate standardized processes in inland ports and that the quality of port processes may be increased. On the other hand some respondents mentioned that standardized training would be needed in logistics in general but despite that, national peculiarities must be taken into account. The European Commission (2009) also mentioned the issue, that the training and education system should rather be adapted on national than on European level. However, cooperation on sector-specific level should be pursued to facilitate cooperation in the inland port sector (European Commission 2009).
12
4.3
S. Pfoser et al.
Personas – Description of Future Employees Types
Three transnational expert rounds were organized in Austria, Hungary and Romania in May 2018. Stakeholders from research and the inland port sector such as representatives from port authorities, shippers or logistics service providers attended the expert rounds. During each expert round, three personas were developed for the top management, the middle management and lower management/operational level. The lower management and operational level have been summarized during the expert rounds since stakeholders argued that they are not sure whether the operational level will be obsolete due to technological advances. The results for each management level are summarized in Table 5. The first line shows the main biographical background, the second line contains the main working conditions of the job and in the third line the required competences and skills are summarized. The last line includes the main chances and challenges for the personas for the three management levels. Expert rounds participants mentioned that the top management positions can also be occupied by women. In fact, expert rounds participants stated that they believe a woman or man can occupy a position in the top and middle management in the future and that these management levels will not be dominated by men in the future. This result is contrary to the results of the survey stated above. Based on the survey 76% of port employees in the top management and 60% of port employees in the middle management are male. Concerning the operational level, expert rounds participants mentioned that the operational level would remain a male domain in the future. These results are in agreement with those obtained in the survey conducted prior to the expert rounds. An interesting result is that in all three transnational expert rounds participants mentioned that each of the three personas need to have IT know-how (e.g. general computer skills). Table 5. Summary of personas developed in the three transnational expert rounds
Biography
Working conditions
‘CEO/Port Manager’ (top management)
‘Operations Manager’ (middle management)
‘Port Worker’ (lower management/operational)
• Male/female • 40 years or older • Work experience in port business/general business: +15 years • Lives close to port (short distance commuter) • Work-life balance is important • Married/divorced with children • Income: *3,000 € (50% fixed and 50% variable part) • High responsibility • Working more than 8 h/day, weekends if necessary, overtime is own responsibility (*10 h overtime per week) • From CIO (Chief Information Officer) to CEO (Chief Executive Officer)
• Male/female • 35 years old • Lives close to port (short distance commuter) • Work-life balance is important • Technical/commercial education with engineering background • Long work experience (*5–10 years) • Married/divorced with children • Income: *1,800 € (70% fixed and 30% variable part) • Flexible working hours (home office) • Responsible for operational (e.g. logistical) activities in port • High availability necessary (24/7) in case of emergencies
• Male • 25–30 years old • Migration background (Austria) • Not married, no children • Lives close to port (short distance commuter) • Secondary education/school-leaving qualification • Income: *1,000€ • Diverse responsibilities which include maintenance, facility management and other operational tasks (e.g. forklift driver) • Shift work with fixed hours
(continued)
Recommendations for Human Resources Development in Danube Inland Ports
13
Table 5. (continued)
Competences/ skills
Chances and challenges
‘CEO/Port Manager’ (top management)
‘Operations Manager’ (middle management)
‘Port Worker’ (lower management/operational)
• Financial skills and competences • Multiple languages: local language, English, second foreign language • IT know-how (ECDL exam, general computer skills) • Leadership skills • Social and networking skills • Logistical knowledge • Sensitivity for innovation (knows which trend is relevant to the port and which is not) • Life long learning and further training to stay up to date (e.g. attending expert rounds, online trainings,…) Chances: • Various development opportunities through collaboration with different industries • Further training possible (lifelong learning) Challenges: • Has to handle a high level of stress • Pressure to reduce costs from various stakeholders • Has to react to market changes
• Technical, commercial and engineering know-how • Multiple languages: local language, English, second foreign language • IT know-how (ECDL exam, general computer skills) • Stress resistant • Social competence • Networking/communication skills
• Language skills: local language and English • IT know-how (general computer skills) • Willing to learn new things (further training) • Networking/ communication skills • Skilled craftsmanship
Chances: • Freedom of design in the job • Possibility to inspire and stimulate colleagues Challenges: • Maintain port operations 24/7 • Stress • Multitasking skills are required
Chances: • Training courses for constant professional development • Interactive work environment Challenges: • Working on shifts, fulfilling different tasks
5 Discussion and Recommendations for Human Resources Development The freight transport sector is often recognized as unattractive labor sector as it is associated with long working hours, poor working conditions and low salaries. Due to the lack of clear career progression and funding sources, especially young persons, which are interested in working in the transport sector, have to obtain the relevant training on their own costs (e.g. driver training costs). Another challenge in the freight transport sector is that almost a third of employees are older than 50 years and will retire in the coming 10 to 15 years. The inability to attract new employees in combination with the current age-distribution of employees in freight transport poses a major challenge to the transport sector (Ecorys et al. 2015). This finding is in agreement with the results obtained in the study. The majority of people employed by port authorities are older than 36 years. Only few port authorities employ students or people who are still in training. Ecorys et al. (2015) indicate that campaigns to attract young people and females to work in freight transport are an important measure to counteract the shortage of employees in the freight transport sector. This measure may also counteract the
14
S. Pfoser et al.
increasing shortage of qualified personnel in the inland port sector. Port authorities themselves can launch these campaigns in order to attract qualified young people and female employees. Thus, port authorities can also make sure to include country or region specific prerequisites, which only apply for their own inland port. In addition, campaigns on international level (e.g. Danube region) launched by politics or other interested groups from the inland ports industry such as non-profit organizations (e.g. EFIP - European federation of inland ports) could increase the awareness of inland ports as an attractive work place for young people. By organizing field trips to inland ports for educational institutions with logistics specialization young people such as students could learn more about the work place inland port. Putz et al. (2018) argue that field trips can be seen as an effective measure to inform students about the possible future career paths in logistics by involving practitioners and hands-on experiences. The majority of employees in the freight transport sector are male. In fact, the share of male employees in the freight transport sector is projected to remain at 80% until 2020 (Christidis et al. 2014). In 2006, 84% of employees in the EU-25 transport sector were male (Schneider et al. 2011). The share of female employees is particular low in the land and water transport sector (Christidis et al. 2014). Even though in recent years the share of female employees in the freight transport sector has increased, the freight transport sector is still dominated by men. This may be due to the fact that the majority of jobs with a risk of accidents and injuries are occupied by men (Schneider et al. 2011). As the results of the study show, the majority of port employees are working on operational level (e.g. forklift driver, berth operator) and thus are confronted with a higher risk of accident and injuries than employees in office jobs. This may be an explanation for the male dominance in the inland port sector. The findings presented by Schneider et al. (2011) and Christidis et al. (2014) are also in accordance with the results of the study: the majority of employees were male (77%). However, as the personas elaborated during the three transnational expert rounds show, participants do not perceive top and middle management positions in inland port authorities as a male domain. In fact, during all three transnational expert rounds participants mentioned that gender is not an issue in the top and middle management (see Table 5). The image of the freight transport sector in general as well as the image of inland ports as a work place need to be improved in the future to attract women (European Commission 2009). Again, specific campaigns to disseminate inland ports as attractive work places can be named as an effective measure to attract women to work in inland ports in the future (Ecorys et al. 2015). By organizing field-trips to inland ports for women (e.g. women’s day in inland ports) including a special program tailored to the target group of women, female participants can have practical insights into potential career paths in inland ports (Putz et al. 2018). Currently, only standardized rules for health and safety issues for professional training of port employees on European level exist (Muntean et al. 2010). As results of the study show, training measures vary on national as well as on inland port level on national level. The European Commission (2009) argues, that harmonized training and education systems on national level are required in the freight transport sector. Transnational cooperation on sector-specific level such as the inland port sector should
Recommendations for Human Resources Development in Danube Inland Ports
15
be pursued to facilitate cooperation in industry. In addition, demographic prerequisites including gender and age should be taken into account when implementing these training measures. Joint training measures such as shared training facilities and the implementation of various learning materials such as online learning may be considered (European Commission 2009). van Hoek (2001) and Gravier and Farris (2008) argue that an adaption of the curriculum in logistics education is needed. Since curricula in logistics education have not been fundamentally changed in the last decades, new requirements from the freight transport industry concerning skills of logistics graduates are not respected (Gravier and Farris 2008; van Hoek 2001). Besides including relevant topics such as digitalization or new trends in the inland port sector, teaching methods need to be adapted in logistics education. Including interactive teaching methods such as presentations and practical teaching methods such as fieldtrips are considered as suitable teaching methods in logistics education (van Hoek 2001). The trend of digitalization has mainly influenced the education sector. Educational resources such as textbooks or course materials are publically available online and free of charge. Due to the trend of digitalization, distance learning is also facilitated (Butcher et al. 2011). An adaption of the curriculum in logistics education including topics relevant for the inland port sector may be an appropriate measure to develop standardized training for port employees in the Danube region. However, national prerequisites should be respected in such standardized training measures, since this was also mentioned by port authorities in the survey and pointed out by the European Commission (2009). As already mentioned at the beginning of this chapter, almost a third of employees in the freight transport sector are older than 50 years and will retire in the coming 10 to 15 years (Ecorys et al. 2015). This can also be supported for the inland port sector by the results of the survey conducted. Lifelong learning measures can be identified as important measures in human resources development to guarantee that current port employees are up-to-date concerning relevant trends and developments in the freight transport sector with special focus on inland ports. Finally yet importantly, providing adaptable online learning materials may be an effective measure to provide learning materials for employees of port authorities at low or now costs. National prerequisites can be included by port authorities themselves (Butcher et al. 2011). As a supplement, field trips or expert rounds with stakeholders from the inland port sector may be suitable to transfer the required knowledge. By allowing logistics students to gain firsthand insights into the various tasks in the inland port sector the learning experience of participants may be enhanced (Butcher et al. 2011; Putz et al. 2018) (Table 6).
16
S. Pfoser et al. Table 6. Recommendations for human resources development in Danube inland ports
Problem description Majority of port employees are older than 35 years and will retire in the next 10–15 years
Recommendations Promote inland ports as attractive work place for young people
Majority of port employees are male, especially top management and operational level are male dominated
Promote inland ports as attractive work place for women
No standardized training for port employees; training measures on individual basis
Development of standardized training for port employees
Target group Internal: Port authorities External: Economic stakeholders (e.g. port companies) Public policy stakeholders (e.g. educational institutions) Community stakeholders (e.g. nonprofit organizations such as EFIP) Internal: Port authorities External: Economic stakeholders (e.g. port companies) Public policy stakeholders (e.g. educational institutions) Community stakeholders (e.g. nonprofit organizations such as EFIP) Internal: Port authorities External: Economic stakeholders (e.g. port companies) Public policy stakeholders (e.g. research institutions, educational institutions) educational institutions)
Measures • Dissemination of career opportunities for young people in inland ports • Field-trips for schools
• Dissemination of career opportunities for women in inland ports • Field-trips for women (women’s’ day)
• Adaption of curriculum in logistics education • Long-life learning measures • Blended learning: online learning in combination with expert rounds
Recommendations for Human Resources Development in Danube Inland Ports
17
6 Conclusion, Limitations and Research Outlook The aim of this paper was to derive recommendations for human resources development in European inland ports. Therefore, a survey in inland ports and three transnational expert rounds were conducted to assess the needs of the inland port sector concerning human resources development. The geographical scope of the study was the Danube region which represents a major economic area in Europe (via donau 2018). In total, eleven port authorities covering 1,487 employees from five countries on the Danube participated in the survey. We found that the majority of the employees in Danube inland ports are male and have a school-leaving qualification. This is also concordant with previous studies conducted by European Commission (2009) and Ecorys et al. (2015). The majority of port employees employed by port authorities participating in the study are older than 36 years entailing the risk of shortage of staff in inland ports in the next decades due to the retirement of current employees. In the future, experts from the inland port sector expect a balance in terms of gender in inland ports. Results show that in the future, further training measures are required by inland ports. In addition, measures to promote inland ports as an attractive work place to young people and women are required. This research contributes to existing knowledge about human resources in the freight transport sector as we point out additional insights about the inland port sector as a work place in the freight transport sector. As results of the study show, inland ports are work places which are dominated by older and male employees. Results of the survey emphasize the need for a higher level of understanding of the work place inland port. Based on the empirical data, we derived three recommendations for internal and external stakeholders to support the development of human resources in inland ports. The first recommendation aims to promote inland ports as a work place for young people. The majority of employees in inland ports is older than 36 years. These employees will retire in the next 10 to 15 years which may lead to a shortage of staff in inland ports. By promoting inland ports as attractive work places among young people such as logistics students by promoting inland ports as work places during lectures (e.g. guest lecture from port authority in school), new qualified employees may be recruited for inland ports. Port authorities, as internal stakeholders, as well as external stakeholders such as educational institutions should be involved to successfully implement this recommendation. The second recommendation focuses on promoting inland ports as work places for women since the majority of employees in inland ports are male. Organizing field trips to inland ports only for women (e.g. a women’s day in inland ports) can be named as a potential measures to implement the second recommendation. Again, internal and external stakeholders should be involved to increase the probability of success. Developing standardized training for port employees in inland ports was identified as the third recommendation in this study. As the results of the study show, no standardized training is provided in Danube inland ports included in the survey and port authorities remarked that they lack adequate training for port employees and partly are planning to adapt their current training measures. This may be achieved by adapting current logistics education curricula and by providing online learning materials, which can be used independently by employees in inland ports. Furthermore, we identified the
18
S. Pfoser et al.
challenges for future employees in inland ports and developed possible scenarios in form of three personas. Those three personas were elaborated based on management levels. The personas aimed to better understand the challenges of future employees in inland ports and to derive measures to meet the training needs of future employees. This study has several limitations, which influence generalizability. Out of seven EU-member Danube states, we received responses from five EU-member states. Thus, results could be different if we would have received data from the missing Danube states Germany and Slovakia. Moreover, the sample size is limited to eleven respondents which may influence validity of the results. A comparison including more Danube inland ports would be interesting to better assess the current needs concerning human resources development. For future surveys, we suggest translating the questionnaires to tackle possible linguistic barriers. For further research, we would recommend the inclusion of inland ports in the Rhine area to provide a comprehensive picture of the current needs concerning human resources development in a broader European region. Moreover, we assume that a comparison of human resource development needs in the Danube and the Rhine region may identify interesting differences and similarities. In addition, a survey on individual level rather than port authority level may be required to better understand the individual needs of employees in inland ports concerning human resources development. By including shippers or companies located in the port area of the considered inland ports similarities and differences may be identified concerning the needs for human resources development in inland ports. Besides training and qualification, an additional focus in future could be on the appreciation of the different jobs. Moreover, there may be country-specific differences in providing the training and qualification measures in cooperation between ports, authorities, universities and training centers, which could be analyzed in future. Finally we suggest that future research should analyze the current funding schemes for human resources development in European inland ports to provide internal and external stakeholders with recommendations for financial funding sources. Acknowledgment. This research is part of the research field ‘sustainable transport systems,’ which was funded by the State of Upper Austria as part of the research program ‘FTI Struktur Land Oberösterreich’. The results presented are part of the project DAPhNE, funded by Interreg Danube Transnational Programme (DTP1-196-3.1). Any opinions, findings, and conclusions or recommendations expressed in this paper are those of the authors and do not necessarily reflect those of the cooperating partners.
References Boockmann, B., Fries, J., Göbel, C.: Specific measures for older employees and late career employment. J. Econ. Ageing 12, 159–174 (2018) Butcher, N., Kanwar, A., Uvalić-Trumbić, S.: A basic guide to open educational resources (OER). Commonwealth of Learning; UNESCO. Section for Higher Education, Vancouver, Paris (2011) CCNR: CCNR Market Observation - Annual report 2017, Strasbourg, France (2017) Christidis, P., Navajas, E., Brons, M., Schade, B., Mongeli, I., Soria, A.: Future Employment in Transport: Analysis of Labour Supply and Demand. Publications Office, Luxembourg (2014)
Recommendations for Human Resources Development in Danube Inland Ports
19
Danube Commission: Market observation for Danube navigation: Results in 2017, Budapest, Hungary (2018) Dolinsek, M., Hartl, S., Hartl, T., Hintergräber, B., Hofbauer, V., Hrusovsky, M., Maierbrugger, G., Matzner, B., Putz, L.-M., Sattler, M., Schweighofer, J., Seemann, L., Simoner, M., Slavicek, D.: Manual on Danube Navigation, 2nd edn. via donau – Österreichische Wasserstraßen-Gesellschaft mbH, Vienna (2013) Dooms, M., Macharis, C. (eds.): A framework for sustainable port planning in inland ports: a multistakeholder approach. In: 43rd Congress of the European Regional Science Association: “Peripheries, Centres, and Spatial Development in the New Europe”, 27th–30th August 2003, Jyväskylä, Finland (2003) Ecorys, Fraunhofer, TCI, Prognos, AUEB-RC/TRANSLOG: Fact-finding studies in support of the development of an EU strategy for freight transport logistics: Lot 1: Analysis of the EU logistics sector (2015) Ellinger, A.E., Ellinger, A.D.: Leveraging human resource development expertise to improve supply chain managers’ skills and competencies. Eur. J. Train. Dev. 38(1/2), 118–135 (2014) European Commission: Transport and logistics: comprehensive sectoral analysis of emerging competences and economic activities in the European Union (2009) European Commission: NAIADES II-Towards quality inland waterway transport: towards quality inland waterway transport (2013) EUSDR: Work programme “education and training, labour market and marginalized communities” (2018) Fawcett, S.E., Magnan, G.M., McCarter, M.W.: Benefits, barriers, and bridges to effective supply chain management. Supply Chain Manag. 13(1), 35–48 (2008) Gowen, C.R., Tallon, W.J.: Enhancing supply chain practices through human resource management. J. Manag. Dev. 22(1), 32–44 (2003) Gravier, M.J., Farris, M.T.: An analysis of logistics pedagogical literature: past and future trends in curriculum, content, and pedagogy. Int. J. Logist. Manag. 19(2), 233–253 (2008) Ishizaki, J., Meguro, K., Ambo, H., Shimada, M., Yamaguchi, S., Hayasaka, C., Komatsu, H., Sekita, Y., Yamadori, A.: A normative, community-based study of Mini-Mental State in elderly adults: the effect of age and educational level. J. Gerontol. Ser. B Psychol. Sci. Soc. Sci. 53(6), 359–363 (1998) LeRouge, C., Ma, J., Sneha, S., Tolle, K.: User profiles and personas in the design and development of consumer health technologies. Int. J. Med. Informatics 82(11), e251–e268 (2013) Marsick, V., Watkins, K.: The learning organization: an integrative vision for HRD. Hum. Resource Dev. Q. 5(4), 353–360 (1994) McLean, G.N., McLean, L.: If we can’t define HRD in one country, how can we define it in an international context? Hum. Resource Dev. Int. 4(3), 313–326 (2001) Meletiou, M. (ed.): Improved port performance through training: the contribution of the International Labour Organization. In: Proceedings of 22nd International Port Conference “Human Resources and Sea Ports Performance”, Alexandria, Egypt, 12–14 March 2006 (2006) Miaskiewicz, T., Kozar, K.A.: Personas and user-centered design: How can personas benefit product design processes? Des. Stud. 32(5), 417–430 (2011) Muntean, M.-C., Nechita, D., Nistor, C., Sarpe, D.: Port management importance in port activities development. In: Proceedings of 3rd WSEAS International Conference on Urban Planning and Transportation (UPT 2010), Corfu Island, Greece, 22–24 July 2010 (2010)
20
S. Pfoser et al.
Nitrini, R., Brucki, S.M.D., Smid, J., Carthery-Goulart, M.T., Renato, A., Areza-Fegyveres, R., Bahia, V.S., Damin, A.E., Formigoni, A.P., Frota, N.A.F., Guariglia, C., Jacinto, A.F., Kato, E.M., Lima, E.E.P., Moreira, D., Nóbrega, A.B., Porto, C.S., Senaha, M.L.H., Silva, M.N.M., Souza-Talarico, J.N., Radanovic, M., Mansur, L.L.: Influence of age, gender and educational level on performance in the Brief Cognitive Battery-Edu. Dement. Neuropsychol. 2(2), 114– 118 (2008) Notteboom, T., Winkelmans, W. (eds.): Stakeholder Relations Management in ports: dealing with the interplay of forces among stakeholders in a hanging competitive environment, Panama (2002) Putz, L.-M., Treiblmaier, H., Pfoser, S.: Using field trips for sustainable transport education. Int. J. Logist. Manag. 29, 1424–1450 (2018) Rodrigue, J.-P., Debrie, J., Fremont, A., Gouvernal, E.: Functions and actors of inland ports: European and North American dynamics. J. Transp. Geogr. 18(4), 519–529 (2010) Rožić, T., Rogić, K., Bajor, I.: Research trends of inland terminals: a literature review. PROMET 28(5), 539–548 (2016) Schneider, E., Irastorza, X., Copsey, S.: OSH in Figures: Occupational Safety and Health in the Transport Sector: An Overview. Publications Office of the European Union, Luxemburg (2011) Trkulja, V.: Die Digitale Kluft: Eine Analyse der Informationsinfrastruktur in BosnienHerzegowina. Information – Wissenschaft und Praxis 59(2), 142–148 (2008) Umesh, U.: A study of the importance of training and development of personnel of cochin port trust, Kerala, India. Rev. Integr. Bus. Econ. Res. 4(1), 245–252 (2014) van Hoek, R.I.: Logistics education: achieving market and research driven skill development. Int. J. Phys. Distrib. Logist. Manag. 31(7/8), 505–519 (2001) via donau: Annual report on Danube navigation in Austria. via donau - Österreichische Wasserstraßen-Gesellschaft mbH, Vienna, Austria (2018) Violante, G.L.: Skill-biased Technical Change. In: Durlauf, S.N., Blume, L. (eds.) The New Palgrave Dictionary of Economics, 2nd edn. Palgrave Macmillan, Basingstoke (2012) Wiegmans, B., Witte, P., Spit, T.: Characteristics of European inland ports: a statistical analysis of inland waterway port development in Dutch municipalities. Transp. Res. Part A: Policy Pract. 78, 566–577 (2015) Witte, P., Wiegmans, B., van Oort, F., Spit, T.: Governing inland ports: a multi-dimensional approach to addressing inland port–city challenges in European transport corridors. J. Transp. Geogr. 36, 42–52 (2014)
System Dynamics Modeling of Logistics Hub Capacity: The Dubai Logistics Corridor Case Study Alberto De Marco1(&), Hussein Fakhry2, Marco Postorino1, Zakaria Mammar2, and Hakim Hacid2 1
Department of Management and Production Engineering, Politecnico di Torino, Turin, Italy [email protected] 2 College of Technological Innovation, Zayed University, Dubai, United Arab Emirates
Abstract. This paper proposes a System Dynamics (SD) modeling and simulation-based approach to support decision making and policy actions to make appropriate and effective investment decisions about the planning of an intermodal air, sea and land logistics hub capacity. The Dubai Logistics Corridor (DLC) is used as a concrete case study. The model offers the necessary decision support that captures the complexity of managing the logistics hub along with overcoming the implicit policy resistance. The paper illustrates the case study model, application, and various case scenario simulations. The model can be used as a predictive tool to encourage decision making and detecting capacity bottlenecks to help in planning and scheduling the capacity investment of a logistics hub. Keywords: Logistics
Transport hub System Dynamics
1 Introduction Several reasons are positioning Dubai as one of the major logistics and transportation hubs in the world. Over the past decade the Emirate of Dubai has emerged as a leading transport and logistics center serving not only the Middle East and North Africa, but also Russia, Europe, Asia and the Far East (Thorpe and Mitra 2011). This has been driven by concerted and farsighted government initiatives that since the mid-1970s have sought to diversify an economy underpinned by oil revenues, but with an otherwise limited domestic resource base. A succession of formal government plans have introduced incentives and inducements aimed at encouraging Free Zone based companies to set-up operations in the region with the aim of fast-tracking the establishment of a modern, service-based economy. The consequent phased development of Dubai’s transport and logistics sector over the past several decades, has culminated in the establishment of a major regional multimodal commercial and transport hub, a socalled ‘Transtropolis’ that includes the Dubai Logistics corridor.
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 21–31, 2020. https://doi.org/10.1007/978-3-030-44783-0_2
22
A. De Marco et al.
Dubai has arisen as a major international multi-modal commercial and logistics hub. This has been driven by bold government plans and incentives to encourage freezone based companies to set-up operations and locate their logistics services (Thorpe and Mitra 2011). Although several development stages of the logistics hub are already in action, the project remains continuously under reinforcement and numerous strategies and policy actions are undertaken to face the increasing demand of logistics capacity. Some of these policies involve investment decisions to sustain the growth of airport and port capacity in response to significant increase in demand. This paper is a contribution to support policy actions for the Dubai logistics hub project as to help public policy makers make appropriate and effective investment decisions about the planning of logistics capacity. This is done via using a System Dynamics (SD) modeling and simulation-based approach offering the necessary decision support that captures the complexity of managing the logistics hub along with overcoming the implicit policy resistance. While the use of SD to model and simulate logistics problems is not new, this work is believed to be unique in combining the three dimensions of air, sea and land logistics and using a concrete case study for simulations: the Dubai Logistics Corridor (DLC). This paper is structured as follows. Firstly, we present the SD approach and list some pertinent works available in logistics and transportation. Secondly, we illustrate the Dubai hub case study model, application, and simulations. Then, we discuss the main results and give the implications. Finally, we draw the conclusions.
2 Literature Analysis: System Dynamics Approach in Logistics SD is a modeling and computer-based simulation approach that helps understand complex systems. SD allows to graphically diagram a system of feedback-based causal loops between interrelated accumulation stocks, flow rates, and auxiliary variables, to define various linear and nonlinear mathematical relations, and to have commercial software packages do the discrete-step computational effort of solving the differential set of equations over a preset time frame (Sterman 2000). As an output to computer simulation, the curve lines of all variables are plotted on a time axis. Model testing is based on historical data and sensitivity analyses. SD lets understand the dynamics of a system, the influence of the various variables to the problem at issue, to support decision making, and test policies through simulations of various case-scenarios. Overall, the efforts for using SD in logistics is done around two main objectives: capacity simulation and policy making (Tako and Robinson 2012). For capacity simulations, different variables can be predicted such as traffic flow rates, storage capacity, quality of service, efficiency, etc. From the policies perspective, the impact can be evaluated, and case-scenarios analyses made about, for instance, the location of new ports or the development of new paths for road transportation. Specifically, some SD-based papers are available in the research area of port transportation that could be easily traced to this study about Dubai Logistics Corridor. For instance, Ruutu (2008) uses SD to forecast the Finnish sea transport national demand and associated capacity. A comparison between the SD approach and time
System Dynamics Modeling of Logistics Hub Capacity
23
series methodology or statistics tools like regression analysis is made to understand the accuracy of the different approaches. Briano et al. (2009) adopt SD decisions for the case of the Voltri terminal Europe located in Genoa, Italy. The purpose of is to design a model for the port’s performance metrics to improving the quality in ports by integration of six sigma and SD. Carlucci and Cira (2009) model through SD a plan for seaport investments. The authors focus on analyzing a small sized seaport. Its main advantage is the ability to linearly depict several relationships occurring amongst the different subjects involved, with increased advantages as opposite to more traditional approaches, like the “Costs-Benefits” mode, or the “Multi-criteria” techniques. Castillo et al. (2004) simulate the decision-making process of vessels carrying merchandise whose final destination is the province of Seville. A forecast is obtained for the port of Seville traffic, highlighting how public investment influences this entrance decision via improvements in the Port of Seville infrastructure and associated cost. De Marco and Rafele (2007) propose a simulation model as a strategic tool for policy making. In particular, the case of logistics and transportations in Piedmont is considered with reference to the localization decision for dry harbor of Genoa. Also, Sebo et al. (1995) develop a SD approach to design the intermodal port of Lewiston, Idaho, and to highlight leverage points, hidden assumptions, second order effects resulting from feedback loops, and system drivers. Intermodals are the interconnections among modes of transport like road, rail, water, and air. The development of an effective and efficient intermodal transportation system requires the identification of barriers to intermodal transportation and the investigation of the impact of proposed changes in infrastructure development, policies, regulations, and planning. A systems approach is necessary to adequately represent the interaction between the sometimes incompatible-concerns of all modes and stakeholders. Finally, dos Santos proposes an SD model to analyze investment policies for the port of Lisbon that could lead to an increase in throughput. Additional objectives include assessing port profits and investments associated with each management policy, as well as their implications to the regional economy. The impact of the port activity on regional employment, trade and GDP is used to measure the beneficial effects associated with each policy. Most papers available in the literature have research objectives like those of this work, such as logistics demand forecast, optimization of port capacity and analysis of investment, which are linked to cost savings, profitability, and long-term competitive advantage. However, out study overcomes some literature gaps. First, it is the first case study reported for the Middle East region as most of the literature is related to European or US-located logistics and transport SD models. Second, unlike the studies already available in the literature, this paper does not focus on port operations only, but it attempts to study a complete logistics hub including both sea, air and land modes of transport.
3 The Dubai Logistics Corridor Dubai Logistics Corridor has been implemented to drastically improve the trade process affecting the Dubai Logistics in its entirety, whether it involves land, air, or sea modes of transport. The Dubai Logistics Corridor is composed of three main parts: the
24
A. De Marco et al.
Jabel Ali port (governed by an institution called DP World), the Al Maktoum airport (reporting to Dubai World Central – DWC), and the Jabel Ali Free Zone Area (JAFZA). In the context of logistics management, when freight moves from one free zone to another it must undergo various long and expensive procedures of custom clearance and legal compliance. However, with the creation of the Dubai Logistics Corridor, goods moving within DP world, JAFZA and DWC, i.e., sea-road-air cargo route, need to go through customs only once at the first point of entry. After that, movement within the corridor will be relaxed as the shipment has complied with the stipulated regulations. This globally unique system has made it possible to process demands more quickly and in a cost-effective manner than ever before. Its innovative policy initiatives spell out that initiating and doing business in Dubai is consistently straightforward and constantly monitored with the advice and guidance of the rulers. Further, the time taken to unload shipment at DP World, clear the containers, and transport them to the Al Maktoum International Airport in DWC would just be a matter of a few hours (Kalli et al. 2013). Prior to the creation of the Dubai Logistics Corridor, a lot of documentation and customs work had to be finalized in a week-long period of time. Thus, The Dubai Logistics Corridor’s business model would help companies reduce their lead times and be able to enjoy more responsive logistics, without compromising their operational efficiency. To create value added to this efficiency, it is crucial to forecast the proper amount of increased capacity investment and anticipate a development schedule. The present work tackles these problems and tries to find a solution.
4 Proposed Approach The research was developed as follows. First, we worked on understanding, structuring, and analyzing how the logistics system works for both the port and airport. This was done through gathering information, freight traffic data, and interviews with local managers including the vice presidents and the logistics managers of both the port and airport operators. Second, we sketched a first-hand conceptual Causal Loop Diagram (CLD) based on system thinking (Forrester 1961). Third, a numerical SD model was created using the Vensim software tool. In compliance with SD principles and practices, the case model was developed to include stock variables, flows, and feedback loops that tackle store inventory management, store labor utilization, and customer demand. The mathematical equations underpinning the stock & flow model, were then developed. After that, the model was tested through analysis of model sensitivity associated with random exogenous variables. After testing, many simulation runs were performed under several scenarios. Finally, we analyzed results and made recommendations. 4.1
Causal Loop Diagram Representation of the Model
Here we present the model via explaining some of the most important feedback loops of the model. Figure 1 gives the CLD representation of the seaport capacity demand
System Dynamics Modeling of Logistics Hub Capacity
25
section. One has to consider that both airport and land transport sections are the same of the seaport section.
Fig. 1. Seaport capacity demand CLD
The CLD can be explained as follows. The Incremental Regional Demand SEA Flow, defined as the monthly increase in capacity demand increase as far as the Regional Demand grows. The incremental growth in capacity demand generates a shortfall in port capacity, called Port Uncovered Capacity. The uncovered capacity is the difference between desired capacity and available capacity. If positive it generates a stock out in the production area. If negative, then there will be excess capacity. As uncovered capacity increases, stakeholders will be encouraged to invest and consequently increase the ‘Additional capacity due to investments’ variable. In turn, the increase in available port capacity generates a reinforcing loop: it decreases Uncovered Capacity and increases the ‘Competitiveness and increased market share’ variable. When available capacity grows, the service offered by the port improves and so does its attractiveness and competitiveness. The interaction between the two airport and seaport model subsections happens through the variables ‘Sea to air demand’ and ‘Air to Sea demand’, respectively. These variables correspond to key factors and assumptions that are fundamental in the model, that is the possibility of neglecting control and travel times between the port and airport using the Free area zones and according to the synergy existing between the seaport and dry port. Because of this synergy there is an increase in the demand for both because considered intertwined, that is, part of the demand for one transits also in the second one generating a greater need for capacity. Figure 2 presents the reinforcing loop affecting the seaport logistics capacity. A positive change in the Incremental Regional Demand causes a variation in the port desired capacity, which is the capacity planned by stakeholders. Obviously, due to
26
A. De Marco et al.
constraints, it does not always meet the available capacity target. Consequently, a shortfall of capacity, called Port Uncovered capacity, incurs. It grows with the growth of the Port Desired Capacity. In turn, it increases the Additional Capacity variable through investments of additional capacity. Not all the discovered capacity is always transformed into additional capacity. Consequently, the additional capacity is added to the available capacity with a delay. A greater available capacity tends to make a higher level of service perceived and, therefore, increases the competitiveness, which is reflected in a further increase in regional demand, thanks to the greater market share taken and the market share of the port. Eventually, this further increases the desired capacity.
Fig. 2. Seaport capacity reinforcing CLD
The seaport logistics reinforcing loop given in Fig. 3, is counteracted by the balancing loop of the Uncovered Capacity. Here the Uncovered Capacity positively influences the Additional Capacity due to investment variable (as capacity increases, the stakeholders will have more incentive to invest to fill this gap). In return, investment that generate capacity will increase Available Capacity, but it will increase with a delay since generating new production capacity is not an instant task and requires considerable time. An increase in the available capacity will in turn decrease the Uncovered Capacity, and here we are back to the initial variable by closing this loop. It can therefore be said that the Uncovered Capacity influences itself, an increase of it determines with the delay a simultaneous decrease.
System Dynamics Modeling of Logistics Hub Capacity
27
5 The SD Model and Case-Scenario Analyses The multiple reinforcing and balancing feedback loops were then converted into a complete SD model, which can be requested by contacting the authors of this work. As a sample, Fig. 3 illustrates the seaport portion of the SD model.
Fig. 3. Seaport section of the SD model used for the simulations
The complete model includes all equations. As an example, the Port Available Capacity is the minimum between the Warehousing, Handling and Vessels Capacity and the Port Uncovered Capacity is the difference between the desired capacity and the available capacity. When it is negative it indicates an excess of capacity.
Fig. 4. Port available capacity simulations
28
A. De Marco et al.
Based on the complete SD model, we run multiple case-scenario analyses using real data collected via interviews and open data sources. In all the simulations, the time horizon is set as long as 96 months, corresponding to 8 years, being 2015 the first year in all simulations. Simulations consider either a cross-demand factor or no cross demand between the seaport or the airport. In other terms, scenarios can be tested under either dependent or independent logistics capacity between the port and the airport. The main results of simulation run for the Port Available capacity is that cross demand generates more demand for both vessels, warehousing and handling capacity than without cross demand. It can therefore be noted that the most expensive simulation (because it generates the greatest capacity) is the one with DEMAND equal to 0.2, which requires an outlay of about 6 billion dollars to generate a surplus of about 12 MTeus. Next, we find the CROSS DEMAND 0.15 and, finally, the ‘no cross demand’ that compared to the first one costs 1 Billion less to generate a capacity less than about 2 Mteus (Fig. 4). The main results of simulation run for the Airport Available capacity are shown in Fig. 5.
Fig. 5. Airport available capacity simulations
In this case the growth rate is important, in fact in the intermediate case of part in the month 0 from almost 1 million Tons and then reach 43 million Tons. The important growth is mainly since the airport is getting ready to meet an important level of capacity for the project is really ambitious. It is sufficient to think that the total of Air cargo’s capacity that is now at the Dubai International Airport (DXB) will be totally transferred to the Al Maktoum Airport. Accordingly, the air cargo, warehousing and handling capacities that form the Available capacity behave in the same way. Warehousing capacity and Handling capacity start from much higher capacity; therefore they start to increase later. Also, in this case it is worth underlining how, once determined which of the three simulations behave more truthfully, it can be used as a policy making tool to evaluate how much and when to invest and above all where to go to act in order to balance the capacity in order to optimize the investments and returns generated by it. It is recalled that a Delay function is used in order to take into consideration the time between the investment expenditure and the actual availability of the related capacity. It can therefore be noted that the most expensive simulation (because it generates greater capacity) is that of the DEMAND 0.2 which requires an
System Dynamics Modeling of Logistics Hub Capacity
29
outlay of about 20 billion dollars to generate a surplus of about 50 MTons. CROSS DEMAND 0.15 instead is a slightly cheaper but less performing solution, with an outlay of 17 billion dollars to generate a capacity of 42 MTons.
6 Validation Figures 6 and 7 report the results obtained from the simulations and compare simulated capacity versus real logistics capacity for both the airport and port systems.
Fig. 6. Airport available real capacity vs simulation forecast
Fig. 7. Port available real capacity vs simulation forecast
30
A. De Marco et al.
As far as the Airport model is concerned, the validation seems to be consistent until 2016, after which the forecasts undergo an important increase that seems not to be followed by reality. In defense of the model, however, DWC has inferred that by 2020 the resulting capacity will reach 16 MTons. A large share of this increase in capacity is due to the acquisition of all the production capacity of Air Cargo at DXB airport, which will consequently become purely a passenger transport provider. It follows that it is not the model that seriously misrepresents the forecast but the airport that is lagging the Master Plan outlined. For this reason, it is difficult to compare the simulations and therefore, to assert which of these seems to be more coherent. For the Seaport simulations, we can see how all three simulations behave in a way that is coherent with respect to reality. More precisely, if we compare the data of 2018 that are the most current today, we can see that the most accurate prediction with the minor error is with CROSS DEMAND 0.15. Depending on this it is possible to state with due precaution that it seems to be a synergy between Airport and Seaport, something confirmed by several articles on the Re-Export of the Dubai Trade.
7 Implications, Limitations and Conclusion This work has twofold implications. First, this model is proposed as a decision support tool to drive enhanced capacity investments in the Dubai Logistics Corridor. Through the proposed simulation model, it is possible to foresee the future seaport and airport capacities. Consequently, this is a predictive tool to encourage decision making. Second, this model detects capacity bottlenecks and can help in capacity investment planning and scheduling. It is also affected by some limitations, such as the assumption of perpetual growth in the time horizon, minimum level of detail of the model, and capacity utilization assumed as an aggregate form of Import-Export and Re-Export. In addition, the model does not include the super-additive relationship between investments that may generate other investments. The work presented in this paper is a unique of its kind by targeting one of the major logistics hubs in the world: the Dubai Logistics Corridor. To allow decision makers sustain the Dubai Logistics Corridor growth, we developed a modeling and computer based simulation approach based on SD to allow to capture the complexity of the logistics system at hand. Different experiments were carried out demonstrating not only the technical viability of the approach, but also its accuracy in term of predicting capacity growth. Future research is directed towards examining other aspects of the Dubai Logistics Corridor such as the appropriateness of connecting Jabel Free Zones to other free zones in the UAE using dedicated corridors.
References Briano, E., Caballini, C., Mosca, M., Revetria, R.: A system dynamics decision cockpit for a container terminal. Int. J. Math. Comput. Simul. 2(3), 55–64 (2009) Carlucci, F., Cira, A.: Modelling a plan for seaport investments through a system dynamics approach. Pomorstvo: Sci. J. Marit. Res. 23(2), 405–425 (2009)
System Dynamics Modeling of Logistics Hub Capacity
31
Castillo, J.I., López-Valpuesta, L., Aracil, M.J.: Dynamising Economic Impact Studies: The Case of the Port of Seville. Centro de Estudios Andaluces, Seville (2004) Forrester, J.W.: Industrial Dynamics. MIT Press (1961) De Marco, A., Rafele, C.: System dynamics simulation: an application to regional logistics policy making. Int. J. Comput. 4(1), 253–260 (2007) Kalli, A.B., Nova, C.F., Mohammadi, H., Sanie-Hay, Y., Al Yaarubi, Y.: The Dubai Logistics Cluster (2013) Ruutu, S.: National sea transport demand and capacity forecasting with system dynamics. M.Sc. Thesis (2008). http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.150.1977 Sebo, D.: A system dynamics approach to intermodalism at the Port of Lewiston (1995). https:// digital.library.unt.edu/ark:/67531/metadc687291/ Sterman, J.D.: Business Dynamics: Systems Thinking and Modeling in a Complex World. McGraw-Hill, Boston (2000) Tako, A., Robinson, S.: The application of discrete event simulation and system dynamics in the logistics and supply chain context. Decis. Supp. Syst. 52(4), 802–815 (2012) Thorpe, M., Mitra, S.: The evolution of the transport and logistics sector in Dubai. Global Bus. Econ. Anthol. 2(2), 1303–1313 (2011)
Towards Intelligent Waterway Lock Control for Port Facility Optimisation Thimo Schindler1(B) , Christoph Greulich2 , Dennis Bode3 , Arne Schuldt2 , Andr´e Decker3 , and Klaus-Dieter Thoben1,3 1
3
BIBA – Bremer Institut f¨ ur Produktion und Logistik GmbH, Hochschulring 20, 28359 Bremen, Germany [email protected] 2 Aimpulse Intelligent Systems GmbH, Fahrenheitstraße 1, 28359 Bremen, Germany https://www.aimpulse.com/ BIK – Institut f¨ ur integrierte Produktentwicklung, Universit¨ at Bremen, Badgasteiner Straße 1, 28359 Bremen, Germany http://www.bik.uni-bremen.de/
Abstract. High tidal ranges pose a significant challenge for affected ports. Waterway locks ensure sufficient water levels but their use often coincides with a loss of water in the harbour basins. As an alternative to energy-intensive pumping stations, it is desirable to fill the port naturally, e.g., by opening the lock gates at high tides. Unfortunately, this is a complex and dynamic scheduling problem due to manifold contributing factors. This paper outlines a novel architecture towards intelligent control for waterway lock operations. The concept employs a multi-agent system to cope with the problem complexity and dynamics. Its software agents represent relevant stakeholders, thereby integrating prediction models derived from machine learning. Keywords: Maritime logistics · Port operations · Waterway locks Data integration · Machine learning · Agent-based modelling
1
·
Introduction
Significant changes in the water level are an important challenge for sea ports exposed to high tidal ranges. Sufficient water levels in the port areas are of great importance for both safe navigation and loading/unloading processes. Waterway locks help ensure an appropriate water level. The frequent use of waterway locks, however, coincides with a respective loss of water. Without a natural influx, the energy-intensive use of pumping stations is therefore usually indispensable. As an energy-saving alternative, it is possible to fill the port naturally at high tides by opening the lock gates. This approach requires that the tide and the c The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 32–41, 2020. https://doi.org/10.1007/978-3-030-44783-0_3
Towards Intelligent Waterway Lock Control for Port Facility Optimisation
33
shipping traffic are taken into account thoroughly. It is a complex and dynamic scheduling problem due to the manifold influence factors such as weather, tide, vessel positions, and traffic planning. Particularly, vessel positions and vessel traffic planning depend on data about a highly distributed problem because vessels should be considered already a few days before they arrive in the port. Furthermore, vessel arrivals can be subject to delays (amongst many other reasons caused by weather and tide). Finally, the water level depends on tide, weather, and lock operations for arriving and departing vessels. A main responsibility of the lockkeeper is to maintain appropriate water levels in order to ensure reliable and safe vessel traffic in the port. Consequently, the keeper can deal only to a minor degree with the energyefficient increase of the water level. The “Tide2Use” project therefore develops an intelligent system to assist lock personnel in increasing also the water level reliably in an energy-efficient way. To this end, the assistance system must integrate up-to-date information, derive appropriate predictions, and propose adequate planning quickly and flexibly. The implementation concept is based on the following methods, mainly from the field of artificial intelligence: 1. Relevant data sources must be identified and integrated. In particular, information about weather, tide, current vessel positions, current vessel traffic planning, and current water levels must be made available. 2. Based on this foundation, machine learning algorithms can derive prediction models, e.g., for vessel arrivals and departures, harbour basin water levels, and matching tidal time windows for energy-efficient water-level increase. 3. A multi-agent system (MAS) that represents the relevant stakeholders will then integrate the manifold data sources and prediction models. Modelling the scheduling problem as a MAS helps coping appropriately with the underlying complexity and dynamics. The remainder of this paper is structured as follows. Section 2 describes the use case of a waterway lock and the aim of reliably keeping a high water level in an energy-efficient way. To this end, Sect. 3 investigates the lock operations as well as involved stakeholders and data sources in more detail. Based on this foundation, Sect. 4 outlines the novel approach towards intelligent control for waterway lock operations. Section 5 discusses risks and how to address them. Finally, Sect. 6 gives a summary and elaborates on next steps.
2
Use Case
The initial use case for the “Tide2Use” project is the port of Bremen in Northern Germany with its waterway lock. While the port authority Hansestadt Bremisches Hafenamt is a local government agency, the port is run, maintained, and continuously developed by bremenports, a state-owned, but privately-organized port operator. The port authority focuses mainly on maintaining the quality of service whereas the port operator is interested in lowering costs and increasing
34
T. Schindler et al.
sustainability. Moreover, the port operator follows a green ports strategy which is a key motivator for this research project. The port at the Weser River is strongly influenced by the tide due to its proximity to the North Sea estuary. The chamber of the port lock has a length of approximately 250 m and can lock ships of the Panamax class. Both heads of the lock can be sealed by sliding gates. Each gate has a set of paddles which can be lifted and lowered to adjust water levels on both sides of the gate. The standard water level of the harbour basin is usually approximately 4.2 m above chart datum and thus significantly higher than the average water level of the river. As a consequence, a single lock procedure can result in a loss of up to 10 cm water height in the harbour basin. The basin can be filled – by pumps and – by opening the gate paddles of the lock in case of high water outside. Currently, the scheduling of the lock is carried out manually with only little assistance. This is a challenging task due to a broad variety of influence factors. Furthermore, the current planning is quite inflexible because many data sources and information are not available at all times. Some data sources are only updated in fixed intervals (e.g., daily), others rely on manual input currently out of the scope of the lockkeeper. The time window in which the river level exceeds the water level of the harbour basin is at maximum two hours in the course of a tidal cycle. It is preferable to balance the water via the lock, as this process does not consume electrical energy to operate the pumps. Furthermore, it causes the water level to rise much faster. Unfortunately, the use of this period to fill the harbour basin is strongly influenced or restricted by safety regulations, environmental conditions, and shipping traffic. Among those factors, shipping traffic is particularly challenging, since it is not only highly dynamic but also prone to human error. Often, delays in expected ship arrivals are reported to the lock on short notice. Other ships even fail to announce their arrival until they almost reach the outer gate. Furthermore, vessels with a high draught cannot manoeuvre on the river during low tides, which severely restricts the degree of freedom for lock scheduling.
3
Waterway Lock Operations
As a foundation for developing an approach towards intelligent lock control, it is particularly important to identify relevant processes, involved stakeholders, and data sources in advance. The following lock operations have a direct influence on the water level of the harbour basin (Fig. 1): 1. locking of ships, during which the water level of the harbour basin can rise or drop depending on the tide outside the harbour, and 2. directly raising the water level by lifting the paddles or utilising a pump.
Towards Intelligent Waterway Lock Control for Port Facility Optimisation
35
Vessels
National Single Window
(shipmaster, clearing agent, pilot)
Registration of seagoing vessels
Lock vessel in
Port authority operations office
Lock vessel out
Lock keeper control room by lifting paddles
Increase port water level Pump station
by pump station
Fig. 1. Waterway lock operations and stake holders
With regard to the optimisation of lock operations, five major and (partly) legally independent stakeholders (Fig. 1) can be identified. Their demands, decision making, and services are crucial for the scheduling process: – The vessels, represented by their respective masters (e.g., shipmasters, clearing agents, pilots), require to enter and leave the port. – The port authority handles registration of vessels and authorises their landing and departure. – The lockkeeper operates the lock and the paddles and initiates measures to maintain the water level. – The pump station operates the pumps which can be used on demand to increase the water level of the harbour basin. – The National Single Window (NSW) serves as an information service where shipmasters register with information which port authorities can retrieve [4]. A vessel can be locked only after it was cleared by the port authority. Clearing of riverboats and barges usually takes only a few minutes while the same process for seagoing vessels is a matter of hours, sometimes even days. Since seagoing vessels can only manoeuvre on the river during high tide, they have to announce themselves via NSW 72 h before arrival (unless the travel time is less than 72 h) and have to be cleared 24 h before arrival. The port authority operations office continuously monitors NSW for new registration requests. After a vessel has been cleared to enter or leave the harbour, it coordinates locking with the lockkeeper when it arrives at the respective gate. A gate can only be moved when the paddles are completely lifted and the water level is the same on both sides of the gate. Additionally, when one of the gates is opened or its paddles lifted, the other gate must be kept sealed. As it can already be assumed from the description, most lock operations are rather time-consuming. For example, lifting the paddles of a gate takes about 8 min, opening or closing a gate takes about 5 min.
Tide forecast River water level Port water level Lock Pump station
Water level prediction
Lock state prediction
Prediction ...
Assistance system
Weather forecast
Data acquisition, processing, and provisioning
AIS
Vessel movement prediction
Information service
Port authority
Information service
T. Schindler et al.
Model for intelligent lock control
36
River/sea Pilot Shipmaster Shipping company
Lock Lockkeeper Pump caretaker
Port Harbourmaster Shipmaster Shipping company Clearing agent
Fig. 2. Architecture for intelligent lock and pump station control
Another responsibility of the lockkeeper is to maintain the water level in the harbour basin between a lower and an upper limit. If the water level needs to be increased while the tide is low, the lockkeeper arranges for the pump station to start pumping. The pump station can raise the water level by 3.5 cm per hour.
4
Architecture for Intelligent Lock Control
The aim is to develop an intelligent control for holding the water level of the harbour basin as energy-efficiently as possible (Fig. 2). The system is intended as an assistance system that supports the lock and pump station personnel in their work. Furthermore, it is intended as an information system to other stakeholders. The architecture of the framework has the following layers: 1. data acquisition and data integration (Sect. 4.1) 2. prediction models derived with machine learning (Sect. 4.2) 3. model and participant integration with software agents (Sect. 4.3) The prediction model, in combination with the multi-agent system, should be able to control the lock and the pumping station in an energy-efficient way. The use of the assistance system shall not impair the movement of vessels. 4.1
Data Acquisition and Data Integration
The data collected during data acquisition originates from different data sources. It is essential for the subsequent modelling that the various information is combined in such a way that it is available to the data-driven model reliably and in sufficient quality. By means of suitable data preparation, such as the removal of
Towards Intelligent Waterway Lock Control for Port Facility Optimisation
37
irrelevant information and outliers, the data can already be manipulated in such a way that it can be optimally fed into the prediction models. Weather data are collected at different locations around the waterway lock and the operating area of the vessels. The German Weather Service (DWD) provides its weather data on a public server. Likewise, the water level information is recorded at different locations within the river and the German Bight. All further data for model generation are recorded by the process participants and exchanged via a REST API [3] and made available to the model in sufficient quality and frequency. This information includes the condition of the lock, the energy consumption of the pumps, and the water levels within the harbour basin. Furthermore, the assistance system requires high-resolution and precise position data of the ships that will likely pass through the waterway lock. Information about planned ship arrivals are collected from different data sources. On the one hand, AIS (Automatic Identification System) data are recorded from AIS receivers. The recorded vessel routes can be used to forecast future ship tracks and arrival times at the lock by data-driven prediction models. On the other hand, information from central port management systems can be accessed to get information about destination port positions and planned arrivals. 4.2
Prediction Models
As presented in [11], measures have already been investigated to improve the planning of lock processes with rule-based algorithms. In addition, the use of machine learning in this case is intended to train an intelligent behavior that takes into account influencing factors for which no sufficiently precise analytical models exist so far. Machine learning methods approximate a sufficiently accurate representation of real-world processes [8]. Artificial neural networks can be used to identify essential and non-obvious features of the underlying process. It is therefore reasonable to assume that this is also possible for the calculation of ship arrival forecasts at the lock due to the existence of complex environmental parameters. For instance, multi-scaled neural networks can help understand and predict vessel movements [7,16]. In addition to the use of geo-positional ship data from AIS data sources as proposed in [15], effects from weather and tide shall be added to the prediction models inputs. Artificial neural networks can functionally map unintuitive causal relationships of the process. Regarding lock control, this will be used to estimate whether a water level increase through the gate makes sense. The data acquired from different sources are dynamically fed into the generated model. Using adaptive algorithms, the fully connected deep network adapts the patterns underlying the real process over its training cycle. This enables predictions about arrival times of vessels at the lock, which are then forwarded to the agent system attached. The following open-source neural-network and framework library is applied in order to derive prediction models from the underlying data sources. Through the high-level application programming interface (API) Keras, it is possible to develop deep neural networks in the Python programming language. The model
38
T. Schindler et al.
predictions are trained on appropriately scaled GPUs with the help of TensorFlow. During the training phase, TensorBoard is used to display and monitor a visualization of the training cycle. Different neural network structures are applied, and the model with the lowest training error is selected depending on its prediction accuracy. The model, which is generated by machine learning methods, should be able to calculate probabilities for the occurrence of different scenarios: – Forecast of vessel routes: what is the probability that a ship will pass the lock at what time? – Water level forecast: how and when will the water level at different locations turn out depending on weather and current conditions? – Lock condition prognosis: the lock condition is predicted, taking into account the necessary environmental parameters. 4.3
Multi-Agent System
The lock scheduling problem demands a solution to cope with high system complexity and dynamics. In particular, potentially conflicting goals of legally independent stakeholders have to be integrated. Multi-agent systems (MAS) allow modelling such complex systems with loosely coupled participants [14]. Agents are autonomous software programs which perceive and interact with their environment and make decisions required to reach their respective goals. In a MAS, agents negotiate, coordinate, form groups, or offer their services to other agents. MAS are applied to solve problems of high complexity and dynamics in many different domains [12]. In particular, they have been applied to many logistics domains such as production logistics [9], container terminal operations [6], container forwarding [13], and groupage traffic [5]. Depending on the software framework, MAS can handle up to millions of concurrently active agents [10]. The assistance system for intelligent lock control is based on the Aimpulse Spectrum agent platform. It is intended to assist the lockkeeper instead of taking direct control over the lock and the pump station. The agent model shall therefore suggest an optimized lock schedule and identify favourable time frames for increasing the water level in an energy-efficient way. While the energy consumption for increasing the water level should be minimised, two potentially conflicting goals have to be taken into account: 1. Waiting times have to be avoided to maintain the quality of service. 2. The water level of the harbour has to be kept within certain limits. Consequently, dedicated agents have to be employed, that will work towards their respective goals. A lock agent will schedule locking operations with vessel agents. Its goals are to minimise the number of locking operations (and thus the water loss) while keeping the waiting time for each vessel below a defined threshold. In order to statistically evaluate the assistance system, the waiting times before and after the introduction are calculated. It shall be ensured that the establishment of the
Towards Intelligent Waterway Lock Control for Port Facility Optimisation
39
system does not increase the waiting time. The lock agent will preferably accept locking requests from groups of vessels which can be locked simultaneously. Consequently, the vessel agents will form groups based on their size, shape, and estimated arrival time, driven by the goal to be locked as soon as possible (based on the respective interaction patterns developed by [13]). The lock agent will provide the current schedule to the lockkeeper as a suggestion of how to schedule lock operations in the real world. A harbour basin agent will pursue the goal of keeping its water level within given limits. To reach this goal, it continuously requests updates from the lock agent and a pump station agent to determine the amount of water that will be added to or removed from the basin due to their scheduled operations. While the pump station agent can only add water to the basin, the lock agent can add or remove water, depending on the tide outside of the harbour. Based on the information given, the harbour basin agent predicts future water levels and takes precautions to ensure that the water level stays within the given limits. If the water level is sinking, the harbour basin agent can auction off the task of increasing the water level to either the pump station agent or the lock agent, who will both make an offer based on the tide, the lock schedule, and the required consumption of energy and time. The framework can automatically add and remove vessel agents as the vessels enter or leave its area of competence. The agents can employ REST APIs to retrieve environment data and predictions. Information exchange between agents is structured by predefined interaction protocols.
5
Risk Management and Evaluation
Parallel to the mandatory continuous process analysis, precise risk identification and quantification is necessary. The risk analysis can assess possible dangers and potential opportunities. When identifying risks, attention must be paid to a sufficiently precise aggregation of individual sub-risks and their possible dependencies to form an overall risk [2]. In case of a port in tide-influenced waters, the most relevant risks are directly related to the port infrastructure: – If the basin is overflowed, • port operations may have to be put on hold (financial loss) and • port infrastructure may suffer severe physical damage. – If the water level falls below its lower limit, • the manoeuvrability of the ships is no longer guaranteed, • ships may run aground and suffer severe physical damage, and • the lack of pressure force can cause the quays to lose stability which leads to severe damage to the structure. – If the lock is not available due to malfunction • ships cannot enter and leave the port, and thus • port operations are jeopardized in general.
40
T. Schindler et al.
Inappropriate decision making can thus potentially cause severe problems. Consequently, the assistance system should not take control over the lock or the pump station directly. Instead, the system shall provide recommendations for the lockkeeper. However, employing the assistance system still leads to additional risks that have to be taken into account: – If data sources are not available or faulty, • they might provide inaccurate data, • which could mislead the assistance system and human operators. – If the assistance system provides faulty recommendations, • the water level may rise or fall beyond its limits and • ship traffic may be impaired. – If the human operators rely too much on the assistance system, • faulty recommendations may lead to inappropriate decisions and • the personnel might be unable to operate the port if the IT fails. In addition to the analysis and evaluation of the risks listed, suitable technical and redundant safety measures and control mechanisms must be implemented. If it is not possible to treat or reduce the identified risks appropriately, severe consequences for the entire port infrastructure and its environment cannot be ruled out. However, successful risk analysis allows complications to be determined at an early stage and essential opportunities to be identified. As a useful property, the agent-based modelling approach is well-suited for both strategy analysis and scenario analysis. Hence, risks can already be evaluated and uncovered early by simulation [1]. The Aimpulse Spectrum [10] framework provides a simulation environment without the necessity for modifications to the agents themselves. To this end, all sources of information can be replaced with matching sets of historical or fictional data. Therefore, the agents can be seamlessly transferred from simulation to the real world.
6
Summary and Next Steps
This paper describes the architecture for an assistance system for intelligent waterway lock and pump station control. To this end, the use case and its processes and participants have been investigated in detail. Based on this foundation, the paper outlines the assistance system framework starting from data acquisition over prediction models derived by machine learning algorithms to the integration in a multi-agent system. Furthermore, the paper identifies potential risks and discusses possible measures to address them. The state of the “Tide2Use” project is as follows. Data sources have already been integrated to a large extent. Currently, the project derives prediction models from the various data sources and implements the multi-agent integration for the assistance system. As a subsequent step, a thorough examination of the prediction and interaction models will be performed by means of agent-based
Towards Intelligent Waterway Lock Control for Port Facility Optimisation
41
simulation. Afterwards, there will be a real-life test run with the assistance system being installed in the lockkeeper control room. The relevant stakeholders participate in all these steps in order to foster acceptance for the novel system right from the beginning. In particular, these stakeholders also include representatives of different port operators to support result transferability early. Acknowledgements. The authors would like to thank their project partners and the anonymous reviewers for their valuable input. The “Tide2Use” project (19H18004) is funded by the German Federal Ministry of Transport and Digital Infrastructure (BMVI) in the “Innovative Port Technologies” (IHATEC) program.
References 1. Boehm, B.W.: Software risk management: principles and practices. IEEE Softw. 8(1), 32–41 (1991) 2. Cottin, C., D¨ ohler, S.: Risikoanalyse. Springer Fachmedien Wiesbaden, Wiesbaden (2013) 3. Daigneau, R.: Service Design Patterns: Fundamental Design Solutions for SOAP/WSDL and RESTful Web Services. Addison-Wesley, Upper Saddle River (2011) 4. European Parliament and Council: Directive 2009/16/EC on Port State Control (2009) 5. Gath, M.: Optimizing Transport Logistics Processes with Multiagent Planning and Control. Springer Vieweg (2016) 6. Henesey, L., Davidsson, P., Persson, J.A.: Agent based simulation architecture for evaluating operational policies in transshipping containers. Auton. Agents MultiAgent Syst. 18(2), 220–238 (2008) 7. Jahn, C., Scheidweiler, T.: Port call optimization by estimating ships’ time of arrival. In: Dynamics in Logistics, pp. 172–177. Springer (2018) 8. Khan, M., Jan, B., Farman, H.: Deep Learning: Convergence to Big Data Analytics. Springer, Singapore (2019) 9. Kirn, S., Herzog, O., Lockemann, P., Spaniol, O.: Multiagent Engineering: Theory and Applications in Enterprises. Springer, Heidelberg (2006) 10. Lorig, F., Dammenhayn, N., M¨ uller, D.J., Timm, I.: Measuring and comparing scalability of agent-based simulation frameworks. In: Multiagent System Technologies, pp. 42–60. Springer (2015) 11. Luy, M.: Scheduling von Schleusungsvorg¨ angen: Algorithmen zur Verkehrsoptimierung am Beispiel des Nord-Ostsee-Kanals. Diplomica Verlag (2018) 12. Macal, C.M., North, M.: Tutorial on agent-based modelling and simulation. J. Simul. 4, 151–162 (2010) 13. Schuldt, A.: Multiagent Coordination Enabling Autonomous Logistics. Springer, Heidelberg (2011) 14. Shoham, Y., Leyton-Brown, K.: Multiagent Systems: Algorithmic, GameTheoretic, and Logical Foundations. Cambridge University Press, Cambridge (2008) 15. Valsamis, A., Tserpes, K., Zissis, D., Anagnostopoulos, D., Varvarigou, T.: Employing traditional machine learning algorithms for big data streams analysis: the case of object trajectory prediction. J. Syst. Softw. 127, 249–257 (2017) 16. Zhang, R., Xie, P., Wang, C., Liu, G., Wan, S.: Understanding mobility via deep multi-scale learning. Procedia Comput. Sci. 147, 487–494 (2019)
On the Influence of Structural Complexity on Autonomously Controlled Automobile Terminal Processes Michael Görges1,2(&) and Michael Freitag2,3 1
2
3
BLG Logistics Group AG & Co. KG, 28203 Bremen, Germany [email protected] Faculty of Production Engineering, University of Bremen, Badgasteiner Str. 1, 28359 Bremen, Germany BIBA – Bremen Institute for Production and Logistics GmbH, Hochschulring 20, 28359 Bremen, Germany
Abstract. Planning of automobile terminal operations is a complex task, which is highly affected by volatile demand fluctuations and unforeseen dynamic events. Autonomous control concepts already showed promising results regarding the terminals logistics performance. Especially, in highly dynamic and complex settings autonomous control copes better with undesired dynamics than conventional yard planning approaches. In this regard, this paper focuses on the influence of structural complexity on the performance of an autonomously controlled automobile terminal. It addresses the terminals size and the vehicle volume as parameters of structural complexity. By using a discrete event simulation model of a generic terminal scenario, this paper analyses the logistics performance of an autonomous control strategy. It shows that autonomous control performs best in situations with a high degree of structural and dynamic complexity. Keywords: Autonomous control terminal
Discrete event simulation Automobile
1 Introduction An increasing trend in vehicle shipment volumes in global vehicle production and distribution networks can be observed. Within these networks, automobile terminals play a key role for the fulfillment of transshipments and technical services related to the customers’ demands. Usually automobile terminals offer handling processes (i.e. loading and unloading of cars from transport carriers), storage processes and technical service processes, which are directly triggered by the terminals customers (OEMs) [1]. This close interrelation with OEMs naturally affects yard planning processes of the automobile terminal. On the one hand, terminals need information about specific planned movement of cars (e.g., cars allocated to particular ship). On the other hand the terminals planning is based on order neutral long-term forecasts. Due to these characteristics (order neutral and order specific planning aspects) the role of automobile terminals in the entire automotive supply chain can be compared to a classical © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 42–51, 2020. https://doi.org/10.1007/978-3-030-44783-0_4
On the Influence of Structural Complexity
43
decoupling point with parallel occurring push and pull processes [2]. Generally, decoupling allows increasing flexibility of supply chains (e.g., terminals may act as short-term buffer of production plant). However, this leads to a higher the complexity for terminals yard planning tasks. In general, yard master planning aims at minimizing driving distances (the distance between the point of car entrance, storage area and its exit point) of cars on the terminal. In order to meet this goal classical master planning approaches assign groups of cars to predefined parking areas (e.g. sorted by manufacturer, model and destination) [3]. In less volatile situation, this assignment leads to good results regarding the realized driving distance. Increasing dynamics (e.g., caused by usage of flexibility potentials) affects this long term orientated planning negatively. It is prone to forecast deviations and unforeseen events [4, 5]. In this context first autonomous control approaches showed promising results regarding the yard assignment under dynamic and volatile conditions in automobile terminals [6, 7]. According to the concept of autonomous control an improved handling of dynamics in systems with high degree of structural complexity is a key element [8]. This paper focuses on the applicability of autonomously controlled processes at automobile terminals. It addresses the impact of increasing structural complexity on autonomous control methods performance. For this purpose, Sect. 2 gives an overview about the planning problems related to automobile terminals. On this basis, Sect. 3 describes the concept of autonomous control and presents first terminal related applications. In order to allow a structured analysis of terminals with different degrees of structural and dynamic complexity, Sect. 4 introduces a scalable generic terminal model and its implementation in a discrete event simulation. Subsequently, Sect. 4 presents an existing autonomous control and a conventional planning method as well as their implementation into the scalable terminal model. Section 5 depicts the simulation results and Sect. 6 gives an outlook with further research directions.
2 Automobile Terminal Planning Processes Planning tasks of automobile terminals are closely related to the physical movements of the vehicles. Figure 1 depicts the elements of the material flow and maps relevant planning tasks to the physical movement of cars according to [3]. The planning tasks can be differentiated according to their specific time scale. According to Fig. 1 yard master planning plays a central role in this cascaded planning process. This task comprises the assignment of incoming car volumes to suitable yard areas. On the one hand, this process aims at increasing the utilization of parking areas and on the other hand on minimizing driving distances of cars on their route between unloading locations, yard position and loading locations. Often yard master planning includes the localization of loading and unloading operations (e.g., general berth positions) [1, 4]. Usually, order neutral forecast give information about incoming vehicle volumes for yard planning purposes. This situation differs for outgoing vehicle volumes. Here terminals have information about particular customers’ orders. Dias et al. [2] describe these parallel order neutral push and customer related pull processes as a classical decoupling point in a supply chain [2]. Decoupling points allows a supply chains to react flexible to market demands fluctuations. However, they may lead to increasing
44
M. Görges and M. Freitag
complex internal dynamics. Classical yard planning addresses the orders’ neutral (forecast-driven) aspect by assigning vehicle volumes to predefined storage areas [5].
Fig. 1. Automobile terminal planning tasks (based on [3])
Volatile demand fluctuations or unforeseen disturbances may affect the planning results negatively and decrease the terminals logistics performance. Planning procedures, which allow a more dynamic assignment of cars to storage areas, may improve the handling of parallel order neutral push customer related pull processes and lead to a better planning performance.
3 Autonomous Control of Logistics Processes The concept of autonomous control aims at improving the handling of dynamic and structural complexity in logistics systems. It postulates a transfer of decision-making capabilities from centralized planning instances to the logistics objects. According to the general idea of autonomous control, intelligent logistics objects are able to interact with others in order to collect information about relevant local system states and to make decisions (e.g., routing decisions) according to their own objectives [9]. By allowing local decision-making, autonomous control aims at generating a positive emergent system behavior, which leads to an improved overall system performance [10]. Due to the locally dispersed decisions of intelligent logistics objects the system performance is less prone to unforeseen events and variations of process parameters [11, 12]. Several implementations of autonomous control strategies can be found for different logistics disciplines (e.g., production logistics [13] or transport logistics [14]). For the area of terminal logistics first autonomous control approaches showed promising results. Böse and Piotrowski [7] propose an agent based approach for
On the Influence of Structural Complexity
45
assigning vehicles to technical services and related storage areas [7]. Görges and Freitag propose a biologically inspired approach, which allows assigning groups of cars to storage areas. Compared to a classical yard assignment this approach improved the logistics performance. Especially under dynamic conditions, this new approach performs best [6]. The analysis of this autonomous control method focused on the impact of internal and external dynamics, but the relation between systems structural complexity and the systems performance is still a topic of interest. In other logistics application autonomous control contributed to the system’s ability to cope with structural complexity [12, 15]. The ability to operate in structural complex logistic systems successfully is a crucial factor for autonomous control methods. Usually, automobile terminals (e.g., Bremerhaven) are characterized by a high degree of structural complexity (i.e., capacity of more than 100,000 vehicles, multiple berths and unloading points like truck slots or rail ramps). Thus, coping with structural complexity is essential for autonomously controlled terminal processes.
4 Automobile Terminal Scenario 4.1
Structural Configuration of the Scenario
In order to address the impact of structural complexity a generic and scalable terminal scenario (analogue to [6]) will be used. Figure 2 shows all elements of this scenario. It consists of nxm parking slots (A11 to Anm ), which are interconnected by driveways. Arriving and departing cars enter via sources (I1 to Ik ) and leave the terminal via sinks (O1 to Oj ). These sources and sinks may be modelled for any kind of transport media (ship, truck or train). On a detailed level, every parking slot is defined by its height (h), is width (w), and the row width (r). The amount of parking rows l in a slot results from these parameters. The height h defines the rows capacity and the row capacities defines the capacity of the parking slot (sum of rows capacities).
Fig. 2. Scalable automobile terminal scenario with variable parking dimensions, different sources and sinks
46
M. Görges and M. Freitag
In this paper the following parameters are used for generating scenarios: The area width is w ¼ 160 m, the area height is h ¼ 75 m and the row width is r ¼ 3 m for each storage area. Hence, the capacity of a single parking slot is 795 for cars with a standard length of 5 m. In order to address the structural complexity, the size of the terminal scenario will be varied (increasing n and m). In particular, the following configurations with increasing sizes will be analyzed: 3 3, 4 4, 5 5, 6 6, 7 7, 8 8 and 9 9 parking areas. The analysis will focus on two typical auto terminals KPIs. The first KPI is the total driving distance of cars. The second KPI is the degree of sorting (i.e., sorting result). It describes the mixture of cars from different categories in one parking area. The higher the sorting result value, the better is the sorting of vehicles. 4.2
Modelling Incoming and Outgoing Volumes
In order to keep the modelling of vehicle volumes as simple as possible, this scenario focuses on vehicles arriving at the terminal via rail and leaving it via ship. There is an OEM for every parallel parking lot. This means a 3 3 size scenario has three OEMs, the 4 4 size scenario has four OEMs and so on. Every OEM in a scenario serves as many shipment destinations as sequential parking lots are available (i.e., three destinations in the 3 3 scenario and four destinations in the 4 4 scenario). In this context, a group or category of cars is defined for the following analysis as the mix of OEMs and destinations. In order to model a realistic incoming behavior for all vehicle groups a sinusoidal arrival function (analogue to [6]) will be used. This allows to model volatile seasonal demand fluctuations. Similar seasonal effects can be observed in arrival volumes of real automobile terminals. Equation (1) shows the underling sine function for every group of vehicles k. The parameter kk determines the mean arrival rate of vehicles in category k, while the amplitude of the sine function is defined by lk . Besides the mean arrival rate, the phase shift uk and the period T determine the dynamic characteristics of this arrival function. The period T has be set to a quarter year. Idk ðtÞ ¼ kkd þ lkd sin
t T
þ ukd
ð1Þ
The following table summarizes all chosen arrival rates of a destination of an OEM in for all size scenarios. Due to the higher terminals capacity the mean arrival rate has to increase with the scenario size. In our study, the mean incoming volume increases by 35 cars per day per OEM. Table 1 shows the volume distribution for all destinations. In order to provide a realistic terminals behavior an initial inventory for all vehicle categories has been modelled. The terminals initial inventory for every class is set arbitrarily to 1000 vehicles. The phase shift is modelled according to the number of destinations of an OEM. It is distributed linear over all categories in order to provide an overall homogeneous influx of cars over time. As in the real world, there are high and low runner destinations. High running destinations have a shorter turnaround time than low running destinations. In all scenarios a high runner destination has a mean terminal turnover time of 10 days and a low runner destination of 30 days. Accordingly, the turnover time vehicle of a particular
On the Influence of Structural Complexity
47
Table 1. Mean arrival rate for all destinations of one OEM for all size scenarios Arrival rates k for scenario 1 2 3 4 5 6 7 8 9 Total mean arrival rate
Scenario 33 44 16.66 13.5 33.33 27 50 40.5 – 54 – – – – – – – – – – 100 135
55 11.33 22.67 34 45 56.67 – – – – 170
66 9.76 19.52 29.29 39.05 48.81 58.57 – – – 205
77 8.57 17.14 25.71 34.29 42.86 51.43 60 – – 240
88 7.64 15.28 22.92 30.56 38.19 45.83 53.47 61.11 – 275
99 6.89 13.78 20.67 27.56 34.44 41.33 48.22 55.11 62 310
OEM is linear distributed between 10 days and 30 days for every destination served by the OEM. These turnover times determine the behavior of outgoing vehicles. 4.3
Conventional Control Method
A simple yard assignment method has been implemented. Due to the scenario settings (described in Sect. 3.2), there is a separate parking lot segment for each destination of each OEM. Based on mean arrival rates and the estimated departure times (i.e., turnaround times) volumes of every vehicle category are assigned to a particular row. Destinations with a high turnover time will be located at a parking area further away from the quayside. Accordingly, there is more free storage space for high running volumes close to the quayside. This way of assignment is comparable to a forecast driven long term yard plan. 4.4
Autonomous Control Method
The autonomous control method used in this evaluation is a pheromone based approach. This approach is inspired by ants’ natural foraging behavior. It aims at marking decision alternatives with artificial pheromones, which can be interpreted by the logistics objects. While searching for food, ants leave evaporating pheromone trails, marking possible routes to food sources. Other ants are attracted by these trails and follow it. Ants following a trail increase the pheromone concentration. The pheromone concentration decreases over time due to the natural evaporation process. This natural concept can be transferred to decision marking in auto terminal scenarios (for a detailed description see also [6]). In particular, cars using this approach try to mark suitable parking rows by leaving artificial pheromones coding information about estimated driving distances between sources, parking rows and possible sinks (i.e. berth places) as well as information about the sorting result and the turnover time. Succeeding cars are able to read all available artificial pheromone information and to decide for a suitable parking row. The following equation describes the calculation scheme for the
48
M. Görges and M. Freitag
pheromone value for a particular car category k for every available row i. As mentioned above, a category represents a combination of cars from one OEM for one destination. Pki
RANK Wik min Wik RANK ðGk Þ di vki ¼ c1 þ c2 Dk þ c3 1 V k þ c4 maxðW k Þ K F
ð2Þ
The pheromone value Pki consists of four terms. Each term focuses on a different target value and can be weighted by a factor c. The first term aims at balancing the estimated turnover time Gk and the estimated driving distances Wik between source and sink of cars belonging to category k. Therefore, this term calculates the ranking position of the estimated distance factor Wik divided by the amount of parking areas F and relates it to the ranking of turnover time of remaining categories. The second term addresses the time of the latest vehicle parked in row i di and the oldest vehicle Dk on the terminal of category k. By doing so, this term tries to address the FIFO principle of a terminal. Vehicles with similar turnover times should stand closely together. The third term focuses on volume of vehicles of category vki in the parking area of row i to the overall volume of vehicles V k belonging to category k. It tries to avoid spreading of vehicle volumes of the same category over many parking areas. The last term addresses the estimated driving distance of cars and aims at realizing short distances on the vehicles route from the source (e.g., truck or rail) to the parking row and finally to the sink (e.g., ship). It is defined as the ratio between the estimated distance Wik based on the moving average and the maximal possible distance for category k regarding all sources, storage areas and sinks. A car deciding for a particular parking row takes all available pheromone values into account and chooses finally the row with the lowest value of Pki . A moving average over the last a cars of a category is used in order to model the evaporation process. In particular Wik and Gk are determined by using a moving average. Two variants of this pheromone-based approach are implemented. In the basic variant cars are able to choose a row out of the set of rows which are available according to the conventional assignment (basic PHE). The alternative variant uses the same principles as described, but cars using this variant may choose a row out of all rows in the scenario without further restrictions (all area PHE).
5 Simulation Results The scalable nxm terminal model has been implemented in a discrete event simulation model. The terminal size has been increased systematically in different simulation runs (see also Sect. 4). All assignment methods described before are implemented in this scenario. Both pheromone-based implementations use the same weighting parameters (c1 ¼ c2 ¼ c3 ¼ 0:1 and c4 ¼ 0:4Þ, which performed well in pretest simulation runs. For reasons of general comparability, a forth assignment method has been applied in terms of a random assignment method. This random method assigns an arbitrarily chosen row to an incoming car. The performance of this method is expected to be the worst. However, it is used as a benchmark and can be seen as a kind of upper bound.
On the Influence of Structural Complexity
49
2000 1500
conv. rand phe (basic)
1000
avg. sorting res. [%]
avg. driving distance [m]
Figure 3 summarizes all simulation results. It depicts the average realized driving distance for all scenarios and for all methods. Moreover, it presents the respective sorting results (proportion of parking lots with two or more vehicle categories over time) for each scenario. Regarding the driving distances Fig. 3 shows that the autonomous control methods lead to a better performance compared to the conventional assignment. Especially, in scenarios with a higher structural complexly the autonomous control (basic PHE) performs better than the conventional method. As expected, the random assignment performs worst. Besides the driving distance, Fig. 3 shows the average sorting results in the simulations runs. As in the real world the conventional method, leads to a very good sorting result, due to its strict predefined assignment to parking areas. Compared to the conventional method the pheromone-based methods performs worse. To the cost of a lower sorting result, both autonomous control methods optimize the total driving distance. In general the pheromone based method helps to cope with increasing structural complexity.
phe (all areas)
500 0
3x3
4x4
5x5
6x6 scenario
7x7
8x8
9x9
3x3
4x4
5x5
6x6 scenario
7x7
8x8
9x9
100 80 60 40 20 0
Fig. 3. Avg. driving distance and sorting results for different size scenarios
In addition, these results show an interesting effect regarding the behavior of the basic pheromone method and the pheromone method for all area assignment. In scenarios with lower structural complexity, the alternative pheromone method (PHE all areas) performs better that the basic variant. This effects changes for scenarios with a higher structural complexity (beginning from a scenarios size 6 6). This can be explained by the underling parameters of the used pheromone method. Especially, the amount of cars for generating moving average (a values) for each Pki may have an effect on the performance. Smaller values may cause quicker changes in the decisions, while bigger values lead to slower adjustments of decisions. Figure 4 confirms this. It shows exemplarily the 3 3 and the 6 6 scenario for both autonomous control methods for increasing a value. For both scenarios (3 3 and 6 6) the basic pheromone method leads to nearly constant results for increasing a values. Due to the lower structural complexity dynamic and the fewer decision alternatives, this parameter does not have an impact on the results. The
50
M. Görges and M. Freitag
phe (basic) phe (all areas)
410 400 390 380 370 360
200
400
600
800 1000 1200 1400 1600 1800 2000 [-]
3x3 scenario
100
phe (basic) phe (all areas)
90 80 70 60 50 40
avg. driving distance [m]
3x3 scenario
420
200
400
600
800 1000 1200 1400 1600 1800 2000 [-]
avg. sorting res. [%]
avg. sorting res. [%]
avg. driving distance [m]
alternative method (PHE all areas) shows a different behavior. In both scenarios an increase of the a value lead to varying results regarding the driving distance and the sorting result. Figure 4 shows that the impact of varying a values is higher in the 6 6 scenario. Here the best avg. driving distance has been realized for a = 1400. A comparison of the basic method and the modified version shows that the alternative version performs best regarding the average driving distance in the 3 3 scenario.
650 640 630 620 610 600 590 580 570 560
6x6 scenario phe (basic) phe (all areas)
200
400
600
800 1000 1200 1400 1600 1800 2000 [-]
6x6 scenario
100
phe (basic) phe (all areas)
90 80 70 60 50 40
200
400
600
800 1000 1200 1400 1600 1800 2000 [-]
Fig. 4. Simulation results for varying a value
Regarding the sorting result, it leads to better results in the 6 6 scenario. These results show, that besides the weighting factors the parameter a can be used to adjust the PHE methods performance.
6 Summary and Outlook This paper presented a biologically inspired autonomous control method for the general assignment of cars at an automobile terminal and compared it with a conventional method. The analysis focused on scenarios with varying degrees of structural complexity. The analysis confirmed that the novel autonomous control method performs best in scenarios with a higher structural complexity (11% driving distance improvement for the basic method in the 6 6 scenario). This analysis showed that the improvements of avg. driving distances lead to lower sorting results. Moreover, the analysis focused on the impact of the pheromone based methods parameters. It showed that an adjustment of the methods underling parameters helps to improve the overall performance of the autonomous control method. Overall, it showed that autonomous control helps to cope with increasing structural complexity in different terminal scenarios. Further research activities will focus on this aspect and investigate methods for dynamically adjusting the moving average parameter in order to optimize the methods results. Another interesting research field is the implementation of similar autonomous control strategies for other logistics object (e.g., ships or trains). Acknowledgements. This research is part of the project “Isabella - Automobile logistics in seaand inland ports: interactive and simulation-based operation planning, dynamic and context-based
On the Influence of Structural Complexity
51
control of device- and load movements”, funded by the German Federal Minis-try of Transport and Digital Infrastructure (BMVI), reference number 19H17003A.
References 1. Mattfeld, D.C.: The Management of Transshipment Terminals: Decision Support for Terminal Operations in Finished Vehicle Supply Chains. Springer Science+Business Media, New York (2006) 2. Dias, J.C.Q., Calado, J.M.F., Mendonça, M.C.: The role of European «ro-ro» port terminals in the automotive supply chain management. J. Transp. Geogr. 18, 116–124 (2010). https:// doi.org/10.1016/j.jtrangeo.2008.10.009 3. Görges, M., Freitag, M.: Dynamisierung von Planugsaufgaben auf Automobilterminals. Ind. 4.0 Manag. 35, 23–26 (2019) 4. Mattfeld, D.C., Orth, H.: The allocation of storage space for transshipment in vehicle distribution. OR Spectr. 28, 681–703 (2006). https://doi.org/10.1007/s00291-006-0051-6 5. Cordeau, J.-F., Laporte, G., Moccia, L., Sorrentino, G.: Optimizing yard assignment in an automotive transshipment terminal. Eur. J. Oper. Res. 215, 149–160 (2011). https://doi.org/ 10.1016/j.ejor.2011.06.008 6. Görges, M., Freitag, M.: Modeling autonomously controlled automobile terminal processes. In: Proceedings of the Hamburg International Conference of Logistics (HICL), Hamburg (2019) 7. Böse, F., Piotrowski, J.: Autonomously controlled storage management in vehicle logistics —applications of RFID and mobile computing systems. Int. J. RF Technol.: Res. Appl. 1, 57–76 (2009) 8. Hülsmann, M., Schloz-Reiter, B., Freitag, M., Wycisk, C., de Beer, C.: Autonomous cooperation as a method to cope with complexity and dynamics?—a simulation based analyses and measurement concept approach. In: Bar-Yam, Y. (ed.) Proceedings of the International Conference on Complex Systems (ICCS 2006), pp. 1–8, Boston, MA, USA (2006) 9. Windt, K., Hülsmann, M.: Changing paradigms in logistics—understanding the shift from conventional control to autonomous cooperation and control. In: Hülsmann, M., Windt, K. (eds.) Understanding Autonomous Cooperation and Control in Logistics, pp. 1–16. Springer, Berlin, Heidelberg (2007) 10. Freitag, M., Herzog, O., Scholz-Reiter, B.: Selbststeuerung logistischer Prozesse - Ein Paradigmenwechsel und seine Grenzen. Ind. Manag. 20, 23–27 (2004) 11. Windt, K., Philipp, T., Böse, F.: Complexity cube for the characterization of complex production systems. Int. J. Comput. Integr. Manuf. 21, 195–200 (2008). https://doi.org/10. 1080/09511920701607725 12. Scholz-Reiter, B., Rekersbrink, H., Görges, M.: Dynamic flexible flow shop problems— scheduling heuristics vs. autonomous control. CIRP Ann. 59, 465–468 (2010) 13. Toshniwal, V., Duffie, N., Jagalski, T., Rekersbrink, H., Scholz-Reiter, B.: Assessment of fidelity of control-theoretic models of WIP regulation in networks of autonomous work systems. CIRP Ann. 60, 485–488 (2011). https://doi.org/10.1016/j.cirp.2011.03.045 14. Rekersbrink, H., Makuschewitz, T., Scholz-Reiter, B.: A distributed routing concept for vehicle routing problems. Logist. Res. 1, 45–52 (2009) 15. Scholz-Reiter, B., Görges, M., Philipp, T.: Autonomously controlled production systems— influence of autonomous control level on logistic performance. CIRP Ann. 58, 395–398 (2009)
Literature Classification on Container Transport Systems for Inter-terminal Transportation Nicole Nellen1(&), Michaela Grafelmann1, Justin Ziegenbein2, Ann-Kathrin Lange1, Jochen Kreutzfeldt2, and Carlos Jahn1 1
2
Institute of Maritime Logistics, Hamburg University of Technology, Hamburg, Germany [email protected] Institute of Technical Logistics, Hamburg University of Technology, Hamburg, Germany
Abstract. The upward trend in global containerized trade is predicted to continue. In addition, increasing ship sizes, growing demand for port-centered value added services and environmental considerations are creating challenges for handling port internal traffic, also referred to as inter-terminal transportation (ITT). Container transports are often carried out by truck, which may lead to congestion in the port area. Due to increasing demand for greater efficiency and for more sustainable and environmentally friendly approaches, new transport solutions receive more attention. In order to strengthen their competitiveness, seaports consider digitalization-driven innovations and technologies and intelligent new concepts to improve the quality and efficiency of port activities, including ITT. The main purpose of the paper is to highlight different approaches to conventional means of land-based transport and future trends. Based on an extensive literature survey, a classification scheme was developed and applied to scientific publications. The classification scheme comprises a multitude of criteria, including methodology and research objective of the publication and the degree of implementation and automation of the discussed approaches. In addition, challenging subject areas that might help the discussion of the topic in the future, are identified. The findings show that solutions discussed in literature often provide conceptual and very theoretic outlines. It is particularly noticeable that the publications do not deal comprehensively with the integration into existing logistics systems. This shows research perspectives in most diverse areas including the effective process integration at nodes and edges of innovative ITT networks. Keywords: Inter-terminal transportation survey Classification
Container transport Literature
1 Motivation With regard to positive trends in the global economy, the United Nations Conference on Trade and Development (UNCTAD) is forecasting a continuous annual growth rate of 3.8% of global maritime trade between 2018 and 2023, predicting the greatest © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 52–61, 2020. https://doi.org/10.1007/978-3-030-44783-0_5
Literature Classification on Container Transport Systems for ITT
53
growth for containerized and dry bulk goods [1]. Moreover, with the introduction of ever larger container vessels, maritime transport chains are becoming more complex. Container vessels enter ports with lower frequency but larger volumes, which causes peaks in various phases of container handling, including truck dispatching [2]. Besides, the growth in truck traffic is also associated with an increasing demand for portcentered, value-added logistics services. In addition, intensive seaport competition and spatial restrictions of ports require efficient means of handling port traffic. Objectives such as minimizing transport delays, time and costs, high capacity utilization rates and minimizing empty runs are pursued [3]. At the same time, environmental protection is becoming more important for sustainable growth in transport and logistics. Due to greater environmental awareness and legal requirements, the commitment to CO2-reduced logistics is increasing, for example by using environmentally friendly modes of transport [4]. Digitalization offers many new opportunities in maritime logistics to increase productivity, efficiency and sustainability. Smart ports as a concept, for example, aim to use modern information technologies to improve planning and management within and between ports. This requires investments in technologies and collaborations needed for information exchange [5]. Ports use these modern technologies to provide better operational control and meet new challenges in maintaining safe, secure and energyefficient facilities that reduce their environmental impact [6]. For example, truck appointment systems at container terminals are used to improve the efficiency of transport operations [7]. In addition, the use of intelligent infrastructure and automation can be taken into account to enable alternative solutions to conventional container transports by truck. At present, neither more efficient and environmentally friendly water-based nor land-based transport modes are used to transfer relatively high quantities within terminals/service facilities. Focus of this paper lies on alternative solutions to conventional land-based transports and highlighting challenges of their integration in seaports. Since waterbased transport is structurally limited in its connectivity of nodes/facilities and moves with quay-cranes are very costly and often represent a bottleneck at terminals, waterbased solutions are not generally suitable. Nevertheless, there are innovative waterbased approaches such as the port feeder barge, which is equipped with its own container crane [8], or waterborne automated guided vehicles [9]. The objective of this work is to provide a structured classification of literature in order to identify relevant and challenging research fields. Therefore, this paper is organized as follows. A conceptual delimitation of ITT is introduced in Sect. 2. Then in Sect. 3, based on an extensive literature survey, we create a scheme to classify relevant literature. General challenging subject areas that might help the discussion of the topic in future research are proposed in Sect. 4, followed by a conclusion in Sect. 5.
2 Introduction to Inter-terminal Transportation Geerlings et al. [10] provide an overview of ports and maritime networks, including basic knowledge of definitions, actors and functions as well as diverse concepts of ports. Steenken et al. [11] describe logistic processes and operations in container terminals including further information on transport and handling equipment.
54
N. Nellen et al.
ITT describes the movement of containers between terminals within a port that serve container ships, railways, inland vessels and other forms of hinterland transport [12]. Further definitions also assign ITT to any type of land- and water-based transport of containers and cargo between organizationally separate areas within a seaport and from and to dry ports [3]. In some ports, the demand for ITT results from separately located container terminals. In addition, there are dedicated rail and inland waterway terminals, which bundle freight from various actors within the port for intermodal transportation. Moreover, there are facilities that offer value-added services, which can include labelling, packaging, inspection and customizing activities [13]. Besides, ITT networks link service facilities such as container repair stations, empty container depots and customs facilities. These procedures can be mandatory and therefore imply ITT [3]. In the following, container transport systems for ITT are characterized by enabling land-based transfers of containerized units between points of origin and destination, which can be defined as nodes that are linked by edges. The network of nodes may include terminals and service facilities within the port as well as inland hubs and dry ports near the port.
3 Literature Review and Classification There are various approaches in literature to design the transport of containers between terminals, ranging from conventional systems to future trends. Several authors investigate different system approaches with the help of simulation or stochastic models, while considering manned and automated systems. Duinkerken et al. [14], Schroër et al. [15] and Gharehgozli et al. [16], inter alia, use simulation to compare different container transport systems. Those systems are based on existing handling equipment for horizontal container transport in container terminals. Duinkerken et al. [14] focus on an object-oriented simulation of multi-trailer systems (MTS), automated guided vehicles (AGV) and automated lift vehicles (ALV) to move containers between two nodes with a maximum distance of 6,000 m. Schroër et al. [15] investigate alternatives for container transport between the terminals Maasvlakte 1 and Maasvlakte 2 in the Port of Rotterdam using discrete event simulation. Gharehgozli et al. [16] also concentrate on MTS, AGV and ALV with a dedicated infrastructure and additionally consider the transport of containers by manned trucks on public roads. Other publications deal with innovative transport systems. For example, van der Heijden [17] developed a concept for unmanned, automatic transport of containers. The containers are loaded onto trailers, which are pulled by electric vehicles guided by a special rail system. In addition, there are publications dealing with the underground transport of containers. Stein et al. [32] present a future traffic scenario in which modules transport the containers autonomously through tunnels. Furthermore, there are publications that give an overview of ITT. Heilig and Voß [3] analyze research works concerning ITT to discuss the recent state-of-the-art and research progress, focusing on publications regarding decision analytics as well as
Literature Classification on Container Transport Systems for ITT
55
innovative information technologies, comprehensively. They consider an assessment of simulation studies as well as optimization and information system approaches. Besides, they suggest future research topics to be linked with ITT like transport scheduling, vehicle routing, collaborative planning and resource sharing, information technologies, green logistics, deployment models and application areas. Shin et al. [18] provide an overview of latest examples and trends related to automated intermodal freight transport system technologies, which mainly aim to relief the infrastructure and to provide a more sustainable, environmentally friendly solution. They give directions for future technical developments. Hu et al. [19] provide a systematic review of existing research on ITT planning in port areas and the hinterland, identifying several research gaps like multi-modality of ITT systems, which are rarely studied. The reviewed papers illustrate the scientific relevance of ITT, however, they tend to cover the technical components of general freight transport systems or the planning and controlling of ITT. This paper, distinguished from the above, focuses mainly on alternative transport systems. Those are either (1) improving transport on the road infrastructure by using other vehicles than conventional trucks, or (2) shifting container transport volumes from road to other traffic modes (e.g. rail); that may be more efficient, eco-friendly, reduce congestion and/or even facilitate improved logistics solutions. 3.1
Classification Scheme
After giving a brief introduction into the topic we analyze relevant publications in the field of container transport systems of ITT in order to extensively reflect the progress of research and the current state of the art. The literature search identified 20 relevant publications between 1995 and 2019, which are presented in Table 2. The classification scheme is separated in ten categories, shown in Table 1: Methodology, Port related, System, Degree of automation, Infrastructure, Dedicated infrastructure, Underground, Maximum yearly movements, Objective of the paper, and Degree of implementation. While Methodology describes the reviewed author’s procedure, Port related examines whether the publications refer to container transports between terminals or other nodes and System describes the type of transport vehicles used. If more than one system is addressed in one publication, those are classified individually. In addition, the Degree of automation (manned or automatic vehicles) and whether the system is designed for the underground transport of containers (Underground) is taken into account. Moreover, the Infrastructure is considered and whether the system requires their own infrastructure (Dedicated infrastructure). Maximum yearly movements is subdivided into rather high, which defines an annual volume of more than 3 million TEU, rather low if it is less. Depending on parameters, means that in the publication the maximum container turnover of the system was not specified and is scalable. Rather, the performance of the system depends on the number of vehicles used, for example. Finally, the underlying objectives of the publication (Objective of the paper) and the degree of maturity of the system (Degree of implementation) are considered.
56
3.2
N. Nellen et al.
Classification Tables
In the following, Table 1 shows the classification categories with the possible specifications. For overview purposes, the individual specifications of the categories are assigned to numbers, which are used in the classification table. Table 1. Classification categories with the possible specifications Category Methodology
# 1 2
Degree of automation
5 1 2
Specification Concept Simulation-based study Mathematical model Yes No Multi trailer system Automated guided vehicle Automated lifting vehicle Conventional trucks Other systems Automated Manned
Objective of the paper
1 2
Increase handling Reduce emissions
3
Relieve infrastructure Reduce costs
3 Port reference System
1 2 1 2 3 4
4
Category Infrastructure
# 1 2
Specification Road Rail
3
Other
1 2 1
Yes No Yes
2
No
Maximum yearly movements
1 2
Degree of implementation
3 1 2
Rather high Depending on parameters Rather low Conceptual Pilot operations (planned) Implemented
Dedicated infrastructure Underground
3
It should be noted that the status of implementation is assessed according to the current status of the respective paper. If no specification of a category is selected, then the information basis was not sufficient for an assignment. In the following table, gray fields symbolize that the specification applies, multiple choice is possible. It becomes evident that different methodologies are used to approach the topic, whereby the publications found mainly relate to ports. However, the processes of integration into the port (connection of nodes, security purposes, etc.) were not particularly taken into account; rather, transport volumes and networks were used as data
Literature Classification on Container Transport Systems for ITT
57
Degree of implementation
Objective of the paper
Maximum yearly movements
Underground
Dedicated infrastructure
Infrastructure
Degree of automation
System
Port reference
Methodology
Table 2. Classification table
1 2 3 1 2 1 2 3 4 5 1 2 1 2 3 1 2 1 2 1 2 3 1 2 3 4 1 23 van der Heijden et al. (1995) Pielage (2001) Ottjes (2002) Hansen (2004) Liu (2004) Duinkerken et al. (2006) Ottjes (2006) Zhang et al. (2006) Vernimmen et al. (2007) Winkelmans (2010) Roh (2011) Roop et al. (2011) Mishra et al. (2013) Nieuwkoop et al. (2014)
Schroër et al. (2014)
Stein et al. (2014) Tierney et al. (2014)
Spruijt et al. (2017) Gharehgozli et al. (2017) Gao et al. (2019)
basis (e.g. Tierney et al. [12], Schroër et al. [15]). The means of transport are often based on existing port equipment such as AGV, ALV and MTS. In addition, there are studies that deal with rail-guided systems that project existing and innovative approaches onto container transport (e.g. van der Heijden et al. [17], Vernimmen et al. [26]).
58
N. Nellen et al.
Automated solutions are often considered, mainly with focus on technical feasibility or efficiency improvement, less on integration in existing processes and connection possibilities. Dedicated infrastructures are often taken into account, whereby a sufficient availability of space in the port area is assumed. In order to address space problems, some publications focused on underground solutions (e.g. Stein et al. [32], Winkelmans [27]). The transport volume that can be handled by a transport system must meet the requirements of individual ports. The maximum yearly movements specification shows that a wide variety of overall volumes can be transported. The publications pursue a broad range of objectives, although it is noticeable that the most common aim is to increase the handling volume. Finally, it can be noted that the vast majority of publications still deal with the problem in a conceptual way.
4 Research Perspectives The examination of the classification table has revealed general gaps in research from which research perspectives are subsequently derived. Therefore, the following section deals generically with research topics, which consider challenges regarding the integration of more efficient and/or autonomous container transport systems for ITT. 4.1
General
Literature research has shown that many systems are in different phases of implementation. Moreover, some systems were not explicitly intended for container transport in port areas but could possibly be adapted to the given problem. Ports are diverse areas of application and adaptation is most likely necessary anyway. The existing infra- and superstructure has to be considered, as well as the integration into existing IT systems. Furthermore, deployment models (e.g. by third-party operators, participating nodes, hauliers) should be taken into account. Safety considerations need to be addressed, also to promote acceptance of a system within the port and wider society. This applies in particular to autonomous processes, especially in mixed traffic with conventional vehicles or at interfaces with nonautomated processes. Additionally, logistic and IT processes need to be secured, since they involve various actors which might need to intensify their cooperation, e.g. by implementing new IT interfaces and exchanging real time data. High investments, construction projects and the elimination of existing structures should be legitimated by the reduction of logistics costs, which improves the competitiveness of a port, or by environmental arguments (e.g. most of the reviewed systems are electrically powered). The connection of a dry port or a hub in the hinterland to ITT systems is currently not sufficiently discussed in literature. The scalability of the solutions is important, as the development of transport volumes is difficult to predict. Moreover, expansion possibilities are beneficial as well as the flexibility to integrate or adapt new intelligent and innovative systems. Therefore, solutions should be evaluated in combination with e.g. collaborative planning techniques.
Literature Classification on Container Transport Systems for ITT
4.2
59
Process Integration at the Nodes
Many of the presented publications do not sufficiently address the connection to existing logistic nodes. However, systems, which simplify the loading and unloading process at the nodes or even make it obsolete, are to be emphasized. Especially due to short transport routes within the port, handling processes have a great impact on overall process costs and time. Large operators, such as the main container terminals within a port, should therefore be equipped with an efficient handling solution and automated gate processes should be considered. Moreover, the integration of existing handling equipment and information systems at the nodes should be taken into account. Especially actors with fewer transport quantities, like relatively small container packing stations or value-added-service facilities, do not necessarily allow a dedicated linkage to ITT systems. Simplified, cost-effective and space-saving handling at the interface at each actors’ node is required. The system should to be integrated as seamlessly as possible into the existing processes. 4.3
Process Integration at the Edges
Some systems aim to separate the container transport from public infrastructure to reduce congestions. There are mainly three approaches: by using elevated forms of transport as well as underground transport and dedicated, non-public, roads. Furthermore, solutions for a more efficient utilization of existing infrastructure are considered (e.g. platooning, the use of AGV on public roads, route optimization). These different approaches require different amounts of available space in the port area and pose different construction challenges. In addition, these approaches intervene to varying degrees in port processes. In addition, it must be questioned how durable the individual systems are and to what extent the infrastructure can be used flexibly. Some of the publications describe transport systems designed to handle a large quantity of node-to-node container transports, which are not generally suitable for shorter route length and lower volumes. Furthermore, it should be borne in mind that dangerous goods or refrigerated containers are also transported in the port. Therefore, it is worthwhile to evaluate the suitability of the concepts and, if necessary, to outline possible adaptations.
5 Conclusions Seaports facing increasing transport volumes and competitive pressure seek for efficient solutions to handle ITT. Many of the publications provide conceptual and very theoretic outlines. By classifying relevant literature, this paper highlights different aspects from publications between 1995 and 2019 and provides a generic outlook on research topics to be considered in the future. It is generally noticeable that the publications do not deal comprehensively with the integration into existing logistics systems. This paper is a first attempt to reveal gaps that hamper successful implementation in the maritime environment. Further publications should focus on individual research perspectives in detail in order to develop comprehensive concepts.
60
N. Nellen et al.
References 1. UNCTAD: Review of Maritime Transport 2018, United Nations (2018) 2. Ramírez-Nafarrate, A., González-Ramírez, R.G., Smith, N.R., Guerra-Olivares, R., Voß, S.: Impact on yard efficiency of a truck appointment system for a port terminal. Ann. Oper. Res. 258(2), 195–216 (2017) 3. Heilig, L., Voß, S.: Inter-terminal transportation: an annotated bibliography and research agenda. Flex. Serv. Manuf. J. 29(1), 35–63 (2017) 4. Bailey, D., Solomon, G.: Pollution prevention at ports: clearing the air. Environ. Impact Assess. Rev. 24(7–8), 749–774 (2004) 5. Heilig, L., Lalla-Ruiz, E., Voß, S.: Digital transformation in maritime ports: analysis and a game theoretic framework. Econ. Res. Electron. Netw. 18(2–3), 227–254 (2017) 6. Molavi, A., Lim, G.J., Race, B.: A framework for building a smart port and smart port index. Int. J. Sustain. Transp. 10(2), 1–13 (2019) 7. Torkjazi, M., Huynh, N., Shiri, S.: Truck appointment systems considering impact to drayage truck tours. Logist. Transp. Rev. 116, 208–228 (2018) 8. Malchow, U.: Port feeder barge: advanced waterborne container logistics for ports. TransNav, Int. J. Mar. Navig. Saf. Sea Transp. 8(3), 411–416 (2014) 9. Zheng, H., Negenborn, R.R., Lodewijks, G.: Closed-loop scheduling and control of waterborne AGVs for energy-efficient inter terminal transport. Transp. Res. Part E: Logist. Transp. Rev. 105, 261–278 (2017) 10. Geerlings, H., Kuipers, B., Zuidwijk, R.: Ports and Networks: Strategies, Operations and Perspectives. Routledge, Abingdon (2017) 11. Steenken, D., Voß, S., Stahlbock, R.: Container terminal operation and operations research-a classification and literature review. OR Spectrum 26(1), 3–49 (2004) 12. Tierney, K., Voß, S., Stahlbock, R.: A mathematical model of inter-terminal transportation. Eur. J. Oper. Res. 235(2), 448–460 (2014) 13. United Nations: Value-added services of logistics centres in port areas. In: Commercial Development of Regional Ports as Logistics Centres, pp. 19–40 (2002) 14. Duinkerken, M.B., Dekker, R., Kurstjens, S.T., Ottjes, J.A., Dellaert, N.P.: Comparing transportation systems for inter-terminal transport at the Maasvlakte container terminals. OR Spectrum 28(4), 469–493 (2006) 15. Schroër, H.J., Corman, F., Duinkerken, M.B., Negenborn, R.R., Lodewijks, G.: Evaluation of inter terminal transport configurations at Rotterdam Maasvlakte using discrete event simulation. In: Winter Simulation Conference, vol. 2014, pp. 1771–1782 (2014) 16. Gharehgozli, A.H., de Koster, R., Jansen, R.: Collaborative solutions for inter terminal transport. Int. J. Prod. Res. 55(21), 6527–6546 (2017) 17. van der Heijden, B., Heere, E., Jongejan, M.: The Combi-Road control system. In: Proceedings of the Second World Congress on Intelligent Transport Systems, vol. 2, pp. 887–890 (1995) 18. Shin, S., Roh, H.S., Hur, S.H.: Technical trends related to intermodal automated freight transport systems (AFTS). Asian J. Shipp. Logist. 34(2), 161–169 (2018) 19. Hu, Q., Wiegmans, B., Corman, F., Lodewijks, G.: Critical literature review into planning of inter-terminal transport: in port areas and the Hinterland. J. Adv. Transp. 2019(2062), 1–15 (2019) 20. Pielage, B.J.: Underground freight transportation. A new development for automated freight transportation systems in the Netherlands. In: 2001 IEEE Intelligent Transportation Systems, pp. 762–767. IEEE, Piscataway (2001)
Literature Classification on Container Transport Systems for ITT
61
21. Ottjes, J., Veeke, H.P., Duinkerken, M.B.: Simulation studies of robotized multi terminal systems. In: Proceedings International Congress on Freight Transport Automation and Multimodality (FTAM) (2002) 22. Hansen, I.: Automated shunting of rail container wagons in ports and terminal areas. Transp. Plan. Technol. 27(5), 385–401 (2004) 23. Liu, H.: Feasibility of using pneumatic capsule pipelines in New York City for underground freight transport. In: Pipeline Engineering and Construction: What’s on the Horizon?, pp. 1– 12 (2004) 24. Ottjes, J., Veeke, H., Duinkerken, M., Rijsenbrij, J., Lodewijks, G.: Simulation of a multiterminal system for container handling. OR Spectrum 28(4), 447–468 (2006) 25. Zhang, J., Ioannou, P.A., Chassiakos, A.: Automated container transport system between inland port and terminals. ACM Trans. Model. Comput. Simul. (TOMACS) 16(2), 95–118 (2006) 26. Vernimmen, B., Dullaert, W., Geens, E., Notteboom, T., T’Jollyn, B., Van Gilsen, W., Winkelmans, W.: A way to cope with growing internal container traffic in the port of Antwerp? Transp. Plan. Technol. 30(4), 391–416 (2007) 27. Winkelmans, W.: Sustainable port development and technological innovation - case study. University of Antwerp, Antwerp (2010) 28. Roh, H.S.: A scheme to build an automated container transport system (AutoCon) between Seoul and Busan. KOTI World Brief 3(31), 2–5 (2011) 29. Roop, S.S., Ragab, A.H., Olson, L.E., Protopapa, A.A., Yager, M.A., Morgan, C.A., Warner, J.E., Mander, J., Parkar, A.S., Roy, S.L.: The freight shuttle system: advancing commercial readiness. Texas Transportation Institute (2011) 30. Mishra, N., Roy, D., van Ommeren, J.C.W.: A stochastic model for inter-terminal container transportation. University of Twente, Department of Applied Mathematics, Enschede (2013). (Memorandum No. 2032) 31. Nieuwkoop, F., Corman, F., Negenborn, R., Duinkerken, M., van Schuylenburg, M., Lodewijks, G.: Decision support for vehicle configuration determination. In: 2004 Proceedings of the International Conference on Networking, pp. 613–618. IEEE (2014) 32. Stein, D., Beckmann, D., Siefer, T., Stein, R.: Automatischer Güterverkehr im Untergrund – Ein verkehrstechnisches Zukunftsszenario. In: Keim, H., Wolf, J. (eds.) Diskussionsbeitrag. Europäische Fachhochschule, Bühl (2014) 33. Spruijt, A., van Duin, J., Rieck, F.: Intralog towards an autonomous system for handling inter-terminal container transport. In: EVS30 Symposium (2017) 34. Gao, Y., Chang, D., Fang, T., Luo, T.: Design and optimization of parking lot in an underground container logistics system. Comput. Ind. Eng. 130, 327–337 (2019)
A Simulation Study of a Storage Policy for a Container Terminal Henokh Yernias Fibrianto1, Bonggwon Kang1, Bosung Kim1, Annika Marbach2, Tobias Buer3, Hans-Dietrich Haasis2, Soondo Hong1(&), and Kap Hwan Kim1 1
Pusan National University, Geumjeong-gu, Busan 46241, Korea [email protected], [email protected], [email protected], {soondo.hong,kapkim}@pusan.ac.kr 2 University of Bremen, Bibliothekstraße 1, 28359 Bremen, Germany {annika.marbach,haasis}@uni-bremen.de 3 German University of Technology in Oman, Halban, Muscat, Germany [email protected]
Abstract. This paper proposes a storage policy for container terminals that handle large numbers of vessels and containers. The storage policy considers the estimated workload at a certain area in a given period; the partition of a storage block into subblocks; the proximities between containers belonging to the same group; the segregation between different groups of containers; and the stack heights of containers. We develop a framework for simulating container repositioning and vehicle congestion and use it to evaluate the yard crane productivity rate, amount of repositioning, and service time of a real-world port terminal. The preliminary result shows that the container terminal operates more efficiently under the storage policy with a bay as a subblock setting. Keywords: Storage policy
Simulation Container terminal
1 Introduction The increasing use of container ships has encouraged container terminals to improve their ability to handle larger numbers of ships and containers more efficiently, and global competition has incentivized container terminal operators to improve their services (shorter vessel turnaround times, faster container unloading, etc.) [1]. One option that promises significant improvement in container terminal operations, according to studies, is that having a management strategy in place can reduce vessel turnaround time [2] and container retrieval time [3], and improve land productivity [4] and overall container terminal productivity [5]. Developing and implementing the most effective strategy, however, may be challenging. The challenges mostly come from the assumptions, such as negligible repositioning operation, negligible truck congestion, and known demand, that are difficult to achieve in practice. In this paper, we propose a policy-based storage management strategy (hereafter storage policy), which is both practical and effective for determining each container’s © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 62–69, 2020. https://doi.org/10.1007/978-3-030-44783-0_6
A Simulation Study of a Storage Policy for a Container Terminal
63
location. The proposed storage policy considers the estimated workload at a certain area in a given period of time, the partition of a storage block into subblocks, the proximities between containers belonging to the same group, the segregation between different groups of containers, and the stack heights of containers. We develop a framework to measure the policy’s key performance indicators: yard crane productivity rate, amount of container repositioning, and service time.
2 Literature Review We briefly review the relevant literature on storage management strategies. These studies identify four specific parameters that affect a storage management strategy: the workload at a particular storage area, vehicle congestion, stack height, and shared space. Jeong, Kim, Woo and Seo [2], who proposed a workload-based yard planning method, showed that considering the workload at each storage block reduces the turnaround times of vessels. Petering [6] presented a storage location assignment system considering the distance between the berth and storage location, yard template, truck congestion, and stack height. Jiang, Lee, Chew, Han and Tan [4] proposed a twostep solution consisting of a template generation which allocates the subblocks for each vessel, and a space allocation and workload assignment which regulates the sharing space between neighboring subblocks. Zhen [7] considered the uncertainty in the number of containers and the amount of truck congestion when allocating subblocks for each vessel. He and Tan [8] investigated a resilient yard template that minimizes the risk of assigning slots that are unavailable because of fluctuations in storage demand. Tan, He and Wang [9] addressed a flexible yard template, considering yard allocation and yard crane deployment, to minimize the total cost of container transportation and YC movement. Jin, Lee and Hu [10] studied a berth and yard template design to balance quay-side workloads caused by vehicle congestion, considering transshipment demand. Li [11] studied the sizes and locations of export container groups, considering peak workloads in yard storage.
3 Storage Policy for a Container Terminal We design a storage policy to define the storage location of an incoming container by sequentially determining the block, subblock, and row location. When deciding the block location, the storage policy considers the workload at each block in the time period when the container is expected to be stored and then selects the block with the least workload to balance the workload among the blocks. After the block has been selected, the storage policy selects the subblock based on the segregation enforcement level and expected proximities. The segregation enforcement level is a parameter with an integer value which limits the number of container groups that may be stored in a subblock. We assume that the segregation enforcement level regulates the amount of repositioning and space utilization. The expected proximity, which is a parameter with a real value ranging from 0 to 1, represents the importance of concentrating containers in a subblock by placing the containers close to
64
H. Y. Fibrianto et al.
other containers in the same group. We also assume that the expected proximity influences the rail-mounted gantry crane (RMGC) movement cost, which affects the time to store/retrieve a container. The selection of the subblock is also dictated by the partition resolution; here, we use a bay as a subblock, a half-bay as a subblock, and a slot as a subblock. The partition resolution complements the segregation enforcement level by regulating the amount of repositioning and space utilization. During the selection of the subblock, the policy uses and updates the reservation data of the groups of containers that occupy each subblock. Finally, the storage policy selects the row location within the subblock. This step is omitted when we use a slot as a subblock. Otherwise, the storage policy selects the row based on the flatness parameter, which is a parameter with an integer value representing the maximum height gap between the highest and lowest stacks in the subblock. A flatness parameter of 1 means that the height in all slots in the subblock must be as equal as possible, and a flatness parameter of 2 or more means that the storage policy will stack containers until the height difference of the highest stack and lowest stack is equal to 2 before the policy selects the lower stack. The policy uses the inventory data of the containers within each slot to determine the selection of rows. Figure 1 illustrates the storage policy flowchart.
Fig. 1. Storage policy flowchart.
In Fig. 2(a), the integers in the blocks are the number of expected containers to be stored in a time period. The block in the first row of the second column should be selected for input containers because it has the least workload. Figure 2(b) shows that the size of the subblock is fixed as a bay, a half-bay, and a slot. Figure 2(c) shows the number of container groups stacked when the size of the subblock is a bay. Figure 2(d) shows the maximum difference between the tiers in a bay according to the flatness level.
Fig. 2. Yard configurations under various parameters.
A Simulation Study of a Storage Policy for a Container Terminal
65
4 Simulation Design We use a simulation to evaluate the storage policy of a real container terminal. The simulation ensures that the container storage and retrieval operations comply with real physical constraints. For instance, each container must be stacked on the ground or on another container, and a container can be retrieved only if there is no container on top of it. Our simulation also considers the congestion between trucks. The simulation framework consists of the simulation system, supervisory system, and database system illustrated in Fig. 3. The supervisory system represents the operating system responsible for managing the terminal’s equipment, infrastructure, and storage, i.e., rail-mounted gantry crane (RMGC), internal and external trucks, truck lanes and intersections, container slots, gates, and quay crane (QC) transfer point. The supervisory system manages the RMGC jobs, internal truck jobs, and container storage locations. The database system records the starting times, starting locations, completion times, and completion locations of RMGC and truck jobs, vessel berthing and leaving times, and tracks all jobs and inventory. We use the three systems to evaluate the container terminal’s performance and to understand how the parameters in the storage policy affect the key performance indicators.
Fig. 3. Simulation framework.
5 Experiment Design and Discussion We base our experiment on an area served by RMGCs in the Busan Port Terminal (BPT) in Korea. As shown in Fig. 4, the area consists of 4 2 blocks with two sets of 4 horizontal blocks parallel to the quay line. Each block has 34/17 bays for storing 20/40 ft containers. For simplicity, we only consider the 40 ft containers; hence, each block has 17 bays and 9 container slots (rows) for each bay. At each slot, containers can
66
H. Y. Fibrianto et al.
be stacked up to 7 tiers. We assume there are 20 internal trucks and 8 QCs, and that one RMGC serves each block. We set the equipment specification similar to the real-world container terminal.
Fig. 4. BPT simulation layout.
We consider a demand pattern with mostly transshipment containers (*75% of all incoming/outgoing containers), where 2 to 5 vessels arrive per day; the number of containers unloaded and loaded from and to each vessel ranges from 50 to 525 containers; and 2 vessels can berth simultaneously. We run the experiment for 30 days. We use the simulation to measure RMGC productivity rate, service time, and the amount of repositioning. We measure the amount of repositioning when processing both loading and import jobs. The RMGC productivity rate affects the container terminal’s overall productivity rate. We measure the average of RMGC productivity rate as the number of containers (cntr.) being stored and retrieved per yard crane per hour. The service time represents the container terminal’s service level from the perspective of the vessel and external truck. We quantify the times required to retrieve a container to the internal truck (loading), store a container from the internal truck (unloading), retrieve a container to the external truck (import), and store a container from the external truck (export). The amount of repositioning represents the inefficiency that is substantially influenced by the storage management strategy. We conduct experiments by changing one variance at a time to investigate the trade-offs between the different parameters. Table 1 summarizes their effects on the container terminal’s performance. For instance, defining a bay as a subblock reduces the need to reposition the YC, i.e., saves time and cost. A comparison of our proposed storage policy with Zhen [7] shows improved performance in the container terminal’s loading, export times, and total numbers of repositioning for loading.
Consignment strategy (modified from Zhen [7])
Proposed approach
1
1
1
1
1
2
3
1
1
Slot
Bay
Bay
Bay
Bay
Bay
Bay
Bay
Segregation
Half-bay
Partition resolution
0.7
0.3
0.5
0.5
0.5
0.5
0.5
0.5
0.5
Proximity
1
1
1
1
5
3
1
1
1
Flatness
11.37
11.40
11.41
11.41
11.39
11.40
11.41
11.40
11.41
11.39
Productivity (Ctrl./ RMGC/hour)
407.32
399.64
394.69
388.53
388.62
400.99
391.36
399.19
398.38
397.06
Loading Time (s)
251.01
261.75
260.41
254.50
256.44
265.18
261.63
262.03
262.36
253.59
Unloading Time (s)
792.70
798.07
801.66
780.33
791.06
818.23
798.65
803.32
811.95
788.31
Import Time (s)
332.76
314.14
313.16
318.50
306.83
321.33
317.72
322.01
324.32
322.98
Export Time (s)
Table 1. Workload balancing storage policy under different parameters.
49768
33932
33777
37223
35404
37658
33201
33903
36770
43569
Total repositioning for loading (cntr.)
17015
20156
20668
17428
18897
20808
20235
20498
20459
17808
Total repositioning for import (cntr.)
A Simulation Study of a Storage Policy for a Container Terminal 67
68
H. Y. Fibrianto et al.
6 Conclusion This paper described a storage policy to improve the performance of large container terminals. The storage policy defined the storage location of an incoming container by sequentially determining the block, subblock, and row location. Four parameters were used to consider the workload of each block. The segregation enforcement level regulated the number of container groups that could share a subblock. The expected proximity controlled the concentration of containers in a group within a certain location. The partition resolution defined the size of a subblock. The flatness dictated how to stack containers in a subblock. A realistic simulation consisting of a simulation system, supervisory system, and database system was used to evaluate the storage policy. Yard crane productivity rate, amount of repositioning, and service time were used as the key performance measures. The result suggested that the container terminal operates more efficiently under the storage policy with a bay as a subblock setting. Future research will develop a parameter tuning guideline for the proposed storage policy. We also will apply the proposed storage policy to different management systems (yard crane dispatching, truck pooling, etc.), measure the improvements in container terminal service, and analyze throughput across the systems. Acknowledgment. This research was supported under the framework of the International Cooperation Program managed by the National Research Foundation of Korea (Project Number: NRF-2016K1A3A1A48954044).
References 1. Lee, C.-Y., Song, D.-P.: Ocean container transport in global supply chains: overview and research opportunities. Transp. Res. Part B: Methodol. 95, 442–474 (2017) 2. Jeong, Y.-H., Kim, K.-H., Woo, Y.-J., Seo, B.-H.: A simulation study on a workload-based operation planning method in container terminals. Ind. Eng. Manag. Syst. 11, 103–113 (2012) 3. Gharehgozli, A., Zaerpour, N.: Stacking outbound barge containers in an automated deepsea terminal. Eur. J. Oper. Res. 267, 977–995 (2018) 4. Jiang, X., Lee, L.H., Chew, E.P., Han, Y., Tan, K.C.: A container yard storage strategy for improving land utilization and operation efficiency in a transshipment hub port. Eur. J. Oper. Res. 221, 64–73 (2012) 5. Petering, M.E., Wu, Y., Li, W., Goh, M., de Souza, R., Murty, K.G.: Real-time container storage location assignment at a seaport container transshipment terminal: dispersion levels, yard templates, and sensitivity analyses. Flex. Serv. Manuf. J. 29, 369–402 (2017) 6. Petering, M.E.: Real-time container storage location assignment at an RTG-based seaport container transshipment terminal: problem description, control system, simulation model, and penalty scheme experimentation. Flex. Serv. Manuf. J. 27, 351–381 (2015) 7. Zhen, L.: Container yard template planning under uncertain maritime market. Transp. Res. Part E: Logist. Transp. Rev. 69, 199–217 (2014) 8. He, J., Tan, C.J.E.O.: Modelling a resilient yard template under storage demand fluctuations in a container terminal. Eng. Optim. 51, 1547–1566 (2019)
A Simulation Study of a Storage Policy for a Container Terminal
69
9. Tan, C., He, J., Wang, Y.J.A.E.I.: Storage yard management based on flexible yard template in container terminal. Adv. Eng. Inform. 34, 101–113 (2017) 10. Jin, J.G., Lee, D.-H., Hu, H.: Tactical berth and yard template design at container transshipment terminals: a column generation based approach. Transp. Res. Part E: Logist. 73, 168–184 (2015) 11. Li, M.-K.J.E.J.o.I.E.: A method for effective yard template design in container terminals. Eur. J. Ind. Eng. 8, 1–21 (2014)
Investigation of Vessel Waiting Times Using AIS Data Janna Franzkeit(&), Hannah Pache, and Carlos Jahn Institute of Maritime Logistics, Hamburg University of Technology, Hamburg, Germany [email protected]
Abstract. The automatic identification system (AIS) enables authorities, shipping companies and researchers all over the world using ever better computer technologies to understand and track vessel movements. This publication focuses on analysing vessels’ waiting times for berth at anchoring places near ports using the example of the port of Rotterdam, Europe’s biggest port. The objective is to define clearly the concept of waiting, i.e. when a vessel waits and when not, and to investigate the amount of waiting vessels and the respective waiting times during a time span of more than two years, using solely AIS data. The indicated anchoring zones in front of the port of Rotterdam, where vessels wait, are clearly detected by visualizing the analysed data. The results of the conducted AIS data analysis show significant differences in waiting times between different vessel types, as well as a correlation between the number of waiting vessels and the average waiting time. The in detail described data pre-processing and statistical analysis are extendable and applicable to other regions and ports all over the world. Additionally, the presented data pre-processing approach is an optimal basis analysis of current waiting conditions and for applying machine learning to AIS data in order to predict future waiting times. Keywords: AIS
Maritime transportation Waiting times
1 Introduction The International Maritime Organization (IMO) published the Safety of Life at Sea (SOLAS) convention in 2002 [6], which requires AIS transmitters to be mandatory since 2004 for all passenger vessels irrespective of size, all vessels over 300 gross tonnage on international voyage, and all cargo vessels over 500 gross tonnage. The existing amount of AIS data is an example for big data, since each obligated vessel automatically sends a signal at intervals ranging from a few seconds to a few minutes depending on the navigational status [7]. Rotterdam is Europe’s biggest port with a handling volume of 289.5 million tonnes dry and liquid bulk in 2018. Liquid bulk accounts for 45.8% of the port’s total handling volume [8]. More than 29.000 vessels visit the port of Rotterdam each year. Usually tankers and bulk carriers do not operate according to fixed time tables in liner shipping and often have longer waiting times before entering the port.
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 70–78, 2020. https://doi.org/10.1007/978-3-030-44783-0_7
Investigation of Vessel Waiting Times Using AIS Data
71
The objective of this paper is the definition of waiting vessels for berth and the analysis of waiting behaviours using AIS (Automatic Identification System) data. The data and the findings are used to conduct a systematic study on waiting times spread over more than two years for Europe’s biggest port Rotterdam. The presented results can improve the planning of shipping routes in order to avoid long waiting times and point out possibilities for improvements in terminal management. Waiting times could be reduced by informing vessels on their route about their time slots at port, enabling the vessels to adjust the sailing speed with foresight. If vessels are able to reduce their sailing speed in a timely manner, fuel and emissions could be saved, as the relationship between speed and fuel consumed is nonlinear (see e.g. [1, 2, 13]). Additionally, the punctuality of vessels is strongly dependent on the prevailing weather conditions en route. Delays of individual large vessels can sometimes affect the punctuality of the entire terminal operation. The paper is structured as follows: Sect. 2 gives a short overview on the existing literature. In Sect. 3 the data used and the data pre-processing are described. The computational results are illustrated in Sect. 4 and the paper concludes with a discussion of the results in Sect. 5.
2 Previous Work AIS data open up opportunities for researchers, vessel owners and authorities that go beyond the safety of vessels. Thus the number of publications including AIS data is rising. For example, Svanberg et al. [12] have identified ten research fields in which the use of AIS data has become indispensable. A brief literature review of publications relevant for this paper is given in the following. In literature the definition of a waiting vessel with explicit values, e.g. for speed or position, is rarely given. The few definitions found, presented in Table 1, differ greatly from each other in terms of speed and vessel position. Although the vessels are defined as waiting, anchoring, or non-sailing, the speed is greater than zero in all definitions found. Causes are currents and winds that move the vessel, as well as the AIS signal itself. Table 1. Multiple definitions of the term waiting vessel. Authors Vessel status Speed Coomber et al. [3] Moored/anchoring 0.90, Tucker-Lewis index (TLI) is 0.913 > 0.90, Adjusted Goodness-of-fit Index (AGFI) is 0.974 > 0.95, and Standardized Root Mean Square Residual (SRMR) is 0.057 < 0.08. The rule of thumb values are taken from literature [4, 8, 14, 19–22].
4 Results and Discussion The SEM results from the previous section show that local industrial development has a positive and significant relationship with logistics and supply chain management, whereas employment and labor market conditions have a negative and significant impact on the local industry. Interestingly, although labor market conditions have a significant impact on the local industry, the results show that the direction of the relation is negative, therefore, the common fear among the local population that the Chinese workforce will take over the market is reflected in the results. Foreign investment from other countries depicted a negative relationship with local industrial development but it is insignificant. People think that other countries may not be that interested in investing in Pakistan. Once there is a boost in the productivity of the local industry due to CPEC, it will promote innovative behavior and technological up-gradation. It could be due to collaboration with already innovative Chinese industries via knowledge transfer. Moreover, the inflow of foreign investment will increase when there is an increase in the volume of international trade; the relationship is positive and significant. But the improvement in logistics & supply chain management due to infrastructural projects would not be a significant factor in attracting foreign investment from other countries. Other internal conditions, such as poor governance, corruption, and law and order could be the important factors for those countries. When the local industry will start working on technological up-gradation and there will be inflow of foreign investment, it will generate employment opportunities for locals because by this time, not only the CPEC related industries but other sectors and industries will also be active in contributing towards economic activity in the country. And people will feel more optimistic about these employment opportunities. Lastly, it is unwise to think of technological up-gradation without incorporating sustainability into it. So, it will have a positive and significant role in promoting sustainable business opportunities in Pakistan. It can be seen that the development of the local industry depends positively on the conditions of logistics and supply chain and international trade but it has a negative relationship with employment and foreign investment. Moreover, foreign investment is insignificant. That is probably justified because people have concerns that although CPEC will improve the local industry however the Chinese workers will take over the labor market. In turn, the local industry will have a positive and significant impact on technological up-gradation which in turn will positively impact sustainability. International trade has a positive influence on foreign investment but the conditions of logistics and supply chain are insignificant to attract any foreign investment. Employment conditions in the labor market are influenced positively and significantly
184
A. Khan et al.
by innovations and foreign investment. Regression results are also presented in Table 2. However, this paper presents some limitations such as limited available literature on the topic and only workers from the Pakistani industries were included during the survey, in future Chinese industrial employees could also be included in this kind of study.
5 Conclusion CPEC as a flagship project of the Belt and Road initiative of China is considered very crucial for regional logistics and supply chain management. But the main concern of this study is to examine its influence on the domestic industry in Pakistan. For this purpose, SEM is used, results show that developments in the logistics sector, labor market, and international trade have a significant role to play in the expansion of the local industry. Pakistan’s government needs to ensure the labor security of locals and execute planned projects in such a way that the local community is involved in it. The process of development in the local industry will affect the innovation and technological status of Pakistan which will result in adopting more sustainable business practices. The results of this study can help the government of Pakistan in designing future policies pertaining to CPEC. It needs to work on governance issues such as corruption and red tape in order to attract foreign investment from other countries which will boost industrial capacity outside the CPEC-related industries and it will create employment opportunities for the locals. The government needs to take locals as well as foreign investors into confidence by disclosing the terms and conditions of the official agreement of CPEC. Moreover, from a future perspective, it is crucial to adopt sustainable innovation techniques. In addition to that, it can also help the managers and industrialists in planning their future business strategies in the wake of CPEC.
References 1. Ashfaq, M., Abid, A.: CPEC: challenges and opportunities for Pakistan. Pakistan Vis. 16(2), 142–169 (2016) 2. Baumgartner, H., Homburg, C.: Applications of structural equation modeling in marketing and consumer research: A review. Int. J. Res. Mark. 13, 139–161 (1996). https://doi.org/10. 1016/0167-8116(95)00038-0 3. Calamur, K.: High traffic, high risk in the strait of Malacca. In: The Atlantic (2017) 4. Civelek, M.E.: Essentials of Structural Equation Modeling. Zea Books, London (2018) 5. Doll, W.J., Xia, W., Torkzadeh, G.: A confirmatory factor analysis of the EUCS instrument. MIS Q. 18, 453–461 (2011) 6. Bete Georgise, F., Klaus-Dieter, T., Hans-Dietrich, H.: Manufacturing Industry Supply Chain Modeling and Improvement in Developing Countries. Universität Bremen (2015) 7. Freitag, M., Bauer, P.C.: Testing for measurement equivalence in surveys: dimensions of social trust across cultural contexts. Public Opin. Q. 77, 24–44 (2013). https://doi.org/10. 1093/poq/nfs064
Influence of Supply Chain Management & Logistics
185
8. Frone, M.R., Tidwell, M.C.O.: The meaning and measurement of work fatigue: development and evaluation of the three-dimensional work fatigue inventory (3d-wfi). J. Occup. Health Psychol. 20, 273–288 (2015). https://doi.org/10.1037/a0038700 9. Govindan, K., Azevedo, S.G., Carvalho, H., Cruz-Machado, V.: Impact of supply chain management practices on sustainability. J. Clean. Prod. 85, 212–225 (2014). https://doi.org/ 10.1016/j.jclepro.2014.05.068 10. Gul, A.: China Welcomes Saudi Plans to Invest in CPEC Project with Pakistan. VoA (2018) 11. Idrees, M., Salman, A., Makarevic, N.: Copyright © 2017 by Sochi state university published in the Russian federation Sochi journal of economy has been issued since 2007. ISSN: 2541–8114. Sochi. J. Econ. 11, 107–115 (2017) 12. Imran, M., ul Hameed, W., ul Haque, A.: Influence of industry 4.0 on the production and service sectors in Pakistan: evidence from textile and logistics industries. Soc. Sci. 7(12), 46 (2018) https://doi.org/10.3390/socsci7120246 13. Masood, Y.: China Economic Corridor Is Debt Reliever for Pakistan. Telegr (2019) 14. Matsunaga, M.: How to factor-analyze your data right: do’s, don’ts, and how-to’s. Int J. Psychol. Res. 3, 97–110 (2010) 15. Mishra, A.A., Shah, R.: In union lies strength: collaborative competence in new product development and its performance effects. J. Oper. Manag. 27, 324–338 (2009). https://doi. org/10.1016/j.jom.2008.10.001 16. Naseem, I., Khan, J.: Impact of Energy Crisis on Economic Growth of Pakistan. Int. j. Afr. Asian stud. 7 (2015) 17. Oyewole, P.: Multiattribute dimensions of service quality in the all-you-can-eat buffet restaurant industry. J. Hosp. Mark. Manag. 22, 1–24 (2013). https://doi.org/10.1080/ 19368623.2011.638418 18. Handfield, R., Straube, F., Pfohl, H.-C., Wieland, A.: Trends and Strategies in Logistics and Supply Chain Management: Embracing Global Logistics Complexity to Drive Market Advantage. DDV Media Group, Bremen (2013) 19. Sahin, M., Todiras, A., Nijkamp, P., Neuts, B., Behrens, C.: A structural equations model for assessing the economic performance of high-tech ethnic entrepreneurs. TI 2013-161/VIII Tinbergen Institute Discussion Paper 20. Sahin, M., Todiras, A., Nijkamp, P., Neuts, B., Behrens, C.: A structural equations model for assessing the economic performance of high-tech ethnic entrepreneurs (2013) 21. Schreiber, J.B.: Reporting structural equation modeling and confirmatory factor analysis. J. Educ. Res. 99, 323–338 (2006) 22. Shahdani, S.M., Khoshkhooy, M.: Structural equation modeling of effective economic and cultural components on energy consumption behavior in urban societies. J. Urban Econ. Manag. 5, 95–115 (2019) 23. Su, Q., Li, Z., Zhang, S.X., Liu, Y.Y., Dang, J.X.: The impacts of quality management practices on business performance: an empirical investigation from China. Int. J. Qual. Reliab. Manag. 25, 809–823 (2008). https://doi.org/10.1108/02656710810898621 24. Ximénez, C.: Recovery of weak factor loadings when adding the mean structure in confirmatory factor analysis: a simulation study. Front. Psychol. (2016). https://doi.org/10. 3389/fpsyg.2015.01943 25. The Global Innovation Index: Creating Healthy Lives—The Future of Medical Innovation. Switzerland, Geneva (2019)
A Disruption Management Model for a Production-Inventory System Considering Green Logistics Hawa Hishamuddin(&), Mohd Azizi Abd Aziz, Noraida Azura Md Darom, Mohd Nizam Ab Rahman, and Dzuraidah Abd Wahab Department of Mechanical and Manufacturing Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Bangi, Malaysia [email protected]
Abstract. Disruptions not only cause negative effects on the supply chain system, but also potential damage on the environment if not managed strategically. Hence, it is crucial for companies to optimise resources efficiently during disruptions to mitigate economic losses and reduce the adverse effects to the environment. This research proposes a recovery model for supply disruptions at the production stage, where integration of environmental consideration in the inventory model is made by including the cost of carbon emissions from logistics activities during the recovery period. LINGO software is used to solve the mathematical model, of which numerical analysis was performed to determine the optimal lot sizing decisions during recovery. The results of the study indicate that the cost of the carbon footprint was closely related to the capacity and distance of the carrier. Additionally, by using large-capacity trucks to reduce delivery frequency of the goods, the advantage of reducing environmental pollution can be achieved. The contribution of this research is a decision support tool for supply chain disruption recovery, while incorporating environmental aspects in the decision-making process. Keywords: Supply chain Logistics
Disruption Recovery Carbon emission
1 Introduction A supply chain (SC) is a complex and dynamic system that is linked to demand, and consists of organizations, mankind, activities, information and sources involved in the movement of products or services. Unexpected and unavoidable events such as machine breakdowns, transportation disruptions, supply disruptions, workers’ strikes and natural disasters might occur that impacts the smooth flow of goods in the system (Hishamuddin et al. 2012). Proper organization of the supply chain is an important step towards ensuring an efficient flow of output and minimizing potential losses due to unexpected disruptions. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 186–195, 2020. https://doi.org/10.1007/978-3-030-44783-0_18
A Disruption Management Model for a Production-Inventory System
187
Sustainable supply chain management (SSCM) is a concept of sustainability that has been introduced into supply chain management. SSCM consists of three components, namely the economy, environment, and society (Elkington 1997; Henriques and Richardson 2004), and is categorized as a combination of the environmental and social components as well as an enhanced dimension and scope of the economy component in a Triple Bottom Line model. In an ideal SSCM, the flow of goods and management orientation from the supplier to the retailer proceeds smoothly, but in reality, the presence of disruptions may result in the opposite to occur. Therefore, nowadays, researchers have focused on studying the integration of sustainable supply chain and disruption management in the development of more resilient systems. The supply chain disruption management (SCDM) field has evolved in the modelling context, from a one-phase system into two phases and beyond (Xia et al. 2004; Hishamuddin et al. 2013). This model is meant to calculate the optimum production and ordering quantities within the recovery period with the overall minimum cost. Xia et al. (2004) examined the window for the optimal disruption recovery time in two stages of production and inventory control changes based on the setup cost, holding cost, production rate and production. This research was explored further by Hishamuddin et al. (2012), who introduced a new real-time recovery model of a production-inventory system subject to disruptions for a single-phase system. Next, Hishamuddin et al. (2013) extended their research to focus on a two-phase model and identified the optimal recovery schedule, which minimized the overall cost of recovery, especially for transportation disruptions. Apart from that, Schmitt et al. (2015) introduced a supply chain simulation model to illustrate the actual situation, where the impact of disruption depends on its location, and higher costs and long-term impacts on the consumption occur as a result of the disruption. Furthermore, Dutta et al. (2016) developed a model for a multi-period CLSC under capacity and demand uncertainty that can be recovered by employing a buy-back offer at the retailer level. Several recovery models have been developed and used practically to reduce damage from disruptions. From the perspective of previous studies, the introduction of environmental impacts into the supply chain decision-making process is closely linked to the management of sustainable supply chains. Based on a review of sustainable supply chain management by Eskandarpour et al. (2015), there are three factors that need to be considered for a sustainable supply chain environment, namely facilities, transport and product design. In giving more emphasis to this, Fahimnia and Jabbarzadeh (2016) claimed that although a trade-off analysis of a designed resilient sustainable supply chain is unable to satisfy product demand when there is a supply chain disruption, it can fulfill the overall market demand at a slight increase in the supply chain cost by adjusting the production and sourcing strategies. Furthermore, the model built by Paksoy (2010) was aimed at optimizing the supply chain network through a proposed quota for the release of carbon dioxide emissions through transportation in the manufacturing process. This was supported by a study conducted by Abdallah et al. (2012), which explained the negative effects of the carbon footprint and ways to mitigate these negative effects. The environmental impacts in the inventory model have also been analysed in the previous literature. In a study conducted by Absi et al. (2013), the costs and constraints of carbon output were also studied in terms of the economy, unity and the state, where
188
H. Hishamuddin et al.
the rate of a fixed carbon output depended on the transport or storage. Next, this study was extended by Absi et al. (2016), where the constraint of using a fixed carbon footprint was analysed along with the actual case. Apart from that, Tseng and Hung (2014) proposed a decision-making model based on a sustainable supply chain design for garment manufacturing by linking the social costs. A number of studies have begun to integrate environmental consideration and cost in SC models. Hammami et al. (2015) proposed a multi-echelon production-inventory model that includes the carbon emissions with lead-time constraints. Next, Gong et al. (2019) proposed a sustainable investment decision model with the use of dynamic programming to optimize cost and minimize emissions in SC processes. Furthermore, Marchi et al. (2019) introduced a SC model for a vendor-buyer green supply chain model with the consideration of environmental and quality issues in production and reworking processes. In addition, Darom et al. (2018) incorporated supply disruption management and environmental consideration in the SC model with mitigation and recovery strategies. As managers nowadays are concerned with both their economic and environmental goals, more studies are still needed in this research area. The objective of this study is aimed at developing a mathematical model to investigate the optimal inventory decisions when a production-inventory system experiences disruptions to the supply chain and to analyse the total carbon emissions from different shipment decisions. The mathematical model was solved using LINGO software, where the numerical parameters and optimal decision variables were identified. The next section will discuss on the model development of the intended system.
2 Model Development 2.1
Model Formulation
The system under consideration is a modified Economic Production Quantity (EPQ) problem. The EPQ is a system for determining the optimal lot size so that manufacturers can produce high profits or low costs (Huang et al. 2016). The model in this research assumed that a disruption occurred in the system and prevented production from proceeding as normal. After the disruption, a recovery time was allocated to allow the recovery schedule to begin, where changes to the production are allowed during this phase. After the recovery window, the system fully recovers and is brought back to the original schedule. The inventory curve consists of two curves, namely a normal curve, where the supply chain works under ideal conditions, and a modified curve, which was formed due to rescheduling after a disruption occurred. Changes to the original curve are allowed in the recovery window, whereby the original schedule is restored at the end of the window. The inventory curve is shown in Fig. 1. The dotted line represents the normal cycle and the disruption is considered to occur on the earlier cycle. Based on the diagram, the recovery window, t0 to t1, indicates that three recovery cycles took place immediately after the disruption occurred.
A Disruption Management Model for a Production-Inventory System
189
Fig. 1. Production-inventory curve
In this inventory curve, the inventory time display was termed Ti and the normal production cycle during the occurrence of the disruption was also considered as a nonfixed decision. The cycle length, n, was equal to the recovery schedule as well as the normal schedule that was considered in this study. After the disruption, production could only be continued when the disruption had been rectified. Partially unfulfilled demand became lost sales, while delayed orders during the recovery period were back ordered. The parameters and notations used in developing the model were as follows: A setup cost for a cycle ($/setup) D demand for each product (unit/year) H annual inventory holding cost ($/unit/year) P production quantity (unit/year) Q lot size in normal schedule (unit) Td disruption time u production downtime for normal cycle (setup time + delay time) (T − Q/P) t0 recovery window starts t1 recovery window ends T production cycle time for normal cycle (Q/D) P production up time for normal cycle (Q/P) B backorder cost per unit time ($/unit/time) L lost sales cost ($/unit) n cycle in recovery window Si shipments for cycle, i, in recovery window (unit) Xi production quantity for cycle, i, in recovery window (unit) Ti production end time for cycle, i, in recovery window (Vi/P) st setup time for cycle d delay time for cycle r safety stock quantity (unit) f penalty function for late recovery in normal schedule Di transportation distance travel SC social cost for carbon emission CO2 emission of carbon from transportation (g/km) qt truck capacity TC transportation unit cost for each shipment ($/shipment)
190
H. Hishamuddin et al.
The model that was developed was a non-linear constrained programming problem. This model was designed to determine the optimum recovery schedule so that the overall cost was minimised and would be able to meet the customer demand based on the constraints created in the system. Moreover, this model applies environmental impact by incorporating carbon effects from transport. The following is a list of the formula used in the improved model. Since this model was related to the production facilities, seven types of costs were associated. The costs involved were the setup cost (TC1), inventory holding cost (TC2), backorder cost (TC3), lost sales cost (TC4), penalty cost (TC5), transportation cost (TC6) and social carbon cost (TC7). The details of these costs are elaborated in the next sub-section. 2.1.1 Objective Function The objective function is the total costs for this model as shown in Eqs. (1) and (2). This overall cost is used to reflect the minimum cost of the recovery period based on the inventory curve in Fig. 1. Based on Eq. (1), TC1 is the setup cost where A is expected cost for each setup, n is the number of cycles in the recovery phase. TC2 is the holding inventory costs that is associated with inventories, where H is the expected cost of holding an amount of inventory per year during the recovery. Next, TC3 is the back order cost based on ordering activities which cannot be fulfilled within the time provided but will be fulfilled later. Whereas, the cost of losing a sale, TC4 is the loss of the profit due to unsuccessful orders. TC5 is the cost penalty cost based on the inventory schedule. The cost of transport, TC6 is the cost of the number of shipments to be done when the production process is completed. The social carbon cost, TC7 is based on multiplying associated values such as recovery cycle numbers with travel distance, truck capacity and the total emission of CO2 with social carbon costs. Finally, the overall total cost, as shown in Eq. (2), is constructed according to the costs mentioned above, respectively. ð1Þ MinTC ðXi ; Si ; nÞ ¼ TC1 þ TC2 þ TC3 þ TC4 þ TC5 þ TC6 þ TC7 1 Xn 2 Xn1 1 ð i Xi þ X2 þ st MinTC ðXi ; Si ; nÞ ¼ A n þ H ðX1 S1 Þ i¼1 2P 2P " # ! n n n X X X Xi2 9Q 6u þ L nQ þ B Xi ðTd þ Ist Þ þ Xi P P i1 i¼1 i¼1 n X Si Di þ f n2 þ ct þ ðn Di CO2 qt SC Þ qt i¼1
ð2Þ
A Disruption Management Model for a Production-Inventory System
191
All the equation above is subjected to the following constraints: n X
Xi ¼
i¼1
n X
n X
ð3Þ
Si
i¼1
Xn ¼ Sn
ð4Þ
Sn ¼ Q
ð5Þ
Si qt
ð6Þ
Xi ¼ P ðnT nSt Td Þ
ð7Þ
i¼1 n X
Xi ¼ nTD
i¼1 n X
nQ
n X
! Xi
ð8Þ
i¼1
Xi iQ þ ði þ 1ÞPu PTd PiSt
ð9Þ
i¼1
The above model can be categorized as a constrained integer nonlinear programming model. Equation (3) ensures that the total production quantity is the same as total shipment and Eqs. (4) and (5) guarantees recovery of the original schedule after n cycles. Equation (6) resembles the constraint for truck capacity and ensures all shipment does not exceed truck capacity. Equation (7) represents the production capacity constraint; whereas Eq. (8) ensures that all demand during the recovery period is accounted for. Lastly, Eq. (9) confirms that all the backorders are non-negative. By solving the above model (2) for Xi, Si, and n subject to the constraints (3)–(9), one can obtain the optimal recovery plan for the two stage supply chain system under disruption. The mathematical model was solved using LINGO software and computational analysis that was performed will be presented in the next section.
3 Results and Discussion 3.1
Numerical Analysis
Table 1. Parameter for test problems Test A H B D
U1 20 1.2 10 400 000
U2 20 1.2 10 400 000
U3 30 1.2 10 400 000
U4 20 2.4 10 400 000 (continued)
192
H. Hishamuddin et al. Table 1. (continued) Test L P Td st qt SC Di TC CO
U1 15 500 000 0.008 0.000057 5000 0.000044 10 2000 150
U2 1 500 000 0.008 0.000057 5000 0.000044 10 2000 150
U3 15 500 000 0.008 0.000057 5000 0.000044 10 2000 150
U4 15 500 000 0.008 0.000057 5000 0.000044 10 2000 15
A numerical analysis was performed to demonstrate the applicability of the model. Table 1 shows the test problems along with the parameters that were relevant to this study. Test 1 was used as the benchmark for this study, and it was repeated with some changes in the parameters. Based on Fig. 2, the TC was high on the first recovery cycle for all the test problems, except for U2. This was because the recovery cycle was forced to recover from the disruption after a cycle and required a higher cost. However, the TC for U2 at the recovery cycle had a low lost sales cost, resulting in the recovery with one cycle only. Additionally, the shape of the curve was convex as the cost of losing sales increased due to the disruption.
Fig. 2. TC vs n
Figure 3 shows the effect of disruption time, Td on total recovery costs (TC) and recovery duration (n), where the total cost, is seen to increase when the disruption time, Td increased. However, the optimal number of recovery cycles, n, increased gradually by one cycle as Td increased. This showed that the longer the Td value, the higher the total recovery cost and recovery duration.
A Disruption Management Model for a Production-Inventory System
193
n
Fig. 3. TC and n vs Td
Figure 4 shows the sensitivity analysis performed to show the effect of truck capacity, qt, on the total recovery costs. The figure also plots the behavior of transportation cost (TC6) and carbon emission costs (TC7) with the change in qt. The total cost and truck capacity of this model shows a convex function. It can be shown that the optimum value of TC occurs when TC6 and TC7 intersect. A possible explanation for this is due to the relationship between truck capacity and carbon emissions with disruption recovery. TC is seen to decrease initially when the capacity of the truck, qt is increased, as trucks that ship with larger capacities have a low shipment frequency compared to trucks with less capacity, therefore recovery costs are reduced. However, a higher truck capacity will emit more carbon dioxide, hence, further increase in truck capacity will eventually increase the total recovery costs. Hence, a tradeoff will be achieved at a certain value of truck capacity that is optimal in both recovery costs and carbon emissions.
Fig. 4. Effect of truck capacity on TC, TC6 and TC7
194
H. Hishamuddin et al.
4 Conclusion Supply chain operations that are affected by disruption and are associated with the environment, such as carbon emissions from logistics, could create an unhealthy environment for current and future generations. Therefore, studies on carbon footprint need to be considered more in supply chain disruption management to ensure a comprehensive approach. Hence, this study explores the integration of environmental aspects into the supply chain disruption recovery model, where a mathematical model was developed to illustrate the problem by considering social carbon costs in a single stage production-inventory system. Other than identifying the optimal recovery cycles, the model is also capable of identifying the appropriate revised production schedule after the occurrence of a disruption. As transportation disruption was considered, the transportation costs and social carbon costs were incorporated to study the release of carbon emissions from related logistics activities. Based on the results of the study, the cost of the carbon emissions was closely related to the distance of the carrier. Hence, the management of the company should take a proactive measure by using large-capacity trucks, especially during long disruptions so that transportation costs and the environmental impact of the disruption can be reduced. This is due to the decrease in the number of shipments when using larger capacity trucks. However, when the size of the shipment lots can already be fulfilled by small-capacity trucks, such trucks should be chosen as they can reduce the carbon emissions and other related costs. The management can implement environmental conservation into the supply chain by taking the appropriate measures to minimize costs. In addition, carbon emissions from transportation activities have the highest environmental impacts. The practices and effects of carbon emissions from these transportation activities are very closely linked, such as facility location selection and supplier’s selection that determines the distance of transmission routes for which this distance is able to determine the carbon emission rate such as the further the transmission distance the higher the emission rate of the carbon. This research can be extended by incorporating certain aspects. An interesting future research would be an extension of the model with multiple stages. Another extension would be to apply the model for different manufacturers by considering the social aspects. Acknowledgment. The authors would like to thank the Ministry of Higher Education for funding this research under the Fundamental Research Grant Scheme FRGS/1/2017/TK03/ UKM/02/3 and Research University Grant GUP-2018-100.
References Abdallah, T., Farhat, A., Diabat, A., Kennedy, S.: Green supply chains with carbon trading and environmental sourcing: formulation and life cycle assessment. Appl. Math. Model. 36(9), 4271–4285 (2012)
A Disruption Management Model for a Production-Inventory System
195
Absi, N., Dauzère-Pérès, S., Kedad-Sidhoum, S., Penz, B., Rapine, C.: Lot sizing with carbon emission constraints. Eur. J. Oper. Res. 227(1), 55–61 (2013) Absi, N., Dauzère-Pérès, S., Kedad-Sidhoum, S., Penz, B., Rapine, C.: The single-item green lotsizing problem with fixed carbon emissions. Eur. J. Oper. Res. 248(3), 849–855 (2016) Darom, N.A., Hishamuddin, H., Ramli, R., Mat Nopiah, Z.: An inventory model of supply chain disruption recovery with safety stock and carbon emission consideration. J. Clean. Prod. 197, 1011–1021 (2018) Dutta, P., Das, D., Schultmann, F., Frohling, M.: Design and planning of a closed-loop supply chain with three-way recovery and buy-back offer. J. Clean. Prod. 135, 604–619 (2016) Elkington, J.: Cannibals With Forks. The Triple Bottom Line of 21st Century, pp. 1–16, April 1997 Eskandarpour, M., Dejax, P., Miemczyk, J., Peton, O.: Sustainable supply chain network design: an optimization-oriented review. Omega (U.K.) 54, 11–32 (2015) Fahimnia, B., Jabbarzadeh, A.: Marrying supply chain sustainability and resilience: a match made in heaven. Transp. Res. Part E: Logist. Transp. Rev. 91, 306–324 (2016) Gong, D.C., Kao, C.W., Peters, B.A.: Sustainability investments and production planning decisions based on environmental management. J. Clean. Prod. 225, 196–208 (2019) Hammami, R., Nouira, I., Frein, Y.: Carbon emissions in a multi-echelon production-inventory model with lead time constraints. Int. J. Prod. Econ. 164, 292–307 (2015) Henriques, A., Richardson, J.: The Triple Bottom Line: Does it All Add Up? Routledge (2013) Hishamuddin, H., Sarker, R.A., Essam, D.: A disruption recovery model for a single stage production-inventory system. Eur. J. Oper. Res. 222(3), 464–473 (2012) Hishamuddin, H., Sarker, R.A., Essam, D.: A recovery model for a two-echelon serial supply chain with consideration of transportation disruption. Comput. Ind. Eng. 64(2), 552–561 (2013) Huang, H., He, Y., Li, D.: EPQ for an unreliable production system with endogenous reliability and product deterioration. Int. Trans. Oper. Res. 24(4), 839–866 (2017) Marchi, B., Zanoni, S., Zavanella, L.E., Jaber, M.Y.: Supply chain models with greenhouse gases emissions, energy usage, imperfect process under different coordination decisions. Int. J. Prod. Econ. 211, 145–153 (2019) Paksoy, T.: Optimizing a supply chain network with emission trading factor. Sci. Res. Essays 5(17), 2535–2546 (2010) Schmitt, T.G., Kumar, S., Stecke, K.E., Glover, F.W., Ehlen, M.A.: Mitigating disruptions in a multi-echelon supply chain using adaptive ordering. Omega 68, 185–198 (2015) Tseng, S.-C., Hung, S.-W.: A strategic decision-making model considering the social costs of carbon dioxide emissions for sustainable supply chain management. J. Environ. Manag. 133, 315–322 (2014) Xia, Y., Yang, M.H., Golany, B., Gilbert, S.M., Yu, G.: Real-time disruption management in a two-stage production and inventory system. IIE Trans. (Inst. Ind. Eng.) 36(2), 111–125 (2004)
A Concept for a Consumer-Centered Sustainable Last Mile Logistics Michael Freitag1,2 and Herbert Kotzab1,3(&) 1
2
University of Bremen, Bremen, Germany [email protected] BIBA – Bremer Institut für Produktion und Logistik GmbH, Bremen, Germany 3 Universiti Utara Malaysia, Sintok, Malaysia
Abstract. The amount of home deliveries of online purchased consumer goods has significantly increased in the past years. The logistics of home delivered goods is however negatively impacting the ecological environment as more packing and additional delivery tours arise. However, the environmental attitude of consumers is at the same time increasing too, which leads to a kind of dilemma of how to combine environmental consciousness with the convenience of online shopping? This paper proposes a methodological approach for designing a consumer-centered sustainable last mile logistics system and its evaluation. The suggested method combines discrete choice experiments with computer simulations which allows a feedback loop in the sustainable design of deliveries to consumers’ homes. Keywords: Sustainability experiment Simulation
Human centered logistics Discrete choice
1 Introduction The share of interactive commerce in Germany is steadily increasing and achieves up to date far a market share of more than 10% of total retail sales (Statista). The sales volumes grew from 1.1 billion € in 1999 to more than 50 billion € in 2018, whereby the main hare of online sold product categories refer to cloths, shoes, tickets for events or books. While these categories show order rates of at least 35%, grocery items are so far not that often ordered online, as the order rates are below 8%. However, the importance of electronic commerce in the area of fast moving consumer goods is increasing and the deliveries of such products (also known as last mile) is going to challenge retailers as well as end users. While retailers try to find a trade-off between convenience and costs, more and more end users are expecting basically an immediate delivery of their online ordered goods, often at the same day. This trend leads to more and more transport processes while simultaneously decreasing the utilization of the transport means (see Schnedlitz et al. 2013). Besides that, more transport results also in more CO2-emissions and last-mile-logistics requires additional packaging due to individualized preferences for picking and packing. Taking the current climate change discussions into account as well as environmental pollution © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 196–203, 2020. https://doi.org/10.1007/978-3-030-44783-0_19
A Concept for a Consumer-Centered Sustainable Last Mile Logistics
197
due to plastic waste, we observe however a changing pattern in consumer behavior as more and more consumers tend to buy regional products and tend to reduce product packaging while shopping. Consequently, consumer face the dilemma of using the advantages and conveniences offered by mobile/electronic commerce and protecting the environment by making a personal contribution for the minimization of CO2-emissions and plastic waste. However, it is nearly impossible for end users to recognize directly the effects of changing consumer behavior on the environment or to even influence it. Normally, end users can opt a lot when it comes to product choices, but nearly nothing when it comes to the selection of sustainable home delivery options. They are mainly able to determine the payment choices as well as the delivery time in terms of standard or express deliveries. When ordering online, retailers are not showing the consequences of increased CO2-emissions due to express deliveries or the costs for additional packaging. Up to date, online consumers do have no transparency when it comes to the visibility of the transport modes that are used for delivering their products. This is very surprising as a current study by PcW (2018) shows that 3 out of 4 grocery shoppers advocate less packaging waste and would prefer returnable packaging. Taking this into account, it is the goal of this paper to present the conceptual ideas of a method that is able to design and to evaluate tailored sustainable consumer logistics options as well as their development potential. The use of this method shall allow consumers to determine their personal preferences in regards to emissions, plastic packaging, delivery times and dates for online ordered groceries in order to get a sustainable tailored home delivery.
2 Current Stage of Consumer-Centered Sustainable Logistics Our understanding of a consumer-centered sustainable last mile logistics concept is embedded within the areas of green/sustainable logistics, consumer/city-logistics as well as food logistics/packaging. In the following we present a current state of the art of the research within these domains. 2.1
Green/Sustainable City Logistics
According to McKinnon et al. (2010), green logistics deals with the following areas: reduction of the externalities through transports, city logistics, consideration of green logistics in business strategy, reverse logistics as well as green supply chain management (SCM). Later, McKinnon et al. (2015) reduce their model of green logistics to the reduction of transport emissions. When it comes the reciprocal effects between innovative logistics concepts and customer requirements, Melkonyan et al. (2017) identify success factors for sustainable logistics systems by the application of causal loop diagrams. Their findings show that transparency in supply chain design and a distinct measurability of sustainability positively affects the conscious choice of sustainable last-mile delivery concepts.
198
M. Freitag and H. Kotzab
While Clausen et al. (2016) look at specific organizational as well as technical measures for realizing green last mile logistics, Dobers et al. (2013) present resourceefficient logistics services and Athanassopoulos et al. (2016) examine the environmental effects of parcel distribution. Clausen et al. (2016) examine also the decoupling of transport for the last mile by the use of consolidation centers as well as the effects of using sustainable transport means such as E-bikes for inner-city deliveries. Hereby, they consider besides economic factors also the impact on the environment by CO2emissions as well as the attractiveness of the concepts for suppliers and recipients. Leyerer et al. (2018) suggest the use of parcel depots for last mile distribution where mathematical optimization suggests the locations as well as the routes between a depot and an end user. Already in the late 1990s, Berg (1999) suggests the so-called Munich model that refers to strategies for traffic reduction, traffic shifts as well as a more environmentalcompatible traffic design. Lohre et al. (2011) suggest hereby the Utrecht cargo-hopper as a positive example for traffic reduction by cooperation. Auffermann (2017) as well as Schönberg and Auffermann (2017) study sustainable urban supplies and the traffic development and Raiber et al. (2014) examined the consequences of electrical drive technology for heavy weight loads and freight goods transport in cities. Bode (2016) presents a city logistics concept that includes an inner-city distribution hub out of which trams are delivering the goods. This has also been already suggested by Lütjen and Piotrowski (2012). Recently, Elbert and Friedrich (2018) present an agent-based simulation which examines the effects of cooperation between logistics service providers for the inner-city freight goods transport. Other approaches for an improved utilization of resources and environmental compatibility within the frame of reference of green logistics and city logistics is presented by Deckert (2016). 2.2
Food Logistics from the Consumer’s Perspective
Nitsche et al. (2016) present future trends in food logistics and show that more and more consumers request larger transparency along the supply chain and increasingly demand sustainable products and sustainable delivery options. Kille et al. (2015) presents the need for alternative sustainable delivery possibilities for online-grocery retailers as Bloemhof et al. (2015) also show the significant impact of food and transport packaging on the sustainability of food logistics. Buchner (2012) focuses on the challenges of food packaging and develops solutions, which include economic, environmental as well as qualitative aspects. Singh et al. (2017) also research in the area of food packaging technology with a focus on plastic and active packaging possibilities. Thereby Trapp et al. (2017) were able to identify – based on environmental accounting/life cycle analysis – that Styrofoam packaging for frozen goods delivered by parcel distributors have the same level of CO2 emissions as a private shopping tour with a car. Both alternatives however emit less CO2 than a consolidated delivery with small deep-freeze vehicles. Meyer and Kotzab (2017) present a current state of consumer logistics research from a distribution channel point of view. Overall, consumer logistics focuses on the analysis of logistics processes from an end-user perspective starting with planning of a shopping trip to the instore processes executed by the consumer and ends with the
A Concept for a Consumer-Centered Sustainable Last Mile Logistics
199
delivery of the products by the end-users to their homes (e.g. Bahn et al. 2015; Meyer et al. 2016, 2017; Teller et al. 2012). Thereby, Galiopoglu et al. (2015) show the consequences for such processes by the use of smart phones which allows nowadays the coupling of individualized shopping behavior with geographical information. These possibilities are summarized under the construct of social, local and mobile (SO-LOMO) commerce. When it comes to individualization, Hüseynioglu et al. (2017) show that this is so far only applied in the acquisition potential of distribution (see Schögel 2012) and when it comes to logistics it is limited to delivery options and delivery windows. 2.3
Critical Reflection on Current State of Research
Overall, we were not able to identify any detailed observation and analysis of individualization possibilities for a sustainable last mile delivery option. Research as well as empirical evidence shows an increased interest as well as demand of consumers for sustainable solutions, however the presently offered possibilities are limited.
3 Method Development 3.1
Objective and Boundary Conditions
The goal of this paper is to present a suggestion for an interdisciplinary tool that allows a tailored design and evaluation of a sustainable tailored consumer logistics system. For this purpose, we define consumer logistics as the logistics of the last mile (see Meyer and Kotzab 2017). This contains online orders, commissioning, packaging and loading of merchandise at the retail store or in a fulfillment center as well as the delivery of the goods by a logistics service provider or the pick-up by the consumer in the store. Another additional option can be seen in a pick-up locker outside a retail store where consumers can get their products also outside the opening hours. We focus on the product category of food as this product category is regularly demanded as well as this category requires additional requirements on packaging and transport. Sustainability is hereby considered as minimization of CO2 emissions as well as minimization of plastic for packaging, loading, transport and delivery as well as for commissioning processes. CO2 emissions result basically from the choice of the transport means for deliveries by logistics service providers as well as from self-pick-up by consumers. Furthermore, the production and recycling as well as thermal recycling of packaging leads to CO2 emissions. A tailored design of consumer logistics is understood as the individual selection of a particular logistics option out of a set of options. This set contains the selection of the source and the sink for the transport, the selection of the logistics service provider, the transport means as well as the transport distance. The evaluation of the chosen option is based on key performance indicators including CO2 footprint, type of packaging and share of plastic, delivery time, delivery date, delivery/deposit location and shipping cost.
200
3.2
M. Freitag and H. Kotzab
Suggested Approach
We suggest a tool which combines the socio-scientific method of discrete choice experiments for the optimal selection for a tailored sustainable consumer logistics option based on the individual consumer preferences with the engineering method of discrete-event simulation for quantifying the chosen option and offering of a feedback possibility based on the quantitative simulation results. With this approach, a consumer gets the opportunity to rethink the beforehand choice and potentially change the logistics option. The basic idea is portrayed in Fig. 1.
Socio-scienƟfic approach
Choice set
EvaluaƟon of delivery opƟons and comparison with chosen preferences
Discrete choice experiments
Consumer
EvaluaƟon SimulaƟon and analysis
Delivery opƟons SpecificaƟon of • Network structure • Parameters • Customer orders • CooperaƟon models SimulaƟon model
Engineering approach
Fig. 1. Interdisciplinary method for designing and evaluating consumer-centered sustainable last mile logistics systems
The socio-scientific perspective shall identify the individual consumer preferences towards indirect logistics costs (CO2 emissions, plastic waste), direct logistics costs (shipping costs) and logistics services (delivery time, time windows, etc.) by utilizing discrete choice experiments. The combinations of these preferences result in a tailored logistics option (retailer, packaging, transport means, transport route, delivery). We suggest examining different customer groups based on a consumer life cycle and household types as suggested by Müller-Hagedorn (1984). From the engineering perspective, these preferred logistics options will be applied to a exemplified logistics network. The results of the individual discrete choice experiments refer to typical customer profiles which will be reproduced until the simulation model has achieved an adequate number of differentiable customers. The model parameters refer thereby to the number of consumers and orders as well as to different cooperation models of the logistics service providers. The customer orders will be processed during the simulation based on the preferred logistics option. This shall identify the impact as well as the dependencies of different consumers with different logistics preferences. The simulation results will show the respondents their CO2 footprint, share of packaging and plastic waste as well as shipping costs, delivery times and delivery dates for all orders during a simulation period. With this, the consumers receive a measurable
A Concept for a Consumer-Centered Sustainable Last Mile Logistics
201
feedback to their preferred logistics option and with this quantification they may be able to rethink their previous decision and to potentially adapt their preferences for a second or even third simulation round. The incremental changes are going to be documented. This simulation-based feedback possibility represents the methodological enhancement of the traditional discrete choice experiment.
4 Conclusion and Outlook This paper suggests an innovative methodological approach for measuring the preferences for tailored sustainable consumer logistics by combining discrete choice experiments with computer-based simulation. Furthermore, the approach includes a feedback possibility in order to give consumers a chance to rethink a previous decision. The simulation model allows a measurable result of a consumer decision in regards to a preferred logistics option and allows an adaption towards increased sustainability by keeping a high degree on individualization. The presented approach is an extension of existing discrete choice experiment approaches by combining computer-based simulation. So far, discrete choice experiments would allow only qualitative statements on a desired last-mile logistics while the pure modelling and simulation of logistics options would be based on theoretical considerations only. Our suggestion allows an inclusion of real consumer preferences as well as a critical reflection and further adaptation of logistics systems on the other side. Of course, there are still some limitations which need to be considered when continuing this research endeavor. First, we need to identify the specific practical requirements as well as challenges which are related with the scale-up of the suggested approach. Second, it is necessary to specify the specific transport means as well as environmental boundary conditions that go into the decision-making process. From a supply chain perspective, it is necessary to further consider – based on the outcomes of the discrete choice experiments – the consequences of consumer-centered sustainable last mile logistics solutions for the demand management processes of the involved retailers and manufacturers as well as logistics service providers. In order to keep the decision sets as simple as possible, our suggestions do not include the basic notions of price elasticity as we consider changing delivery fees not as a part of the product price. Nevertheless, there might be an effect of changing fees on the overall buying behavior. Finally, it is also necessary to expand the notion of sustainability to more than only carbon emissions and packaging waste. The choice sets may also include e.g. decision possibilities towards compensation programs.
References Athanassopoulos, T., Dobers, K., Clausen, U.: Reducing the environmental impact of urban parcel distribution. In: Zijm, H., Klumpp, M., Clausen, U., Hompel, M.T. (eds.) Logistics and Supply Chain Innovation, pp. 159–181. Springer, London (2016) Auffermann, C.: Nachhaltige Stadt- und Verkehrsentwicklung. BVL Magazin 3, 40 (2017)
202
M. Freitag and H. Kotzab
Bahn, K.D., Granzin, K.L., Tokman, M.: End-user contribution to logistics value co-creation: a series of exploratory studies. J. Mark. Channels 22(1), 3–26 (2015) Berg, C.: City-Logistik – Das Münchner Modell. ILV-Institut Für Logistik und Verkehrsmanagement GmbH, München (1999) Bloemhof, J.M., van der Vorst, J.G.A.J., Bastl, M., Allaoui, H.: Sustainability assessment of food chain logistics. Int. J. Logist. Res. Appl. 18(2), 101–117 (2015) Bode, W.: Neue City-Logistik-Konzepte und -Techniken für mehr Nachhaltigkeit per City-GVZ und eStore. In: Deckert, C. (ed.) CSR und Logistik - Spannungsfelder Green Logistics und City-Logistik, pp. 281–291. Gabler Verlag, Wiesbaden (2016) Buchner, N.: Verpackung von Lebensmitteln: Lebensmitteltechnologische, verpackungstechnische und mikrobiologische Grundlagen. Springer Verlag, Berlin (2012) Clausen, U., Geiger, C., Pöting, M.: Hands-on testing of last mile concepts. Transp. Res. Procedia 14, 1533–1542 (2016) Deckert, C. (ed.): CSR und Logistik - Spannungsfelder Green Logistics und City-Logistik. Gabler Verlag, Wiesbaden (2016) Dobers, K., Klukas, A., Lammers, W., Laux, M., Mauer, G., Schneider, S.: Optimisation approaches for resource-efficient logistics services. In: Clausen, U., ten Hompel, M., Klumpp, M. (eds.) Efficiency and Logistics. Springer, Heidelberg (2013) Elbert, R., Friedrich, C.: Simulation-based evaluation of urban consolidation centers considering urban access regulations. In: 2018 Winter Simulation Conference (WSC), pp. 2827–2838. IEEE, Gothenburg (2018) Galipoglu, E., Kotzab, H., Pöppelbuß, J.: Multi-Channel-Systeme im Wandel. WiSt 44(8), 434– 441 (2015) Hüseyinoğlu, I., Galipoğlu, E., Kotzab, H.: Social, local and mobile commerce practices in omnichannel retailing: insights from Germany and Turkey. Int. J. Retail Distrib. Manag. 45(7/8), 711–729 (2017) Kille, C., Manner-Romberg, H., Miller, J., Müller-Steinfahrt, U., Symanczyk, W., Veres-Homm, U., Weber, N.: Abschlussbericht E-Commerce - Herausforderungen und Lösungen für den Logistikstandort Hamburg. Nürnberg (2015) Leyerer, M., Sonneberg, M.O., Breitner, M.H.: Decision Support for urban e-grocery operations (2018) Lohre, D., Bernecker, T., Gotthardt, R.: Praxisleitfaden zur IHK-Studie “Grüne Logistik” Umsetzungsbeispiele und Handlungsempfehlungen aus der Praxis. Industrie- und Handelskammer, Stuttgart (2011) Lütjen, M., Piotrowski, J.: City Logistik - Intelligenter Güterverkehr per Straßenbahn. Ind. Manag. 28(2), 47–50 (2012) McKinnon, A., Browne, M., Piecyk, M., Whiteing, A.: Green Logistics: Improving the Environmental Sustainability of Logistics, 3rd edn. Kogan Page, London (2015) McKinnon, A., Cullinane, S., Browne, M., Whiteing, A.: Green Logistics: Improving the Environmental Sustainability of Logistics, 2nd edn. Kogan Page, London (2010) Melkonyan, A., Krumme, K., Gruchmann, T., De La Torre, G.: Sustainability assessment and climate change resilience in food production and supply. Energy Procedia 123, 131–138 (2017) Meyer, J., Kotzab, H.: Consumer logistics and consumer value (-Co-) creation. In: Large, R., Kramer, N., Radig, A.-K., Schäfer, M., Sulzbach, A. (eds.) Logistikmanagement, Beiträge zur LM 2017 (LM 2017), pp. 313–322. Eigenverlag Lehrstuhl für Allg. BWL, Logistik- und Beschaffungsmanagement, Stuttgart (2017) Meyer, J., Kotzab, H., Teller, C.: Shopper logistics processes in a store-based grocery-shopping environment. In: Freitag, M., Kotzab, H., Pannek, J. (eds.) Dynamics in Logistics, pp. 313– 323. Springer, Bremen (2017)
A Concept for a Consumer-Centered Sustainable Last Mile Logistics
203
Meyer, J., Kotzab, H., Teller, C.: Shopper logistics processes - a critical state-of-the-art from a viewpoint of consumer value and convenience. In: Ojala, L., Töyli, J., Solakivi, T., Lorentz, H., Laari, S., Lehtinen, N. (eds.) Nofoma 2016 - Proceedings of the 28th Annual Nordic Logistics Research Conference, pp. 430–443, Turku (2016) Müller-Hagedorn, L.: Die Erklärung von Käuferverhalten mit Hilfe des Lebenszykluskonzeptes. Wirtschaft und Studium 13(11), 561–569 (1984) Nitsche, B., Figiel, A., Straube, F.: Zukunftstrends in der Lebensmittellogistik Herausforderungen und Losungsimpulse. Universitätsverlag der TU Berlin, Berlin (2016) Raiber, S., Spindler, H., Feldwieser, M.: Zusammenfassung: Elektrischer Schwerlastverkehr im urbanen Raum (2014). https://www.iao.fraunhofer.de/images/iao-news/schwerlastverkehr.pdf. Accessed 28 Feb 2019 Schönberg, T., Auffermann, C.: Neue Wege der urbanen Versorgung - Einhaltung der Kühlkette als große Herausforderung. BVL Magazin 3, 8–9 (2017) Schnedlitz, P., Lienbacher, E., Waldegg-Lindl, B., Waldegg-Lindl, M.: Last Mile: Die letzten – und teuersten – Meter zum Kunden im B2C E-Commerce. In: Crockford, G., Ritschel, F., Schmieder, U.-M. (eds.) Handel in Theorie und Praxis – Festschrift zum 60. Geburtstag von Prof. Dr. Dirk Möhlenbruch, pp. 249–273. Springer, Wiesbaden (2013) Schögel, M.: Distributionsmanagement: das Management der Absatzkanäle. Vahlens Handbücher der Wirtschafts- und Sozialwissenschaften, München (2012) Singh, P., Wani, A.A., Langowski, H.: Food Packaging Materials: Testing & Quality Assurance. CRC Press, Boca Raton (2017) Teller, C., Kotzab, H., Grant, D.B.: The relevance of shopper logistics for consumers of storebased retail formats. J. Retail. Consum. Serv. 19(1), 59–66 (2012) Trapp, M., Lütjen, M., Castellanos, J.D.A., Jelsch, O., Freitag, M.: Life cycle assessment for frozen food distribution schemes. In: Jahn, C., Kersten, W., Ringle, C.M. (eds.) Proceedings of the Hamburg International Conference of Logistics (HICL), pp. 267–284. Epubli Verlag, Hamburg (2017)
The Omnichannel Retailing Capabilities Wheel: Findings of the Literature Bastian Mrutzek1(&), Herbert Kotzab1,2, and Erdem Galipoglu1 1
University of Bremen, 28359 Bremen, Germany 2 Universiti Utara Malaysia, Sintok, Malaysia
Abstract. Omnichannel retailing is an approach for boundless channel integration of offline channels and the virtual world. Customers can buy their products among different touchpoints simultaneously, regardless of time and location. In this setting, many retailers aim to set up single channel solutions for their warehouse & delivery management, organization- and IT-systems. Some retailers implement this better than others that leads to the question which capabilities are necessary to manage omnichannel retailing? In order to do so, a content-based literature analyze of 63 academic papers identified by a Scopus literature search have been conducted. The ordinary capabilities refer to forecasting of market/ customer requirements, implementation of omnichannel settings, provision of assortments and access to channels. The dynamic omnichannel retailing capabilities include an understanding of customer and market developments, integration and coordination of various channel activities, interaction with end-users, creation of an omnichannel environment as well as network and an innovation ability. The major contribution of this paper is the linkage between omnichannel retailing and the dynamic capability view (DCV) by proposing a set of two ordinary and six dynamic omnichannel retailing capabilities. Keywords: Omnichannel retailing review
Dynamic capability view Literature
1 Introduction Omnichannel retailing is a relatively new approach for boundless channel integration of offline channels and the virtual world where customers can use multiple channels simultaneously [1, 2]. Customers can buy their products among different touchpoints, regardless of time and location [1]. Omnichannel retailing refers to the management of diverse channels of distribution, touchpoints which optimizes the customer experience and the channel performance [3]. Touchpoints allow contact and communication between end-users and retailers, thus guaranteeing the combination between the advantages of physical stores with the improved information level of online shopping [3]. This is getting more intense since local, mobile and social commerce (also known as SO-LO-MO, see [4] or [5]) has been established due to the increasing acceptance and use of smartphones, social media, mobile internet and location-based technologies. The use of multiple channels in order to distribute products/services to different customer groups is not a new phenomenon (e.g. [6]). Such multi-channel distribution © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 204–214, 2020. https://doi.org/10.1007/978-3-030-44783-0_20
The Omnichannel Retailing Capabilities Wheel
205
refers to parallel and rather uncoordinated management of multiple individual distribution channels (e.g. [7]). This means that these channels do not interact and are not integrated [1]. A first integration attempt is made by cross-channel retailing, which aims at executing a synergetic channel strategy (see [8]). Taking these developments into account it appears obvious that technology is becoming an important factor for any retail channel strategy. This means that it is required to identify capabilities and resources to provide successful strategies. Thus, the purpose is to identify capabilities by using the notion of the DCV [9–11]. The reason for doing so lies in our argumentation that omnichannel retailing can be characterized by following the definition of the six dimensional model of service innovations [12]. They state that “a service innovation is a new service experience or service solution that consists of one or several of the following dimensions: new service concept, new customer interaction, new value system/business partners, new revenue model, new organizational or technological service delivery system” [12]. The concept of omnichannel retailing contains several of these dimensions, e.g. creating new interaction possibilities for the customer through usage of different channels simultaneously [3]. Retailing enterprises who want to establish this new and innovative supply of service on the market need (dynamic) capabilities regarding technology, personnel, organization and culture [12]. The DCV takes these changing circumstances into account. The advanced solution of this approach is the possession of dynamic capabilities with unique and hard to copy characteristics which help to facilitate the firm’s unique asset base and therefore, secure the achieved competitive advantage in the long run [10, 13]. The major functions of the DCV framework refer to the sensing and seizing of opportunities as well as threats with the solution of transforming assets and inputs of the firm. Thereby, Teece [11] sees the technological and developmental opportunities as the focus of sensing (and shaping) capabilities. Processes, which detect slightly changes in customer needs and in targeted markets are also important [14, 15]. Seizing opportunities consists of the investment strategy of a firm by figuring out when, where and how much resources are necessary to meet the sensed opportunities [11]. The last element of this framework is the permanent adjustment and readjustment of tangible and intangible input as well as the protection of assets [10]. Continuing research divides capabilities of the firm into ordinary and dynamic capabilities, where ordinary capabilities are also known as static [16] or zero-level [17] capabilities. Ordinary capabilities consist of combinations of workforce, facilities as well as equipment, recurring routines and coordination of administrative processes - to do things right. Contrary to the rarely inimitable ordinary capabilities, dynamic capabilities achieve the status as most of the time inimitable. The focus of dynamic capabilities lies in doing the right things at the right time. Recognition of key developments or trends by the top management is one of the main functions of this scheme. The dynamic capabilities core competence is the orchestration and leadership of entrepreneurial processes, learning capacities as well as strategies which help to meet changing customer needs and exploit technological and business opportunities [18, 19]. Consequently, this paper is examining which ordinary and dynamic capabilities are necessary for being and becoming a successful omnichannel retailing company and therefore, the following two research questions are developed: (a) What are the
206
B. Mrutzek et al.
ordinary capabilities that comprise basic operative competences for daily omnichannel retail processes? and (b) Which dynamic capabilities distinguish a competitive and successful from an unsuccessful and competitively weak omnichannel retailer? We are going to answer these questions by the means of the literature review of 63 academic papers that were published between 2008 and 2018. The result of our analysis refers to a conceptual model of eight ordinary and dynamic omnichannel retailing capabilities which distinguish between successful and less successful omnichannel retailers. These omnichannel retailing capabilities extend the notions of the DCV to the field of retailing. Moreover, the pairing with the approach of the six-dimensional model of service innovations by Hertog et al. [12] enlarges the applicability of the DCV approach on different topics. The remainder of the paper is organized as follows: After an introduction, section two explains our methodological approach for gathering and analyzing our data. The results of our literature review are presented in section three. The paper closes with a critical reflection of our results and an outlook for future research.
2 Literature Review The approach of a literature review is identifying content in a special research area by screening and analyzing documents in a way that is structured and reproducible [20]. The focus of this literature review lies in identifying and determining capabilities which are necessary to implement and operate omnichannel retailing. The literature review is based on the approach of Denyer and Tranfield [21] and is divided into different steps. The first step is the formulation of research questions, which is documented in the introduction of this paper. The purpose of this approach is the detection of studies and articles where capabilities can be identified which are necessary and helpful for implementing omnichannel retailing. Identification of Papers. The introduction of smart mobile devices made it possible for customers to shop anywhere as well as anytime. Accordingly, the relevance of omnichannel retailing in the literature is paired with this development [22] and a time period between 2008 and 2018 is set. The identification of papers, as the second step of the literature review, was conducted by using Scopus, due to its large abstract and citation database of peer-reviewed literature. We used the following terms as combinations included in the title, abstract and/or keywords of a paper: “omni chann*” OR “omni-chan*” OR “omnichan*” AND “retail*”. Moreover, the search string was limited to the subject areas of “BUSI”, “COMP”, “ENGI” or “DECI”. In a second search the term “retail*” was replaced by the term “cap*”. Paper Selection and Evaluation. The selection process as the third step is made up of filtering relevant from irrelevant papers by reading the titles, keywords and abstracts. The search string delivered 153 studies and only paper which were classified as appropriate in terms of the specific disciplines of logistic, marketing and strategic management were selected for a further examination. To ensure the reliability of the selection process, three researcher independently read and interpreted the identified papers [23]. Following the notion of [24], each coder was equipped with clear
The Omnichannel Retailing Capabilities Wheel
207
instructions alongside the coding scheme. The results were compared and disagreements amongst researchers discussed. The researchers then agreed on a final screening result that includes 63 relevant articles for answering the formulated questions. Analysis and Synthesis. The last step of the literature review is the analysis and synthesis of the conducted review. In doing so, an approach of consolidation is realized and similar findings are merged, whereby a more profound result will be generated. The following descriptive synthesis is subdivided into second-order categories (constructs) which divide ordinary and dynamic capabilities. These categories contain several firstorder categories (indicators) with items (variables) which can be assigned to specific papers and articles where they are described.
3 Descriptive Synthesis of Results The fourth chapter outlines the results of the literature review. It is divided into the notions of ordinary and dynamic capabilities of Teece [9]. Figure 1 summarizes the set of our identified ordinary and dynamic capabilities for the area of omnichannel retailing.
Indicators (Variables) Policy (transparency among all channels) Realloca on (dynamic structural change within internal processes) Digitaliza on (contactless payment)
Constructs
Tools (app with social media connec vity)
Omnichannel Environment
Customer Loyalty (social media response & customer tailored marke ng ac vi es)
One-Direc onal Communica on (publicizing omnichannel retailing poten als)
Network (crea ng innova ve logis c networks)
Innova veness
Restructuring (realignment of the supply chain) Harmoniza on (supply chain comprehensive inventory management)
CapabiliƟes
Customer Interac on
Supply Chain Management
Dynamic
Integra on & Coordina on
Ordinary
Technology (supervision systems)
Availability
Transport (different pick-up and delivery op ons)
Informa on Management (data security) Transfer (customer loyalty transfer - offline online)
Informa on (product informa on distribu on) Service (free Wi-Fi in stores)
Efficiency (batching effects)
Technology (digital assistants)
Moderniza on (introducing a reward system to manage decentraliza on) Agility (reac ng ini a vely on market shi s)
Omnichannel Retailing
Channel Integra on (data integra on) Omnichannel Selec on (selec ng appropriate channels)
Innova ons (innova ve technologies for crea ng unique selling points) Infrastructure (installing innova ve logis c processes)
Channel Capabili es (crosschannel fulfillment)
Bi-Direc onal Communica on (using customer-generated content)
Innova on Management (adap ng innova ons from different industries into enterprise processes)
Customer / Market Understanding
Implementa on
Customer / Market Behavior (mobile apps to receive data)
Logis c Service Level ( meliness, feed speed)
Heterogeneous Customer Demands (SO-LO-MO approach)
Digital Means (global website)
Customer Loyalty (customer rela onship management) Data Analysis (cloud services)
Supply of Products (supply assurance) Forecas ng of Demand (bigdata analysis)
Return Management (simplifying the process) Strategy (joint forecasts) Pricing (harmonizing channels)
Fig. 1. Ordinary and dynamic omnichannel retailing capabilities wheel
208
B. Mrutzek et al.
The wheel depicts important ordinary and dynamic capabilities that companies partially need to set up to operate as an omnichannel retailer. The wheel is buildup of circles, starting from the inside and getting more detailed going further steps to the outside. The following documentation of our findings is organized upon this graphic. 3.1
Ordinary Capabilities for Omnichannel Retailing
Ordinary capabilities are needed to fulfill routinely performed operations [9]. In the context of omnichannel retailing we identified operative competences that can be separated into the two constructs of availability and implementation of processes. Availability. This construct of ordinary capabilities contains six indicators of availability, where each of these indicators are defined by their variables. Technological availability implies supervision and production scheduling systems [25]. Information in terms of availability is an continuous product information distribution [14, 26, 27]. Services regarding ordinary capabilities is the provision of free Wi-Fi in local stores [15, 28, 29] and the supply of products means the assurance that products are available online as well as offline [30]. Moreover, different pick-up and delivery options should be available to meet transportation expectations of the customers. Forecasting of demand is a standardized process for retailers, in terms of omnichannel retailing that means the analyzation of big data sets [31]. Implementation. The second set of ordinary capabilities identified is referred to the implementation of services. Logistic service level implementation in omnichannel retailing is an overall inventory and assortment configuration as well as accurate timeliness and feed speed [32, 33]. Digital means include the implementation of digital services and social media presence [5], whereas a strategy in this context is often relevant for long-term problems like a more downstream oriented logistic and relationship management concerns building a division for omnichannel management [34]. Pricing in this context affects capabilities to deal with gaps of online and offline prices, which have to be harmonized [35]. Return management in the present of implementation and ordinary capabilities regards the installation of simplified terms of return [28]. These ordinary capabilities are the foundation of the following dynamic capabilities, which refer to activities of doing the right things at the right time [9]. Moreover, they are characterized by cognitive abilities and leadership as well as orchestration skills consisting of signature processes to create sustainable competitive advantages. 3.2
Dynamic Capabilities for Omnichannel Retailing
In respect of omnichannel retailing we condensed the following six constructs: customer/market understanding, integration and communication, customer interaction, omnichannel environment, supply chain management as well as innovativeness. Customer and Market Understanding. The first construct within dynamic capabilities is the understanding of customers and market situations, especially recognizing trends and needs is central to this part. The customer and market behavior includes the utilization of mobile apps to create individual customer profiles [30, 36–41] as well as
The Omnichannel Retailing Capabilities Wheel
209
data analysis to detect trends and provide cloud services [42–44] and the transfer of knowledge of experiences regarding the customer and market situations [33]. Heterogeneous customer demands can be controlled through the usage of methods like the SoLo-Mo approach [4, 45] and the customer loyalty as a significant competitive parameter can be strengthened through integrated customer relationship management [46]. Integration and Coordination. The second construct is divided into six indicators. Channel integration represents the data integration for multiple channels into one software [1, 47–50], whereas omnichannel selection is referring to the selection of appropriate channels [51] and information management is focusing on data security over all channels and touchpoints [52]. Facing information and data issues, the technology represents a major part of omnichannel retailing, e.g. digital assistants [53, 54]. Efficiency in terms of dynamic capabilities incorporates batching effects of all channels [55]. Moreover, transfer of knowledge in the class of integration and coordination regards the ability to transfer customer loyalty from offline to online channels and vice versa [33, 56, 57]. Customer Interaction. The third construct includes four indicator categories. Tools in terms of customer interaction applies to e.g. the installation of an app with social media connectivity and real time data [58–61]. Customer interaction is highly responsible for customer loyalty and therefore, social media responsibility and customer tailored marketing activities are important [58, 62–66]. Moreover, interaction with customers include communication. That can take place one-directional by publicizing omnichannel retailing potentials [67, 68] as well as bi-directional through the usage of customer-generated content [69, 70]. Omnichannel Environment. The fourth dynamic capability construct contains five indicator categories which are supporting the company to create an omnichannel environment. In the policy there must be defined, that transparency among all channels have to be the goal [71] and the channel capabilities should include cross-channel fulfillment [72], whereas new channels have to be set up to create touchpoints for a seamless shopping experience [73]. To change the environment towards an more omnichannel typical environment reallocation of resources and capabilities in terms of dynamic structural change within internal processes have to be done [51, 74, 75]. Digitalization in terms of this construct includes the installation of contactless payment methods [76, 77]. Supply Chain Management. The sixth construct of supply chain management refers to the business partners-oriented parts of omnichannel retailing. To assure a seamless shopping experience the network in terms of the creation of innovative logistic networks in the supply chain [78] and restructuring referring to a realignment of the supply chain [33] is necessary. Through harmonization, e.g. a supply chain comprehensive inventory management [79], and modernization, e.g. the introduction of a reward system to manage decentralization [42], the company can react on market shifts and innovate the business processes. Therefore, agility is another important capability under the aspect of supply chain management [80].
210
B. Mrutzek et al.
Innovativeness. The last construct consists of three indicator categories. Innovation management applies to the utilization of existing IT structures, the adaption and transfer of innovations [81]. Innovations open up the creation of unique selling points through innovative technologies [51] and infrastructure under consideration of the construct of innovativeness comprises the installation of innovative logistic processes to utilize the potential of omnichannel retailing [80].
4 Conclusion The purpose of this paper was to identify ordinary as well as dynamic capabilities that omnichannel retailers need to possess in order to achieve a superior and sustainable competitive advantage. Our literature review led to a set of two ordinary and six dynamic omnichannel retailing capabilities which both were further divided into several subsets of activities. The gathered ordinary capability categories include the need to supply products and services simultaneously and uniformly on every channel while providing diverse delivery options as well as a logistic service level with consistent and safe handling of goods. Dynamic omnichannel capabilities involve the capability to understand customers and markets by using software, e.g. digital assistants, to analyze developments and to identify trends. Moreover, the integration and coordination of channels, e.g. with supporting software for distribution systems and the digitalization, by installing monitors in stores as well as free WI-FI, can help to foster customer interaction. This would guarantee customer access to mobile devices in stores whereby real time data can be gathered, analyzed and used for customer tailored marketing activities as well as fluent customer brand interactions through different touchpoints. Therefore, an omnichannel strategy is necessary which refers to knowledge in fields of handling increased complexity and the alignment of varying channels and customer needs. This includes innovative supply chain capabilities to introduce new infrastructure and transport options as well as a harmonization of all partners inside the chain. Ultimately, the ability to innovate the structures and processes as well as the integration and management of innovation is as important as sensing market trends and customer needs. The identified capabilities in this paper which contain definitive processes, abilities and requirements indicate the variety of capabilities which are necessary to manage omnichannel retailing successfully. Future research could direct their focus on validating the identified set of ordinary and dynamic omnichannel retailing capabilities empirically by conducting case studies with multichannel retailing companies. Furthermore, an examination of various retail segments, within the scope of omnichannel retailing, could be valuable by identifying specifications and comparing it with the presented frame of reference.
The Omnichannel Retailing Capabilities Wheel
211
References 1. Beck, N., Rygl, D.: Categorization of multiple channel retailing in multi-, cross-, and omnichannel retailing for retailers and retailing. J. Retail. Consumer Serv. 27, 170–178 (2015). https://doi.org/10.1016/j.jretconser.2015.08.001 2. Piotrowicz, W., Cuthbertson, R.: Introduction to the special issue information technology in retail: toward omnichannel retailing. Int. J. Electron. Commerce 18, 5–16 (2014). https://doi. org/10.2753/JEC1086-4415180400 3. Verhoef, P.C., Kannan, P.K., Inman, J.J.: From multi-channel retailing to omni-channel retailing: introduction to the special issue on multi-channel retailing. J. Retail. 91, 174–181 (2015). https://doi.org/10.1016/j.jretai.2015.02.005 4. Yumurtacı Hüseyinoğlu, I.Ö., Galipoğlu, E., Kotzab, H.: Social, local and mobile commerce practices in omni-channel retailing: insights from Germany and Turkey. Int. J. Retail Distrib. Manag. 45, 711–729 (2017) 5. Galipoglu, E., Kotzab, H., Pöppelbuß, J.: Multi-Channel-Systeme im Wandel. WiSt Wirtschaftswissenschaftliches Studium. 44, 434–441 (2015). https://doi.org/10.15358/03401650-2015-8-434 6. Lynch, J., Alba, J.: Interactive home shopping: consumer, retailer, and manufacturer incentives to participate electronic marketplaces. J. Market.: Q. Publ. Am. Market. Assoc. 61, 38–53 (1997) 7. Ahlert, D., Hesse, J., Jullens, J., Smend, P.: Multikanalstrategien: Konzepte, Methoden und Erfahrungen. Springer, Heidelberg (2003) 8. Schwerdt, Y.: Multi-channel, cross-channel und omni-channel retailing–Schwerdt-blog. Online verfügbar unter zuletzt aktualisiert am 26, 2013 (2013). http://schwerdtblog. absatzwirtschaft.de/2013/02/26/multi-channel-cross-channel-und-omnichannel-retailing/ 9. Teece, D.J.: The foundations of enterprise performance: dynamic and ordinary capabilities in an (economic) theory of firms (2014) 10. Teece, D.J.: Explicating dynamic capabilities: The nature and microfoundations of (sustainable) enterprise performance. Strateg. Manag. J. 28, 1319–1350 (2007). https://doi. org/10.1002/smj.640 11. Teece, D.J., Pisano, G., Shuen, A.: Dynamic capabilities and strategic management. Strateg. Manag. J. 18, 509–533 (1997). https://doi.org/10.1002/(SICI)1097-0266(199708)18:7% 3c509:AID-SMJ882%3e3.0.CO;2-Z 12. den Hertog, P., van der Aa, W., de Jong, M.W.: Capabilities for managing service innovation: towards a conceptual framework. J. Serv. Manag. 21, 490–514 (2010). https:// doi.org/10.1108/09564231011066123 13. Ambrosini, V., Bowman, C.: What are dynamic capabilities and are they a useful construct in strategic management? Int. J. Manag. Rev. 11, 29–49 (2009) 14. Bell, D., Gallino, S., Moreno, A.: Showrooms and information provision in omni-channel retail. Prod. Oper. Manag. 24, 360–362 (2015) 15. Frishammar, J., Cenamor, J., Cavalli-Björkman, H., Hernell, E., Carlsson, J.: Digital strategies for two-sided markets: a case study of shopping malls. Decis. Support Syst. 108, 34–44 (2018) 16. Collis, D.J.: Research note: how valuable are organizational capabilities? Strateg. Manag. J. 15, 143–152 (1994) 17. Winter, S.G.: Understanding dynamic capabilities. Strateg. Manag. J. 24, 991–995 (2003) 18. Eisenhardt, K.M., Furr, N.R., Bingham, C.B.: CROSSROADS—microfoundations of performance: balancing efficiency and flexibility in dynamic environments. Organ. Sci. 21, 1263–1273 (2010)
212
B. Mrutzek et al.
19. Teece, D.J.: Dynamic capabilities: routines versus entrepreneurial action. J. Manag. Stud. 49, 1395–1401 (2012) 20. Vom Brocke, J., Simons, A., Niehaves, B., Riemer, K., Plattfaut, R., Cleven, A.: Reconstructing the giant: on the importance of rigour in documenting the literature search process. In: Ecis, pp. 2206–2217 (2009) 21. Denyer, D., Tranfield, D.: Producing a systematic review. In: The Sage Handbook of Organizational Research Methods, pp. 671–689 (2009) 22. Galipoglu, E., Kotzab, H., Teller, C., Yumurtaci Hüseyinoglu, I.Ö., Pöppelbuß, J.: Omnichannel retailing research – state of the art and intellectual foundation. Int. J. Phys. Dist. Log. Manag. 48, 365–390 (2018). https://doi.org/10.1108/IJPDLM-10-2016-0292 23. Creswell, J.W.: Research Design: Qualitative, Quantitative, and Mixed Methods Approaches. Sage Publications, Thousand Oaks (2009) 24. Krippendorff, K.: Content Analysis: An Introduction to its Methodology. SAGE Publications, Thousand Oaks (2004) 25. Bernon, M., Cullen, J., Gorst, J.: Online retail returns management: Integration within an omni-channel distribution context. Int. J. Phys. Distrib. Logist. Manag. 46, 584–605 (2016) 26. Grewal, D., Roggeveen, A.L., Nordfält, J.: Roles of retailer tactics and customer-specific factors in shopper marketing: substantive, methodological, and conceptual issues. J. Bus. Res. 69, 1009–1013 (2016) 27. Pascoe, M., Wright, O., Winzar, H.: Using best-worst scaling to reveal perceived relative importance of website attributes. Asia Pac. J. Market. Logist. 29, 393–408 (2017) 28. Rodríguez-Torrico, P., Cabezudo, R.S.J., San-Martín, S.: Tell me what they are like and I will tell you where they buy. An analysis of omnichannel consumer behavior. Comput. Hum. Behav. 68, 465–471 (2017) 29. Cook, G.: Customer experience in the omni-channel world and the challenges and opportunities this presents. J. Direct Data Digit. Market. Practice 15, 262–266 (2014) 30. Gallino, S., Moreno, A., Stamatopoulos, I.: Channel integration, sales dispersion, and inventory management. Manag. Sci. 63, 2813–2831 (2016) 31. Balakrishnan, J., Cheng, C.-H., Wong, K.-F., Woo, K.-H.: Product recommendation algorithms in the age of omnichannel retailing–An intuitive clustering approach. Comput. Ind. Eng. 115, 459–470 (2018) 32. Hübner, A., Wollenburg, J., Holzapfel, A.: Retail logistics in the transition from multichannel to omni-channel. Int. J. Phys. Distrib. Logist. Manag. 46, 562–583 (2016) 33. Ishfaq, R., Defee, C.C., Gibson, B.J., Raja, U.: Realignment of the physical distribution process in omni-channel fulfillment. Int. J. Phys. Distrib. Logist. Manag. 46, 543–561 (2016) 34. Hansen, R., Sia, S.K.: Hummel’s digital transformation toward omnichannel retailing: key lessons learned. MIS Q. Executive 14(2), 51–66 (2015) 35. Kireyev, P., Kumar, V., Ofek, E.: Match your own price? Self-matching as a retailer’s multichannel pricing strategy. Market. Sci. 36, 908–930 (2017) 36. Dai, B., Pelton, L.E.: Exploring consumers’ skincare retail patronage. J. Retail. Consumer Serv. 43, 269–277 (2018) 37. Riemann, U.: Analysis of cloud services on business processes in the digitalization of the consumer product industry. In: Web-Based Services: Concepts, Methodologies, Tools, and Applications, pp. 1351–1387. IGI Global (2016) 38. Ma, S.: Fast or free shipping options in online and Omni-channel retail? The mediating role of uncertainty on satisfaction and purchase intentions. Int. J. Logist. Manag. 28, 1099–1122 (2017) 39. Bell, D.R., Gallino, S., Moreno, A.: Offline showrooms in omnichannel retail: demand and operational benefits. Manag. Sci. 64, 1629–1651 (2017)
The Omnichannel Retailing Capabilities Wheel
213
40. Jin, M., Li, G., Cheng, T.C.E.: Buy online and pick up in-store: design of the service area. Eur. J. Oper. Res. 268, 613–623 (2018) 41. Lee, C.K.H.: A GA-based optimisation model for big data analytics supporting anticipatory shipping in Retail 4.0. Int. J. Prod. Res. 55, 593–605 (2017) 42. Wiener, M., Hoßbach, N., Saunders, C.: Omnichannel businesses in the publishing and retailing industries: synergies and tensions between coexisting online and offline business models. Decis. Support Syst. 109, 15–26 (2018) 43. Bai, H., McColl, J., Moore, C.: Luxury retailers’ entry and expansion strategies in China. Int. J. Retail Distrib. Manag. 45, 1181–1199 (2017) 44. Wang, S.-T.: Integrating KPSO and C5. 0 to analyze the omnichannel solutions for optimizing telecommunication retail. Decis. Support Syst. 109, 39–49 (2018) 45. Mirsch, T., Lehrer, C., Jung, R.: Transitioning to an Omnichannel Approach: A Dynamic Capability Perspective (2016) 46. Ieva, M., Ziliani, C.: Mapping touchpoint exposure in retailing: implications for developing an omnichannel customer experience. Int. J. Retail Distrib. Manag. 46, 304–322 (2018) 47. Herhausen, D., Binder, J., Schoegel, M., Herrmann, A.: Integrating bricks with clicks: retailer-level and channel-level outcomes of online–offline channel integration. J. Retail. 91, 309–325 (2015) 48. Marchet, G., Melacini, M., Perotti, S., Rasini, M., Tappia, E.: Business logistics models in omni-channel: a classification framework and empirical analysis. Int. J. Phys. Distrib. Logist. Manag. 48, 439–464 (2018) 49. Murfield, M., Boone, C.A., Rutner, P., Thomas, R.: Investigating logistics service quality in omni-channel retailing. Int. J. Phys. Distrib. Logist. Manag. 47, 263–296 (2017) 50. Brynjolfsson, E., Hu, Y.J., Rahman, M.S.: Competing in the Age of Omnichannel Retailing. MIT, Cambridge (2013) 51. Kim, H., Lee, D., Ryu, M.: An optimal strategic business model for small businesses using online platforms. Sustainability 10, 579 (2018) 52. Miell, S., Gill, S., Vazquez, D.: Enabling the digital fashion consumer through fit and sizing technology. J. Glob. Fashion Market. 9, 9–23 (2018) 53. von Briel, F.: The future of omnichannel retail: a four-stage Delphi study. Technol. Forecast. Soc. Change 132, 217–229 (2018) 54. Parise, S., Guinan, P.J., Kafka, R.: Solving the crisis of immediacy: how digital technology can transform the customer experience. Bus. Horiz. 59, 411–420 (2016). https://doi.org/10. 1016/j.bushor.2016.03.004 55. Fornari, E., Fornari, D., Grandi, S., Menegatti, M., Hofacker, C.F.: Adding store to web: migration and synergy effects in multi-channel retailing. Int. J. Retail Distrib. Manag. 44, 658–674 (2016) 56. Swoboda, B., Weindel, J., Schramm-Klein, H.: Crosswise and reciprocal interdependencies within retailers’ multichannel structures. Int. Rev. Retail Distrib. Consumer Res. 26, 347– 374 (2016) 57. Kemperman, A., van Delft, L., Borgers, A.: Omni channel fashion shopping. In: Successful Technological Integration for Competitive Advantage in Retail Settings, pp. 144–167. IGI Global (2015) 58. Vazquez, D., Dennis, C., Zhang, Y.: Understanding the effect of smart retail brand– Consumer communications via mobile instant messaging (MIM)–an empirical study in the Chinese context. Comput. Hum. Behav. 77, 425–436 (2017) 59. Gao, F., Su, X.: Omnichannel retail operations with buy-online-and-pick-up-in-store. Manag. Sci. 63, 2478–2492 (2016) 60. Taylor, D.G., Levin, M.: Predicting mobile app usage for purchasing and informationsharing. Int. J. Retail Distrib. Manag. 42, 759–774 (2014)
214
B. Mrutzek et al.
61. Pollamarasetty, S., Potti, R.: Omni-channel retailing: enriching customers’ shopping experience. In: Handbook of Research on Strategic Supply Chain Management in the Retail Industry, pp. 233–249. IGI Global (2016) 62. Blom, A., Lange, F., Hess Jr., R.L.: Omnichannel-based promotions’ effects on purchase behavior and brand image. J. Retail. Consumer Serv. 39, 286–295 (2017) 63. Gupta, A.: Customer service: a key differentiator in retailing. In: Handbook of Research on Strategic Supply Chain Management in the Retail Industry, pp. 75–86. IGI Global (2016) 64. Hutchinson, K., Donnell, L.V., Gilmore, A., Reid, A.: Loyalty card adoption in SME retailers: the impact upon marketing management. Eur. J. Market. 49, 467–490 (2015) 65. Cavender, R., Kincade, D.H.: A luxury brand management framework built from historical review and case study analysis. Int. J. Retail Distrib. Manag. 43, 1083–1100 (2015) 66. Rosenmayer, A., McQuilken, L., Robertson, N., Ogden, S.: Omni-channel service failures and recoveries: refined typologies using Facebook complaints. J. Serv. Market. 32, 269–285 (2018) 67. Saghiri, S., Wilding, R., Mena, C., Bourlakis, M.: Toward a three-dimensional framework for omni-channel. J. Bus. Res. 77, 53–67 (2017) 68. Yurova, Y., Rippé, C.B., Weisfeld-Spolter, S., Sussan, F., Arndt, A.: Not all adaptive selling to omni-consumers is influential: the moderating effect of product type. J. Retail. Consum. Serv. 34, 271–277 (2017) 69. Kang, J.-Y.M.: Showrooming, webrooming, and user-generated content creation in the omnichannel era. J. Internet Commer. 17, 145–169 (2018) 70. Lapoule, P., Rowell, J.: Using social media to support trade shows: developing the capabilities. South Asian J. Bus. Manag. Cases 5, 88–98 (2016) 71. Fulgoni, G.M.: “Omni-Channel” retail insights and the consumer’s path-to-purchase: how digital has transformed the way people make purchasing decisions. J. Advertising Res. 54, 377–380 (2014) 72. Luo, J., Fan, M., Zhang, H.: Information technology, cross-channel capabilities, and managerial actions: evidence from the apparel industry (2015) 73. Huré, E., Picot-Coupey, K., Ackermann, C.-L.: Understanding omni-channel shopping value: a mixed-method study. J. Retail. Consum. Serv. 39, 314–330 (2017) 74. Rippé, C.B., Weisfeld-Spolter, S., Yurova, Y., Dubinsky, A.J., Hale, D.: Under the sway of a mobile device during an in-store shopping experience. Psychol. Market. 34, 733–752 (2017) 75. Picot-Coupey, K., Huré, E., Piveteau, L.: Channel design to enrich customers’ shopping experiences: Synchronizing clicks with bricks in an omni-channel perspective–the direct Optic case. Int. J. Retail Distrib. Manag. 44, 336–368 (2016) 76. Reis, J., Amorim, M., Melão, N.: Omni-channel service operations: building technologybased business networks. In: 2017 IEEE International Conference on Service Operations and Logistics, and Informatics (SOLI), pp. 96–101. IEEE (2017) 77. Burnes, B., Towers, N.: Consumers, clothing retailers and production planning and control in the smart city. Prod. Plann. Control 27, 490–499 (2016) 78. Larke, R., Kilgour, M., O’Connor, H.: Build touchpoints and they will come: transitioning to omnichannel retailing. Int. J. Phys. Distrib. Logist. Manag. 48, 465–483 (2018) 79. Zhang, M., Ren, C., Wang, G.A., He, Z.: The impact of channel integration on consumer responses in omni-channel retailing: the mediating effect of consumer empowerment. Electron. Commer. Res. Appl. 28, 181–193 (2018) 80. Wollenburg, J., Hübner, A., Kuhn, H., Trautrims, A.: From bricks-and-mortar to bricks-andclicks: logistics networks in omni-channel grocery retailing. Int. J. Phys. Distrib. Logist. Manag. 48, 415–438 (2018) 81. Tambo, T.: Fashion retail innovation: about context, antecedents, and outcome in technological change projects. In: Fashion and Textiles: Breakthroughs in Research and Practice, pp. 233–260. IGI Global (2018)
Sustainable Retail Supply Chain Management – A Bibliometric Viewpoint Kristina Petljak1 and Herbert Kotzab2,3(&) 1
Faculty of Economics and Business, University of Zagreb, Zagreb, Croatia 2 University of Bremen, Bremen, Germany [email protected] 3 Universiti Utara Malaysia, Sintok, Malaysia
Abstract. In this paper, we examine the research domain of sustainable retail supply chain management (SRSCM). Sustainability is a multi-dimensional construct trying to balance social, environmental as well as economic objectives simultaneously. By using citation and co-citation analysis we are able to illustrate the intellectual foundation and roots of this field of research. Our results show that the research domain of SRSCM is embedded in a number of various research sub disciplines including sustainability, retail, logistics, operations and supply chain management as well as the consumer side of supply chain. Keywords: Retail supply chain management analysis HistCite VOSviewer
Sustainability Bibliometric
1 Introduction One of the most mentioned topics today is the increasing concern about environmental and social issues in food production, distribution and consumption. From a corporate social responsibility (CSR) perspective, food and agribusiness companies are frequently subject to sustainability interests and there is an increasing need for them to respond to the challenges and obligations posed by environmental and social issues (Souza-Montera and Hooker 2017). Companies are facing rapid changes due to the growing concern and rising awareness among consumers of e.g. traceability in the food chain, the origin of raw materials and safety, environmental impacts of products and processes as well as societal issues such as welfare (e.g. Morgan et al. 2018). Customers, governments, non-governmental organizations, the media and wider society are also demanding from companies to provide an open and well-substantiated account on how they operate, what their impact on society is, and how they are minimizing negative impacts and saving scarce natural resources (Hingley et al. 2013). Over the past couple of decades, as the growth of retailers has led to the fundamental shift in marketplace power from manufacturers to retailers (Arnold 2002) and as retailers became closer to consumers, they are now taking the leadership role in sustainability as well (Wiese et al. 2015). Some of the greatest sustainability issues lie within supply chain and its operations as sustainability questions and answers are more © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 215–224, 2020. https://doi.org/10.1007/978-3-030-44783-0_21
216
K. Petljak and H. Kotzab
and more shifting from the single-firm level to supply chains and networks (Hartmann 2011). Nowadays retailers are demonstrating sustainability dedication not only on their web pages (Kotzab et al. 2011) or at the store level (Jones et al. 2011), but more and more on a supply chain level as well (Maloni and Brown 2006; Petljak et al. 2018). Research has so far only partially examined sustainability within the retail domain. Research looked mainly on food consumption patterns (Carlsson-Kanyama 1998; Carrigan and de Pelsmacker 2009), environmental labels and eco or organic labels (de Snoo and van de Ven 1999; Kaiser and Edwards-Jones 2006), ethical trade (Browne et al. 2000), resource efficiency view of the consumer and retailer (Ogle et al. 2004), general sustainability in retailing (Jones et al. 2007), certification (Binnekamp and Ingenbleek 2008), retailers’ corporate social responsibility (Biloslavo and Trnavcevic 2009; Erol et al. 2009; Kolk et al. 2010), sustainable development (Balan 2009), sustainable business relations (Reynolds et al. 2009), green logistics (Jarosz 2008; de Brito et al. 2008) or green retail (food) supply chain management (Petljak et al. 2018). Taking all these issues into account, we see a need to identify the intellectual foundation of SRSCM. Contrary to the objectives of a systematic or content-based literature reviews, which identify, evaluate and integrate the main findings of individual studies based on research questions (see e.g. Seuring and Gold 2012; Durach et al. 2017), the goal of this paper is to determine the roots (= ‘intellectual foundation’) of SRSCM, because of its growing importance as research field per se. By reviewing the intellectual foundation of SRSCM research with the means of bibliometric methods (see next section), we are able to identify the most important/influential publications and their interrelationships. The results of our analyses provide not only thematic trends in these publications but also present those papers which have impacted research in the field significantly (see e.g. White and McCain 1998).
2 Methodology For performing citation and co-citation analysis, we apply the software tools HistCite 12.03.17 (Garfield 2009) and VOSviewer 1.6.11 (van Eck and Waltman 2010). Both software packages analyse and visualise bibliometric results which help to identify the most important work on a topic as well as their timely development. We collected data from the Web of Science Core Collection and gathered all academic journal articles that were published between 1955 and 2018 with the following search string: TS = ((green*OR environ*OR social*OR sustain*) AND retail*) in any of the title, abstract, or author-supplied keyword fields, which resulted with altogether 1393 articles. After a qualitative check of the sample articles (including keyword analysis and abstract check) we deducted articles which were marginally connected with the topic of interest. So, our search resulted in 708 records from 284 journals. This final sample represents more than 2,000 authors and over 38,000 citations with more than 12,000 citation links.
Sustainable Retail Supply Chain Management – A Bibliometric Viewpoint
217
3 Results and Discussion Figure 1 shows the citation relations of the 25 most cited papers in our sample based on the local citation score (LCS) in a chronological manner.
Fig. 1. Citation relations of the 25 most cited SRSCM papers as indicated by HistCite (for the circle number see Appendix 1)
The publication date of these papers spans from 2005 to 2017. The citation relations include 25 nodes and (only) 11 links with a citations’ strength range between 5 to 16. The pattern of the citation relationships shows that 14 out of these 25 papers have no relation at all. There is one larger citation cluster including seven papers and two citation dyads. The large citation cluster evolves over the papers written by Dong et al. (2016, 390) about sustainability investments, Choi (2013, 196) on local sourcing and fashion quick response system, Choi and Chiu (2012, 134) on implication of newsvendor models for sustainable fashion retailing, Shen (2014, 253) on sustainable fashion supply chains and on the perception of fashion sustainability in online community (Shen et al. 2014, 214), Caniato et al. (2012, 136) on environmental sustainability in fashion supply chain, while the two citation dyads are connecting works done by Kolk et al. (2010) and Björklund et al. (2016, 79) which are more evolved and connected with green logistics and Chkanikova (2015, 350) and Hatanaka et al. (2005, 334), papers about food labeling and third party certification. Figures 2 and 3 show the results for our co-citation analyses for articles as well as for journals.
218
K. Petljak and H. Kotzab
Fig. 2. Visualisation of identified co-citation article clusters (based on VosViewer)
Based on the bibliographic analysis conducted in the VosViewer, we can conclude that the field of sustainable retail supply chain management in terms of article cocitation can be organized into four different clusters as presented in Table 1. Table 1. 32 most co-cited SRSCM papers (alphabetical order) Red Cluster 1: Sustainable supply chain management
Green Cluster 2: Marketing management and metrics quality papers
Blue Cluster 3: Operations management Yellow Cluster 4: Food waste management
Carter and Rogers (2008), de Brito et al. (2008), Eisenhardt (1989), Kleindorfer et al. (2005), Linton et al. (2007), Porter (2006), Rao and Holt (2005), Seuring and Müller (2008), Srivastava (2007), Vachon and Klassen (2008), Yin (2003) Ajzen (1991), Anderson and Gerbing (1988), Carrigan et al. (2004), Fornell and Larcker (1981), Hu and Bentler (1999), Laroche et al. (2011), Nunally (1978), Podsakoff et al. (2003), Roheim et al. (2011), Vermeir and Verbeke (2006) Benjaafar et al. (2013), Dong et al. (2016), Hua (2011), Liu (2012), Savasakan et al. (2006), Savaskan et al. (2004), Swami and Shah (2013) Barney and Wright (1998), Gustavsson et al. (2011), Parfitt et al. (2010)
The green Cluster ‘Sustainable Food/Marketing/Retail/Consumer’ is dominated by the research articles from the marketing and psychology field, predominately empirical articles which were then used for the explanations of the methodological part of the research paper. The yellow cluster is represented by the food waste research stream. It looks the smallest, but it is gaining momentum. The third cluster, represented in red, ‘Supply Chain/Operations/Logistics/ Management-oriented’ is the core of SRSCM and includes topics around sustainable supply chain management, green supply chain management, green retail, green consumer behavior, green logistics and green retail operations in general, from the assortment level to the store level and collaborations in managing sustainable supply chain. The final cluster in blue presents the influence of management theory and is ‘OR/Production/Transport-oriented’.
Sustainable Retail Supply Chain Management – A Bibliometric Viewpoint
219
Figure 3 shows a visualization of the identified co-citation journal clusters, which structure is similar to the previously described co-citation article clusters.
Fig. 3. Visualisation of identified co-citation journal clusters (based on VosViewer)
We see substantial dominance of jointly using marketing, food and retail related journals and sustainability journals in general (see Table 2). What is interesting to note is the high-quality level of the journals, so that we may conclude that the knowledge base of SRSCM takes its knowledge from the best research outlets, so its intellectual foundation is based on solid ground. Table 2. 30 most co-cited SRSCM journals (alphabetical order) 30 most co-cited journals (alphabetical order) Green Cluster ‘Sustainable British Food Journal; Business Strategy and the Food/Marketing/Retail/Consumer’ Environment; Ecological Economics; Food Policy; International Journal of Consumer Studies; International Journal of Life Cycle Assessment; International Journal of Retail and Distribution Management; Journal of Business Ethics; Journal of Business Research; Journal of Cleaner Production; Journal of Consumer Research; Journal of Marketing; Journal of Retailing; Resources, Conservation and Recycling (continued)
220
K. Petljak and H. Kotzab Table 2. (continued)
30 most co-cited journals (alphabetical order) Red Cluster ‘Supply Academy of Management Journal; Academy of Chain/Operations/Logistics/ManagementManagement Review; International Journal of oriented’ Production and Operations Management; International Journal of Physical Distribution and Logistics Management; Journal of Operations Management; Journal of Supply Chain Management; Strategic Management Journal; Supply Chain Management: An International Journal Blue Cluster ‘OR/Production/TransportEuropean Journal of Operational Research; oriented’ International Journal of Production Economics; International Journal of Production Research; Management Science; Production and Operations Management; Sustainability; Transportation Research Part E: Logistics and Transportation Review
4 Conclusions The results of our study show that SRSCM is embedded in a number of various sub disciplines including sustainability, retail, logistics, operations and supply chain management but also consumer research. SRSCM research seems to be still in an early stage of a research life cycle as many cited references are related to methodological issues in regards to literature reviews. The research domain is empirically founded, which is documented by the many sources referring to various empirical methodological approaches (both qualitative case studies and quantitative survey studies). The citation clusters themselves can be seen as representative foundations for specific research streams. Thereby we were able identify a stream of food waste that seeks its theoretical base in the resource-based view. The results of the journal co-citation analysis show that the most frequently used journals are highly reputable. Further we see a definition base that can be used for identifying the correct terminology in the area of green, environmental and sustainable operations and supply chain management. Another stream deals with green consumer behavior and is characterized by end-user price formation issues for environmental-friendly products. And the more OR-specific citation cluster represents closed-loop supply chain stream in combination with carbon footprint and emission trading references. So far, a specific SRSCM paper has not been yet identified amongst the most cited papers.
Sustainable Retail Supply Chain Management – A Bibliometric Viewpoint
221
Appendix 1 Top 25 Citations as identified by HistCite1
References Ajzen, I.: The theory of planned behavior. Organ. Behav. Hum. Decis. Process. 50(2), 179–211 (1991) Anderson, J.C., Gerbing, D.W.: Structural equation modeling in practice: a review and recommended two-step approach. Psychol. Bull. 103(3), 411–423 (1988) Arnold, S.J.: Lessons learned from the world’s best retailers. Int. J. Retail. Distrib. Manag. 30 (11/12), 562–570 (2002) Bălan, C.: Carbon-footprint policy of the top ten global retailers: contribution to sustainable development. Amfiteatru Econ. J. 12(27), 52–65 (2009) Barney, J.B., Wright, P.M.: On becoming a strategic partner: the role of human resources in gaining competitive advantage. Hum. Resour. Manag. 37, 31–46 (1998) Benjaafar, S., Li, Y., Daskin, M.: Carbon footprint and the management of supply chains: insights from simple models. IEEE Trans. Autom. Sci. Eng. 10(1), 99–116 (2013) Biloslavo, R., Trnavčevič, A.: Web sites as tools of communication of a “green” company. Manag. Decis. 47(7), 1158–1173 (2009) Binnekamp, M., Ingenbleek, P.: Do “good” food products make others look “bad”? Br. Food J. 110(9), 843–864 (2008)
1
Underlined numbers indicate circle numbers in Fig. 1; LCS = Local citation score; GCS = Global citation score.
222
K. Petljak and H. Kotzab
Björklund, M., Forslund, H., Isaksson, M.P.: Exploring logistics-related environmental sustainability in large retailers. Int. J. Retail. Distrib. Manag. 44(1), 38–57 (2016) Browne, A.W., Harris, P.J.C., Hofny-Collins, A.H., Pasiecznik, N., Wallace, R.R.: Organic production and ethical trade: definition, practice and links. Food Policy 25(1), 69–89 (2000) Caniato, F., Caridi, M., Crippa, L., Moretto, A.: Environmental sustainability in fashion supply chains: an exploratory case based research. Int. J. Prod. Econ. 135(2), 659–670 (2012) Carlsson-Kanyama, A.: Food consumption patterns and their influence on climate change: greenhouse gas emissions in the life-cycle of tomatoes and carrots consumed in Sweden. Ambio 27(7), 528–534 (1998) Carrigan, M., de Pelsmacker, P.: Will ethical consumers sustain their values in the global credit crunch? Int. Mark. Rev. 26(6), 674–687 (2009) Carrigan, M., Szmigin, I., Wright, J.: Shopping for a better world? An interpretive study of the potential for ethical consumption within the older market. J. Consum. Mark. 21(6), 401–417 (2004) Carter, C., Rogers, D.: A framework of sustainable supply chain management: moving toward new theory. Int. J. Phys. Distrib. Logist. Manag. 38(5), 360–387 (2008) Chkanikova, O.: Sustainable purchasing in food retailing: interorganizational relationship management to green product supply. Bus. Strategy Environ. 25(7), 478–494 (2016). https:// doi.org/10.1002/bse.1877 Choi, T.-M., Chiu, C.-H.: Mean-downside-risk and mean-variance newsvendor models: implications for sustainable fashion retailing. Int. J. Prod. Econ. 135(2), 552–560 (2012) Choi, T.-M.: Local sourcing and fashion quick response system: the impacts of carbon footprint tax. Transp. Res. Part E: Logist. Transp. Rev. 55, 43–54 (2013) de Brito, M., Carbone, V., Blanquart, C.M.: Towards a sustainable fashion retail supply chain in Europe: organisation and performance. Int. J. Prod. Econ. 114(2), 534–553 (2008) de Snoo, G.R., van de Ven, G.W.J.: Environmental themes on ecolabels. Landsc. Urban Plan. 46 (1–3), 179–184 (1999) Dong, C., Shen, B., Chow, P.-S., Yang, L., Tong, C.: Sustainability investment under cap-andtrade regulation. Ann. Oper. Res. 240(2), 509–531 (2016) Durach, C.F., Kembro, J., Wieland, A.: A new paradigm for systematic literature reviews in supply chain management. J. Supply Chain Manag. 53, 67–85 (2017) Eisenhardt, K.: Building theories from case study research. Acad. Manag. Rev. 14(4), 532–550 (1989) Erol, I., Cakar, N., Erel, D., Sari, R.: Sustainability in the Turkish retailing industry. Sustain. Dev. 17, 49–67 (2009) Fornell, C., Larcker, D.: Evaluating structural equation models with unobservable variables and measurement error. J. Mark. Res. 18(1), 39–50 (1981) Garfield, E.: From the science of science to Scientometrics visualizing the history of science with HistCite software. J. Inf. 3, 173–179 (2009) Gustavsson, J., Cederberg, C., Sonesson, U., van Otterdijk, R., Meybeck, A.: Global food losses and food waste: extent, causes and prevention, p. 29. Food and Agriculture Organisation of the United Nations (FAO), Rome (2011) Hartmann, M.: Corporate social responsibility in the food sector. Eur. Rev. Agric. Econ. 38(3), 297–324 (2011) Hatanaka, M., Bain, C., Busch, L.: Third-party certification in the global agrifood system. Food Policy 30(3), 354–369 (2005) Hingley, M., Lindgreen, A., Reast, J., Forsman-Hugg, S., Katajajuuri, J., Riipi, I., Mäkelä, J., Järvelä, K., Timonen, P.: Key CSR dimensions for the food chain. Br. Food J. 115(1), 30–47 (2013)
Sustainable Retail Supply Chain Management – A Bibliometric Viewpoint
223
Hu, L.-T., Bentler, P.M.: Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct. Equ. Model. 6(1), 1–55 (1999) Jarosz, L.: The city in the country: growing alternative food networks in Metropolitan areas. J. Rural Stud. 24, 231–244 (2008) Jones, P., Comfort, D., Hillier, D.: Sustainability in the global shop window. Int. J. Retail. Distrib. Manag. 39(4), 256–271 (2011) Jones, P., Comfort, D., Hillier, D.: Sustainable development and the UK’s major retailers. Geography 92(1), 41–47 (2007) Kaiser, M.J., Edwards-Jones, G.: The role of ecolabeling in fisheries management and conservation. Conserv. Biol. 20, 392–398 (2006) Kleindorfer, P.R., Singhal, K., Wassenhove, L.N.: Sustainable operations management. Prod. Oper. Manag. 14, 482–492 (2005) Kolk, A., Hong, P., van Dolen, W.: Corporate social responsibility in China: an analysis of domestic and foreign retailers’ sustainability dimensions. Bus. Strategy Environ. 19, 289–303 (2010) Kotzab, H., Munch, H., de Faultrier, B., Teller, C.: Environmental retail supply chains: when global Goliaths become environmental Davids. Int. J. Retail. Distrib. Manag. 39(9), 658–681 (2011) Laroche, M., Bergeron, J., Barbaro-Forleo, G.: Targeting consumers who are willing to pay more for environmentally friendly products. J. Consum. Mark. 18(6), 503–520 (2011) Linton, J.D., Klassen, R., Jayaraman, V.: Sustainable supply chains: an introduction. J. Oper. Manage. 25(6), 1075–1082 (2007) Maloni, M.J., Brown, M.E.: Corporate social responsibility in the supply chain: an application in the food industry. J. Bus. Ethics 68(1), 35–52 (2006) Morgan, J., Olynk Widmar, N., Wilcoxc, M., Croney, C.: Perceptions of agriculture and food corporate social responsibility. J. Food Prod. Mark. 24(2), 146–162 (2018) Nunnally, J.C.: Psychometric Theory, 2nd edn. McGraw-Hill, New York (1978) Ogle, J.P., Hyllegard, K.H., Dunbar, B.H.: Predicting patronage behaviors in a sustainable retail environment: adding retail characteristics and consumer lifestyle orientation to the beliefattitude-behavior intention model. Environ. Behav. 36(5), 717–741 (2004) Parfitt, J., Barthel, M., Macnaughton, S.: Food waste within food supply chains: quantification and potential for change to 2050. Philos. Trans. R. Soc. B: Biol. Sci. 365(1554), 3065–3081 (2010) Petljak, K., Zulauf, K., Štulec, I., Seuring, S., Wagner, R.: Green supply chain management in food retailing: survey-based evidence in Croatia. Supply Chain Manag.: Int. J. 23(1), 1–15 (2018) Podsakoff, P.M., MacKenzie, S.B., Lee, J., Podsakoff, N.P.: Common method biases in behavioral research: a critical review of the literature and recommended remedies. J. Appl. Psychol. 88(5), 879–903 (2003) Porter, M.E., Kramer, M.R.: The link between competitive advantage and corporate social responsibility. Harvard Bus. Rev. 84(12), 78–92 (2006) Rao, P., Holt, D.: Do green supply chains lead to competitiveness and economic performance? Int. J. Oper. Prod. Manag. 25(9), 898–916 (2005) Reynolds, N., Fischer, C., Hartmann, M.: Determinants of sustainable business relationships in selected German agri-food chains. Br. Food J. 111(8), 776–793 (2009) Roheim, C.A., Asche, F., Santos, J.I.: The elusive price premium for ecolabelled products: evidence from seafood in the UK market. J. Agric. Econ. 62, 655–668 (2011) Savaskan, R.C., Bhattacharya, S., Van Wassenhove, L.N.: Closed-loop supply chain models with product remanufacturing. Manage. Sci. 50(2), 239–252 (2004)
224
K. Petljak and H. Kotzab
Savaskan, R.C., Van Wassenhove, L.N.: Reverse channel design: the case of competing retailers. Manage. Sci. 52(1), 1–14 (2006) Seuring, S., Müller, M.: From a literature review to a conceptual framework for sustainable supply chain management. J. Clean. Prod. 16(15), 1699–1710 (2008) Seuring, S., Gold, S.: Conducting content-analysis based literature reviews in supply chain management. Supply Chain Manag.: Int. J. 17(5), 544–555 (2012) Shen, B., Zheng, J.-H., Chow, P.-S., Chow, K.-Y.: Perception of fashion sustainability in online community. J. Text. Inst. 105(9), 971–979 (2014) Shen, B.: Sustainable fashion supply chain: lessons from H&M. Sustainability 6, 6236–6249 (2014) Souza-Monteiro, D., Hooker, N.: Comparing UK food retailers corporate social responsibility strategies. Br. Food J. 119(3), 658–675 (2017) Srivastava, S.K.: Green supply-chain management: a state-of-the-art literature review. Int. J. Manag. Rev. 9, 53–80 (2007) Swami, S., Shah, J.: Channel coordination in green supply chain management. J. Oper. Res. Soc. 64(3), 336–351 (2013) Vachon, S., Klassen, R.D.: Environmental management and manufacturing performance: the role of collaboration in the supply chain. Int. J. Prod. Econ. 111, 299–315 (2008) Van Eck, N.J., Waltman, L.: Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics 84(2), 523–538 (2010) Vermeir, I., Verbeke, W.: Sustainable food consumption: exploring the consumer “attitude– behavioral intention” gap. J. Agric. Environ. Ethics 19(2), 169–194 (2006) White, H.D., McCain, K.W.: Visualizing a discipline: an author co-citation analysis of information science, 1972–1995. J. Am. Soc. Inf. Sci. 49, 327–355 (1998) Wiese, A., Zielke, S., Toporowski, W.: Sustainability in retailing – research streams and emerging trends. Int. J. Retail. Distrib. Manag. 43(4/5) (2015) Yin, R.K.: Case Study Research: Design and Methods, 3rd edn. Sage Publications, Thousand Oaks (2003)
Part III: Distributed and Collaborative Planning and Control
Autonomous Production Control Methods - Job Shop Simulations Ziqi Zhao1 , Oliver Antons2(B) , and Julia C. Arlinghaus3 1
2
Research Center for Modern Logistics, Graduate School at Shenzhen, Tsinghua University, Shenzhen 518055, China Chair of Management Science, RWTH Aachen University, 52072 Aachen, Germany [email protected] 3 Chair of Production Systems and Automation, Otto-von-Guericke University Magdeburg, 39106 Magdeburg, Germany
Abstract. With the development of Industry 4.0 and the Internet of Things, autonomous production control is regarded as a feasible and promising approach for meeting the increasing challenges of complexity and flexibility. To implement autonomous production control methods in practice, a deeper understanding of their characteristics is necessary. This research provides a comparative perspective on existing methods. We study selected autonomous production control methods under various scenarios, and derive insights for the design of such systems in industrial practice. Keywords: Autonomous production control · Autonomous control Industry 4.0 · Production planning and control
1
·
Introduction
Confronted with the challenge of high complexity and volatility, manufacturers desire a higher degree of agility and flexibility in production planning to stay competitive. Autonomous production control appears to be a promising approach, as it enables logistic objects to process information and to render and execute decisions on their own [22]. It is able to handle dynamic and complex production circumstances by distributed and flexible coping of complexity. Through the development of Cyber-Physical Production Systems (CPS), 5G and Internet of things (IOT), autonomous production control has increasing potential and practical significance. Several autonomous production control methods have been introduced in last 20 years. Previous studies have shown that in certain settings autonomous production control can achieve logistics targets better than conventional production planning and control approaches [11]. Many open questions regarding the characteristics of autonomous production methods remain, and knowledge on these can significantly ease their implementation in practice. In this paper, we review the performance of selected autonomous c The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 227–235, 2020. https://doi.org/10.1007/978-3-030-44783-0_22
228
Z. Zhao et al.
production control methods through simulation, evaluating their performance characteristics. The remainder of this paper is structured as follows. In the next section, we provide an overview of selected autonomous production control methods. Subsequently, we describe the deployed methodology as well as our simulation results. The last section gives a conclusion on our main results as well as avenues for future research.
2 2.1
State of the Art Autonomous Production Control
Production planning and control is an essential tool for any manufacturer. It addresses the allocation of resources to jobs and the subsequent creation of a schedule. In order to measure the performance of PPC, several indicators are used such as throughput time (TPT), work in progress (WIP), delay rate and utilization. Traditionally, a production schedule is created by a central planning authority. Autonomous production control however is based on a different approach: Every entity within the system, i.e. machines, resources, products, is equipped with a certain degree of intelligence. Through addition of computational power and connectivity combined with either a distributed control approach or a centralized control entity, each entity is able to monitor its environment and coordinate with other entities with the manufacturing system. This approach enables the system to be more agile and react quickly to any disturbance at its source. Thus, the system exhibits a high degree of flexibility, and is able to continuously adapt its schedule on a machine level [2]. 2.2
Existing Autonomous Production Control Methods
Windt et al. divided autonomous production control methods into three approaches: rational, bounded rational, and combined strategies, which were derived from behavioral economics [20]. Examples of rational methods are QLE and DLRP. In these methods, objects exchange relevant information and decide according to future system states anticipation [4]. Biologically methods, such as Ant, PHE, Bee Foraging and Chemotaxis, belong to rational methods and use aggregated data from past events [12]. Scholz-Reiter et al. classified autonomous production control methods into two categories: local information methods and information discovery methods [12]. Local information methods gather and process only local information, such as QLE and PHE. Information discovery methods can collect relevant information from other objects, but not cover the whole system, such as DLRP. 2.3
Research Gap
There have already been studies on autonomous production control methods, comparing different approaches. For example, Scholz-Reiter et al. conducted a
Autonomous Production Control Methods
229
simulation study in 2009, but only compared three different autonomous control methods, namely QLE, DUE and PHE [11]. Windt et al. compared the performance of autonomous production control methods in two scenarios (with and without machine failure) by simulation [21]. As far as we know, the latest research in this field was conducted in 2011 by Becker et al. comparing six methods in 4 simulation scenarios: standard, full flexibility (suspend processing sequence to increase decision alternatives), increased load (increase processing time by 10%), both of full flexibility and increased load [2]. Table 1. Autonomous production control methods Methods
c.f Year Key idea
Holonic manufacturing
[6] 1996 Machines bid to get jobs and get punished for delays
Market based
[17] 2000 Parts carry a shopping list of work needed to be done, parts auction for access to the machines
Ant
[3] 2001 Ants choose machines based on pheromone concentration
Pheromone based approach (PHE)
[1] 2006 Average throughput time is used as a pheromone
Due date method
[14] 2007 QLE and choose the most urgent due date in queue
Distributed logistics routing protocol (DLRP)
[19] 2007 Machines communicate best routes
Queue length estimator [13] 2007 Compares estimated waiting time at buffers (QLE) Bee foraging
[13] 2008 Based on the routes of previous parts
AMS-SCA
[9] 2012 Based on a swarm of cognitive and adaptive agents
Potential field (PF)
[8] 2012 The state of potential field depends on the attractiveness of the resource providing the service
Pheromone based coordination (PBC)
[18] 2012 The pheromone quantum of manufacturing cell is calculated inversely proportional to the cost, which guarantees a minimal cost to process the orders
Sudo
[16] 2013 A part agent chooses a machine, by the length of a job list and the conveyance cost
Integrated APC
[5] 2017 An integrated method considering order release, sequencing and capacity control to meet due date
Direct workload (DWL)
[4] 2018 Jobs are allocated only to the valid machine with the lowest workload
230
Z. Zhao et al.
We expand on this by evaluating newly developed methods after 2011 in different simulation scenarios.
3 3.1
Methodology Simulation Setup
The main simulation scenario is a make-to-order shop floor with five production stages. Within each stage, 6 functional equivalent machines are available as also reviewed by Schipper et al. [10]. These machines are denoted by 1, . . . , 30. However, these machines behaviour is not identical: Their processing times vary slightly, and their production cost is adjusted accordingly. This structure is visualized in Fig. 1.
Fig. 1. Flexible network setup
In order to create a simulation that allows us to compare the aforementioned methods, we work with the following seven assumptions: 1. Each product needs to be processed in every stage, in ascending order. 2. The arrival of jobs follows a Poisson distribution with parameter λ, and the processing time of the machines follows an exponential distribution with parameter μi , i ∈ {1, . . . , 30}. 3. The lot size per arrival is assumed to be one, which equals the capacity of all the machines. Machines process parts according to the first-come-first-served principle. 4. No limit on queue length. 5. Besides the machines, other resources are available at all times. 6. Setup times are included in the processing time, and transportation times between machines are negligible. 7. The cost of each processing process can be expressed as ci , i ∈ {1, . . . , 30}. The parameters for the six machines within a stage are not equal, μi is set at the start for all simulation runs by adding a random offset to the base value for the machines of that stage. This setup defines our standard scenario. We extend this setup by considering the possibility of machine breakdown. This results in two scenarios, for which the selected autonomous production control methods will be compared.
Autonomous Production Control Methods
3.2
231
Selected Autonomous Production Control Methods
As a baseline for the simulation, we included the classical autonomous production control method of QLE. Furthermore, four new methods developed after 2010 were chosen: AMSSCA, PF, DWL and PBC, as they have not been reviewed in the previously mentioned studies. Additionally, they fulfill further criteria such as feasible implementation and comparability as well as a similar scope. Consequently, we compare the performance of methods for local decision making on dispatching, similar to most approaches [5]. We do not consider the autonomous production control method of Grundstein et al. as its scope is much wider in comparison, including dispatching, queue processing and capacity control [5]. In the remainder of this section we introduce each chosen method briefly. The QLE method compares the full queue length of all viable processing paths, and chooses the shortest one. The queue length of a machine is given by the estimated total operation time of all parts within the queue of the machine. As such, this method uses expected information, and tries to minimize the corresponding expected waiting time. The AMS-SCA method chooses the optimal machine based on a pheromone markers value among the available machines. The pheromone value pi calculation considers the executing ability, processing time and machining cost of the corresponding machine i. pi =
q , q ∈ {0, 1} , αt + αc = 1 αt ∗ Mti /Mt0 i + αc ∗ Mci /Mc0 i
q denotes the executing ability of the machine i regarding the requested task. If a machine does not meet the requirements of the task, q = 0, and the pheromone is 0. Otherwise, q = 1. Mti and Mci represent the total time and machining cost of the task t at the machine i, respectively. Mt0 i and Mc0 i are the minimum total time and machining cost of the task t respectively in ideal situation. The factors αt and αc are the weight of the machining time and cost respectively. The weight can be changed to meet different goals of the company [9]. The key idea of the DWL method is to balance the workload of machines. The DWL method extends the QLE method. As an important difference, each machine has a workload limit, and a job is only allocated to a machine if the resulting workload of this allocation is within this limit. The workload considers expecting processing time of both jobs in the queue and the jobs being currently processed at the machine [4]. The PF method chooses the optimal machine by its attractiveness among the alternative machines. If a machine is broken now, its attractiveness is 0. 1 , where Otherwise, its attractiveness equals to ai = (1+W aitingT ime)∗Distance W aitingT ime refers to the expected waiting time if the job is assigned to machine i. Distance refers to the distance between the current position of the job and the machine i. This method aims to minimizes the throughput time by reducing both travel and waiting time. In the PBC method, the pheromone value is calculated inversely proportional to the cost. This cost consists of processing cost, storage cost and tardiness cost.
232
Z. Zhao et al.
As such, this method takes due dates into consideration. It calculates a tardiness cost per machine, which is proportional to the length of the resulting due date delay.
4
Results
An overview of the performance of all methods in the standard scenario is given in Fig. 2. The mean utilization of all methods, visualised in Fig. 2(a) shows how close the studied methods are. With an arrival rate parameter of 3, the utilization
(a) mean utilization
(b) mean processing time
(c) mean waiting time
(d) delay rate
(e) WIP
(f) TPT standard deviation
Fig. 2. Performance indicators for the standard scenario
Autonomous Production Control Methods
233
is above 90% for all methods, decreasing for all methods when more and more jobs arrive. Notably, this decrease is worse for the DWL method, which can be explained by this methods tendency not to assign jobs to the slowest machines. In respect to the mean processing time though, a notable better performance can be seen for the DWL method. Regarding the remaining performance measures such as mean waiting time, delay rate, work in progress in throughput time the methods perform quite similar, with PBC often falling slightly behind. This is most likely due to the fact that it aims to pursue the lowest cost while remaining within the due date, hence it prefers the slow machines. This also results in the biggest WIP and delay rate, c.f. Figs. 2(e) and (d). Taking the ranking for the studied metrics into a weighted measure of performance, the following method ranking emerges: DWL > PF > AMS-SCA > SCA > QLE > PBC. Lastly, Fig. 3 gives a comprehensive overview on the relative performance characteristics of each method in the standard scenario (Fig. 3(a)) as well as when introducing machine failure (Fig. 3(b)). For each performance dimension, the best performings method is used as reference point with a relative performance of one. Notably, these methods react differently to the introduction of machine failure. It highlights the superiority of DWL in time dimensions, the superiority of PBC regarding costs, as well as the acceptable performance of AMS-SCA in both areas. Furthermore, the high flexibility of PF in high workload scenarios with machines breakdown becomes apparent.
(a) standard scenario: λ = 3.5/ without (b) breakdown: λ = 3.5/ failure rate=10−5 / fixed process breakdown/ fixed process
Fig. 3. Overview on relative performance indicators
5
Conclusion
This paper gives an overview of four recently developed autonomous production control method. From a comparison of these methods in a job shop scenario we identify key characteristics and performance capabilities. Subsequently, these
234
Z. Zhao et al.
results can aid in a decision process for the application of autonomous control methods, depending on desired performance dimensions as well as time and cost constraints. This simulation however only covers a glimpse of possible applications for autonomous production control methods. For example, these methods can be used not only individually on the shop floor, but also coupled with central planning [15] or in combination, allowing online-switching with each other in order to adapt to different situations [7]. Furthermore, the dependence of autonomous production control performance on the underlying production network remains unclear. Also, the influence of the production networks size (i.e. number of machines, etc.) entices further research.
References 1. Armbruster, D., et al.: Autonomous control of production networks using a pheromone approach. Phys. A Stat. Mech. Appl. 363(1), 104–114 (2006) 2. Becker, T., Windt, K.: A comparative view on existing autonomous control approaches: observations from a simulation study. In: H¨ ulsmann, M., Scholz-Reiter, B., Windt, K. (eds.) Autonomous Cooperation and Control in Logistics, pp. 275– 289. Springer, Heidelberg (2011) 3. Cicirello, V.A., Smith, S.F.: Ant colony control for autonomous decentralized shop floor routing. In: Proceedings 5th International Symposium on Autonomous Decentralized Systems, pp. 383–390. IEEE (2001) 4. Fernandes, N.O., Martins, T., Carmo-Silva, S.: Improving materials flow through autonomous production control. J. Ind. Prod. Eng. 35(5), 319–327 (2018). https:// doi.org/10.1080/21681015.2018.1479895 5. Grundstein, S., Freitag, M., Scholz-Reiter, B.: A new method for autonomous control of complex job shops - integrating order release, sequencing and capacity control to meet due dates. J. Manuf. Syst. 42, 11–28 (2017). https://doi.org/10.1016/ j.jmsy.2016.10.006 6. M´ arkus, A., Vancza, T.K., Monostori, L.: A market approach to holonic manufacturing. CIRP Ann. 45(1), 433–436 (1996) 7. Mueller, D., et al.: Complexity-oriented evaluation of production systems for online-switching of autonomous control methods. In: Clausen, U., Langkau, S., Kreuz, F. (eds.) Advances in Production, Logistics and Traffic. Lecture Notes in Logistics, pp. 246–264. Springer, Cham (2019). https://doi.org/10.1007/978-3-03013535-5 18 8. Pach, C., et al.: An effective potential field approach to FMS holonic heterarchical control. Control Eng. Pract. 20(12), 1293–1309 (2012). https://doi.org/10.1016/j. conengprac.2012.07.005 9. Park, H.-S., Tran, N.-H.: An autonomous manufacturing system based on swarm of cognitive agents. J. Manuf. Syst. 31(3), 337–348 (2012). https://doi.org/10.1016/ j.jmsy.2012.05.002 10. Schipper, M.A., Chankov, S.M., Bendul, J.: Synchronization emergence and its effect on performance in queueing systems. Proc. CIRP 52, 90–95 (2016). https:// doi.org/10.1016/j.procir.2016.07.016 11. Scholz-Reiter, B., G¨ orges, M., Philipp, T.: Autonomously controlled production systems—influence of autonomous control level on logistic performance. CIRP Ann. 58(1), 395–398 (2009). https://doi.org/10.1016/j.cirp.2009.03.011
Autonomous Production Control Methods
235
12. Scholz-Reiter, B., Rekersbrink, H., G¨ orges, M.: Dynamic flexible flow shop problems–scheduling heuristics vs. autonomous control. CIRP Ann. 59(1), 465– 468 (2010). https://doi.org/10.1016/j.cirp.2010.03.030 13. Scholz-Reiter, B., Jagalski, T., Bendul, J.C.: Autonomous control of a shop floor based on bee’s foraging behaviour. In: Kreowski, H.J., Scholz-Reiter, B., Haasis, H.D. (eds.) Dynamic in Logistics, pp. 415–423. Springer, Heidelberg (2008) 14. Scholz-Reiter, B., et al.: Analysing the dynamics caused by autonomously controlled logistic objects. In: Proceedings of the 2nd International Conference Changeable, Agile Reconfigurable and Virtual Production-CARV 2007. Citeseer (2007) 15. Schukraft, S., et al.: Strategies for the coupling of autonomous control and central planning: evaluation of strategies using logistic objectives achievement and planning adherence. In: 2015 IEEE International Conference on Systems, Man, and Cybernetics, pp. 121–126. IEEE (2015) 16. Sudo, Y., Matsuda, M.: Agent based manufacturing simulation for effcient assembly operations. Proc. CIRP 7, 437–442 (2013) 17. Vollmer, L.: Agentenbasiertes Auftragsmanagement mit Hilfe von PreisLiefertermin-Relationen. VDI-Verlag (2000) 18. Wang, L., et al.: Pheromone-based coordination for manufacturing system control. J. Intell. Manuf. 23(3), 747–757 (2012) 19. Wenning, B.L., et al.: Autonomous control by means of distributed routing. In: H¨ ulsmann, M., Windt, K. (eds.) Understanding Autonomous Cooperation and Control in Logistics, pp. 325–335. Springer, Heidelberg (2007) 20. Windt, K., Becker, T.: Applying autonomous control methods in different logistic processes-a comparison by using an autonomous control application matrix. In: 2009 17th Mediterranean Conference on Control and Automation, pp. 1451–1455. IEEE (2009) 21. Windt, K., et al.: A classification pattern for autonomous control methods in logistics. Logist. Res. 2(2), 109–120 (2010). https://doi.org/10.1007/s12159-010-0030-9 22. Windt, k., et al.: A generic implementation approach of autonomous control methods in production logistics. In: IEEE ICCA 2010, pp. 629–633. IEEE (2010)
Individual Predictive Maintenance Approach for Diesel Engines in Rail Vehicles Hendrik Engbers1(&), Simon Leohold1, and Michael Freitag1,2 1
2
BIBA – Bremer Institut für Produktion und Logistik GmbH at the University of Bremen, Bremen, Germany {eng,leo,fre}@biba.uni-bremen.de Faculty of Production Engineering, University of Bremen, Bremen, Germany
Abstract. An essential requirement for climate-friendly and sustainable transport and logistics services is the cost improvement of rail freight services. Maintenance in rail freight is a major cost driver. Therefore, the goal is to reduce costs by digitalization and state-of-the-art maintenance approaches. In this paper, we present an approach for an individual predictive maintenance system for diesel engines of rail vehicles. It is a data-driven approach that leverages data characteristics from historical and current time series data from the Engine Control Unit (ECU). The proposed methodology applies a meta-learning technique to select the most suitable forecasting model for each engine component in order to determine the time of failure. The meta-learning technique allows the methodology to be applied to other engine series. As a result main fault classes of an engine type have been identified and the corresponding potential based on the analysis of historically corrective and preventive measures are presented. Further analysis shows that the lifetime of turbochargers and the injection system are insufficiently exploited. Keywords: Predictive Maintenance recommendation
Meta-learning Algorithm
1 Introduction 1.1
Motivation
As the German government is committed to providing climate-friendly and sustainable transport and logistics services, the distribution of transport volume is to be shifted significantly in favor of rail freight from 18.6% in 2017 up to 25.0% by 2030. An essential requirement for this is to improve the competitiveness of rail freight services [1]. Since maintenance is a major cost-driver there are attempts to reduce costs by digitization and further automation of processes [2]. Currently, maintenance of rail vehicle powertrains is predominantly done by preventive maintenance strategies. As part of these approaches, the probability of engine failures is reduced by regular, timebased maintenance [3]. However, due to this strategy components are replaced regardless of their actual condition. Consequently, even technically faultless engines are not available for the duration of the maintenance processes [4]. Particularly during shunting operations, the individual workload profiles of rail vehicles can vary © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 236–244, 2020. https://doi.org/10.1007/978-3-030-44783-0_23
Individual Predictive Maintenance Approach for Diesel Engines
237
considerably. Therefore, vehicles that are used exclusively for shunting show different characteristics in terms of component wear compared to others, that are also used in mainline service. For this reason, preventive replacement of engines or single-engine components after a pre-defined number of operating hours is not efficient. In addition, there are high expenses for personnel and required technical infrastructure for maintenance. Altogether savings potentials of up to 20% are realistic through the digitalization of maintenance in this area. Predictive Maintenance (PM) therefore has a major role to play in increasing the competitiveness of rail freight transport [3]. 1.2
Objective
The objective of this article is to provide the foundations for an individual predictive maintenance of diesel engines in rail vehicles. According to this approach, failures of engine components are predicted with sufficient lead time and accuracy, so that an optimized maintenance planning and execution with a low risk of failure is achieved. A particular challenge related to the presented use case is the development of a model that is not only suitable for the forecast of failures of a specific engine but is also able to adapt to each individual engine of a vehicle fleet [4]. The methodology to be developed is to select a suitable forecasting method automatically. This way, the methodology should be transferable to other engine series without additional expert effort. In the following section, relevant preliminary works and the demand for further research are presented. In Sect. 3, we describe our proposed solution approach. Section 4 presents the main failure classes and their economic potentials due to an individual PM strategy. Section 5 summarises and outlines the further proceeding.
2 State of the Art 2.1
Maintenance Strategies for Diesel Engines in Rail Freight
The goal of maintenance tasks is maintaining the intended condition of a product. That includes the prevention of incidents and the analysis of failure and malfunction behavior in order to improve the detectability of incidents [5]. Strategies for achieving maintenance targets can be subdivided into reactive and preventive strategies. Preventive strategies, as shown in Fig. 1 are further split into periodically preventive, condition-based and predictive strategies [6]. Reactive maintenance strategies neglect any maintenance measures or inspections. Repairs are only carried out in the event of failure. This leads to high consequential costs and downtimes. On the other hand, the goal of periodic preventive maintenance is to minimize disturbances and failures by preventive measures [6]. In rail freight, the latter approach has been put into practice by extensive regulations and schedules for vehicle maintenance. This means that maintenance measures are carried out after a defined interval of use, for example, after a certain number of operating hours [3]. However, the disadvantage of this strategy is that maintenance is carried out on schedule, regardless of the individual condition of the machinery. Therefore useful lifetimes are not fully utilized. In addition, the object to maintain the operation is not met during maintenance, although operation would still be possible without any technical restrictions [6].
238
H. Engbers et al.
Fig. 1. Maintenance strategies [6] adapted
Condition-based maintenance strategies try to further exploit the lifetime of a product by using diagnostic systems. If critical deviations of sensor values are registered, measures for maintenance are triggered. Predictive strategies try to identify potential faults and forecast their further progression so that an efficient maintenance planning and execution can take place [6, 7]. Due to increasing availability of sensor data, predictive maintenance approaches are gaining attention in railway industry [7]. A variety of techniques can be used to create failure forecast models. The following section compares essential approaches and points out the demand for meta-learning approaches. 2.2
Methods for Predicting the Remaining Useful Life for Preventive Maintenance
Preventive maintenance strategies are based on the remaining useful life (RUL) assumptions of individual components. Periodic preventive maintenance is therefore based on either laboratory analysis or field tests to determine a minimum lifespan under specified conditions. Condition-based maintenance aims to detect critical conditions with enough RUL for preventive maintenance, whereas in predictive maintenance an up-to-date forecast of the RUL is available throughout the entire lifecycle. Forecasts can be done via model-based, knowledge-based or data-driven methods or a combination of them [8–11]. Model-based methods use complex mathematical descriptions of the physical properties. The models are typically developed in the design phase of the product and demand a high level of expert knowledge. However, because the models still involve significant simplifications of the real-world behavior, large data sets and tests are necessary for validation. When a validated model is available, they typically outperform other methods regarding accuracy [10]. Knowledge-based maintenance methods use experience, computational intelligence (CI) and expert domain knowledge to pre-design rule sets for maintenance decision making. Data-driven approaches are useful when reliable sensor data is available, whereas a mathematical model is not. Machine learning algorithms are used to detect patterns in historic condition monitoring data to build behavior models for forecasting. Therefore, data-driven methods provide a good trade-off between complexity and precision [8–11].
Individual Predictive Maintenance Approach for Diesel Engines
239
Prior works for rail systems have mainly focused on other rail car components such as trucks and bearings. There have been some works on similar marine diesel engines such as Bocchetti et al. [12] using model-based predictive maintenance. Xu et al. [13] analyzed the performance of neural networks for time-series forecasting failures and reliabilities in automotive engine systems. Freitag et al. applied a dynamic maintenance approach to generators of an offshore wind farm [14]. The railcar engine itself to our knowledge has not been subject to predictive maintenance with data-driven approaches. 2.3
Meta-learning for Selecting and Parameterizing Appropriate Forecasting Models
In the field of data mining and machine learning, there is a differentiated understanding of the concept of meta-learning [15]. In this paper, meta-learning refers to a methodology for using knowledge from known problems to develop efficient models in new application areas [16]. A comprehensive overview on this topic can be found in [15]. Because of the large number of different algorithms and required parameters of forecasting models, the solution space is often too large to evaluate all algorithms based on trial and error principle [17]. Therefore meta-learning can be used as a method to select and optimize the most suitable forecasting model [15]. For example, Gomes et al. [18] used a hybrid learning approach to select the optimal parameters of a Support Vector Machine (SVM). Reif et al. [19] increased the accuracy and reduced the runtime of a genetic algorithm for selecting suitable parameters for an SVM, as a starting point of Particle Swarm Optimization. Meta-Learning was also successfully used in the model formulation for time series analysis [20–22]. Kück et al. [23] also used a neural network and performance assessments as meta-learners to select models for predicting time series. However, no approaches for the selection of algorithms and their parameters for the forecasting of engine failures were found in the literature.
3 Solution Approach To achieve the described objective, as shown in Fig. 2, a data-driven approach is pursued. Based on the knowledge of the operating behavior of engines and the corresponding data characteristics of the transmitted sensor values as well as the related performance evaluations of historical time series, the most suitable forecasting model is selected. The following paragraphs focus each on one essential aspect of designing a machine learning framework and its specific meanings for the proposed methodology. The framework aspects are based on the machine learning canvas from [24], which has proven to be a helpful focusing tool when handling machine learning tasks.
240
H. Engbers et al.
Fig. 2. Individual predictive maintenance approach
Sources of Data. The most important data source is the ECU, which controls and monitors the engines functions [25]. The ECU enables access to a large number of sensor values which can provide conclusive information about the current status of the system. In addition, an engine test bench, several test engines with faulty components and external sensors are available for the acquisition of further measured values that are not monitored by the ECU. This dataset is then used to develop data characteristics that are subsequently transferred to a meta database, as shown in Fig. 2. In addition, we use data from the field of vehicles with reported engine problems as supplementary training data. Model Decisions. The individual forecasting model is used to determine the time and probabilities of failures of major components of the engine. This way, it is possible to aggregate the health status of the engine and carry out an optimal maintenance planning. Forecasting Points and Methodology. A continuous monitoring of the sensor values and the transmission as a real-time data stream is desirable, as this would allow a quick intervention in the case of an urgent predicted failure. For technical reasons, however, this is not feasible. Therefore, it is the approach to limit the acquisition and analysis of operational data to certain periods to determine the current health status of the engine. Predominantly measured values will be taken in the period from the departure from the station until the target speed of the vehicle is reached. If there is a stable internet connection, the recorded sensor data is transmitted, and the respective state of health is aggregated. If several of these states are recorded over time, then a trend diagram, as shown in Fig. 3, results which is used to estimate the remaining useful life of a component.
Individual Predictive Maintenance Approach for Diesel Engines
241
Fig. 3. Remaining useful life estimation procedure [26]
Description of the Machine Learning Task. Machine learning techniques shall be used in two areas of the concept: For learning dependencies between the component health and ECU sensor data and learning the relationship between state of the engine and the eligibility of different prediction methods. The set of ECU data is filtered for relevant time series and summarized as features as input for the learning algorithm aiming to reduce training time and increase performance. To achieve accurate RUL predictions at any time, it is first necessary to have a detailed estimation of the components health status. The components health status can be represented by a health index (HI) from 1 being in best shape and 0 meaning the end of useful life. The mapping from raw sensor data of the ECU to a HI is a regression problem and can be solved with suitable regression techniques such as neural networks. The HI data needed for model training can be determined in a laboratory analysis of the components. Learning the eligibility of different prediction methods from engine meta-data is a multiclass classification problem and can be solved with classification techniques such as decision tree, neural network classifier or support vector machine. Features. Features make up the input data for the regression and classification models and are selected and extracted from the raw sensor data of the ECU. The main goal of feature selection and extraction for machine learning is a more efficient representation of the relevant information needed to estimate the model outcome. Good features can reduce necessary training time and increase performance. Feature selection describes the process of defining a smaller set of features from the original feature space. The goal is, to only keep the features which are essential to the models outcome. This can be achieved either manually by using domain knowledge or by analytically scoring the impact of each original feature to the outcome of the model. The advantage of feature selection is that the original features remain unchanged and therefore, its influence on the outcome stays transparent. On the downside is the complete loss of information
242
H. Engbers et al.
from discarded signals. If all raw signals have significant impact on the models outcome, feature selection might be inappropriate. Feature extraction is an aggregation or transformation of features to a smaller subset to keep all the relevant information while reducing dimensionality. If for example from two signals only the difference of them has an impact on the outcome than the difference can be used as a feature with the advantage that no relevant information is lost while reducing dimensionality. On the downside, the lost (for the outcome irrelevant) information during the transformation cannot be retrieved and therefore, the impact of a single original signal on the outcome is unclear. For most machine learning problems, feature selection and feature extraction are both being used sequentially, e.g., in a first step signals without a major impact on the outcome are being discarded and in a second step, the remaining features are being combined to further reduce dimensionality [27]. Feature selection and extraction techniques shall be used to identify sensors, which data can be used to predict the health index for each component as well as to identify how those signals correlate to build up the meta database to use this knowledge for setting a predictive maintenance tasks for similar components in similar engines.
4 Economic Potential Due to Predictive Maintenance of Diesel Engines in Rail Freight In the rail freight industry, the technical possibilities for the application of predictive maintenance strategies already exist due to the progressive digitalization and automation. However, the economic feasibility of this approach should first be assessed separately for individual engine components. For this reason, we first analyzed the cost structure for maintenance of engine components in order to select promising components for model development. Figure 4 shows the relative maintenance costs of the components of diesel engines used in shunting operations and mainline services.
Fig. 4. Relative maintenance costs by engine component
Turbochargers (50%) and the injection system (26%) cause the largest part of the expenditure for maintenance of the engine series under consideration. Figure 4 shows
Individual Predictive Maintenance Approach for Diesel Engines
243
the costs incurred in one year for preventive and corrective measures. An extensive analysis has shown that the proportionate material costs for preventive measures for turbochargers are 94% and for the injection system 81%. It can be deduced from the data that the very high proportion of preventive costs, e.g. the premature replacement of turbochargers, means that the useful life of the components is insufficiently exploited. Spreading the operating time through individual predictive maintenance, therefore, seems particularly promising for turbochargers and the injection system.
5 Conclusion We have presented a data-driven approach for the predictive maintenance of diesel engines in rail freight transport. We select the most suitable forecasting model for the respective engine component using a meta-learning procedure. The forecast values are used to derive the health status of the engine. For this purpose, suitable data characteristics and performance evaluation are combined in a meta-database. The experience gained this way is used to enable a high forecasting quality and to transfer the methodology to other engine types without further expert effort. In addition, the potential analysis has shown that especially the exhaust gas turbochargers and the injection system of the considered engine series are also economically promising for predictive maintenance. For this reason, we will focus on the turbocharger and the injection system in the further development of the methodology. Acknowledgements. The authors would like to thank the BAB – Bremer Aufbaubank for their support within the project IPM – Individual prediction of engine failures of rail vehicles (funding code FUE0611B).
References 1. Bundesministerium für Verkehr und digitale Infrastruktur (BMVI): Innovationsprogramm Logistik 2030 (2019) 2. Ahmad, R., Kamaruddin, S.: An overview of time-based and condition-based maintenance in industrial application. Comput. Ind. Eng. (2012). https://doi.org/10.1016/j.cie.2012.02.002 3. Bobsien, S.: Digitale Instandhaltungsstrategie für produktiven Güterverkehr (2018) 4. Reichel, J., Müller, G., Haeffs, J.: Betriebliche Instandhaltung. Springer, Heidelberg (2018) 5. DIN Deutsches Institut für Normung e.V.: DIN 31051, Grundlagen der Instandhaltung (2012) 6. Matyas, K.: Ganzheitliche Optimierung durch individuelle Instandhaltungsstrategien. In: Industrie Management, Jahrgang 18, Nr. 2, S. 13–16 (2002). ISSN 1434–1980 7. Roland Berger study: Rail supply digitization (2017) 8. Okoh, C., Roy, R., Mehnen, J., Redding, L.: Overview of remaining useful life prediction techniques in through-life engineering services. Procedia CIRP (2014). https://doi.org/10. 1016/j.procir.2014.02.006 9. Pecht, M., Kang, M. (eds.): Prognostics and health management of electronics. Fundamentals, machine learning, and internet of things. Wiley, Hoboken (2018)
244
H. Engbers et al.
10. Gugulothu, N., TV, V., Malhotra, P., Vig, L., Agarwal, P., Shroff, G.: Predicting remaining useful life using time series embeddings based on recurrent neural networks (2017). http:// arxiv.org/pdf/1709.01073v2 11. Sankavaram, C., Pattipati, B., Kodali, A., Pattipati, K., Azam, M., Kumar, S., Pecht, M.: In: 2009 IEEE International Conference on Automation Science and Engineering, CASE 2009, Bangalore, India, 22–25 August 2009. IEEE, Piscataway (2009) 12. Bocchetti, D., Giorgio, M., Guida, M., Pulcini, G.: A competing risk model for the reliability of cylinder liners in marine Diesel engines. Reliab. Eng. Syst. Saf. (2009). https://doi.org/10. 1016/j.ress.2009.01.010 13. Xu, K., Xie, M., Tang, L.C., Ho, S.L.: Application of neural networks in forecasting engine systems reliability. Appl. Soft Comput. (2003). https://doi.org/10.1016/S1568-4946(02) 00059-5 14. Freitag, M., Oelker, S., Lewandowski, M., Murali, R.: A concept for the dynamic adjustment of maintenance intervals by analysing heterogeneous data. AMM (2015). https://doi.org/10. 4028/www.scientific.net/AMM.794.507 15. Lemke, C., Budka, M., Gabrys, B.: Metalearning: a survey of trends and technologies. Artif. Intell. Rev. (2015). https://doi.org/10.1007/s10462-013-9406-y 16. Gabbay, D.M., Siekmann, J., Brazdil, P., Giraud-Carrier, C., Soares, C., Vilalta, R.: Metalearning. Springer, Heidelberg (2009) 17. Ali, A.R., Gabrys, B., Budka, M.: Cross-domain Meta-learning for Time-series Forecasting. Procedia Comput. Sci. (2018). https://doi.org/10.1016/j.procs.2018.07.204 18. Gomes, T.A.F., Prudêncio, R.B.C., Soares, C., Rossi, A.L.D., Carvalho, A.: Combining meta-learning and search techniques to select parameters for support vector machines. Neurocomputing (2012). https://doi.org/10.1016/j.neucom.2011.07.005 19. Reif, M., Shafait, F., Dengel, A.: Meta-learning for evolutionary parameter optimization of classifiers. Mach. Learn. (2012). https://doi.org/10.1007/s10994-012-5286-7 20. Lemke, C., Gabrys, B.: Meta-learning for time series forecasting and forecast combination. Neurocomputing (2010). https://doi.org/10.1016/j.neucom.2009.09.020 21. Prudêncio, R.B.C., Ludermir, T.B.: Meta-learning approaches to selecting time series models. Neurocomputing (2004). https://doi.org/10.1016/j.neucom.2004.03.008 22. Wang, X., Smith-Miles, K., Hyndman, R.: Rule induction for forecasting method selection: meta-learning the characteristics of univariate time series. Neurocomputing (2009). https:// doi.org/10.1016/j.neucom.2008.10.017 23. Kück, M., Crone, S.F., Freitag, M.: Meta-learning with neural networks and landmarking for forecasting model selection an empirical evaluation of different feature sets applied to industry data. In: 2016 International Joint Conference on Neural Networks (IJCNN), Vancouver, BC, Canada, 24–29 July 2016, pp. 1499–1506. IEEE (2016). https://doi.org/10. 1109/IJCNN.2016.7727376 24. Dorard, L.: The machine learning canvas. Design better machine learning systems. Keep teams of scientists, engineers and managers focused on the same objectives (2019). https:// www.louisdorard.com/machine-learning-canvas. Accessed 10 Sept 2019 25. Kohler, J.: Motortechnologie. Elektronisches Motormanagement: Intelligente Regelung und Steuerung des Motors (2014) 26. Chen, Z., Cao, S., Mao, Z.: Remaining useful life estimation of aircraft engines using a modified similarity and supporting vector machine (SVM) approach. Energies (2018). https://doi.org/10.3390/en11010028 27. Khalid, S., Khalil, T., Nasreen, S.: A survey of feature selection and feature extraction techniques in machine learning. In: 2014 Science and Information Conference (SAI), London, UK, 27–29 August 2014, pp. 372–378. IEEE (2014). https://doi.org/10.1109/SAI. 2014.6918213
Modelling Autonomous Production Control: A Guide to Select the Most Suitable Modelling Approach Oliver Antons1(B) and Julia C. Arlinghaus2 1
Chair of Management Science, RWTH Aachen University, 52072 Aachen, Germany [email protected] 2 Chair of Production Systems and Automation, Otto-von-Guericke University Magdeburg, 39106 Magdeburg, Germany
Abstract. This paper studies and compares Minimal Models, Linear Programming and Discrete-event Simulation as approaches to model Production Planning and Control with regard to their ability to include the concept of autonomous control. After a brief explanation of autonomous control in production planning, the three aforementioned concepts are introduced in detail. We derive their benefits and drawbacks for different scenarios, and subsequently give advice when to deploy each method, applicable for researchers and practitioners alike. Keywords: Autonomous production control · Production Planning and Control · Minimal models · Linear programming · Discrete-event simulations
1
Introduction
Production Planning and Control (PPC) has become an increasingly difficult challenge for European industries. Driven by many factors its complexity is ever increasing (e.g. rising globalization and variant diversity), while a greater degree of agility in PPC is simultaneously desired. These new requirements and various technological advancements have led to the holistic vision of future manufacturing called Industry 4.0: Intelligent and interconnected production machines as well as smart products form a network, and consequently are able to interact and coordinate within themselves, allowing for emergent behavior. Such emergent behavior results in an efficient coordination and subsequent production control, while simultaneously attaining high agility as well as resilience against possible disturbances. However, the question remains of how to control and coordinate all actors (intelligent machines and smart products) within the aforementioned network of intelligent objects. Purely centralized control approaches suffer from the sheer amount of relevant data within the network and lack the desired agility, while a purely decentralized control approach suffers from the intrinsic selfish c The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 245–253, 2020. https://doi.org/10.1007/978-3-030-44783-0_24
246
O. Antons and J. C. Arlinghaus
behavior of machines and products. This dilemma has led to the assumption, that an approach combining decentralized and centralized elements yields the best results in terms of agility and productivity [19]. Such combined approach is often referred to as hybrid approach, which is characterized by the degree of autonomy its actors posses. This autonomy can be viewed as scale on [0, 1]: In a centralized approach, there is no autonomy, hence this approach has an autonomy degree of zero. In a completely decentralized approach every entity has full autonomy, reflected by an autonomy degree of one. A hybrid approach has an autonomy degree between zero and one. For a subset of PPC, the superiority of a hybrid approach with an autonomy degree of 0.4 has already been proven [5]. Therefore, the question how to model hybrid control approaches with varying degrees of autonomy is of great interest - since such model allows to evaluate the benefits of autonomy without the need of invasive tests on real-world systems. Various models have been used in production research to model production planning in the last decades, most notable minimal models, linear programming and discrete-event simulations. Each modelling approach however has different characteristics and its own domain, but there exists no guide addressing the suitability of models to include the concept of autonomy. In the following sections, we will describe three approaches for modelling autonomy in PPC and compare their benefits and drawbacks. From this comparison, we derive advice how to model autonomy in PPC for researchers and practitioners alike.
2
Modelling of Production Planning and Control
Within PPC the assignment of production orders to machines and the corresponding allocation of resources such as processing time and raw materials is a core task and of vital importance. Any approach of production planning needs to take this into consideration, in order to create a schedule for the production environment, and enable subsequent control of its execution on the manufacturing level. A model deployed to study autonomous production control in a PPC setting henceforth requires the ability to derive such a scheduling. However, the scope of a suitable model can vary considerably. In production research, many different approaches have been established over the last decades. For one, minimal models have originated in theoretical physics [4,9] and long been adopted in production research in order to study the behavior and properties of manufacturing systems [6,10,20]. They are characterized by a very narrow scope, only considering parameters deemed essential of the researched behavior. This approach is of course far from production practice, as it does not burden itself with a large numerical parameter space, but consequently allows to focus on the essential aspects. PPC has also been an interesting research area for operations research, which uses mixed integer linear programs (MILP) among others to express the objective and constraints of job scheduling in form of a mathematical model. Such models often address a selected area of PPC, such as a job shop scenario [1,11,22]. However, while generally being able to provide an
Modelling for Autonomous Production Control
247
optimal solution, the resulting problem complexity for real world scenarios often leads to impracticable computing times [3,17]. Another approach are so called discrete-event simulations (DES): Contrary to minimal models, discrete-event simulations are not necessarily as rigid as mathematical models. They allow to simulate production practice, without the need to rely on invasive real-world experiments and observations. In the field of manufacturing DES has long been established as a useful tool for multiple purposes, such as planning and decision support [13]. Among others, its application includes areas such as capacity planning [14] and supply chain network planning [16]. These modelling approaches can be viewed as representatives for different classes of modelling, and differ significantly regarding the scope these models are intended for as well as the level of abstraction required. Consequently, their applicability differs regarding modelling of autonomy in PPC. We review these three approaches for three different modelling scenarios. Starting with Conception, we refer to the modelling of fundamental mechanism required for autonomous entities within PPC to coordinate with each other. As such, a high degree of abstraction is necessary, and only the absolutely required entities are modeled. We refer to the next scenario as Theory & Development, which reduces the previously applied abstraction and takes a greater number of entities into consideration. A multitude of autonomous entities with PPC are considered and their coordination is modeled. However, such model still remains an abstraction and only contains a small subset of the entities of a real world PPC system. In the last scenario, referred to as Practice & Implementation, the abstraction is reduced as much as possible and the model comes close to a real world implementation of autonomous entities within PPC. Within the remainder of the paper, we review these three approaches and evaluate whether they are able to reasonably model autonomy within each modelling scenario.
3
Minimal Models
A minimal model only contains the essential information necessary to model a real world situation in an abstracted way. Moreover, a minimal model is deterministic by nature, and consequently does not contain random factors. The main task of PPC in its capacity of production planning lies in the assignment of jobs to machines, and thereby creating a schedule for a manufacturing plant. In order to study this problem on a molecular level, we will utilize a minimal model that assumes a discrete time horizon. Such an approach allows the modelling of autonomy in PPC in great detail since the decision process for each autonomous entity can easily be defined and controlled. Several authors used this models possibility to map various approaches, such as a complete decentralized control, in which each entity only considers local information in its decision process; a hybrid approach where an entity can weight both local and global information or a fully centralized approach, where an entity does not receive information but orders from a single controller [5,8,12].
248
O. Antons and J. C. Arlinghaus
With these assumptions, this assignment problem can be expressed as a graph coloring dynamics (GCD) model [5]. Let G := (V, E) a graph with vertex set G and edge set E. Every vertex v ∈ V represents a production job, and every edge (v, w) ∈ E, v, w ∈ V expresses a conflict between the production jobs v and w such that they can not be processed on the same machine. The amount of machines available is denoted as K ∈ N, and to solve this GCD a coloring CG = {C1 , . . . , C|V | } is needed which ensures that each pair of conflicting jobs is assigned to different machines, i.e. ∀v, w ∈ V, (v, w) ∈ E : Cv = Cw . For every graph G there exists a minimal amount of colors necessary to have a conflict-free coloring. This number is called chromatic number of G and typically denoted by λG . Given a conflict-free coloring CˆG for the graph G, we can easily generate a assignment with the desired properties if λG ≤ K. By assigning each color to one manufacturing machine, and subsequently assigning all jobs, which corresponding vertex is of that color, to the associated manufacturing machine. This process creates a valid schedule, taking all restrictions into consideration. An example can be seen in Fig. 1, featuring a graph, with an conflict-free coloring in Fig. 1b. According to this solution, jobs 1, 7, 10, 22 will be scheduled on machine one and so forth. However, the structure of this problem offers the potential to distributed approaches: By granting a certain degree of autonomy to each vertex, we can enable the vertices to negotiate their color between themselves such that a conflict-free coloring emerges. On a PPC-level this can be seen as adding a level of intelligence to products (through computing power and communication interfaces), such that these product can communicate with each other as well as the production machines, independently determining a valid scheduling. This approach has been studied by [5] and [2].
(a) Unsolved Graph
(b) Solved Graph
Fig. 1. Exemplary graph with V = 24, K = λ = 6
Thus this modelling approach offers the ability to vary the degree of autonomy given to the systems actors such as products and machines. However, while this minimal model approach offers many benefits it does not come without drawbacks. Due to its minimalistic approach, it appears to be very far from industrial practice and since it only considers a small time frame, it is very difficult to create a holistic approach based on a minimal model (Table 1) .
Modelling for Autonomous Production Control
249
Table 1. Applicability of minimals models for autonomous PPC Application Conception Theory & Development Practice & Implementation
Method
Minimal models ++
4
+
—
Linear Programming
The core of the GCD approach is the graph coloring problem itself, which has been extensively studied in OR, an research area of mathematics focused on optimization. While the problem of finding an conflict-free coloring can be easily formulated as a mixed-integer problem (MIP), however it is an N P-hard problem [18]. Linear programming is most commonly used in PPC in order to derive an optimal scheduling if feasible for the problem size [7,17,22]. For larger problems, heuristic approaches based on linear programs generally are the tool of choice. A simple formulation based on the graph G introduced in the previous chapter is displayed in P 1. The objective function (1) is to minimize zero, since we just want a valid assignment of vertices to colors (translating into an assignment of products to machines). Constraint (2) ensures that every vertex has exactly one color, while constraint (3) ensures that conflicting vertices do not have the same color. (4) ensures that all decision variables are binary. (P 1)
min 0 s.t.
xiv = 1
(1) ∀v ∈ V
(2)
xiv1 + xiv2 ≤ 1
∀(v1 , v2 ) ∈ E ∀i ∈ M
(3)
xiv ∈ {0, 1} M := {1, . . . , K}
∀i ∈ M ∀v ∈ V
(4)
i∈M
Due to the aforementioned complexity of the problem, increasing the inputsize of the problem (the graph G) is a major difficulty, and centralized solution approaches, such as a central MIP-solver, can reach their computational limit quite fast. Splitting this problem as described in the previous chapter into a vertex perspective can also be expressed as a MIP, as seen in P 2. This MIP needs to be solved by every vertex v, with the objective to minimize the color conflicts with its neighborhood δ(v). This is expressed in (5), with Snc being the coloring of vertex v’s neighborhood, i.e. 1 if vertex n has color c Snc = 0 otherwise.
250
O. Antons and J. C. Arlinghaus
(P 2)
min
xvc Snc
(5)
c∈C n∈δ(v)
s.t.
xvc = 1
(6)
c∈C
xvc ∈ {0, 1} C := {1, . . . , K}
∀c ∈ C
(7)
In order to solve the graph coloring, this problems needs to be solved for every vertex v ∈ V . It is important to note, that for each of these problems the colors of all vertices but the one are parameters, and only one vertices color is modeled as a decision variable (xvc ) while the color of the other vertices is a parameter Snc . This highlights the importance of coordination between these vertices, because a conflict-free coloring can only be derived if the objective functions yields zero for all vertices. Furthermore, this codependency of all vertex problems (c.f. Snc ) requires coordination in between them: Either all vertex problems are solved consecutively each iteration, or only a randomized subset of vertex problems is solved every iteration alternatively. In order to model autonomy within the PPC, both this coordination between these molecular problems as well as their objective function give opportunities. For example, the coordination between the vertices (i.e. their corresponding MILPs) could be split into groups, with their size scaling with regard to the degree of autonomy. Furthermore, the objective function of said MILP could consider further global information, which could be weighted against the coloring conflicts with the corresponding vertex neighbors [2]. Hence, linear programming can be used to model autonomy in production control. Due to its mathematical formulation however, it can easily become far from production practice, and its scaleability, especially regarding large scale modelling of autonomous entities, leaves open questions as of yet (Table 2). Table 2. Applicability of linear programming for autonomous PPC
Method
Application Conception Theory & Development Practice & Implementation
Linear programms +
5
++
+
Discrete-Event Simulation
The concept of a discrete-event simulation puts an emphasis on a rather long time horizon instead of only a single time frame, contrary to a minimal model. This time horizon is split into discrete steps as a necessary simplification in order to control the magnitude of the model (unlike a continuous-event simulation).
Modelling for Autonomous Production Control
251
This concept allows all objects and methods within the model to take the current simulation time (i.e. the discrete time step) into consideration. Furthermore, it is possible to create a timeline, both for the system itself as well as every single object within. In order to model PPC as a DES, all production jobs as well as machines are modeled as objects in the simulation, similar to all decisionand organization entities, while every possible action of these objects would be mapped as a method. Moreover, DES typically contains an event list during runtime which can keep track of events that have not been simulated at the current runtime, but are bound to be simulated due to the current state. Contrary to minimal models, DES generally addresses stochastic factors, allowing for random influences within the model. In the context of PPC, this feature allows us to consider a random distribution for the arrival of jobs and the processing time on machines, rather than using a static approach. This naturally allows to model the PPC problem much closer to practice. In a 2014 study, Negahban and Smith concluded an increase in DES for research on PPC by 400% between 2002–2013 compared to 1969–2002, showing the vastly increased recognition of DES for modelling of PPC [15]. The authors furthermore detected a similar increase regarding optimization as goal of DES within the PPC context. While the first applications of DES for production planning were often used to create a scheduling based on rules [21], resulting in a heuristic rather then optimization, the introduction of a greater degree of intelligence to the simulations objects in order to enable autonomous behavior and coordination, leading subsequently to the emergence of an optimal scheduling. The first application of DES in order to study autonomous control within the PPC context has been carried out by Scholz-Reiter et al., comparing DES to continuous system dynamics [20]. Compared to a minimal model, DES has significant differences such as the inherent ability to include stochastic factors. While this comes in handy when modelling closely to practice, these random factors can hugely influence the outcome of a simulation, whereas in a deterministic model this is not the case. Furthermore, another advantage of DES lies in its wider approach, allowing to model many objects, thus being applicable to many real-world scenarios (Table 3 ). Table 3. Applicability of discrete-event simulation for autonomous PPC Application Method
Conception Theory & Development Practice & Implementation
Discrete-event simulation —
6
+
++
Conclusion
This paper presented three different approaches to model autonomous production control, with minimal models allowing to study the process on a molecular
252
O. Antons and J. C. Arlinghaus
level while discrete-event simulation allows to study process in a broader way, closer to practice. Linear programming is placed between these two approaches, as it can mathematically model a wide range of approaches. While the approaches differ quite a bit, they all allow us to study important aspects of PPC. As such, they are essential tools for researches and practitioners alike. Given the characteristics of each method, as noted in the previous sections, we suggest to select which method to apply depending on scope of goals as shown in Table 4. While this research provides suggestion on choosing an appropriate method to study autonomy in PPC, it can be significantly enhanced by considering more of the many available models rather than three representatives. Hence, further research can expand on this comparison and possible introduce a quantitative metric. Moreover, there still remain open questions regarding limits of scaleability, computational complexity and the degree of desired autonomy within PPC. Table 4. Applicability of various methods in order to model autonomy within the context of Production Planning and Control Application Method
Conception Theory & Development Practice & Implementation
Minimal models
++
+
—
Linear programms
+
++
+
+
++
Discrete-event simulation —
References 1. Al-Ashhab, M.S., et al.: Job shop scheduling using mixed integer programming. Int. J. Mod. Eng. Res. 7(3), 7 (2017) 2. Antons, O., Bendul, J.: Decision making in industry 4.0 – a comparison of distributed control approaches. In: Studies in Computational Intelligence, vol. 853, pp. 329–339 (2019). https://doi.org/10.1007/978-3-030-27477-1 25 3. Auer, P. et al.: A new heuristic and an exact approach for a production planning problem. In: Cottbus Mathematical Prepints (2019) 4. Bak, P.: How Nature Works: The science of Self-Organized Criticality. Springer, Heidelberg (1947). https://doi.org/10.1007/978-14757-5426-1 5. Blunck, H. et al.: The balance of autonomous and centralized control in scheduling problems. Appl. Netw. Sci. 3.1 (2018). https://doi.org/10.1007/s41109-018-0071-6 6. Chryssolouris, G., et al.: Flexibility and complexity: is it a trade-off? Int. J. Prod. Res. 51(23–24), 6788–6802 (2013). https://doi.org/10.1080/00207543.2012.761362 7. Gomes*, M.C., Barbosa-P´ ovoa, A.P., Novais, A.Q.: Optimal scheduling for flexible job shop operation. Int. J. Prod. Res. 43(11), 2323–2353 (2005). https://doi.org/ 10.1080/00207540412331330101 8. Hadzhiev, B., et al.: A model of graph coloring dynamics with attention waves and strategic waiting. Adv. Complex Syst. 12(6), 549–564 (2009) 9. Kauffman, S.A.: The origins of order: self-organization and selection in evolution. In: Stein Daniel, L. (ed.) Spin Glasses and Biology, pp. 61–100. World Scientific Publishing Co. Pte. Ltd., Singapore (1970)
Modelling for Autonomous Production Control
253
10. Koren, Y., Hu, S.J., Weber, T.W.: Impact of manufacturing system configuration on performance. CIRP Ann. 47(1), 369–372 (1998) 11. Ku, W.Y., Christopher Beck, J.: Mixed integer programming models for job shop scheduling: a computational analysis. Comput. Oper. Res. 73, 165–173 (2016) 12. Madureira, A., et al.: Negotiation mechanism for self-organized scheduling system with collective intelligence. Neurocomputing 132, 97–110 (2014). https://doi.org/ 10.1016/j.neucom.2013.10.032 13. Manikas, A., Gupta, M., Boyd, L.: Experiential exercises with four production planning and control systems. Int. J. Prod. Res. 53(14), 4206–4217 (2014). https:// doi.org/10.1080/00207543.2014.985393 14. Montreuil, B., et al.: Holistic modelling, simulation and visualisation of demand and supply chains. Int. J. Bus. Perform. Supply Chain Model. 7(1), 53–70 (2015) 15. Negahban, A., Smith, J.S.: Simulation for manufacturing system design and operation: literature review and analysis. J. Manuf. Syst. 33(2), 241–261 (2014). https:// doi.org/10.1016/j.jmsy.2013.12.007 16. Rix, J., Haas, S., Teixeira, J.: Virtual prototyping: virtual environments and the product design process. In: IFIP Advances in Information and Communication Technology (1995). https://doi.org/10.1007/978-0-387-34904-6 17. R¨ osl¨ of, J., et al.: An MILP-based reordering algorithm for complex industrial scheduling and rescheduling. Comput. Chem. Eng. 25(4–6), 821–828 (2001). https://doi.org/10.1016/S0098-1354(01)00674-3 18. Schindl, D.: Some new hereditary classes where graph coloring remains NP-hard. Discrete Math. 295(1–3), 197–202 (2005). https://doi.org/10.1016/j.disc.2005.03. 003 19. Scholz-Reiter, B., G¨ orges, M., Philipp, T.: Autonomously controlled production systems–influence of autonomous control level on logistic performance. CIRP Ann. 58(1), 395–398 (2009). https://doi.org/10.1016/j.cirp.2009.03.011 20. Scholz-Reiter, B., et al.: Modelling dynamics of autonomous logistic processes: discrete-event versus continuous approaches. CIRP Annals 54(1), 413–416 (2005). https://doi.org/10.1016/S0007-8506(07)60134-6 21. Vaidyanathan, B.S., Miller, M., Park, Y.H.: Application of discrete event simulation in production scheduling. In: Proceedings of the 1998 Winter Simulation Conference (1998) ¨ uven, C., Ozbak´ ¨ 22. Ozg¨ yr, L., Yavuz, Y.: Mathematical models for job-shop scheduling problems with routing and process plan flexibility. Appl. Math. Model. 34(6), 1539– 1548 (2010). https://doi.org/10.1016/j.apm.2009.09.002
Predictive Manufacturing Tardiness Inference in OEM Milk-Run Operations Antônio G. N. Novaes1, Orlando F. Lima Jr.1(&), José Eduardo Souza De Cursi2, Jaime Andres Cardona Arias1, and José Benedito Silva Santos Jr1 1
2
Universidade Estadual de Campinas, Campinas, SP, Brazil [email protected] Institut National Des Sciences Appliquées, Rouen, France
Abstract. In an OEM milk-run pickup operation over a road network, the manufacturing of components by suppliers is subject to varying tardiness levels on order release dates. Such faults are traditionally diagnosed and treated with a “fail and fix” strategy (FAF), when a failure is recognized as a sudden disruption problem. In practice, quite often a degradation phase occurs in the manufacturing process before a disruption happens. But, within the Industry 4.0 paradigm, it is necessary to prevent faults that may occur at some time in the future, changing the traditional FAF response to a robust predicting and preventing strategy. In such a context, faults must be forecasted in a dynamic way, over a Big Data basis, and the resulting forecasts must be released at once to the logistics agent to allow him to review his milk-run collecting program in due time, thus leading to a better integrated performance. An approximate method to forecast tardiness levels in supplier’s production, intended to help the related logistic operators to reschedule their services in due time, is proposed and illustrated with a case study. Keywords: OEM supply chain
Milk-run Tardiness inference
1 Introduction Large industries generally use a capacity-oriented planning and scheduling framework to integrate multi-echelon manufacturing networks. Of particular interest is the coordination of production scheduling of finished and intermediate products in the operational planning level. Flexible production coordination [1, 2] represents a complex planning and operational challenge when the production facilities are geographically separated, because the timing of inter-facility transfers of intermediate products between plants and the transport vectors are vital elements in the overall process, in terms of cost, reliability, lead time and service level [3]. However, leading industry firms often do not adequately plan the production scheduling of their manufacturing facilities together with the logistics operators in a simultaneous and integrated way. Instead, it is common to schedule them in a somewhat weak integrated form. When the intermediate processes are located at different geographic sites, the lack of coordination among production operations and logistics services often leads to tardiness effects when releasing sub-products or components to the logistics agent [3]. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 254–262, 2020. https://doi.org/10.1007/978-3-030-44783-0_25
Predictive Manufacturing Tardiness Inference
255
Along the Industry 4.0 evolution, effective integration of production and logistics scheduling is becoming mandatory. Quite a number of articles have been published on integrated scheduling models of production and outbound distribution, as in [4, 5], among others. But great part of the existing models on integrated scheduling of production and distribution are deterministic and static [5], with all the relevant parameters known in advance. According to [5], most probabilistic models are, in fact, real-time versions of deterministic problems, in which all the parameters associated with a manufacturing order are not known until it effectively happens. However, most realworld problems occur in dynamic environments, where unpredictable real-time events may cause changes in the scheduled plans. Examples of such real-time disruptions include machine failures, rush orders, rework, road traffic congestions, etc. When the production facilities are geographically separated, the integrated production and distribution scheduling models usually include vehicle routing problems. Due to continuing developments in telecommunication, computing and information technologies, dynamic real-time vehicle routing and scheduling problems have received special research attention in [6–10]. One important point to consider in the solution of dynamic scheduling and routing problems, is the definition of real-time control policies and procedures, starting with the analysis of the effects of disruptions (machine breakdowns, job cancellation, rush orders, traffic delays, etc.) and followed by the establishment of corrective solutions in a dynamic way [11]. In the present industrial context, it is of particular interest the Original Equipment Manufacturer (OEM) supply chain setting, in which leading industries acquire components from suppliers to assemble them into final products to be sold to end costumers. Computer producers and automobile industries are typical OEM examples. In accordance, stakeholders are forced to replace the traditional receiving, inspection and storage pattern by commit-to-delivery business agreements. To deal with this challenge in the evolving Industry 4.0 framework, all participants must develop integrated production and transport scheduling strategies as to reduce negative tardiness effects when supplying the main OEM production line [9, 10]. Associated with the OEM production process, logistics operators usually adopt the milk-run scheme to collect components and parts from supplying partners, and transport them to the central firm production line, over planned routes. Trucks are dispatched at specified time periods, visiting a number of suppliers to collect components or sub-products [12–14]. Failure of on-time delivery of components by suppliers in a milk-run collecting and transporting process can generate delivery time uncertainty at the OEM’s production line, leading to overstocked inventory, shop floor disturbances, penalties levied to suppliers and logistics operators, or even breakdown of the central production line of the leading OEM industry. From the supply chain perspective, tardiness of order deliveries will result in not only financial penalties, but also damages to company’s image and loss of market share. Given the magnitude of actual supply chains and highquality standards, there is a strong need to monitor supplying delays and quickly identify and mitigate potential problems. Moreover, the continuous pressure to reduce resources and pressure to cut costs, further increases the need for the development of procedures and tools that can quickly and efficiently address these potential supply chain disruptions. Thus, the objective of this work is to analyse and propose means of reducing the negative effects of unforeseen tardiness occurrences in the supplier’s
256
A. G. N. Novaes et al.
production lines, by implementing robust fault-forecasting procedures within the supplier’s premises, whose results are to be conveyed to the logistics operators to be used in reviewing his collecting plans, accordingly. Getting robust tardiness forecasts from the suppliers with enough lead time, the logistics agents will use such information to change, in due time, his milk-run plans, even assigning auxiliary vehicles to perform the collecting tasks if necessary. Although the logistics operator does not participate in the supplier’s shop floor decision process, he nevertheless needs robust forecasts on tardiness occurrences as to review his milk-run plan in due time. This will allow him to better integrate his tasks with the OEM supply chain objectives. In fact, a disruption in the supplier production line can be regarded as a fault in the logistics process, since it may impair the time constraint imposed by the OEM leader company to the delivery of components in the central OEM assembling line. Of course, if the supplier is able to repair his machine in advance, an immediate message must be sent to the logistics operator to review his plan at once.
2 Order Release Tardiness Effects on OEM Milk-Run Operations A disruption in an OEM supplier manufacturing production line can provoke a failure in the on-time delivery of a component order to the milk-run logistics operator. This disruption is regarded as a fault in the logistics process since it can impair the time constraint imposed by the OEM leader company for the delivery of components in its assembling line. Faults in on-time delivery of OEM orders are diagnosed and traditionally treated with a “fail and fix” strategy (FAF), when a failure is recognized as a sudden disruption event. But usually there is a degradation process before an unexpected disruption occurs. Thus, within the Industry 4.0 paradigm, it is necessary to prevent faults that may occur at some time in the future, changing the traditional FAF response into a predicting and preventing strategy, by using advanced analytical technologies. Fault prognostic methods [15, 16] have been developed and proposed to predict the performance and life expectancy of systems. In fact, in the Industry 4.0 era, the production has to transform itself into predictive manufacturing, continuously anticipating corrective actions to avoid disruptions in the system [17, 18]. 2.1
Order Delivery Tardiness Forecast
Traditionally, logistics operators face unexpected delivery delays when getting components from a supplier, along the milk-run tour operation. In a previous work [9], it was assumed that the logistics operator would realize that a component order accomplishment is late when the collecting vehicle gets to the supplier premises. In order to mitigate the negative effects of such tardiness, [9] analysed some different strategies when an order delay occurs. In one alternative, the truck driver would wait at the supplier plant until the components are ready. Depending on the situation, this strategy can lead to milk-run cycle times above the upper limit imposed by the OEM industry leader. Another strategy was to let the vehicle wait up to a time Tup for collecting the components. When this time limit is surpassed, the truck would continue
Predictive Manufacturing Tardiness Inference
257
its milk-run sequence without the components of the faulty supplier, and a special vehicle will be sent to perform the collecting task and transfer its content to the assembling OEM production line. Particularly within this strategy, the Tup ¼ 0 alternative is also a possibility. Among other objectives in [9], the optimizing model searched for the value of Tup that would minimize the total cost and, at same time, maintain the expected milk-run cycle time under the desired upper limit. But in fact, within the Industry 4.0 paradigm, it is advisable to predict order failures that may occur at some time in the future, changing the traditional FAF approach into a preventing strategy. Accordingly, the manufacturing system in each OEM supplier must have a robust fault forecasting method, from which the logistics agent would receive recurrent order-tardiness prognostics that will allow him to review his operation plan quickly, thus mitigating unnecessary delays along the milk-run tours. One example of such predictive fault detection approach, regarding machinery breakdown situations, is [18]. The authors developed a machine performance degradation assessment method that allows manufacturing operators to foresee future machine failures, anticipate corrective actions, and predict order delays in the production line. 2.2
An Approximate Tardiness Forecast Example
A simplified OEM supplier manufacturing framework was simulated in order to estimate tardiness levels in component order accomplishment time, which generates delays in the subsequent milk-run operation. Each milk-run collecting shift starts at 8 am at the OEM assembling plant, when special empty containers are loaded into the vehicle. The time limit to deliver components to the OEM production line is 4 pm, and there is only one milk-run cycle per working day in this case. At 3 pm the truck is expected to be at the supplier’s premises to collect the components. First, the empty containers are discharged from the truck at the supplier plant, and then the loaded containers with components are set in the truck. The analysis was performed on a data sample of 950 sequential milk-run cycles. The following available variables were: (a) q - machine breakdown frequency; (b) Tq machine breakdown time, if any; (c) Dt - machine downtime; (d) T- the programmed time that the vehicle is to be at the supplier plant to collect components; (e) q - the supplier manufacturing workload, expressed as the rate of production load by its productive capacity ð0 q 1Þ; (f) Tard – resulting tardiness value related to component order accomplishment time. In this example, T ¼ 3 pm, but it will vary for each supplier in the route. Machine breakdown frequency is represented by the binary variable q; where q ¼ 1 if a machine breakdown occurs at a particular supplier in the analysed milk-run shift, and q ¼ 0 otherwise. It was assumed in this example that the probability of occurring more than one machine breakdown in a same milk-run shift is negligible. The probability of having a machine breakdown at the analysed supplier factory is 47.3%, meaning there is a machine breakdown, on the average, once every two shifts. The machine downtime probability distribution is shown (see Fig. 1), with an expected value of 1.68 h and a standard deviation of 2.49 h.
258
A. G. N. Novaes et al.
Fig. 1. Cumulative frequency distribution of machine downtime
The relation between T and Tq is an important element in this example because, if Tq > T, there will be a zero tardiness value in the collecting task, at the analysed supplier, since the machine breakdown will occur after the collecting visit, when the vehicle has already collected the components from the supplier in cause. On the other hand, [19] state that a manufacturing workload around 0.7 represents a relaxed situation in the production process. A moderate load is about q ≅ 0.8, whereas a utilization rate of q ≅ 0.9 corresponds to a heavy workload. It is generally expected that, as q comes closer to unit, negative effects on the shop floor tend to increase, in terms of labour efficiency, materials and tools availability, equipment failures, number of setups, reworking (quality problems), etc. In this situation, a heavier workload tends to produce higher tardiness occurrences. The tardiness analysis has been divided into two groups. The first one containing the collecting milk-run shifts that did not show a machine breakdown in the particularly analysed supplier. The second, containing the milk-run shifts that presented a machine breakdown. 2.3
Milk-Run Shifts Without Machine Breakdown
From the total of 950 milk-run shifts of the data sample, 501 cases corresponded to cycles without machine breakdown. The only explaining variable of tardiness in this case is q and varies as a function of the subsequent milk-run shifts along time (see Fig. 2). It starts with q ffi 0:7, goes up to q ffi 0:9; and gets back to q ffi 0:7 at the end of the data series. The correlation coefficient between Tard and q was only 0.11, meaning that the latter variable is not statistically significant to explain tardiness occurrence in this case. In fact, it is possible that other factors preclude the interrelation between the two attributes. For instance, how the supplier measures his productive workload, uses different equipment types on the shop floor, etc., are possible
Predictive Manufacturing Tardiness Inference
259
intervening factors. Thus, it would be advisable that the research staff apply additional efforts into digging more data and more information on the subject, before making a final judgement whether the workload rate is, in fact, a tardiness explaining variable, or not. In other words, this type of analysis must rely on robust qualitative information as well, in addition to quantitative data.
Fig. 2. Variation of q along the sequential milk-run shifts
2.4
Milk-Run Shifts with Machine Breakdown
A sub-total of 449 milk-run cycles presented a machine breakdown. Variable q is equal to the unit, since a machine breakdown always occurs in this case, and therefore its participation as an explaining variable remains implicit. Since in this example one is just examining one unique supplier, with T fixed, the explaining effect of this attribute is not apparent. Later in the research, when getting the values of T for the other suppliers, it will be possible to investigate the possible effect of this variable on tardiness values. This will be done in a further application of the model. Since variable q was discarded from the regression analysis, the explaining variables were Tq and Dt: A better regression form was obtained by substituting Dt1:5 for Dt; leading to the following equation Tard ¼ a þ b Tq þ cDt1:5
ð1Þ
Table 1 shows the regression results obtained with the Statistica software. The test of significance t; with 446 degrees of freedom and showing all the p-values less than 0.05, indicates that all variables and the intercept are statistically significant.
260
A. G. N. Novaes et al. Table 1. Regression results for dependent variable tardiness ðTard Þ. R ¼ 0:927; R2 ¼ 0:859 N ¼ 449 Coefficients Eq. (1) Coefficient Std. error Student t (446) Intercept a ¼−1.1715 −0.0795 −14.73 b ¼ 0.2674 0.0154 17.38 Tq c ¼ 0.1978 0.0041 48.37 D1:5 t
The probability distributions of the observed values of Tard, together with the probability distribution of the same variable obtained with the model, both for the milkrun shifts with machine breakdown are shown (see Fig. 3). For the observed values of Tard; the expected value and the standard deviation are 1.14 and 2.34 respectively, and the estimated similar values obtained with the model are 1.36 and 1.99 respectively. It is observed that the model distorts the probability distribution a little bit in the range 0\Tard\1hr, but it satisfactorily fits the data for Tard 1hr, which is the critical region for practical purposes (Fig. 4).
Fig. 3. Tardiness probability distribution, observed values.
It is expected that, with a greater number of variables and of better quality, it will be possible to fit a better inference model, letting the logistics operator to get manufacturing tardiness forecasts in order to anticipate operating decisions, with regard to his milk-run tasks. In addition, it would be necessary to get a tardiness forecasting model for the milk-run shifts without machine breakdown (Sect. 2.3).
Predictive Manufacturing Tardiness Inference
261
Fig. 4. Tardiness probability distribution, model estimates.
3 Conclusion and Research Prospects It has been proposed and discussed in this paper an approximate method to reduce the negative effects of unforeseen tardiness occurrences in the supplier’s production lines, by implementing fault-forecasting procedures. The basic idea is to convey the resulting tardiness information from the supplier to his logistics partner, as to allow the latter to change, in due time, his milk-run plan, in order to mitigate the effects of disruptions in the shop floor of the central OEM company. In terms of future research, and apart from improving the approximate model described in Sect. 2.2, the work team is now focusing on a Bayesian modelling approach, in which the tardiness inference process successively uses interim model results to get posterior distributions, with the objective of dynamically improving the causal inference among the operational parameters that modify the variables of interest. Such approach reflects a typical inverse problem framework [20], in accordance with the Industry 4.0 context, in which Big Data information becomes extensively available in association with IoT sensors and actuators. Acknowledgements. This research has been supported by the Brazilian CNPq Foundation, Projects 302412/2016-6 and 470899/2013-1.
References 1. Hopp, W.J., Iravani, S.M., Xu, W.L.: Vertical flexibility in supply chains. Manag. Sci. 56(3), 495–502 (2010) 2. Hajji, A., Gharbi, A., Kenne, J.P., Pellerin, R.: Production control and replenishment strategy with multiple suppliers. Eur. J. Oper. Res. 208, 67–74 (2011)
262
A. G. N. Novaes et al.
3. DeMatta, R., Miller, T.: Production and inter-facility transportation scheduling for a process industry. Eur. J. Oper. Res. 158, 72–88 (2004) 4. Selvarajah, E., Steiner, G.: Approximation algorithms for the supplier’s supply chain scheduling problem to minimize delivery and inventory holding costs. Oper. Res. 57(2), 426–438 (2009) 5. Chen, Z.L.: Integrated production and outbound distribution scheduling: review and extensions. Oper. Res. 58(1), 130–148 (2010) 6. Larsen, A.: The dynamic vehicle routing problem. Ph.D. Dissertation, Technical University of Denmark (2000) 7. Yang, J., Jaillet, P., Mahmassani, H.: Real-time multivehicle truckload pickup and delivery problems. Transportation Science 38(2), 135–148 (2004) 8. Güner, A.R., Murat, A., Chinnam, R.B.: Dynamic routing for milk-run tours with time windows in stochastic time-dependent networks. Transp. Res. Part E 97, 251–267 (2017) 9. Novaes, A.G., Lima Jr., O.F., Luna, M., Bez, E.T.: Mitigating supply chain tardiness risks in OEM milk-run operations. In: Freitag, M., Kotzab, H., Pannek, J. (eds.) Dynamics in Logistics: Proceedings of the 5th Conference LDIC 2016, vol. 1, pp. 141–150. Springer, Cham (2016) 10. Novaes, A.G., Lima, O.F., Montoya, G.M.: Forecasting manufacturing tardiness in OEM milk-run operations within the industry 4.0 framework. In: Freitag, M., Kotzab, H., Pannek, J. (eds.) Dynamics in Logistics – Proceedings of the 6th International Conference LDIC 2018, pp. 305–309. Springer, Heidelberg (2018) 11. Ouelhadj, D., Petrovic, S.: A survey of dynamic scheduling in manufacturing systems. J. Sched. 12, 417–431 (2009) 12. Chuah, K.H.: Optimization and simulation of just-in-time supply pickup and delivery systems. Ph. D. Dissertation, University of Kentucky (2004) 13. Brar, G.S., Saini, G.: Milk run logistics: literature review and directions. In: Proceedings of the World Congress of Engineering 2011. London (2011) 14. Novaes, A.G., Bez, E.T., Burin, P.J., Aragão, D.P.: Dynamic milk-run OEM operations in over-congested traffic conditions. Comput. Ind. Eng. 88, 326–340 (2015) 15. Basseville, M., Nikiforov, I.V.: Detection of Abrupt Changes: Theory and applications. Prentice-Hall, New Jersey (1993) 16. Muenchhof, M., Beck, M., Isermann, R.: Fault-tolerant actuators and drive: structures, fault detection principles and applications. Ann. Rev. Control 33, 136–148 (2009) 17. Xu, Y., Sun, Y., Wan, J., Liu, X., Song, Z.: Industrial big data for fault diagnosis: taxonomy, review, and applications. IEEE Access 5, 17368–17380 (2017) 18. Tran, V.T., Pham, H.T., Yang, B.S., Nguyen, T.T.: Machine performance degradation assessment and remaining useful life prediction using proportional hazard model and support vector machine. Mech. Syst. Signal Process. 32, 320–330 (2012) 19. Branke, J., Mattfeld, D.C.: Anticipation and flexibility in dynamic scheduling. Int. J. Prod. Res. 43(15), 3103–3129 (2005) 20. Koscianski, A., Souza de Cursi, J.E.: Physically constrained neural networks and regularization of inverse problems. In: 6th World Congress of Structural and Multidisciplinary Optimization. Rio de Janeiro (2005)
Implementation of a Total Cost of Ownership Model for Last-Mile Logistics as a Constraint Satisfaction Problem Bernd Nieberding1(B) and Johannes Kretzschmar2 1
University of Applied Sciences Erfurt, 99085 Erfurt, Germany [email protected] 2 Friedrich Schiller University Jena, 07743 Jena, Germany [email protected]
Abstract. To validate the benefits of shared-use and cargo-sharing applications in urban delivery, we describe a model of the dispatching processes and their necessary factor combinations of production, with respect to the total costs of ownership (TCO) and feasibility. Further, we show how to implement this model as Constraint Satisfaction Problem (CSP) in an experimental declarative software tool to give the dispatcher a tool for fast decision making. Keywords: Total costs of ownership · Constrained Satisfaction Problem · Cargo-sharing · Shared-use · Urban delivery
1
Introduction
Urban delivery is a proper field for a sustainable implementation of electric mobility, see [1] and [2]. For practical applications the range and economic efficiency of electric vehicles plays a crucial role. To this, new operational concepts are necessary and, as a consequence, a redesign and optimization of existing logistic chains. This implies dispatching processes with new types of vehicles such as cargo cycles or mobile depots, see [3] and [4]. Further, new business models are necessary to extend the utilization of vehicle capacities and operation times. Main goal of these new models has to be a spreading of the higher capital costs of electric vehicles on a greater number of heads and services to benefit from their lower operational costs, with the aim of minimizing the total costs of the provided services, see [5]. Lacks of productivity of vehicles in current dispatching processes are given by a missing transparency of unused capacities between different companies, with respect to operational times and cargo loading. To scoop these potentials, the field of sharing economy offers two promising approaches, so-called shared-use and cargo-sharing. The concept of shared-use is strongly related to the wellknown concept of car-sharing. While in usual car-sharing models the user is an c The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 263–273, 2020. https://doi.org/10.1007/978-3-030-44783-0_26
264
B. Nieberding and J. Kretzschmar
individual person, shared-use is a cross-company sharing of vehicles for services of different branches with complementary operation times in subsequent shifts. Cargo-sharing aims to the problem of unused load capacities in urbandelivery. In this context, cargo-sharing means the exploitation of available freight capacities of a vehicle belonging exclusively to one service provider with transport request from another one. This paper presents a model to validate the benefits of shared-use and cargosharing concepts in the area of media and pharma logistics, which takes all factors of the production process into account. Aim of this model is to validate each dispatching process and its necessary factor combination monetarily, e.g. total cost of ownership (TCO), and with respect to feasibility. In order to extend the application of the TCO model beyond straight forward cost calculation, we show how to implement it as Constraint Satisfaction Problem (CSP) [6] in an experimental declarative software tool.
2 2.1
TCO Model Production Factors
To archive the necessary key figures for our TCO model we roughly follow the structured classification of determinants for the production factor human work presented in [7] and generalize this concept to the factors transport vehicle, human, good and hub. In this model each factor is described by determinants, describing the factor itself, and dependencies, describing the behavior of a factor in combination with other factors. The determinants are divided into two classes. The first class named individual parameters consists of the subclasses costs, e.g. vehicle costs or salary, and performance, e.g. load capacity or necessary driver license. The second class named situative parameter consists of the subclasses operation, e.g. energy consumption or muscle strength, and ambience, e.g. capital interests or overhead costs. Using the determinants each factor can be associated with key figures such as costs per kilometer or costs per day to monetarily validate the production processes, see Sect. 2.2. On the other hand the determinants characterize each factor with specific needs if the factor is used in combination with other factors. For example, a driver can only use vehicles, which meet his driving license, or goods, which require cooling devices, can only be transported by appropriate vehicles. These dependencies are described in Sect. 2.3. Based on the key figures of factors given by the determinants and on the dependencies of the combination of production factors it is now possible to evaluate, if a specific service or distribution of goods can be performed in the desired process. 2.2
Cost Functions
The total costs of a dispatching tour route R are given by Ctotal (R) = Cvehicle (R) + Cdriver (R) + Chub (R) ,
(1)
Implementation of a Total Cost of Ownership Model
265
where Cvehicle (R) are the partial total costs of a vehicle, Cdriver (R) are the partial total costs of a driver and Chub (R) are the partial total costs of a hub. The partial total costs of the vehicle on route R, with length d (R) and time tvehicle (R) := td (R) + tprep (R) + tcharge (R) ,
(2)
=:tdriver (R)
including times td (R) for driving and dispatching and times tprep (R) for vehicle preparation or tcharge (R) for battery charging, are calculated through fix toll Cvehicle (R) = cvar vehicle · d (R) + cvehicle · tvehicle (R) + Cvehicle (R) ,
(3)
fix where cvar vehicle are the vehicle costs per kilometer and cvehicle are the vehicle costs toll per day or hour. The total toll costs Cvehicle (R) are calculated as ⎧ km ⎪ ⎨d (R) · ctoll distance-based toll, toll time Cvehicle (R) = Ctoll (R) (4) time-based toll, ⎪ ⎩ 0 toll-free, time where ckm toll is the toll-rate per kilometer and Ctoll (R) is the partial total toll-rate per time-period. The partial total costs of the driver on route R, with operation time tdriver (R), including times for driving and dispatching and times for vehicle preparation, see (2), are calculated through
Cdriver (R) = tdriver (R) · cdriver · pdriver (R) ,
(5)
where cdriver are the costs per day or hour and pdriver (R) is a tour-depending allowance, e.g. operation time in night-shift. To keep the notation simple, we assume that a dispatching route uses only one hub and is served by only one driver. Other scenarios can be modeled by adding additional routes with other hubs and drivers. Let V be a volume measure and W be a weight measure of shipments. Let Vhub (R) and Whub (R) be the total measures of shipments of route R, which are stored in a hub. Further, let Vhub be the maximal volume capacity and Whub be the maximal weight capacity of a hub and let
Vhub (R) Whub (R) , phub (R) := max (6) Vhub Whub be the fraction of the hub utilization of route R. Then, the partial total costs of the hub on route R are can be calculated through
fix var toll , (7) + Chub + Chub Chub (R) = phub (R) Chub fix var are the total fixed costs per day, Chub are the total variable costs per where Chub toll day and Chub are the total toll costs per day. Due to the utilization of mobile hubs, which may be daily re-located, the differentiation between total fixed costs per day and total variable costs per day is useful.
266
B. Nieberding and J. Kretzschmar
Notes and Comments. In Eqs. (3), (5) and (7) linear cost functions were assumed. Depending on the number, length and duration of dispatching services per vehicle and day other cost functions might be useful, such as step-wise increasing functions or mixed functions with linear and step-wise increasing parts, to sharp the overall calculation. The rates for fixed and variable costs, e.g. cvar vehicle or cdriver , can be calculated according to [8] and [9]. 2.3
Feasible Production Combinations
Let v be a vehicle of a fleet V, w a driver n from a set of drivers W, h a hub from a set of (mobile) hubs H and S (R) = i=0 Si (R) be a set of unique delivery loads on a tour route R, where S0 (R) is the initial load of the vehicle and Si (R), with i ≥ 1, are the loads transferred from the hub to the vehicle at the i-th visiting. Then the tuple or combination of factors (R, S (R) , v, w, h) is called feasible if the following conditions or dependencies are satisfied: Capacity Condition: Let Vvehicle be the maximal volume capacity and Wvehicle be the maximal weight capacity of a vehicle. Then, the capacity condition between the delivery shipments of route R using vehicle v is satisfied, if for all 0 ≤ i ≤ n holds W (s) ≤ Wvehicle (v) and V (s) ≤ Vvehicle (v) . (8) s∈Si (R)
s∈Si (R)
The capacity condition between the delivery shipments of all routes R stored in hub h is given through phub (R) ≤ 1, (9) R
with phub (R) defined as in (6). For practical applications the volume-based conditions need several improvements to fit specific volume dimensions of shipments and storage spaces of vehicles and hubs. Range Condition: Currently, the utilization of electric vehicles with limited distance ranges, long charging times in combination with the current state of charging infrastructure and battery systems implies strong restrictions of the operation-distance range dvehicle of a vehicle v on a route R, given by d (R) ≤
m
λi dvehicle (v),
(10)
i=0
where m is the number of re-charging or re-fueling on route R and λi ∈ [0, 1] is the percentage of the i-th fueling with respect to distance range.
Implementation of a Total Cost of Ownership Model
267
License Condition: Let lvehicle be the necessary driving license for a vehicle v and Ldriver be the set of driving licenses owned by a driver w. Then, the license condition is (11) lvehicle (v) ∈ Ldriver (w). Moreover, if the driver is also responsible for moving a hub h, with necessary license lhub , the condition is extended to {lvehicle (v), lhub (h)} ⊆ Ldriver (w).
(12)
Storing Condition: The width range of new transport vehicles from regular trucks to small cargo cycles in combination with shipments with specific needs such as weather protection, access protection or cooling, implies several conditions for storing of shipments in a vehicle or hub. Let ashipment be the class of necessary access protection, rshipment be the class of necessary weather protection and eshipment be the class of necessary access protection for a shipment s. Further, let Avehicle , Ahub be the sets of available access protection classes, Rvehicle , Rhub be the sets of available weather protection classes and Evehicle , Ehub be the sets of available cooling classes of a vehicle v and a hub h. Then, the storing conditions are satisfied if for all s ∈ Si (R) , i ≥ 1, holds ⎞T ⎛ ashipment (s) ⎝ rshipment (s) ⎠ ∈ Avehicle (v) ∩ Ahub (h) × Rvehicle (v) ∩ Rhub (h) eshipment (s)
(13)
× Evehicle (v) ∩ Ehub (h), and for all s ∈ S0 (R) holds ⎞T ⎛ ashipment (s) ⎝ rshipment (s) ⎠ ∈ Avehicle (v) × Rvehicle (v) × Evehicle (v). eshipment (s)
(14)
Comfort Condition: Similar to storing conditions, the width range of desired new transport vehicles provide different levels of comfort for the driver, especially weather protection, use of physical strength and time consumption in the delivery process. Let rdriver be the class of necessary weather protection, bdriver be the class of available physical strength and gdriver be the class of necessary delivery comfort for a driver w. Further, let Rvehicle be the set of available weather protection classes, Bvehicle be the set of necessary physical strength classes and Gvehicle be the set of available comfort classes of a vehicle v. Then, the comfort conditions are satisfied if ⎞T ⎛ rdriver (w) ⎝bdriver (w)⎠ ∈ Rvehicle (v) × Bvehicle (v) × Gvehicle (v). (15) gdriver (w)
268
B. Nieberding and J. Kretzschmar
Loading Condition: Heavy shipments may require additional transport equipment such as fork lifts. Let fshipment be the necessarily required equipment for a shipment s, Fvehicle , Fhub be the sets of available equipment for a vehicle v or hub h and Fdriver be the set of equipment, which can be used by a driver w. Then, the loading condition is satisfied if for all s ∈ Si (R) , i ≥ 1, holds fshipment (s) ∈ Fvehicle (v) ∩ Fhub (h) ∩ Fdriver (w),
(16)
and for all s ∈ S0 (R) holds fshipment (s) ∈ Fvehicle (v) ∩ Fdriver (w).
(17)
Time Condition: Using shared use application the time-depending availability of vehicles may be limited by already planed services. Therefore, conditions are necessary to ensure a fitting of the desired service route to the existing vehicle schedule. Let τ : D → R≥0 be an isometry, with q ∈ D → t ∈ R≥0 , where D is the set of dates. Let qstart , qend ∈ D be the boundaries of a time window for an available vehicle v. Then the following conditions must be satisfied: τ (qstart (v)) + tvehicle (R) ≤ τ (qend (v)) ,
(18)
tdriver (R) ≤ tmax (w),
(19)
and where tmax is the maximal available operation time of a driver w. Regular Driver Condition: Let W (R) ⊆ W be the set of drivers, which serve the dispatching tour route R regularly. Then, the regular driver condition for driver w is satisfied if w ∈ W (R) . (20)
3
Constraint Based TCO Implementation
Independently of the level of automated planning of the dispatching process a fast decision making of the dispatcher to deal with disturbances plays a crucial role. To this applications are necessary to give the dispatcher all important information of optional planning possibilities in short time. A promising approach to this problem is given by reformulating the TCO model as a Constraint Satisfaction Problem (CSP), like presented as a first concept in [10]. A (CSP) consists of a set of variables that are related to each other by formal expressions. A solver then processes these statements and searches for an assignment for all variables where all expressions become valid. The arithmetic and conditional formalism of the TCO model described above is obviously a solid foundation for getting processed by a CSP solver. As a CSP implementation, the model can be utilized multidirectionally, process parameter ranges and ensure consistency.
Implementation of a Total Cost of Ownership Model
3.1
269
TCO MODEL as a PROLOG CSP
The CSP application was implemented in Sicstus Prolog [11] by means of the clp(q, r) [12] library. This library component enables the formulation of arithmetic constraints with variables in the real and rational numbers domain. For example, the calculation formula for the overall costs of tires for a vehicle ctire vehicle =
year ctire-kit vehicle · dvehicle , dtire vehicle
(21)
tire tire-kit with annual mileage dyear vehicle , tires mileage dvehicle and tires cost rate cvehicle , results in the following Prolog statement:
calcVeh24(C_veh_tire, D_veh_year, D_veh_tire, VarCalcVeh36) :% accessing Constraint Variables in Prolog List Structures ( VarCalcVeh36 = [C_veh_tireKit]; VarCalcVeh36 = [C_veh_tireKit, C_tire, N_veh_tire], calcVeh36(C_veh_tireKit, C_tire, N_veh_tire)), % arithmetic constraint clpr statement clpr: {C_veh_tire = C_veh_tireKit * D_veh_year / D_veh_tire}.
Every formula from the TCO model is represented by an accordingly numbered Prolog predicate. The object variables get grounded in the rule head and deconstructed into the corresponding elements in the body. A concluding clpr statement holds the actual arithmetic constraint. The conditions from Sect. 2.3 are modeled accordingly. Furthermore, we took particularly advantage of the descriptive capability of Prolog. Whereas, for example, the relationships of driver license requirements are modeled on a set theoretic basis in the TCO model above Sect. 2.3, Prolog facilitates a predicate centered implementation. This enabled the propositional way of declaring having, demanding licenses as well as the complex relationships of various license classes, as illustrated in the following short example: % defining driver and staff objects as well as their relationship driverL(dlA). driverL(dlB). subL(dlA,dlB). staff(bob). hasLicense(bob,dlB). % Prolog rule for reasoning subsuming driver licenses hasLicense(X,SubL) :- staff(X), driverL(SubL), subL(SubL,DlX), hasLicense(X,DlX).
We think, this kind of declarative representation is suitable for several determinants like qualifications, health issues or comfort demands. Finally, the CSP implementation in Prolog comprises all influencing factors, calculation formulas and conditional constraints uniformly formalized. In this way, the algorithmic division into factor exploration, cost calculation and evaluation of the conditions can be omitted. The homogeneous representation allows the integration of further domain aspects, as well as the interchangeability of the model or model components.
270
3.2
B. Nieberding and J. Kretzschmar
CSP-TCO Application
The Prolog resolution calculus allows the computation of valid variable allocations and also provides the implementation of meta-predicates for managing solutions and determining a cost optimum. Though the common access to a Prolog knowledge base via read-eval-print console is not user-friendly. We therefore implemented the CSP solver and CSP model as component of a Java-based software tool as illustrated in Fig. 1, which supplies an adequate graphical user interface (GUI) for a logistics service provider. The GUI enables various functionalities regarding the adjustment of TCO parameters, allocating tours, managing vehicle fleets and the calculation as well as prediction of costs. Each functionality is then formulated as a corresponding Prolog query by the query processor which interacts with the Prolog interface. The query processor also receives the reply by the Prolog interpreter and instantly updates affected GUI elements. The Prolog knowledge base holds the TCO model as described above and can be automatically filled with various enterprise data by a CSV-based data importer.
Fig. 1. Software architecture of the TCO CSP application containing a GUI frontend for accessing the Prolog constraint database as well as a data importer for integration into logistics enterprise systems
3.3
Usecases
The implementation of the above described software tool is currently still under development. By now, the tool offers the following use cases which are a substantial basis for more sophisticated functionalities in automatically supporting TCO-centered strategic decisions in logistics. Cost Calculation: Of course, the tool provides the simple application of the intended TCO model. Every change of a parameter by the user is evaluated regarding the current TCO model. If the model is consistent, free variables are going the get grounded with values or ranges, if possible. If every influence factor is grounded, the system automatically calculates the overall costs. In case of an inconsistent input, the tool immediately reports an error.
Implementation of a Total Cost of Ownership Model
271
Parameter Analysis: Like mentioned above in Sect. 3, it is possible to use the TCO model in a multidirectional way when implemented as CSP. The tool allows to enter specific or constrained costs and then calculates the impact on the corresponding input parameters. This enables a parameter analysis depending on interim or total costs. If there is no interest for a specific TCO optimum, this use case provides a decision scope for a logistics dispatcher in allocating resources or acquiring a vehicle fleet. The tool therefore supports extended decision-making beyond the influences factors of the TCO model like for example probability of failure regarding vehicles or staff. Inter-parameter Constraints: Besides explicit or constrained values, the tool allows to enter complex arithmetical statements and dependencies between parameters, because all inputs get transformed into Prolog query constraints anyway. This is useful, when some influencing factors can not determined due unclear data situations or incomplete knowledge. Besides, it helps to keep the model simple and transparent by adding small calculations directly in the GUI input element. It is also possible to implement small models, like for example the prediction of an energy price trend by a arithmetic formula. Semantics: We discovered, that the method of building Prolog queries and processing these regarding a GUI form has an interesting side effect. The query processor operating the Prolog interface implicitly connects the parameters of the TCO model with their semantics due the interpretation of the logic programming. By checking the satisfiability and consistency after every user input, the tool is also able to exclude non-accomplishable parameter assignments. In this way, the tool not only automatically detects user input errors by consistency check but also helps to prevent errors in a highly complex domain like logistics without any further algorithmic adaptions.
4
Simulation
In this section we give a short simulation to show the benefits of the previously developed TCO model. The considered tour data consists of five newspaper delivery tour routes operated by two different vehicles in the same area, e.g. the delivery points and quantity of deliveries are similar on each tour, see Table 1. A sheer consideration of the individual parameters of the vehicles, e.g. the fixed and variable vehicle costs, leads to a favoritism of vehicle B, because both values are lower than Vehicle A and therefore promise lower tour costs. An evaluation of the tour costs using the TCO model takes situative parameters into account. Here, the vehicle design of vehicle A enables the driver a closer approach to a delivery point and a faster entering and exiting of the vehicle during the delivery process, which decreases the average tour duration from 5.05 to
272
B. Nieberding and J. Kretzschmar Table 1. Key figures and parameters of the used vehicles and tours. Key figures
Vehicle A Vehicle B
Fixed costs (EUR/h) Variable costs (EUR/km)
1.64 0.12
1.59 0.11
Tour distance (km) Tour duration (h) Tour deliveries (pc)
∅ 44.20 ∅ 4.53 ∅ 580
∅ 43.70 ∅ 5.05 ∅ 576
Tour vehicle total costs (EUR) ∅ 12.70
∅ 13.00
4.53 h. As a cross-effect the averaged total costs of the tour using vehicle A is 0.3 EUR or 2.3% lower than vehicle B. Regarding driver costs in the tour costs the cross-effect increases and the the total costs using vehicle A are 8.7% lower than vehicle B.
5
Conclusions
We have introduced a TCO model to describe all necessary cost functions and dependencies for combinations of production factors in the field of last-mile or urban distribution. Further, we described the implementation of the TCO model as a Constrained Satisfaction Problem using Prolog to support the decision making process of the dispatcher with a GUI. Further research steps are feedback-loops in the TCO model for an accurate estimation of the factor determinants which are mostly affected by shared-use applications, e.g. annual mileage or operation hours per day. Also the implementation of the developed GUI into in existing telematic software will be done in the near future. Acknowledgements. This work was supported by the German Federal Ministry for Economic Affairs and Energy, funding code 01ME17001D. The responsibility for the content of this publication lies with the author.
References 1. Adler, U., Gather, M., Apfelst¨ adt, A., Franz, S., Fuchs, J., Gottschall, K. L¨ uttmerding, A.: Parameter eines adaptiven Reichweitenmodells und elektromobile Applikationen in Fuhrparks. In: Abschlussbericht SMART CITY LOGISTIK, Erfurt (2016) 2. Vogel, M.: Elektromobilit¨ at in gewerblichen Anwendungen. In: Begleit- und Wirkungsforschung Schaufenster Elektromobilit¨ at (BuW) Ergebnispapier Nr. 9. Deutsches Dialog Institut GmbH. Frankfurt am Main (2015) 3. Verlinde, S., Macharis, C., Milan, L., Kin, B.: Does a mobile depot make urban deliveries faster, more sustainable and more economically viable: results of a pilot test in Brussels. Transp. Res. Procedia 4, 361–373 (2014)
Implementation of a Total Cost of Ownership Model
273
4. Navarro, C., Roca-Riu, M., Furi´ o, S., Estrada, M.: Designing new models for energy efficiency in urban freight transport for smart cities and its application to the Spanish case. Transp. Res. Procedia 12, 314–324 (2016) 5. Lerch, C., Kley, F., Dallinger, D.: New business models for electric cars: a holistic approach. In: Working Paper Sustainability and Innovation, No. S5/2010 6. Apt, K.: Principles of Constraint Programming. Cambridge University Press, Cambridge (2003) 7. Corsten, H., G¨ ossinger, R.: Produktionswirtschaft - Einf¨ uhrung in das industrielle Produktionsmanagement, 14th edn. De Gruyter Oldenbourg, Berlin (2016) 8. Fiedler, J.: Fahrzeugrechnung und Kalkulation. In: Praxis des Controllings in Speditionen. BSH, Frankfurt am Main (2007) 9. Wittenbrink, P.: Transportkostenmanagement im Straßeng¨ uterverkehr Grundlagen - Optimierungspotenziale - Green Logistics, 1st edn. Gabler, Wiesbaden (2011) 10. Kretzschmar, J., Johlke, M., Rossak, W.: A TCO analysis tool based on constraint systems for city logistics. In: Proceedings of The Seventh International Conference on Advances in Vehicular Systems, Technologies and Applications (VEHICULAR18), Venice, Italy, pp. 37-38 (2018) 11. Carlsson, M., Fruehwirth, T.: SICStus Prolog User’s Manual 4.3. Books On Demand - Proquest (2014) 12. Holzbaur, C.: OFAI clp(q, r) Manual. Austrian Research Institute for Artificial Intelligence, Vienna (1995). TR-95-09
A Theoretical Framework Assessment Proposal for a Complexity Degree Measurement on a Supply Chain Network Silvio Luiz Alvim1(&), Jose Benedito Silva Santos Jr.2, and Carlos Manuel Taboada Rodriguez1 1
2
UFSC - DEPS/CTC, Trindade, Florianópolis, SC 88010-970, Brazil [email protected] Unicamp - LALT, Albert Einstein 951, Campinas, SP 13083-852, Brazil
Abstract. A complex system is usually defined as an environment that the current processes and activities do not allow a simple approach to the overall management tasks. Despite the distinct types of complexity present on a supply chain (SC), the management level needs a more comprehensive understating of how does it impact the SC performance, identifying the supply chain complexity (SCC) degree and applying the appropriate approach to support the decisionmaking and business risk assessment. Based on an exploratory research methodology, a theoretical framework proposal was developed and applied in a test case to validate potential use in the SC management network based on a literature review of the most recent works on SCC models. The preliminary findings were promising, and it is encouraging further investigation and model improvements. Keywords: Supply chain
Complexity Framework
1 Introduction A complex system is usually defined as an environment that the current processes and activities do not allow a simple approach on the overall management tasks. Nowadays, the terminology “complexity” encompasses a new group of topics (information theory, chaos theory, system theory, cybernetic and risk management) that extends the concept to be understood as supply chain complexity (SCC) [1]. Were not found unified concepts of SC as complex systems, however, several definitions of complexity can be found. The first category is the structural complexity, in which there are several elements in a system and multiple interconnections within these elements. The second category is related to the functional complexity, composed by the dynamic resultant from the interrelation movement among these elements. The third category of complexity corresponds to modeling complexity. It is noteworthy that can occur problems resulting from the calculation of knowledge and the goals of conflict, due to the difficulty of standardization, and that these problems are closely related [2]. Blecker, Kersten and Meyer define supply chain complexity as the framework that combines volume and type of interrelated transactions, activities and processes across © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 274–285, 2020. https://doi.org/10.1007/978-3-030-44783-0_27
A Theoretical Framework Assessment Proposal
275
the supply chain (SC), integrating restrictions and uncertainties in accordance to these processes, transactions and activities take place [1]. Many approaches to measure the SCC could be found in the literature, e.g., information theory, non-linear dynamics, axiomatic theory, and specialist’s panel, among others [3]. Although the identification of the complexity attributes on a SC is essential on a decision-making process, the current business environment requires to measure it related to costs and operational performance indicators, in order to identify improvement and needed mitigations actions [4]. Despite of the distinct types of complexity present in a SC, the management level needs a more comprehensive understanding how does it impact the SC performance, identifying the SCC degree and apply the appropriate approach to support the decision-making and business risk assessment [5]. These issues motivated several publications, presenting models that aim to identify the type of complexity, complexity drivers, logistical complexity, cognitive decisionmaking and risk assessment in the SCC environment. Extending the findings of Manuj and Sahin [6], that presents the SC and SC decision-making complexity model, the research question (RQ) that drives this work is: How to measure, in a pragmatic way, the degree of complexity on a SC network, considering the business environment inputs (internal and external to the SC), evaluating the relevant business dimensions (finance, customer perception, operational performance, process development and risks) to set the right actions to deal with the SCC?
The study presented in this paper is the first step, that aims to answer the RQ based on a theoretical framework assessment model proposal, limited on the qualitative inputs, and its application analyzed using trough a real test case to assess the approach potential. The work from Manuj and Sahin [6], approaching the SCC model, was reviewed. The main contribution of the study of these authors was to follow the analysis of the variables do BSC extended to model [6], proposing a new scheme for measuring complexity through an inter-relationship between these two models, creating a basis for future researches. Within the following phase, the theoretical model will be required to be validated on a mathematical basis, case studies, application and simulation. The paper content is structured as following: Sect. 2 presents the literature review background; Sect. 3 shows the methodology approach applied; Sect. 4 presents the theoretical assessment framework proposal; Sect. 5 shows the test case and the preliminary results; finally, on Sect. 6, the conclusions and future research opportunities are presented.
2 Literature Review Background 2.1
Supply Chain Complexity
The complexity is one of the relevant issues mentioned on the SC literature. The SCC definitions present different principles but, in general, the common understanding is that SCC is a multifaceted phenomenon caused by several sources. Risk and uncertainty, technological complicatedness, organizational practice, large number of
276
S. L. Alvim et al.
suppliers, mix of products, and manufacture processes multiplicity can be identified as key elements. Obviously, it is not easy to identify what just defines the SCC and which effects are key for an adequate decision-making process in the SC. The challenge is how to identify a clear complexity framework concept and the relevant variables that help to manage a complex decision-making process into the SC [7]. Data collected process is key, when used as a proactive attitude to react properly, answering tactical or operational non-planned issues that occur over the SC [8]. On the other hand, strategies that anticipate uncertainties, will be able to dampen turbulence and disruptions when operating on a complex environment [9]. 2.2
Drivers in Supply Chain Complexity
The SC operations in a complex environment is constantly at risk, that can be classified as a positive or, a negative risk. For example, in a dual sourcing strategy, the negative effect of supply disruption could be converted in a positive effect, allowing to the chain do a selfbalance cost, in case that, one of the suppliers increase the price due an own internal failure. Decision-making complexity must increase as the sequence flexibility for the parts in the production batch increases [10]. The flexibility is required to attend the high demand spikes, production mix, special situations, resources and capacity, quality model, and other dynamic product assortment. Supplier delays has a direct impact on a complex decision-making process and always challenge the tradeoff between SC costs and service level to customers. The Table 2 presents the main drivers in SC complexity. Table 1. Drivers in supply chain complexity Source: adapted from [12] and [13]. # Driver Concept 1. Uncertainty The lack of predictability and reliability of demand and of supply chain in processes. SC sources: customer demand, SC processes, and market conditions 2. Variability Unexpected, large and variable changes in requirements over time. Difference among planned and actual elements in the chain. SC source: SC processes behavior 3. Multiplicity Complexity covered by a combination of several elements such as product profile (raw material, semi-finished or finished goods), processes synchronization, act globally vs. locally, stakeholders alignment, clear targets definitions, etc. SC source: high quantity of components and multilevel structures contributes to an increase on SCC 4. Size Relative number or volume of products or activities. SC source: minimal order quantity, production batches, lot sizes, number of suppliers, and product assortment 5. Speed Required responsiveness across the supply chain in terms of throughput times, delivery times and frequencies. SC source: Short product life cycle, service level agreements, and real time track & trace 6. Diversity Hybrid, homogeneity and heterogeneity systems (supplier, product, transportation modal). SC source: the level of customization of products and services offered to customers
A Theoretical Framework Assessment Proposal
277
The literature presents distinct models to identify the complexity of SC (see Sect. 2.1). The Table 2 shows the relevant models with a concept summary related to the drivers in SCC (see Sect. 2.2). The “model application effort” dimension was added as a seventh driver, in order to complement the analysis. Table 2. Supply chain complexity models vs. drivers in SCC Author Year Concept summary [12]
[1]
[6]
SCC drivers Uncertainty Variability Multiplicity Size Speed Diversity Effort
1998 The SCC triangle ● is composed by three elements: (a) deterministic chaos; (b) parallel interactions; (c) demand amplification. Some variables were inputted impacting the chaos (SC decision-making process; SC Control System) ● 2005 The complexity has two dimensions: (a) Internal organization supplier-customer interface and dynamic environment, and (b) Organizational aspects uncertainty and product technological intricacy ● 2011 The model composition is (a) antecedents to SCC: supply chain complexity, supply chain decisionmaking complexity, moderators to the link between SCC, and supply chain decision-making complexity, and
●
●
●
N/A
N/A
++
●
●
N/A N/A
N/A
+++
●
●
●
N/A
++
●
(continued)
278
S. L. Alvim et al. Table 2. (continued)
Author Year Concept summary
[14]
[15]
[16]
SCC drivers Uncertainty Variability Multiplicity Size Speed Diversity Effort
(b) moderators to the link between supply chain decision-making complexity and outcomes, and the performance outputs 2012 The complexity is ● determined by the following main parameters: (a) number of elements and interrelations that set the system; (b) the degree of uncertainty that enters the system; (c) supplier leverage on the customerrequested product variety; and (d) geographical components that act on the system ● 2013 The method estimates the operational (too called dynamic) complexity associated with the different stages of the SC, where each stage is identified by the alteration of data or material 2015 There are two sub- ● streamers path in SCC: (a) SC as a complex adaptive system that have a capability to learn and adapt to the environment, or (b) SC as social network, using social network analysis method to understand how relational tier are
●
●
●
●
●
N/A
●
N/A
N/A
+++
●
●
+++
N/A
+++
N/A N/A
(continued)
A Theoretical Framework Assessment Proposal
279
Table 2. (continued) Author Year Concept summary
[17]
[2]
SCC drivers Uncertainty Variability Multiplicity Size Speed Diversity Effort
composed and how this connection affects the social capital, convergence, resource, and contamination in SC and networks ● 2016 In accordance to the authors: the SCC is the detail degree of complexity and dynamic complexity. The SCC drivers can generate within the business unit (BU): internal drivers; or external drivers of decisionmaking processes and environmental factors. Disregarding this type or origin, the SCC can occur due to the commercial strategy or improper business practices (complexity dysfunctional) ● 2018 To [2] does not exist a unified concept of SC as Complex system. Complexity definitions are considered: (a) structural complexity; (b) functional complexity; and (c) modelling complexity
●
●
N/A N/A
N/A
++
●
●
●
●
++
●
Source: authors. Legend: Model application effort: High (+++); Medium (++); Low (+); N/A (Not Applicable).
280
S. L. Alvim et al.
The summary of the models was evaluated in a cross reference (Table 1 vs. Table 2). As a result, it was found that the most adequate indication for this study is the model presented by Manuj and Sahin [6]. 2.3
SC Risk Management
Tang defines SC risk management as the coordination and collaboration between various partners across the chain, in order to ensure the operations continuity and the business profitability [18]. Manuj and Mentzer broadened the concept of SC for global operations of risk management, defining it as being the process of identification and assessment of business risks and their losses through the application of appropriate strategies to coordinate logistics operations on a sustainable basis between different partners CS [19]. The type and the maturity level of the relationships among the various SC links can influence the operations performance. The information sharing level between the SC partners, the service level agreements and the delimitation of the work scope and responsibilities, represent the main relevant elements that would influence the uncertainty level and, therefore, the disruption risks related to that specific SC [20–22].
3 Methodology An exploratory research approach was applied on this research development, and in order to validate the assessment framework developed, a test case was carried out in a real example. The exploratory approach is recommended when a better understanding or a clarification on a given topic is required. It also allows flexibility to the research development and to validate the proposed assumptions [23]. Based on a literature review of the SCC and decision-making process subjects, an assessment framework proposal was developed and applied on a test case, in order to validate the model’s applicability. On Sect. 4 is described the steps to apply the SCC assessment framework development.
4 Supply Chain Complexity Theoretical Assessment Framework The SCC theoretical assessment framework proposal was developed through the 5 steps stated below and it is presented on Fig. 1.
A Theoretical Framework Assessment Proposal
281
Fig. 1. Supply chain complexity decision-making integrated framework. Source: adapted from [6].
Step 1: identify which model could be adopted as a SCC assessment execution, aligned with the decision-making key drivers. The model presented by Manuj and Sahin [6] was extend and adapted to this work (see Fig. 1). Step 2: develop an evaluation measurement to the respective SC complexity degree drivers: Uncertainty, Variability, Multiplicity, Size, Speed, and Diversity (see Sect. 2.2, Table 2). Step 3: identify the relevant variables, qualitative and quantitative, to be considered on the decision-making process related to the SCC degree definition, using the Balanced Score Card (BSC) drivers as a reference and adding the risk management dimension on the scheme [24]; Step 4: choose the appropriate tool to the decision-making process based on the SCC degree (at this phase limited to a theoretical base platform); Step 5: develop an implementation plan (actions and priorities) based on the assessment framework proposal (follow phase: mathematical basis, case studies and results analysis, application and simulation).
282
S. L. Alvim et al.
5 Test Case and Preliminary Results An organization, so-called ABC (computer manufacturing), faced an inventory issue: raw material amount of *U$K800, as a potential excess and obsolescence due to the product life cycle. Problem Statement The ABC Brazil plant need to convert this excess in saleable product (*1,980 servers), export and delivery on time (until 90 days) to the USA Distributor, assuring the sales out before the product life cycle status change to obsolete. Variables Identification Inventory cost: potential write plus additional raw material investment to conclude the product conversion from raw material: U$K235. Total at risk: U$M 1,035. Logistics costs projection: transportation and warehouse materials handling. Customer (Distributor): receipt and sale of all products before the obsolete life cycle start. Scenarios Scenario 1: No actions and keep the inventory write-off. Business impact of *U$K800. Complexity element: finance. Scenario 2: Work on the solution to avoid the potential business impact. Complexity elements: financial, product life cycle, logistics services and customer satisfaction. SCC Theoretical Assessment Framework Application Scenario 1 represents the real finance impact in the business, without actions. Scenario 2 represents the most appropriate alternative chosen by the supply chain team, applying the assessment framework developed. The preliminary result was discussed and evaluated by the ABC SC specialists. In this context, the complexity elements considered to the framework evaluation were: (1) Finance - with a total investment of U$M 1,035 and revenue projection of U$M 1,609, the complexity level was classified by the specialists as high; (2) Customer - considering an express shipment to the customer warehouse, with enough time to convert the inventory in sales out, the complexity level was defined as high; (3) Internal business process - inbound and outbound logistics and production will absorb the additional workload, and the complexity was classified as high; (4) Learning and growth - the transportation modal definition (cost optimization and delivery performance) was classified as low complexity level, as per the ABC SC team know-how; (5) Risk - the operational risks identified was medium: quality of SC information, volatile demand, and sales assurance in the USA market (product life cycle).
A Theoretical Framework Assessment Proposal
283
Procedure Complexity elements input to determine the SCC degree based on assessment framework model: 1 – low complexity; 3 – medium complexity; and 5 – high complexity. Based on the ABC SC team perception, the complexity degree was assigned to the key elements identified on the practical application. These inputs enable the decision-maker to compare the complexity elements differences, and take the actions aligned by the scale complexity and the business strategy. The complete process construct, inputs related to this case, score calculation and analyses are represented on Fig. 2.
1-Complexity Inputs
Uncertainty Size Emergence Technology Learning and adaptation Customer requirements Organizational diversity Changes Demand fluctuation Product & Services complexity Finance Globalization Structure IT Interoperability Security Speed Variability Logistics structure Lack of information Synchronization Risk
2- SC Complexity
3- SC Complexity Elements
Problem
Internal factors
Variables Inventory Cost
Structure Finance Product SKU Process People Resource Forecast Inventory Distribution model
Material-800k Scrap Complexity
Life Cycle Product
Logistics Services
Customer Satisfaction
Complexity Scenarios Analysis Financial Customer
Operational
Development
Risk
Scenario 1 Scenario 2
External factors
Economy Innovation Environment conditions Supplier Risk CompetitionCustomer request
4- Fundamental x Variable x Tool
5 - Outcome: Decision-making Top Critical Element: Finance (score: 4,55)
Moderators: Business strategies- Human facilitators
Fig. 2. Complexity elements in the framework construct and application. Source: authors.
6 Conclusions This work aimed to develop a theoretical framework assessment model to support the decision-making process on SC network complex environment on a more assertive and pragmatic way. Although the literature presents several studies on SC complexity, this research has combined new elements (complexity score determination based on the specific complex elements of the SC, BSC Drivers) to guide the decision-making process under a complex SC environment.
284
S. L. Alvim et al.
The test case application evidences the importance of the manager role as a guidance to the team during the process to validate the complexity elements measurement, using the proposal matrix elements degree. Does not only the score weight is decisive: the combination of the specialists understanding and experience, about the relevant variables on a complex SC, and the relation with the business strategy are key to support the decision-making process. The model proposed and presented in this research is limited to one scenario with only one supplier and one customer. As a future step, could be developed a more robust framework, adding quantitative values to be evaluated. Despite that, the preliminary framework results show that is possible to improve the SCC decision-making with a more collaborative approach with the other business areas, assuring a complete identification of the relevant variables to be included on the assessment framework. A future framework development would encompass the product life-cycle dimension and how many nodes must be considering on the complexity analysis.
References 1. Blecker, T., Kersten, W., Meyer, C.M.: Development of an approach for analyzing supply chain complexity. In: Blecker, T., Friedrich, G. (Ed.): Mass Customization: Concepts - Tools - Realization. Proceedings of the International Mass Customization Meeting, pp. 47–59 (2005) 2. Ivanov, D.: Structural Dynamics and Resilience in Supply Chain Risk Management, pp. 91– 114. Springer, Cham (2018) 3. Chryssolouris, G., Efthymiou, K., Papakostas, N., Mourtzis, D., Pagoropoulos, A.: Flexibility and complexity: is it a trade-off? Int. J. Prod. Res. 51(23–24), 6788–6802 (2013) 4. Aelker, J., Bauernhansl, T., Ehm, H.: Managing complexity in supply chains: a discussion of current approaches on the example of the semiconductor industry. Procedia CIRP 7, 79–84 (2013) 5. Brandon-Jones, E., Squire, B., Van Rossenberg, Y.G.: The impact of supply base complexity on disruptions and performance: the moderating effects of slack and visibility. Int. J. Prod. Res. 53(22), 6903–6918 (2015) 6. Manuj, I., Sahin, F.: A model of supply chain and supply chain decision-making complexity. Int. J. Phys. Distrib. Logist. Manage. 41(5), 511–549 (2011) 7. Modrak, V., Helo, P.T., Matt, D.T.: Complexity measures and models in supply chain networks. Complexity 2018 (2018) 8. Montoya, G.N., Santos Jr., J.B.S., Novaes, A.G., Lima, O.F.: Internet of Things and the risk management approach in the pharmaceutical supply chain. In: International Conference on Dynamics in Logistics, pp. 284–288 (2018) 9. Marchi, J.J., Erdmann, R.H., Rodriguez, C.M.T.: Understanding supply networks from complex adaptive systems. BAR-Braz. Adm. Rev. 11(4), 441–454 (2014) 10. Efstathiou, J., Calinescu, A., Blackburn, G.: A web-based expert system to assess the complexity of manufacturing organizations. Rob. Comput. Integr. Manuf. 18(3–4), 305–311 (2002) 11. Raaidi, S., Bouhaddou, I., Benghabrit, A.: Is supply chain a complex system? In: MATEC Web of Conferences, EDP Sciences, p. 00018 (2018) 12. De Leeuw, S., Grotenhuis, R., Van Goor, A.R.: Assessing complexity of supply chains: evidence from wholesalers. Int. J. Oper. Prod. Manage. 33(8), 960–980 (2013)
A Theoretical Framework Assessment Proposal
285
13. Wilding, R.: The supply chain complexity triangle: uncertainty generation in the supply chain. Int. J. Phys. Distrib. Logist. Manage. 28(8), 599–616 (1998) 14. Gerschberger, M., Manuj, I., Freinberger, P.: Investigating supplier-induced complexity in supply chains. Int. J. Phys. Distrib. Logist. Manage. 47(8), 688–711 (2017) 15. Wu, Y.R., Huatuco, L.H., Frizelle, G., Smart, J.: A method for analysing operational complexity in supply chains. J. Oper. Res. Soc. 64(5), 654–667 (2013) 16. Bode, C., Wagner, S.M.: Structural drivers of upstream supply chain complexity and the frequency of supply chain disruptions. J. Oper. Manage. 36, 215–228 (2015) 17. Aitken, J., Bozarth, C., Garn, W.: To eliminate or absorb supply chain complexity: a conceptual model and case study. Supply Chain Manage. Int. J. 21(6), 759–774 (2016) 18. Tang, C.S.: Perspectives in supply chain risk management. Int. J. Prod. Econ. 103, 451–488 (2006) 19. Manuj, I., Mentzer, J.T.: Global supply chain risk management strategies. Int. J. Phys. Distrib. Logist. Manage. 38, 192–223 (2008) 20. Bask, A.H.: Relationships among TPL providers and members of supply chains: a strategic perspective. J. Bus. Ind. Mark. 16, 470–486 (2001) 21. Fantazy, K.A., Kumar, V., Kumar, U.: An empirical study of the relationships among strategy, flexibility, and performance in the supply chain context. Supply Chain Manage. 14, 177–188 (2009) 22. Zelbst, P.J., Green Jr., K.W., Sower, V.E., Reyes, P.: Impact of supply chain linkages on supply chain performance. Ind. Manage. Data Syst. 109, 665–682 (2009) 23. Malhotra, N., Birks, D.: Marketing Research: An Applied Approach, 3rd edn. Pearson Education, Harlow (2007) 24. Kaplan, R.S., Norton, D.P.: The Balanced Scorecard: Translating Strategy into Action, 1st edn. Harvard Business Review Press, Brighton (1996)
Supply Chain Integration: A Bibliometric Analysis Herbert Kotzab1,2 1
, Ilja Bäumler2(&)
, and Paul Gerken2
Universiti Utara Malaysia, Sintok, Malaysia 2 Universität Bremen, Bremen, Germany [email protected]
Abstract. This paper examines the intellectual foundation of supply chain integration (SCI) and presents the most influential papers and authors of this research domain. The paper displays a visualization of the results of citation and co-citation analysis by using the software packages HistCite and VosViewer. The results of our analyses display a profound theoretical foundation of SCI embedded in a clearly defined theoretical field driven by the dynamic capability, relational as well as resource-based view. SCI research is further driven by empirical-quantitative research looking at the effects of SCI on firm performance. Keywords: Supply chain integration Co-citation analysis
Foundation Citation analysis
1 Introduction 1.1
Starting Points of Consideration
Supply Chain Management (SCM) refers to the internal and external integration of business processes in order to increase customer value (e.g. Frohlich and Westbrook 2001) and becomes therefore important since competition changed from company against company to company’s network against another company’s network (see Kotzab et al. 2015). This integration of business processes goes in a forward direction (downstream to customers) as well as into a reverse direction (upstream to suppliers) with the aim to optimize a whole entity instead of a single part of the chain (Cooper et al. 1997; Cooper and Ellram 1993; Heikkilä 2002). Overall, the construct of integration plays a major role in the whole SCM-discussion as Mouritsen et al. (2003) already pinpointed by identifying SCI as the key prerequisite of SCM. This is supported by Frohlich and Westbrook (2001) who refer to the strategic importance of SCI by highlighting the value of integrating suppliers, manufacturers and customers. 1.2
Research Objectives and Methodology
In order to be successful in a business environment that has been changed by globalization and digitalization, value creation processes must be considered beyond the respective company boundaries and analyzed in their entirety. This requires the integration of the business processes of all participants in a supply chain (SC) © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 286–298, 2020. https://doi.org/10.1007/978-3-030-44783-0_28
Supply Chain Integration: A Bibliometric Analysis
287
(Lee et al. 2004). However, this requires the integration and coordination of the various actors (Li et al. 2009). So far, a lot of research (e.g. Fabbe-Costes and Jahre 2008; Flynn et al. 2010; Gimenez et al. 2012; Leuschner et al. 2013; Prajogo and Olhager 2012; Wong et al. 2011) examined the impact of SCI on performance (of the firm) and identify thereby a positive relationship between these two variables. On the contrary research concludes, that there is lack of clear evidence and inconclusive results whether SCI improves SC performance (Chavez et al. 2015; Fabbe‐Costes and Jahre 2007). However, the goal of this paper is to go into depth with the construct of SCI by identifying and mapping the intellectual foundation of SCI research domain. Thereby, we diagnose the most influential works, portray their interrelationships and reveal citation clusters/themes which research regularly draws upon (e.g. White and McCain 1998). For doing this, we collected and analyzed data from the Web of Science Core Collection (for a detailed description of the search methodology see Table 1).
Table 1. Search Methodology Refinement step 1. Search TOPIC (= in title, abstract and author supplied keyword) “supply chain integration” in Web of Science Core Collection (no time limitation) 2. Refined by Web of Science Categories: “Management” OR “Business” OR “Economics” OR “Operations Research Management Science” 3. Refined by Document Type: “ARTICLE” 4. Refined by Source Titles: Select only Journals with 2 Star ABS Journal Rating or higher OR 2nd best score or higher in at least 9 out of the 12 rankings in the Harzing list (38 Journals – see Appendix 1)
Results 7.506 4.065 2.967 1.717
We used the latest Harzing List (Harzing 2019) as well as the 2018 ABS Journal Guide for selecting our relevant journals (Chartered Association of Business Schools 2018). Thereby we used only those journals of the ABS categories “Operations and Technology Management” and “Operations Research and Management Science” which have at least a 2-star rating result or those which have the 2nd best score or higher in at least 9 out of the 12 rankings in the Harzing list. The final data set consists of 1.717 articles by more than 3.000 authors. This sample includes more than 53.000 cited references which were further examined with two bibliometric software tools namely HistCite (Garfield 2009a) and VOSviewer (Eck and Waltman 2010) in order to receive both, analytical as well as visualized results for citation and co-citation analyses. The remainder of the paper is as follows. After introducing the research objective and the general methodological considerations, chapter 2 presents the theoretical as well as practical importance and aspects of SCI. Thereafter we document in chapter 3
288
H. Kotzab et al.
the results of our bibliometric analyses and identify the roots and intellectual foundation of SCI. The paper closes with a discussion of our findings following a statement about theoretical consequences for the SCI community and an outlook for future research.
2 What is Supply Chain Integration? SCI is defined as the degree to which a supply chain actor enters into strategic cooperation with other supply chain actors and to what extent they control inter- and intra-organizational processes in a collaborative manner (Jayaram et al. 2010; Schoenherr and Swink 2012; Wiengarten et al. 2016). The aim is thereby to achieve the most effective and efficient flow of products, information and finance, so that maximum added value is offered to the end customer (Flynn et al. 2010). Obviously, SCI concepts consider the flow of materials and information along a value chain. Ideally, the boundaries between the activities of the respective organizations should flow smoothly into each other and no longer be separated for specific organizations. As already mentioned, there are two directions of integration, forward integration of physical flows of goods from suppliers to customers and backward integration of the data flow from customers to suppliers (Prajogo and Olhager 2012). Furthermore, SCI can be divided into internal and external integration where internal integration is based on the consolidation and synchronization of internal company processes and external integration refers to the cross-company merging of inter-organizational strategies and processes (Flynn et al. 2010). In order to achieve an optimal level of SC integration the following six different dimensions have to be considered (Stank et al. 2001): customer service, internal integration, material and service supplier, technology and planning, measurement and relationship. Another distinction in SCI is given by Mouritsen et al. (2003) who differ between information integration and organizational integration. A high degree of informational SCI is characterized by increased logistics-related communication between the individual actors as well as improved coordination of an organization’s logistics activities between suppliers and customers (Schoenherr and Swink 2012). On the one hand, a high degree of information integration can achieve a variety of benefits. These can be reduced product or service costs, the creation of a sustainable competitive advantage, reduced complexity, reduced lead times and increased flexibility in production and delivery. In addition, higher reliability, better inventory management and a better understanding of the end customer’s needs can be achieved (Korpela et al. 2017; Stank et al. 2001). This allows manufacturers to respond more flexibly to individual customer needs, delivery times can be shortened and inventory can be minimized, which contributes significantly to the efficiency of a SC. On the other hand, a low degree of
Supply Chain Integration: A Bibliometric Analysis
289
information integration leads to the so-called bullwhip effect, which has been widely discussed for several years (Lee et al. 1997). Since it is not feasible for companies alone to establish end-to-end information integration along the SC, companies must establish collaborative relationships. The advantages of SC integration with the help of digital technologies results in reduction of transaction cost due to less inter- and intra-corporate exchange (Korpela et al. 2017). The exchange of information using IT systems makes it possible to disseminate more information within a shorter time (Prajogo and Olhager 2012). In order to achieve an increase in the overall performance, companies should therefore focus on information integration. This is achieved by sharing critical information, both strategically and operationally, within a SC network by means of IT (Prajogo and Olhager 2012). For an effective integration of business processes, it is essential to share, for example, tracking data or customer demand information electronically between organizations along the SC. The organizational integration is therefore important as the supply chain is considered to be a whole entity and decision-making is not carried out from an individual supply chain actor’s point of view but from the whole supply chain perspective (see e.g. Cooper et al. 1997). Successful organizational integration requires a high degree of mutual trust between the supply chain actors (see e.g. Skjoett‐Larsen 2000). Overall, organizational integration is seen as the facilitator of sharing activities between the members of a supply chain (Mouritsen et al. 2003). Taking all these aspects into account, we consider SCI as an accepted research domain or area within the field of SCM which allows to further examine the roots of this particular domain.
3 Results Since 1995, more than 1,700 SCI-specific papers were published in the 38 journals, that we examined. Since 2010, more than 100 papers (except in 2012 and so far in 2019) were annually issued. The most productive authors (in terms of number of publications) are Zhao XD (28 papers), Huo BF (23 papers), Gunasekaran (18 papers), Jayaram (16 papers) and Wong CWY (15 papers). The most important institutions in terms of number of publications are Michigan State University (57 papers), Hong Kong Polytech University (52 papers), Arizona State University (32 papers) as well as Politecnio Milano and Zheijiang University (with each 28 papers). In Fig. 1 we can see the historiography of the citation relations of the 30 most cited papers of our sample based on the local citation score (LCS = Number of citations to the paper from within the collection; see Garfield (2009b)) as identified by the HistCite software (see also Appendix 2). The publication years of these papers span from 1997 to 2012 and their citation relations show 30 nodes with 95 links with a minimum citation count of 41 and a maximum of 345.
290
H. Kotzab et al.
Fig. 1. Citation relations of the 30 most cited CCC papers as indicated by HistCite
Overall, nearly all papers represent kind of two citation networks, where Frohlich and Westbrook (2001) is the connector. There is one isolated reference (Carter and Rogers 2008) that presents a framework for sustainable SCM, while the remaining 29 papers deal either with particular SCI aspects on operational or overall performance of a firm or with SCM/Logistics issues. Figures 2 and 3 present the results of a co-citation analysis for references and for sources as identified by the VOSviewer software tool. The distances between the respective objects in both figures relate to the similarity of the objects (Eck 2011). We are able to identify in both cases three clusters. The first cluster ‘Methodology/Theory’ refers to 11 red-dotted publications at the right side of Fig. 2 and includes solely papers in regards to processing quantitative empirical research (e.g. structural equation modeling) and ensuring valid results as well as papers representing a clear theoretical position, here dynamic capabilities, relational view as well as the resource-based view (Barney 1991; Dyer and Singh 1998; Teece et al. 1997). The second cluster ‘SCM/SCI’ (11 green dotted, left side of Fig. 2) represents papers rather dealing with general aspects of SCI as well as with supplier integration, supply chain collaboration and the bullwhip effect. It also includes one methodological paper related to case study research. The third cluster ‘SCI-Performance linkage’ (8 blue dotted, upper part of Fig. 2) includes mainly papers which deal with the examination on the effects of SCI on performance. The majority of the papers in clusters 2 and 3 stem from the Journal of Operations Management being one of the most prestigious journals in the area. Interestingly enough we are able to identify direct citation linkages between all clusters, which can also be verified by the co-citation patterns of the 30 most cited journal outlets (see Fig. 3).
Supply Chain Integration: A Bibliometric Analysis
291
Fig. 2. Network of the 30 most co-cited articles (see also Table 2)
Table 2. 30 most co-cited papers (alphabetical order) Red Cluster 1: ‘Methodology’
Green Cluster 2: ‘SCM-SCI’
Blue Cluster 3: ‘SCI-Performance’
Anderson and Gerbing (1988); Armstrong and Overton (1977); Barney (1991); Dyer and Singh (1998); Fornell and Larcker (1981); Hu and Bentler (1999); Nunnally (1978); Podsakoff et al. (2003); Podsakoff and Organ (1986); Rai et al. (2006); Teece et al. (1997) Chen and Paulraj (2004); Eisenhardt (1989); Fisher (1997); Frohlich and Westbrook (2001); Lee et al. (1997); Pagell (2004); Petersen et al. (2005); van der Vaart and van Donk (2008); Vickery et al. (2003); Mentzer et al. (2001); Stank et al. (2001) Devaraj et al. (2007); Droge et al. (2004); Flynn et al. (2010); Koufteros et al. (2005); Narasimhan and Kim (2002); Rosenzweig et al. (2003); Swink et al. (2007); Wong et al. (2011)
Also, here we see three interlinked journal clusters out of which one contains more than half of the journals (17) and is consequently larger than the two others (representing 9 and 4 journals). The red dotted cluster at the left side of Fig. 3 stands for management journals including marketing, information systems, product innovation, organization, strategic management as well as operations management journals.
292
H. Kotzab et al.
Fig. 3. Network of the 30 most co-cited publication outlets (see also Table 3)
Table 3. 30 most co-cited journals (alphabetical order) Red Cluster ‘Management oriented SCI’
Green Cluster ‘ProductionSCI’
Blue Cluster ‘Logistics/SCM-SCI’
Academy of Management Journal; Academy of Management Review; Administrative Science Quarterly; Decision Science; Harvard Business Review; Industrial Marketing Management; Information Systems Research; Journal of Business Research; Journal of Management; Journal of Marketing Research; Journal of Marketing; Journal of Operations Management; Journal of Production & Innovation Management; Management Science; MIS Quarterly; Organisation Science; Strategic Management Journal European Journal of Operational Research; International Journal of Production & Operations Management; International Journal of Production Economics; International Journal of Production Research; Journal of Cleaner Production Omega; Production and Operations Management; Production Planning and Control; Supply Chain Management: An International Journal; International Journal of Operations & Production Management International Journal of Physical Distribution & Logistics Management; International Journal of Logistics Management; Journal of Business Logistics; Journal of Supply Chain Management
Supply Chain Integration: A Bibliometric Analysis
293
The strongest journal here is represented by the Journal of Operations Management. Opposite to this cluster on the right side of Fig. 3 (green dotted journals) are production planning, operations research and production economics journals representing the manufacturing/production perspective of integration. Interestingly enough it contains also one more SCM-specific journal. The four journals (blue dotted) in the upper part of Fig. 3 represent the leading logistics/SCM-specific journals.
4 Discussion and Conclusion The most influential work for the domain of SCI is the work by Frohlich and Westbrook (2001; citation count of 345) who basically started the discussion on SCI. The next 11 papers though refer to the theoretical as well as methodological foundation of the SCI research domain, which is clearly characterized as being quantitative-empirical. The theoretical fundament of SCI research is found in the resource-based, relational and dynamic capabilities view of the organization. This means that the overall foundation of SCI is found inside the firm which is looking for the adequate external setting for successfully integrate upstream and downstream. The dominating research question relates to the examination of the effects of SCI on the performance of the firm or a specific firm function. Researchers in the field of SCI use mainly the Journal of Operations Management as their knowledge hub, followed by the International Journal of Production Economics, the International Journal of Production and Operations Management and Management Science which all represent a clear Operations Management view. The first logistics/SCM-specific journal comes with Supply Chain Management: An International Journal. However, the domain of SCI has a broad journal fundament from the leading academic management journals as well as leading specific area journals. Our findings offer several positive consequences for the scientific SCI community: The results can confirm, contradict or even suggest notions for beginners or experts in the field of SCI by providing the intellectual foundation in terms of authors, papers, journals and thematical citation clusters of SCI research. Thus, one can easily distinguish, relate and prioritize findings in literature research. Furthermore, our paper provides insight into the relevant SCI research communities by providing a list of appropriate journals. As bibliometric analyses are built upon available data as well as the constraints by the authors’ literature search, our paper has some limitations. First, even though we eliminated typos and other errors (e.g. different journal labels for the same journal) and improved the data quality of the data set, it cannot be concluded that the elimination of every possible typographical difference or mistake was achieved. Although, different available automatic algorithmic based correction packages (e.g. algorithms provided by “OpenRefine” or “VOSviewer”) were used, some citations still had to be corrected and equalized manually. Second, due to the nature of quantitative analyses, like bibliometric analysis, simplification and generalization is achieved, in most cases, at the expense of information preciseness. Beside the possibility to measure the relatedness of papers on the basis of the co-citations, as we did, it is also possible to measure the relatedness of papers on the basis of the number of words that occur in both documents. Second approach might reveal further insight in the field of SCI. This could be subject to future research.
294
H. Kotzab et al.
Appendix 1 Included Journals based on Harzing Quality List (2019) and Academic Journal Guide (2018) No.
Source title (sorted in alphabetical order)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
Annals of Operations Research Business Process Management Journal Computers Operations Research Decision Sciences European Journal of Industrial Engineering European Journal of Operational Research IEEE Transactions on Engineering Management Industrial and Corporate Change Industrial Marketing Management Information Management Information Systems Research Interfaces International Journal of Computer Integrated Manufacturing International Journal of Electronic Commerce International Journal of Operations Production Management International Journal of Physical Distribution Logistics Management International Journal of Production Economics International Journal of Production Research International Journal of Quality Reliability Management International Journal of Technology Management Journal of Business Logistics Journal of Business Research Journal of Management Information Systems Journal of Operations Management Journal of Product Innovation Management Journal of Purchasing and Supply Management Journal of Strategic Information Systems Journal of Supply Chain Management Journal of the Operational Research Society Management Science Manufacturing Service Operations Management Omega International Journal of Management Science Production and Operations Management Production Planning Control Research Policy Strategic Management Journal Supply Chain Management an International Journal Total Quality Management Business Excellence
Supply Chain Integration: A Bibliometric Analysis
Appendix 2 30 most cited papers based on the local citation score (LCS)
295
296
H. Kotzab et al.
References Anderson, J., Gerbing, D.W.: Structural equation modeling in practice: a review and recommended two-step approach. Psychol. Bull. 103, 411–423 (1988). https://doi.org/10. 1037/0033-2909.103.3.411 Armstrong, J.S., Overton, T.S.: Estimating nonresponse bias in mail surveys. J. Mark. Res. 14, 396–402 (1977). https://doi.org/10.2307/3150783 Barney, J.: Firm resources and sustained competitive advantage. J. Manag. 17, 99–120 (1991). https://doi.org/10.1177/014920639101700108 Chartered Association of Business Schools: Academic Journal Guide 2018 (2018) Chavez, R., Yu, W., Gimenez, C., Fynes, B., Wiengarten, F.: Customer integration and operational performance: The mediating role of information quality. Decis. Support Syst. 80, 83–95 (2015). https://doi.org/10.1016/j.dss.2015.10.001 Chen, I.J., Paulraj, A.: Towards a theory of supply chain management: the constructs and measurements. J. Oper. Manag. 22, 119–150 (2004). https://doi.org/10.1016/j.jom.2003.12. 007 Cooper, M.C., Ellram, L.M.: Characteristics of supply chain management and the implications for purchasing and logistics strategy. Int. J. Logist. Manag. (1993). https://doi.org/10.1108/ 09574099310804957 Cooper, M.C., Lambert, D.M., Pagh, J.D.: Supply chain management: more than a new name for logistics. Int. J. Logist. Manag. (1997). https://doi.org/10.1108/09574099710805556 Devaraj, S., Krajewski, L., Wei, J.C.: Impact of eBusiness technologies on operational performance: The role of production information integration in the supply chain. J. Oper. Manag. 25, 1199–1216 (2007). https://doi.org/10.1016/j.jom.2007.01.002 Droge, C., Jayaram, J., Vickery, S.K.: The effects of internal versus external integration practices on time-based performance and overall firm performance. J. Oper. Manag. 22, 557–573 (2004). https://doi.org/10.1016/j.jom.2004.08.001 Dyer, J.H., Singh, H.: The relational view: cooperative strategy and sources of interorganizational competitive advantage. Acad. Manag. Rev. 23, 660–679 (1998). https://doi.org/10. 2307/259056 van Eck, N.J.: Methodological Advances in Bibliometric Mapping of Science. Erasmus Research Institute of Management (ERIM), Erasmus University Rotterdam, Rotterdam (2011) van Eck, N.J., Waltman, L.: Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics 84, 523–538 (2010). https://doi.org/10.1007/s11192-009-0146-3 Eisenhardt, K.M.: Building theories from case study research. Acad. Manag. Rev. 14, 532 (1989). https://doi.org/10.2307/258557 Fabbe-Costes, N., Jahre, M.: Supply chain integration and performance: a review of the evidence. Int. J. Logist. Manag. 19, 130–154 (2008). https://doi.org/10.1108/09574090810895933 Fabbe-Costes, N., Jahre, M.: Supply chain integration improves performance: the Emperor’s new suit? Int. J. Phys. Distrib. Logist. Manag. 37, 835–855 (2007). https://doi.org/10.1108/ 09600030710848941 Fisher, M.: What Is the Right Supply Chain for Your Product? Harvard Business Review (1997) Flynn, B.B., Huo, B., Zhao, X.: The impact of supply chain integration on performance: a contingency and configuration approach. J. Oper. Manag. 28, 58–71 (2010). https://doi.org/ 10.1016/j.jom.2009.06.001 Fornell, C., Larcker, D.F.: Evaluating structural equation models with unobservable variables and measurement error. J. Mark. Res. 18, 39–50 (1981). https://doi.org/10.2307/3151312
Supply Chain Integration: A Bibliometric Analysis
297
Frohlich, M.T., Westbrook, R.: Arcs of integration: an international study of supply chain strategies. J. Oper. Manag. 19, 185–200 (2001). https://doi.org/10.1016/S0272-6963(00) 00055-3 Garfield, E.: From the science of science to Scientometrics visualizing the history of science with HistCite software. J. Inf. Sci. Sci.: Concept. Model. Sci. 3, 173–179 (2009). https://doi.org/10. 1016/j.joi.2009.03.009 Garfield, E.: HistCite Help (2009b) Gimenez, C., Sierra, V., Rodon, J.: Sustainable operations: their impact on the triple bottom line. Int. J. Prod. Econ. Sustain. Dev. Manuf. Serv. 140, 149–159 (2012). https://doi.org/10.1016/j. ijpe.2012.01.035 Harzing, A.-W.: Journal Quality List (57th ed.) (2019) Heikkilä, J.: From supply to demand chain management: efficiency and customer satisfaction. J. Oper. Manag. 20, 747–767 (2002). https://doi.org/10.1016/S0272-6963(02)00038-4 Hu, L., Bentler, P.M.: Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct. Equ. Model.: Multidiscip. J. 6, 1–55 (1999). https:// doi.org/10.1080/10705519909540118 Jayaram, J., Tan, K.-C., Nachiappan, S.P.: Examining the interrelationships between supply chain integration scope and supply chain management efforts. Int. J. Prod. Res. 48, 6837– 6857 (2010). https://doi.org/10.1080/00207540903358329 Korpela, K., Hallikas, J., Dahlberg, T.: Digital supply chain transformation toward blockchain integration. Presented at the Hawaii International Conference on System Sciences (2017). https://doi.org/10.24251/HICSS.2017.506 Kotzab, H., Teller, C., Grant, D.B., Friis, A.: Supply chain management resources, capabilities and execution. Prod. Plan. Control 26, 525–542 (2015). https://doi.org/10.1080/09537287. 2014.927932 Koufteros, X., Vonderembse, M., Jayaram, J.: Internal and external integration for product development: the contingency effects of uncertainty, equivocality, and platform strategy. Decis. Sci. 36, 97–133 (2005). https://doi.org/10.1111/j.1540-5915.2005.00067.x Lee, H.L., Padmanabhan, V., Whang, S.: Information distortion in a supply chain: the bullwhip effect. Manag. Sci. 50, 1875–1886 (2004). https://doi.org/10.1287/mnsc.1040.0266 Lee, H.L., Padmanabhan, V., Whang, S.: The Bullwhip effect in supply chains. Sloan Manag. Rev. 38, 93–102 (1997) Leuschner, R., Rogers, D.S., Charvet, F.F.: A meta-analysis of supply chain integration and firm performance. J. Supply Chain. Manag. 49, 34–57 (2013). https://doi.org/10.1111/jscm.12013 Li, G., Yang, H., Sun, L., Sohal, A.S.: The impact of IT implementation on supply chain integration and performance. Int. J. Prod. Econ. Spec. Issue Oper. Strat. Supply Chain. Manag. 120, 125–138 (2009). https://doi.org/10.1016/j.ijpe.2008.07.017 Mentzer, J.T., DeWitt, W., Keebler, J.S., Min, S., Nix, N.W., Smith, C.D., Zacharia, Z.G.: Defining supply chain management. J. Bus. Logist. 22, 1–25 (2001). https://doi.org/10.1002/j. 2158-1592.2001.tb00001.x Mouritsen, J., Skjøtt-Larsen, T., Kotzab, H.: Exploring the contours of supply chain management. Integr. Mfg Syst. 14, 686–695 (2003). https://doi.org/10.1108/ 09576060310503483 Narasimhan, R., Kim, S.W.: Effect of supply chain integration on the relationship between diversification and performance: evidence from Japanese and Korean firms. J. Oper. Manag. 20, 303–323 (2002). https://doi.org/10.1016/S0272-6963(02)00008-6 Nunnally, J.C.: Psychometric Theory. McGraw-Hill Series in Psychology, 2nd edn. McGrawHill, New York (1978)
298
H. Kotzab et al.
Pagell, M.: Understanding the factors that enable and inhibit the integration of operations, purchasing and logistics. J. Oper. Manag. 22, 459–487 (2004). https://doi.org/10.1016/j.jom. 2004.05.008 Petersen, K.J., Handfield, R.B., Ragatz, G.L.: Supplier integration into new product development: coordinating product, process and supply chain design. J. Oper. Manag. 23, 371–388 (2005). https://doi.org/10.1016/j.jom.2004.07.009 Podsakoff, P.M., MacKenzie, S.B., Lee, J.Y., Podsakoff, N.P.: Common method biases in behavioral research: a critical review of the literature and recommended remedies. J. Appl. Psychol. 88, 879–903 (2003). https://doi.org/10.1037/0021-9010.88.5.879 Podsakoff, P.M., Organ, D.W.: Self-reports in organizational research: problems and prospects. J. Manag. 12, 531–544 (1986). https://doi.org/10.1177/014920638601200408 Prajogo, D., Olhager, J.: Supply chain integration and performance: The effects of long-term relationships, information technology and sharing, and logistics integration. Int. J. Prod. Econ. Adv. Optim. Des. Supply Chains 135, 514–522 (2012). https://doi.org/10.1016/j.ijpe.2011. 09.001 Rai, A., Patnayakuni, R., Seth, N.: Firm performance impacts of digitally enabled supply chain integration capabilities. MIS Q. 30, 225–246 (2006). https://doi.org/10.2307/25148729 Rosenzweig, E.D., Roth, A.V., Dean, J.W.: The influence of an integration strategy on competitive capabilities and business performance: an exploratory study of consumer products manufacturers. J. Oper. Manag. 21, 437–456 (2003). https://doi.org/10.1016/S02726963(03)00037-8 Schoenherr, T., Swink, M.: Revisiting the arcs of integration: cross-validations and extensions. J. Oper. Manag. 30, 99–115 (2012). https://doi.org/10.1016/j.jom.2011.09.001 Skjoett-Larsen, T.: Third party logistics – from an interorganizational point of view. Int. J. Phys. Distrib. Logist. Manag. (2000). https://doi.org/10.1108/09600030010318838 Stank, T.P., Keller, S.B., Closs, D.J.: Performance benefits of supply chain logistical integration. Transp. J. 41, 32–46 (2001) Swink, M., Narasimhan, R., Wang, C.: Managing beyond the factory walls: effects of four types of strategic integration on manufacturing plant performance. J. Oper. Manag. 25, 148–164 (2007). https://doi.org/10.1016/j.jom.2006.02.006 Teece, D.J., Pisano, G., Shuen, A.: Dynamic capabilities and strategic management. Strat. Manag. J. 18, 509–533 (1997). https://doi.org/10.1002/(SICI)1097-0266(199708)18:7% 3c509:AID-SMJ882%3e3.0.CO;2-Z van der Vaart, T., van Donk, D.P.: A critical review of survey-based research in supply chain integration. Int. J. Prod. Econ. 111, 42–55 (2008). https://doi.org/10.1016/j.ijpe.2006.10.011 Vickery, S.K., Jayaram, J., Droge, C., Calantone, R.: The effects of an integrative supply chain strategy on customer service and financial performance: an analysis of direct versus indirect relationships. J. Oper. Manag. 21, 523–539 (2003). https://doi.org/10.1016/j.jom.2003.02.002 White, H.D., McCain, K.W.: Visualizing a discipline: an author co-citation analysis of information science, 1972–1995. J. Am. Soc. Inf. Sci. 49, 327–355 (1998). https://doi.org/10. 1002/(SICI)1097-4571(19980401)49:4%3c327:AID-ASI4%3e3.0.CO;2-4 Wiengarten, F., Humphreys, P., Gimenez, C., McIvor, R.: Risk, risk management practices, and the success of supply chain integration. Int. J. Prod. Econ. 171, 361–370 (2016). https://doi. org/10.1016/j.ijpe.2015.03.020 Wong, C.Y., Boon-itt, S., Wong, C.W.Y.: The contingency effects of environmental uncertainty on the relationship between supply chain integration and operational performance. J. Oper. Manag. 29, 604–615 (2011). https://doi.org/10.1016/j.jom.2011.01.003
Development of a Decision Support Model for Managing Supply Chain Design Problems in Global Service Supply Chains Juri Reich1,2(&), Tina Wakolbinger1, and Aseem Kinra2 1
Vienna University of Economics and Business, Vienna, Austria [email protected] 2 University of Bremen, Bremen, Germany
Abstract. Most large service manufacturing companies outsource support as well as core activities to remain competitive in a globalized economy. Managers are faced with a complex decision problem as they must find a balance between maximizing the benefit from business process outsourcing while keeping risk as low possible. Many outsourcing initiatives failed in the past. Research suggests that one of the main reasons is poor decision making with an insufficient regard of all relevant factors, especially qualitative ones. We look at the problem from a supply chain design perspective and further develop the application of decision support methods originally developed with a focus on product manufacturing, to the manufacturing of services. We review existing approaches and develop our own conceptual decision framework. Keywords: Decision support Supply chain design Outsourcing Service supply chain
Business processes
1 Introduction Large-scale global outsourcing emerged in manufacturing after world war two, continued through information technology and now becomes increasingly prevalent for business processes [1], particularly in service industries [2]. Global supply chain management emphasizes the management of globally dispersed supply chains that emerge from such outsourcing and offshoring activities. Managers need to decide how to organize cooperation between different entities and where to locate which parts of the supply chain. Global supply chain design decisions are highly complex as they are characterized by conflicting goals and a vast number of heterogeneous input factors. A good understanding of decision problems and decision support tools is crucial to the success of Business Process Outsourcing BPO in global supply chains. Businesses tend to underestimate the impact of outsourcing and offshoring decisions and often base their decisions on simplistic quantitative considerations, mainly cost [3]. In contrast, qualitative criteria are often ignored, even though they have been found to be more important than quantitative ones [4] – possibly because quantitative information is easier to handle, whereas processing qualitative factors requires special know-how and sophisticated analysis tools. It is thus no surprise that recently the terms “reshoring” © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 299–308, 2020. https://doi.org/10.1007/978-3-030-44783-0_29
300
J. Reich et al.
and “backshoring” gained prominence. Srai and Ane [5] found that the most common reason for backshoring are quality concerns, supporting the notion that backshoring decisions are usually not long-term adjustments to changed conditions in the business environment, but rather costly short-term corrections of poor offshoring decisions – often due to insufficient consideration of qualitative factors [3]. Mani, Barua and Whinston [6] found that the success of business process network design correlates with the performance of information systems used by a company. Ellram et al. [2] found that the extent by which service companies engage in BPO to gain competitive advantage correlates with the usage of decision tools to control these operations. An improvement of decision support tools and higher degree of usage by managers could have a significantly positive effect on the performance of service companies. Johnston [7] suggests that service research should be integrated with operations management to develop new methods to improve performance and efficiency of service supply chains. The purpose of our research is to explore how mathematical optimization methods from the Operations Management domain that were developed with a product manufacturing focus can be employed in a services context to solve global supply chain (network) design problems that arise in BPO. This paper addresses the following research question: “How can existing supply chain management optimization methods be enhanced and adapted to create an effective and easy to use decision support tool for global business process outsourcing in service industries?” In this paper we propose the development of a decision support model that consists of three well-established methods combined in an innovative way. We start with a short literature review of similar studies in the service supply chain context, and bring out the gaps. We then lay out the design of a new decision support framework and finally conclude with proposed steps for further model development, including application and validation.
2 Literature Review There is some literature related to the use of decision analysis in service supply chains with a focus on the design of business process networks. However, it does not cover the complexity of the decision analysis in its entirety. Iannou, Karakerezis and Mavri [8] created a linear programming model to optimize performance of a banking network. The model decides on the number and location of branches as well as on the range of services offered by each. Their approach gives a precise mathematical solution to the network design problem as it clearly states which branches to open where and with what service offering. It thus integrates facility location, service selection and capacity allocation. However, it has some limitations. The model conducts a single objective optimization and offers no possibility for tradeoffs. It only measures “overall performance” instead of financial indicators such as maximum profit, which would potentially make the results more convincing to managers. It also only incorporates quantitative performance indicators and does not include qualitative information.
Development of a Decision Support Model for Alleviating Supply
301
Piplani and Saraswat [9] created a mixed-integer linear programming model to minimize cost in a multinational reverse-logistics network offering repair and refurbishment services for electronic devices. The model covers the entire value chain of activities from return, over repair and refurbishment to re-selling or disposal. It decides on locations for each activity based on cost factors as well as service demand and market price for refurbished products. It contains multiple stages and several possible product flows – devices can be re-sold, sent back to the customer, or disposed of. Just as the prior model however it is a single-objective approach and despite covering an explicitly global problem, it does not include any factors related to transnational complexity, such as worker skill, language or communication. Henao et al. [10] developed a mixed-integer linear programming tool that addresses the problem of variability and seasonality. Employees are hired long-term and paid continuously, while demand for their specific skill set is often volatile. The program minimizes total personnel cost by deciding on the optimum set of skills that each employee is taught. Total cost is comprised of training cost, cost of employee shortage as opportunity cost of lost sales and cost of employee surplus as idle work hours. Each employee has a fixed time capacity that is allocated to serve demand in the departments they are trained for. This model is a straightforward “translation” of the classic capacitated production allocation problem from product to the service manufacturing sector. Just as the previous two approaches it only includes a single objective and does not consider any qualitative factors. Several similar approaches that employ mathematical optimization methods from SCM to business process network design exist. However, none of these include multiple objectives or consider qualitative information, neither do they provide visualizations or show different options and trade-offs. Research of Schuff et al. [11] among pioneers of the decision support domain shows a general disappointment about how little acceptance their tools receive in business. They conclude that decision support systems must become more transparent, visual, interactive and usable. The systems perceived as the most effective by managers are transparent regarding their underlying mechanisms and allow to explore the entire decision space, instead of only delivering a single, presumably optimum solution. As business complexity increases, data visualization becomes more important. Furthermore, systems must become more interactive and involve the users in their design and setup [11]. In the next section we will propose the development of a decision support framework for global business process network design that includes abovementioned features.
3 Model Development 3.1
Introduction
We recognize the challenge to design decision support tools and information systems suited to solving BPO problems. We look at the task from an SCM point of view that is traditionally focused on quantitative optimization methods. We consider business processes as value chains for non-physical goods. A request triggers the creation of a
302
J. Reich et al.
“case”, which is passed on and worked upon throughout a chain of back office activities until it is ready and creates value for the customer. These back-office tasks do not depend on a specific location and require no direct customer interaction. Thus, they are potential candidates for offshoring and outsourcing. At the same time, claims processing is regarded as the traditional “moment of truth” in any client relationship. It is essential to perform it at the highest quality level to not jeopardize customer satisfaction. Managers must design business process networks to supply the customer with the combined efforts of multiple actors. On a tactical level, they must decide which resource delivers what service at which place and time to whom. On a strategic level, they must figure out how to leverage the opportunities provided by globalization to design their business process network to be as cost-efficient as possible, yet keep quality standards high. They need to find a balance between maximizing value by leveraging the expected benefits of offshoring and minimizing risk by mitigating its possible downsides [12]. Such decision problem is complex as it includes conflicting goals and requires dealing with a magnitude of quantitative and qualitative factors as well as decision and information variables. 3.2
Basic Setup
The design of the decision support framework is now laid out. It shall give a precise mathematical solution to a global business process network design problem by allocating process activities to different locations and calculating the total network cost. It should also act as a decision support tool in the literal sense by making transparent the trade-off between risks and benefits, considering all relevant quantitative and qualitative information and calculating the entire range of efficient outcomes. Managers should be able to discuss and select their preferred configuration based on their expertise and strategy. We assume three different back office processes which are triggered by customer demand. Which activities are required to be performed and how many minutes the activity takes on average, depends on the process. From the past we know an expected demand for each process and thus also the demand for each activity. We also know the average duration per activity and process. We now need to assign this demand to a set of possible locations. These include the headquarters as well as third-party shared service centers. The locations differ in costs, which include variable costs per time unit, including wages and taxes, assumed collaboration and friction costs which are incurred if an unfinished case is moved from one location to another for the next activity, fixed costs that are incurred if a location is used at all and activity set-up costs which are incurred if a location is to offer a certain activity, for example for software licenses and employee training. The set-up is similar to a logistics network problem from SCM, but instead of an unfinished product being transported through production facilities, a case is routed via arches through shared services centers that perform a range of activities (Fig. 1).
Development of a Decision Support Model for Alleviating Supply
303
Fig. 1. Basic structure of our problem
Our framework is based on a mixed-integer linear programming (MILP) model. This method allows us to precisely depict and solve a complex network design problem. The target function reflects our first goal, minimizing cost through crossborder labor arbitrage. Our second goal, risk minimization is reflected by another method, the analytical hierarchy process (AHP) of Saaty [13]. The AHP allows us to incorporate qualitative information to describe and rate all relevant risk factors. It has no limitations regarding data types. Qualitative and quantitative information can be assessed simultaneously. Its result is a risk score for each possible location. The AHP score is integrated as a soft constraint into the MILP and gradually raised. This way the program calculates all efficient trade-offs between cost and risk. All tradeoffs connected form the Pareto frontier, which represents the range of efficient solutions to a multi-objective decision problem. Managers can use the Pareto frontier to discuss the relationship between cost and risk. They can conduct sensitivity and scenario analysis. And most important, managers can now see the entire solution space, based on which they can discuss and select the network configuration best suited to their corporate strategy. Below we will describe the model in detail (Fig. 2). 3.3
The Mixed-Integer Linear Program (MILP)
(1)
Fig. 2. The MILP target function
304
J. Reich et al.
The target function minimizes the sum of all four cost types. The first is labor costs, described by the instances of all activities performed for all processes in all locations, multiplied by the duration of each activity per process and the wage costs in the location it is performed in. The second type is collaboration costs that are incurred when a case is moved from one location to another. This can include any friction such as time loss or need for increased managerial effort. The estimated friction cost per case that is moved from one location to another is multiplied by the number of cases moved between these locations across all activities and processes. The third element is the activity setup cost which is incurred whenever a location is offering capacity for a certain activity. The last element is the fixed location cost which is incurred whenever a location offers any capacity at all (Table 1). Table 1. Sets and parameters Item P A L a′ a″ l′ l″ D (p, a) K (a, l) X (p, a″, l″) W (a, l) T (p, a) C (l′, l″) Act (a, l) Fix (l) RISK (l)
Description Processes Activities Locations Previous activity (upstream) Subsequent activity (downstream) Location of previous activity Location of subsequent activity Demand D of instances of activity a performed in process p Maximum capacity K in minutes of location l to perform activity a Instances X of activity a performed in location l for process p Hourly wage W for activity a in location l Time in minutes T required for activity a in process p Collaboration cost C when moving a case from location l′ downstream next to location l″ Setup cost for activity a in location l Fixed cost for location l Risk associated with outsourcing business processes to location l reflected as an AHP preference score (the lower the score, the higher the risk)
X
Xp;a;l ¼ Dp;a
ð2Þ
Xp;a;l Tp;a Ka;l
ð3Þ
l
X p
Xp;a0 ;l0 ¼
X l
Xp;a00 ;l00
ð4Þ
Development of a Decision Support Model for Alleviating Supply
Yl M
X
Ya;l M
p;a
X p
305
Xp;a00 ;l00 0
ð5Þ
Xp;a00 ;l00 0
ð6Þ
8Xp;a0 ;a00 ;l0 ;l00 2 N
ð7Þ
8Ya;l 2 f0; 1g
ð8Þ
8Yl 2 f0; 1g
ð9Þ
Constraint (2) ensures that all demand D of activity a for process p is satisfied in any of the locations l. Constraint (3) ensures that the working time capacity of location l for activity a is not exceeded. Constraint (4) requires all instances of a process worked on in a location to be passed on to any location for the next activity, if there is one. Constraint (5) ensures that if a location is to accommodate any activity, the binary Variable Y(l) must be set to 1, incurring the fixed costs for this location. Constraint (6) does the same for enabling a location to offer a certain activity, incurring an activity setup cost. Constraint (7) requires the decision variable X to be a positive integer, while constraints (8) and (9) require decision variables Y to be binary. 3.4
The Analytical Hierarchy Process (AHP)
The AHP is used to assess all risk factors related to offshoring and outsourcing activities. The decision criteria should be selected and hierarchically arranged in collaboration with the decision-maker to make sure that the risk assessment reflects the characteristics of the specific problem, industry and environment. We looked at existing AHP applications in the BPO context [14, 15], to create an exemplary AHP. Such model could also be used as the initial basis for discussion with decisionmakers in an actual application. It is depicted in Fig. 3. We are aware that these factors might to some extent be overlapping and interrelated. We also assume that the individual risk level per location is independent of the designed network and thus disregard scale effects or learning. Therefore, we do not assume any correlation between capacity allocated to a certain location and its risk factor.
Fig. 3. An exemplary AHP describing outsourcing/offshoring risk
306
J. Reich et al.
After having selected and hierarchically arranged the criteria, the AHP is conducted together with the managers. First the individual importance of risk criteria and their parent criteria is determined through pairwise comparison of items on the same hierarchical level. Then each alternative is compared to all other alternatives regarding all lowest-raking risk factors. The AHP requires expertise on how each decision alternative performs regarding each criterion. In larger organizations this knowledge is widely dispersed. Thus, prior to the AHP it is necessary to gather this data and prepare it for managers to get a holistic picture. Only then managers can make informed judgements. An alternative to this sequential approach is to directly involve a wide range of experts by using group-AHP techniques [16]. After the pairwise comparisons are made, the individual AHP preference score can be calculated for each location. 3.5
Integrating AHP and MILP to Calculate the Pareto Frontier
The AHP score is integrated into the MILP by calculating the overall network-wide AHP risk score weighted by activity throughput. The greater share a location takes of the overall work performed, the higher is the impact of its individual risk score on the network’s overall risk score. After having calculated the initial minimum cost solution, we gradually raise our desired risk score “Risk Target” as a soft constraint for the cost minimization, until the highest possible AHP score is reached, representing the solution with the minimum risk and highest cost, given the constraints (Fig. 4).
(10) Fig. 4. Soft constraint to integrate AHP score into the MILP
All solutions representing trade-offs between benefit and risk are visualized on a Pareto front. Decision makers can now see, discuss and select their preferred configuration (Fig. 5).
Fig. 5. The Pareto frontier (exemplary)
Development of a Decision Support Model for Alleviating Supply
307
4 Conclusion Decision support tools should be made easy to use and understand, otherwise adoption rates and acceptance by managers remain low. There is great potential in improving existing model-based decision support methods to make them more effective and increase their usage and acceptance in practice. Special focus should be put on usability and transparency, while at the same time keeping the underlying techniques as simple as possible. This paper has proposed the design of an SCM decision support framework for service industries. It is constructed to support strategic process outsourcing design decisions with a time horizon of several years. We believe that it can be developed into an effective tool to improve decision-making in global BPO problems. It gives a precise solution to the network configuration problem while incorporating conflicting goals as well as quantitative and qualitative data. It is based on three well-established methods combined in a new way. Compared to existing methods to merge linear programming and AHP, the Pareto frontier approach does not attempt to calculate one single presumably optimum solution, but instead creates transparency over the full range of efficient trade-offs. This way managers can use the result as the base for discussion and select a configuration that fits their strategic goals. As the next step, we will test our conceptual model on a real-life business process outsourcing problem. This will help us test our assumptions of decision support performance criteria, as well as the effectiveness of the approach itself.
References 1. Clott, C.B.: Perspectives on global outsourcing and the changing nature of work. Bus. Soc. Rev. 109(2), 153–170 (2004) 2. Ellram, L.M., Tate, W.L., Billington, C.: Understanding and managing the services supply chain. J. Supply Chain Manage. 40(3), 17–32 (2004) 3. Lampón, J.F., Lago-Peñas, S., González-Benito, J.: International relocation and production geography in the European automobile components sector: the case of Spain. Int. J. Prod. Res. 53(5), 1409–1424 (2015) 4. Kinkel, S., Maloca, S.: Drivers and antecedents of manufacturing offshoring and backshoring—a German perspective. J. Purch. Supply Manage. 15(3), 154–165 (2009) 5. Srai, J.S., Ané, C.: Institutional and strategic operations perspectives on manufacturing reshoring. Int. J. Prod. Res. 54(23), 7193–7211 (2016) 6. Mani, D., Barua, A., Whinston, A.: An empirical analysis of the impact of information capabilities design on business process outsourcing performance. MIS Q. 34(1), 39–62 (2010) 7. Johnston, R.: Service operations management: return to roots. Int. J. Oper. Prod. Manage. 19(2), 104–124 (1999) 8. Ioannou, G., Karakerezis, A., Mavri, M.: Branch network and modular service optimization for community banking. Int. Trans. Oper. Res. 9(5), 531–547 (2002) 9. Piplani, R., Saraswat, A.: Robust optimisation approach to the design of service networks for reverse logistics. Int. J. Prod. Res. 50(5), 1424–1437 (2012)
308
J. Reich et al.
10. Henao, C.A., Ferrer, J.C., Muñoz, J.C., Vera, J.: Multiskilling with closed chains in a service industry: a robust optimization approach. Int. J. Prod. Econ. 179, 166–178 (2016) 11. Schuff, D., Paradice, D., Burstein, F., Power, D.J., Sharda, R. (eds.): Decision Support: An Examination of the DSS Discipline, vol. 14. Springer, New York (2010) 12. Harland, C., Knight, L., Lamming, R., Walker, H.: Outsourcing: assessing the risks and benefits for organisations, sectors and nations. Int. J. Oper. Prod. Manage. 25(9), 831–850 (2005) 13. Saaty, T.L.: The Analytical Hierarchy Process, Planning, Priority. Resource Allocation. RWS Publications, Pittsburgh (1980) 14. Boardman Liu, L., Berger, P., Zeng, A., Gerstenfeld, A.: Applying the analytic hierarchy process to the offshore outsourcing location decision. Supply Chain Manage. Int. J. 13(6), 435–449 (2008) 15. Yang, D.H., Kim, S., Nam, C., Min, J.W.: Developing a decision model for business process outsourcing. Comput. Oper. Res. 34(12), 3769–3778 (2007) 16. Saaty, T.L.: Group decision making and the AHP. In: Golden, B.L., Wasil, E.A., Harker, P.T. (eds.) The Analytic Hierarchy Process, pp. 59–67. Springer, Heidelberg (1989)
Shelter Site Selection and Allocation Model for Efficient Response to Humanitarian Relief Logistics Panchalee Praneetpholkrang(B) and Van-Nam Huynh School of Advanced Science and Technology, Japan Advanced Institute of Science and Technology, Nomi, Ishikawa 923-1292, Japan {panchalee,huynh}@jaist.ac.jp
Abstract. Shelter sites selection and allocation is the most critical part of humanitarian relief logistics and affects the success of disaster management strategy. This paper presents shelter location allocation model to efficiently respond to relief logistics. The mathematical model is developed to minimize the total cost which combines fixed cost for establishing shelters, transportation cost, and service cost. The model is tested with two scenarios i.e. capacitated shelter and uncapacitated shelter, then compared with the existing location and allocation plan. The Genetic Algorithm is used to solve the model. The numerical experiment with the case study of the flood in Tha Uthae, Surat Thani, Thailand is employed to demonstrate the application of the proposed model. The results obtained from this proposed model clearly outperform the current plan. Moreover, the sensitivity analysis is conducted to observe how the cost structure changed when the parameter is adjusted. The obtained results are constructive for decisionmakers to determine the appropriate strategies for disaster management. Keywords: Humanitarian logistics · Emergency logistics · Discrete facility location · Optimization · Cost structure · Disaster management
1
Introduction
In the past, disasters have caused enormous damage to humanity and have had a tremendous impact on the economy. According to a report by the Centre for Research on the Epidemiology of Disasters (CRED), natural disasters in 2018 resulted in 67,572 deaths, affected 198.8 million people, and caused $166.7 billion worth of economic losses worldwide. Floods had the most drastic effects on people when compared with other types of disasters, and the Asian continent was most affected by catastrophes [7]. Due to severely damaging effects, many academicians turned their attention to the study of which involves preparedness, mitigation and effective response to catastrophes in order to reduce c The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 309–318, 2020. https://doi.org/10.1007/978-3-030-44783-0_30
310
P. Praneetpholkrang and V.-N. Huynh
its devastating impacts [15]. Furthermore, it leads to an increase in humanitarian relief logistics research. Humanitarian logistics aims to relieve the victims’ suffering through the processes of evacuating the victims from affected areas to safe places by planning, and implementing and controlling the flow and storage of products with the given financial budget [2,6,23]. To achieve the purposes of humanitarian logistics, the optimization models are proposed to cope with diverse important criteria such as effectiveness, efficiency, or equity. These criteria are demonstrated through either single objective optimization or multi-objective optimization which incorporates various criteria simultaneously. For effectiveness view, the response time, transportation distance, coverage, and reliability are determined. Equity is considered according to fairness in accessing resources, while efficiency focuses on controlling the set of costs [21]. In this context, cost criteria cannot be ignored since it involves the investment of money, funds, as well as private donors [10]. Apart from helping the victims escape natural catastrophes, finding the necessary facilities, especially temporary shelters for victims who cannot stay at their homes, is important to consider before the disasters occur [19]. Without the appropriate methods required to determine the characteristics of the environment requiring relief, organizations may make ad-hoc decisions which lead to high costs, waste of resources, sluggish response, and failure to satisfy demands [2]. Therefore, decision-making regarding shelter sites selection and allocation is the most critical part of humanitarian relief logistics since it influences securement, equity, efficiency, effectiveness, and affects the success of the strategy [2,5,13,16,24]. Moreover, accommodation for victims should be provided adequately and according to the standards for evacuation shelters. The necessary resources should include portable restrooms, temporary kitchen, temporary warehouse, vehicles for mobilizing the victims, and staffs for assisting the victims during their stay in the shelters [13]. By considering the aforementioned issues, this research aims to propose mixed integer nonlinear programming to define the optimal number of shelters, to select the shelters for victims, and to assign the affected area to the appropriate shelters with the appropriate cost. The model corresponds to both capacitated and uncapacitated shelters. The obtained results from the model are compared with the current shelter site selection plan announced by the Department of Disaster Prevention and Mitigation. The repeatedly flooded areas in Tha Uthae Subdistrict, Surat Thani, Thailand is selected as the case study of shelter location and allocation problem.
2
Related Work
Facility location models are categorized into continuous facility location and discrete facility location. For continuous location, the facilities are allowed to be placed anywhere within the planning continuous area, whereas discrete location, also referred to as discrete space, permits the finite set of potential facilities that are predestined to be selected [3,4,18,20]. Facility, in humanitarian logistics, includes emergency medical centers, warehouse or distribution centers, and
Shelter Site Selection
311
emergency shelters. Research related to site selection for placing emergency medical locations and warehouses has been widely conducted, but research on shelter site selection is still fairly unexplored [19]. To evaluate the relief effort, cost efficiency is the criteria that decision makers use to determine resource utilization. Moreover, significant factors such as capacity, budget, and transportation modes are also considered [21]. There are existing literature on facility location that were employed in the humanitarian logistics field. Horner et al. [11] present a GIS-based method for selecting the special needs shelter for elderly casualties during a hurricane, with the objective of minimizing transportation cost. In this regard, shelters’ capacity and the desired number of shelters to be located are identified. Lin et al. [14] propose the location allocation model for placing the temporary depot around the earthquake-affected area. The purpose is to minimize transportation and penalty costs caused by unmet demand, delayed delivery, and service unfairness among demand points. Hu et al. [12] study shelter site selection for response to the earthquake in Beijing, China. In their research, the bi-objective model is used for minimizing the distance and total cost of shelter construction. The Nondominated Sorting Genetic Algorithm is employed to improve both effectiveness and efficiency performances. Ahmadi et al. [1] formulate the mixed integer nonlinear programming to determine the location of depots and define routing for last mile transportation. With regard to location decision, they aim to minimize the traveling time, the penalty cost of unmet demand, and the fixed cost of opening the depot. The proposed model is applied in the case study of the earthquake in San Francisco. Aforementioned literature reveals that the facility location problem in the context of humanitarian relief logistics received attention from researchers and was applied to identify the appropriate location in preparation and response to various kinds of disasters. Other than efficiency-based facility location models which take into account the cost of operation, there are various research papers focused on the responsiveness and effectiveness criteria (i.e. time, coverage, and distance [5,9,17]). However, the study of efficient location allocation of shelter is still fairly unexplored when compared with other facilities. Cost efficiency is the noteworthy criteria that reflects the use of resources and is the indicator to measure how well limited resources are utilized. Moreover, cost criteria also helps decisionmakers to plan and allocate the budget for response and prepare sufficiently. Therefore, this study aims to present the shelter location allocation model to decide which shelters should be selected, and which affected areas should be assigned to the appropriate shelters.
3
Methodology
In this section, the mathematical model is proposed for shelter site selection and allocation. The objective of this model is to minimize the total cost which includes the fixed cost of opening the shelters, the transportation cost of victim mobilizing, and the service cost during the victims’ stay in the shelter; this
312
P. Praneetpholkrang and V.-N. Huynh
is calculated based on the cost of staff hiring. The shelter site selection and allocation are determined both in aspects of “capacitated shelter” and “uncapacitated shelter” for providing the alternative for decisionmakers. First, the candidate shelters are predetermined. The data that are used to formulate the model includes the number of victims in each affected area, traveling distance, shelters’ capacity, fixed cost for opening shelters, transportation cost, and duration of the disaster. The model assumes that the victims in each affected area are mobilized as an entire unit and not separately assigned to different shelters, that all affected areas are faced with the disaster at the same time, that the number of victims and location of the candidate shelters are fixed, and that the vehicles used in evacuation process are homogenous. In this study, the Genetic Algorithm is employed to solve the proposed model since it avoids getting trapped with the local optimal solution, and successfully used to deal with many location and allocation problems. Set I Set of affected area i J Set of candidate shelter j Parameters dij cj hi fj M α β γ T
Distance between affected area i and candidate shelter j Capacity of the candidate shelter j Number of victims in area i Fixed cost of opening the shelter j Maximum acceptable distance between affected area and shelter Constant coefficient of transportation cost per kilometer per person Wage per person for hiring staff to work in the shelter Ratio of the required staff per victims Duration of the disaster occurrence
Decision variable Xj 1, if candidate shelter j is selected or otherwise 0 Yij 1, if affected area i is assigned to candidate shelter j or otherwise 0 Zij The victim in area i is assigned to candidate shelter j The model can be formulated as follows:
min
j∈J
xj fj + α
i∈I j∈J
dij yij hi + βT
Zij i∈I
γ
(1)
Shelter Site Selection
313
Subject to
yij
=
1,
∀i∈I
(2)
yij dij yij zij
≤ ≤
xj , M
∀i∈I,j∈J ∀i∈I ,j∈J
(3) (4)
≤
cj xj
∀j∈J
(5)
zij
=
hi
∀i∈I
(6)
xj yij
∈ ∈
{0, 1} {0, 1}
j∈J
i∈I
j∈J
∀j∈J ∀i∈I ,j∈J
(7) (8)
The objective function (1) is to select the shelter that generates minimum total cost which includes the fixed cost for opening the shelter, transportation cost, and service cost. Constraint (2) identifies that each affected area will be entirely assigned to only one shelter. Constraint (3) restricts each affected area to be allocated to only selected shelters. Constraint (4) ensures that the distance between affected area and selected shelter does not exceed the maximum acceptable distance. Constraint (5) ensures that the number of assigned victims does not exceed the capacity of selected shelter. Constraint (6) identifies the constraint of the number of victims in each affected area. Constraints (7) and (8) states the binary variable in the model.
4
Case Study
The flood case study in Tha Uthae, Surat Thani of Thailand is applied to the proposed model for selecting the shelters as well as assigning the victims to the shelter with the minimum total cost. The terrain of Tha Uthae is a lowland and repeatedly faces flooding during the rainy season. Normally, the Department of Disaster Prevention and Mitigation, Ministry of Interior is the agency that decides the evacuation shelters for each community based on their administrative area. A majority of the candidate shelters are schools, city halls, or temples. However, the assigned shelters are rather decentralizing than centralizing. The historical data of the 2011 floods in Tha Uthae which was gathered by Surat Thani National Statistical Office [22] are used in this study. There were 10 affected areas, 20 potential shelters, and 5,076 victims that suffered due to the floods. The distance between affected areas and candidate shelters are estimated from the given coordinates using the Euclidean distance. In this study, the maximum acceptable distance for each route is assumed to be 10 km. The vehicles that are used for victim transportation belong to the Royal Thai Army. The truck capacity is 12 persons and the fuel consumption rate 8 km per liter. For shelter capacity, the schools that are utilized as shelter can contain 2,000 victims, as suggested by JICA [9]; other shelters that are not the schools can
314
P. Praneetpholkrang and V.-N. Huynh
accommodate 500 victims. Although there are no construction costs since existing facilities are used as shelter, related costs for opening the shelters still need to be included, such as costs for portable toilets, tents to use as a temporary kitchen, medical center, and warehouse [8], which hereafter are defined as the fixed cost. To estimate the service cost that occurs when serving the victims during the time they reside in the shelter, the cost of staff hire is determined. Although assisting the victims is volunteer work, government staffs are still paid by their agencies. In this case, the standard wage of 380 Thai Baht per person per day is taken into account. The number of required staff is 1 staff per 50 victims [8]. The average duration of the disaster occurrence based on historical data is 6 days. Tables 1 and 2 are the parameters that are used in the numerical experiment. Table 1. Affected area and number of victims Affected area
A1 A2 A3 A4 A5 A6 A7 A8 A9 A10
No. of victims (person) 750 540 400 800 650 250 350 450 500 386
Table 2. Candidate shelters, capacity, and fixed cost
5 5.1
Candidate shelters
Capacity Fixed cost (THB)
S1, S3, S5, S6, S8, S10, S11, S14, S15, S16, S17, S19, S20
500
114,000
S2, S4, S7, S9, S12, S13, S18
2,000
144,000
Results and Discussion Computational Results
Table 3 shows the results generated by the proposed model. The number of selected shelters, shelter allocation, and total cost of both capacitated and uncapacitated shelters are compared with the current shelter assignment announced by the government agency. “Capacitated shelter” considers the restrictions of capacity and that the maximum acceptance distance does not exceed 10 km. There are 5 selected shelters—S7, S12, S18, S19, and S20—to serve the victims. The shelter utilization rates are 40%, 89.3%, 77%, 90%, and 100% respectively and the total cost is 899,471 Thai Baht. For “uncapacitated shelter”, the capacity in constraint 5 is ignored. It reveals that only small shelters which generate cheaper costs and are located within the acceptable distances of 10 km are chosen. There are 3 selected shelters include S3, S6, and S20. The number of selected shelters and the total cost are less than that of the capacitated shelters. Since the objective function is not bound by
Shelter Site Selection
315
the capacity restriction, the model then seeks to select a few shelters which are located in the acceptable distance to minimize the total cost. The total cost of 3 selected shelters is 580,891 Thai Baht. However, uncapacitated shelters would be difficult to employ in a practical manner due to overabundantly assigning the victims to particular shelters, which leads to congestion and will eventually affect the victims’ welfare. Both capacitated and uncapacitated shelters are compared to the current shelter assignment planned by the government sector. The numerical experiment reveals that the service cost of all plans remain constant, as shown in Table 3, since the number of victims is not changed and all victims are rescued. Moreover, it is evident that the current plan fails to achieve cost efficiency because there are 10 shelters that are selected and allocated based on their administrative area. The shelter allocation is decentralized and causes the setup cost to be unavoidably higher. Likewise, the total cost obtained from the proposed model, both capacitated shelter and uncapacitated shelter is lower than the current plan as 40.02% and 61.26% respectively. Table 3. The result of case study with acceptable distance not over 10 km
5.2
Affected area
Capacitated shelter Uncapacitated shelter Current plan
A1
S12
S20
S12
A2
S18
S6
S13
A3
S18
S6
S14
A4
S7
S6
S6
A5
S12
S20
S15
A6
S18
S20
S16
A7
S18
S20
S17
A8
S19
S3
S18
A9
S20
S20
S19
A10
S12
S20
S7
Setup cost (THB)
660,000
342,000
1,260,000
Transportation cost (THB) 6,911
6,331
7,036
Service cost (THB)
232,560
232,560
232,560
Total cost (THB)
899,471
580,891
1,499,596
Sensitivity Analysis
The sensitivity analysis is conducted to demonstrate how parameters influence the objective function and the model. Here, the maximum acceptable distances (constraint 4) are set between 10–30 km to allow the numerical experiment to be more flexible. In the case of capacitated shelter, it is the most cost efficient when the maximum acceptable distance does not exceed 25 km. It is required to select 5 shelters to serve the victims. Relaxing the maximum acceptable distance results in an increase in the transportation cost. On the contrary, the fixed cost
316
P. Praneetpholkrang and V.-N. Huynh
of opening the shelter does not increase as the relaxed distance is extended. Meanwhile, the relaxation of distance will not significantly affect the service cost since the constraint strictly ensures that all victims are served thoroughly (see Fig. 1). For uncapacitated shelter, it shows that, as the maximum acceptable distance is relaxed, the fixed cost of selected shelters decreases. This is because the relaxation of the acceptable distance means that the cheapest shelter can be found and selected without considering the limitation of the shelters’ capacity. Since total cost is dominated by fixed cost, it leads the total cost to decrease as the maximum acceptable distance is relaxed (see Fig. 2).
Fig. 1. Sensitivity analysis for capacitated shelter with distance 10–30 km
Fig. 2. Sensitivity analysis for uncapacitated shelter with distance 10–30 km
Shelter Site Selection
6
317
Conclusion and Future Research
This study presents the mathematical model for shelter site selection and allocation for efficient response to relief logistics during the disaster. The model is formulated as mixed integer nonlinear programming and solved by Genetic Algorithm in order to achieve cost minimization. The proposed model is tested with the real world case study of the floods in Tha Uthae, Surat Thani, Thailand. The comparisons of the results obtained from this model (i.e. capacitated and uncapacitated shelter and current shelter allocation plan announced by the government) are shown. The comparison indicates that, when using the proposed model, the obtained results outperform the current shelter allocation plan. This study has positive implications for the decisionmakers to develop the appropriate strategies for future work, the model should be extended to large scale optimization. Additionally, uncertainty, such as demand uncertainty or transportation network disruption caused by the disaster, should be determined when formulating the model.
References 1. Ahmadi, M., Seifi, A., Tootooni, B.: A humanitarian logistics model for disaster relief operation considering network failure and standard relief time: a case study on San Francisco district. Transp. Res. Part E: Logist. Transp. Rev. 75, 145–163 (2015). https://doi.org/10.1016/j.tre.2015.01.008 2. Balcik, B., Beamon, B.M.: Facility location in humanitarian relief. Int. J. Logist. Res. Appl. 11(2), 101–121 (2008). https://doi.org/10.1080/13675560701561789 3. Boloori Arabani, A., Farahani, R.Z.: Facility location dynamics: an overview of classifications and applications. Comput. Ind. Eng. 62(1), 408–420 (2012). https:// doi.org/10.1016/j.cie.2011.09.018 4. Boonmee, C., Arimura, M., Asada, T.: Facility location optimization model for emergency humanitarian logistics. Int. J. Disaster Risk Reduct. 24(January), 485– 498 (2017). https://doi.org/10.1016/j.ijdrr.2017.01.017 5. Boonmee, C., Asada, T., Arimura, M.: A bi-criteria optimization model for hierarchical evacuation and shelter site selection under uncertainty of flood events. J. East. Asia Soc. Transp. Stud. 12, 251–268 (2017) 6. Boonmee, C., Ikutomi, N., Asada, T., Arimura, M.: An integrated multi-model optimization and fuzzy AHP for shelter site selection and evacuation planning. J. Jpn. Soc. Civ. Eng. Ser. D3 (Infrastruct. Plan. Manag.) 73(5), I 225–I 240 (2017). https://doi.org/10.2208/jscejipm.73.i 225 7. Centre for Research on the Epidemiology of Disasters (CRED): Natural Disasters 2018 (2018). https://www.cred.be/sites/default/files/CREDNaturalDisaster2018. pdf 8. Department of Disaster Prevention and Mitigation. Ministry of Interior, Thailand: Evacuation and Temporary Shelter Management Manual (2011). http:// www.openbase.in.th/files/book immigrantion%20center.pdf 9. G¨ ormez, N., K¨ oksalan, M., Salman, F.S.: Locating disaster response facilities in Istanbul. J. Oper. Res. Soc. 62(7), 1239–1252 (2011). https://doi.org/10.1057/jors. 2010.67
318
P. Praneetpholkrang and V.-N. Huynh
10. Gutjahr, W.J., Nolz, P.C.: Multicriteria optimization in humanitarian aid. Eur. J. Oper. Res. 252(2), 351–366 (2016). https://doi.org/10.1016/j.ejor.2015.12.035 11. Horner, M.W., Ozguven, E.E., Marcelin, J.M., Kocatepe, A.: Special needs hurricane shelters and the ageing population: development of a methodology and a case study application. Disasters 42(1), 169–186 (2018). https://doi.org/10.1111/disa. 12233 12. Hu, F., Yang, S., Xu, W.: A non-dominated sorting genetic algorithm for the location and districting planning of earthquake shelters. Int. J. Geogr. Inf. Sci. 28(7), 1482–1501 (2014). https://doi.org/10.1080/13658816.2014.894638 13. Kongsomsaksakul, S., Yang, C.: Shelter location-allocation model for flood evacuation planning. J. East. Asia Soc. Transp. Stud. 6(1981), 4237–4252 (2005). http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.134.5534&rep=rep1& type=pdf 14. Lin, Y.H., Batta, R., Rogerson, P.A., Blatt, A., Flanigan, M.: Location of temporary depots to facilitate relief operations after an earthquake. Socio-Econ. Plan. Sci. 46(2), 112–123 (2012). https://doi.org/10.1016/j.seps.2012.01.001 15. Loree, N., Aros-Vera, F.: Points of distribution location and inventory management model for Post-Disaster Humanitarian Logistics. Transp. Res. Part E: Logist. Transp. Rev. 116(May), 1–24 (2018). https://doi.org/10.1016/j.tre.2018.05.003 16. Manopiniwes, W., Irohara, T.: A review of relief supply chain optimization. Ind. Eng. Manag. Syst. 13(1), 1–14 (2014). https://doi.org/10.7232/iems.2014.13.1.001. http://koreascience.or.kr/journal/view.jsp?kj=SGHHEA&py=2014&vnc=v13n1 &sp=1 17. Manopiniwes, W., Nagasawa, K., Irohara, T.: Facility location alternatives between expected and worst case time performance in humanitarian relief logistics. J. Jpn. Ind. Manag. Assoc. 66(2E), 142–153 (2015) 18. Mark, D.S.: What you should know about location modeling. Naval Res. Logist. 55(January 2008), 283–294 (2008). https://doi.org/10.1002/nav.20284 ¨ Kara, B.Y.: Shelter site location under multi-hazard scenar19. Ozbay, E., C ¸ avu¸s, O., ios. Comput. Oper. Res. 106, 102–118 (2019). https://doi.org/10.1016/j.cor.2019. 02.008 20. Pitu, M.B., Richard, F.L.: Discrete Location Theory. Wiley, Canada (1990) 21. Rodr´ıguez-Esp´ındola, O., Gayt´ an, J.: Scenario-based preparedness plan for floods. Nat. Hazards 76(2), 1241–1262 (2015). https://doi.org/10.1007/s11069-014-15442 22. Surat Thani National Statistical Office: No Title (2012). http://surat.old.nso.go. th/surat/flood/data/longlub.pdf%0D%0A 23. Thomas A., Kopczak, L.: From Logistics to Supply Chain Management: The Path Forward in the Humanitarian Sector (2005) 24. Verma, A., Gaukler, G.M.: Pre-positioning disaster response facilities at safe locations: an evaluation of deterministic and stochastic modeling approaches. Comput. Oper. Res. 62, 197–209 (2015). https://doi.org/10.1016/j.cor.2014.10.006
Part IV: Modeling, Simulation, and Optimization
A Hypercube Queuing Model Approach for the Location Optimization Problem of Emergency Vehicles for Large-Scale Study Areas Felix Blank(&) University of Würzburg, 97070 Bavaria, Germany [email protected]
Abstract. Emergency service systems provide essential services to people in need. Most of them have to operate under uncertainty and in complex environments. Their locations have to be chosen in a way that all or at least most of the incoming demand can be covered within a justifiable response time. In this paper, a hypercube queuing approach is presented that locates a high amount of emergency units within a large-scale study area. Due to the hypercube restrictions, computational times increase with the number of servers that are located. Therefore, an algorithm for the aggregation of demand areas is presented. To find an at least appropriate solution, a genetic algorithm is applied. It can be shown that computational times can be lowered significantly while the solution error is minimal. Furthermore, average response times for the emergency service system decrease with the location of additional servers in the study area. Keywords: Hypercube
Emergency service system Queuing
1 Introduction Emergency service systems (ESS) are systems that provide immediate help to people in need. Possible areas of operation include fire-rescue, on-site medical care as well as emergency services in the case of man-made or natural disasters. Most ESS are subject to certain requirements regarding their coverage of the respective area and their response times. Due to the inherent uncertainties and the spatial distributions of demand in the system, the location decisions are not straightforward and have to be taken with respect to the underlying study area and the spatial distribution of the demand. Therefore, locations have to be chosen in a way that, even in remote parts of the study area, all or at least most of the incoming demand can be covered within a justifiable response time. A huge body of existing literature deals with the location optimization problem by using various models and techniques. In this paper, a model is presented that incorporates methods of the hypercube queuing model (HQM) into a location problem for large-scale study areas and location decisions. The model can be used to optimize location decisions of ESS in large-scale study areas like the location of ambulances or other emergency vehicles in a city area. In order to deal with the computational efforts that are required for the location of many © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 321–330, 2020. https://doi.org/10.1007/978-3-030-44783-0_31
322
F. Blank
emergency units, an aggregation algorithm (AA) is proposed. The paper is organized as follows: In Sect. 2, a brief literature review discusses existing contributions to the location of ESS and the HQM. Section 3 contains the formulation of the model as well as the AA. The applied metaheuristic, the exemplary study area as well as the computational results are presented in Sect. 4. Section 5 includes the conclusion as well as future research directions.
2 Literature Review Facility location models are used by private and public organizations to determine optimal or near-optimal locations for their entities, like warehouses and, in the case of ESS, emergency stations or fire departments. Comprehensive reviews of facility location models for ESS can be found in Brotcorne et al. [1], Caunhye et al. [2] and more recently in Farahani et al. [3]. In general, facility location models for ESS can be divided into coverage models as well as p-median models. While the first group intends to locate facilities (also called servers in the case of ESS) in a way that maximizes the coverage over a demand area, the second group tries to minimize the distance between the demand points and the servers. 2.1
Coverage and P-Median Models
The very first contributions to the ESS location problem used static and deterministic inputs to obtain demand locations while ignoring factors like changing demands or other inherent dynamics of such systems. Toregas et al. [4] formulated the LocationSet-Covering-Model that required all demand being fulfilled within a pre-determined time frame. Church and ReVelle [5] proposed the Maximum-Coverage-LocationProblem (MCLP) that determines the location of each emergency unit in a way which maximises the covered space of each part of the study area. Daskin and Stern [6] stated an extension to the MCLP that maximizes the number of demand areas that is covered more than once. Hogan and ReVelle [7] formulated models that maximize backup coverage while Gendreau et al. [8] developed a model that uses two distinct time constraints. The p-median model was originally stated by Hakimi [9]. Calvo and Marks [10] used it to locate multi-level health facilities. Carson and Batta [11] determined a dynamic ambulance positioning strategy with the help of a p-median model. 2.2
Hypercube Queuing Model
The HQM is a markovian finite-state model and was initially stated by Larson [12] as a combination of queuing theory, facility location and analysis. It was then used to evaluate the performance of an underlying system. Based on a given set of server locations, the HQM can be used to derive certain performance measures that can be used to evaluate the decision-making. Larson [13] later introduced the approximate HQM that reduced computational difficulties while incorporating the original model. Since then the HQM has been widely applied and extended. This includes the better estimation of service rates [14], multiple dispatch of servers [15], modeling of co-located
A Hypercube Queuing Model Approach
323
servers [16] and customer-dependent service rates [17]. More recent extensions focus on the incorporation of waiting lines and customer preferences [18, 19]. Since the basic HQM and its extensions are descriptive models, they cannot be used to obtain optimal locations of facilities or servers [20]. The performance metrics that can be obtained by solving the basic HQM therefore need to be embedded in an optimization process [21, 22]. Batta et al. [23] and Saydam and Aytug [24] used the HQM in combination with the maximum expected coverage location problem (MEXCLP) to evaluate the performance of the derived locations. Galvão et al. [25] relax the server independence assumption of the maximum availability location problem with the help of the HQM. Geroliminis et al. [26] develop the spatial queuing model (SQM) and introduce server specific service rates as well as the allocation of demand areas to the responsible servers while minimizing the average response time of the overall system. Geroliminis et al. [27] develop a method for larger scale systems and propose a districting algorithm by reducing the steady states. Boyaci and Geroliminis [28] present two different models that also reduce the state-space by aggregating servers. Iannoni et al. [29] state a hypercube approximation algorithm to consider large numbers of emergency units. Akdogan et al. [30] propose different possible formulations of service rates in a SQM-based emergency service location study. The use of the HQM can lead to more precise performance measures as well as the incorporation of server unavailability and backup structures. To the author’s best knowledge, only [27, 28] and [29] consider large-scale ESS design while using HQM based methods. The existing body of literature reduces computational efforts mostly by reducing the state space. In large study areas the computational times do not only solely depend on the number of servers that are considered, but also on the number of demand areas. Since server responsibilities have to be checked for each state and demand area, increasing demand areas in a study area also increases computational times significantly. Due to the advances in computational power, larger number of emergency vehicles can nowadays be analysed without necessarily compromising the steady-statespace. Some of the required assumptions, like symmetrically located servers, identical workloads or homogenous demand, can reduce the accuracy of the solutions found by the model. In this paper, an approach that does not reduce the steady-state-space, but builds on dynamic formation of super demand areas is presented.
3 Large-Scale SQM Consider a study area of J individual demand areas (atoms). In order to serve the incoming demand, several servers N have to be located within the study area. It is assumed that not every server can be sent to each atom. Therefore, each server has primary and lower level response areas that are determined with respect to the spatial distribution of the demand. The servers can only be busy or available and thus have only two, binary coded, states. If we consider a five-server system, in which the first, third and fifth server is busy, the corresponding state can be denoted as 10101. This generates 2N different states for the system that are the vertices of the hypercube and are named Ba . The probabilities of each state are derived from an equation system that balances the flows between the separate hypercube states. The underlying equation
324
F. Blank
system is constructed by formulating one equation for each state that includes all upward and downward tranisitions. An upward transition happens for all incoming demand calls from the relevant demand areas and a downward transition happens whenever the server that differs between the two states completes its service. The resulting probabilities of the equation system describe the likelihood of each state of the hypercube queuing model. The occurring demand is defined as a call for emergency help and happens solely at the center of each atom. It is assumed to be independent and not per definition identical between the atoms while following a time homogenous Poisson distribution. Whenever a demand (call) enters the system, the available servers are checked for availability and the closest available server is then dispatched. After completion of service, the dispatched server returns to the base location. If no server is available, the incoming call is lost to the system. 3.1
Model Formulation
The optimization model can then be formulated as follows: XN XJ
min T ¼
n¼1
j¼1
pnj tnj
ð1Þ
Subject to: XJ
fy j¼1 j j
X i2Wj
Ccov
ð2Þ
xi yj 8 j 2 J
ð3Þ
xi ¼ N
ð4Þ
XI i¼1
xi ; yj 2 ½0; 1 8 i 2 I; j 2 J
ð5Þ
P PfBb g½
P kab þ lab a a þ ¼1 Ba 2 CN : d Ba 2 CN : dab ab ¼ 1 P P N ¼ P B P B þ f gl f gk a a ab 8 b ¼ 0; 1; . . .; 2 1 ab a a þ B a 2 C N : d ¼1 Ba 2 CN : dab ab ¼ 1
ð6Þ X2N 1 a¼0
P pnj ¼ f j
Ba 2Enj
PfBa g ¼ 1
PfBa g
1 Pf2N1 g
8 j 2 J, n 2 N
ð7Þ ð8Þ
A notation similar to [30] is used. J describes the set of regions, N the number of servers to be located while I is the set of potential location sites. Wj is the set of
A Hypercube Queuing Model Approach
325
locations covering atom j. xi , yj are binary variables that show whether location site i is chosen or atom j is covered. Ccov is a pre-defined coverage value that takes a value equal or less to 1. k is the system wide demand and f j the demand fraction of atom j. pnj is the fraction of dispatches server n sends to atom j, tnj is the travel time of server n to atom j. CN denotes the vertices of the N-dimensional hypercube. kab and lab are the upward and downward transition rates of the system, e.g. the transition rates that lead to a change of the system from state a to state b corresponding to the respective vertices Ba and Bb of the N-dimensional hypercube while PfBa g is the associated steady-stateprobability of vertex a. kab is the demand for a service offered by the ESS and lab is the þ service rate for the requested service. dab and d ab are the upward and downward Hamming distances from state a to state b and describe the difference in notation þ Þ between state 10000 and state between the two states. For example, the difference ðdab 10001 is one. Enj describes the set of states in which server n is the nearest available for region j. The model is controlled over the decision variables xi and yj with tnj as an input. Constraint (2) ensures that a certain pre-defined coverage level ðCcov 1Þ is met. Constraint (3) controls the decision variable yj with respect to the coverage of Atom j. Constraint (4) guarantees that only the pre-defined number of vehicles is located. Constraint (6) specifies the equation system that is necessary to derive the steady-stateprobability of the HQM. Each equation defines the balance of flows of one hypercube state. The sum of the probabilities of all hypercube states is equal to one (Constraint (7)). Since per definition incoming demand calls can be lost due to unavailable servers, the sum of the fraction of dispatches from all servers to one demand area can be lower than one. The denominator of (8) normalizes the fraction of dispatches under the consideration that not all servers are busy (the steady-state Pf2N1 g). Each fraction of dispatch pnj describes the probability of a dispatch for server n to demand area j. This probability is then multiplied in (1) with the travel time of server n to demand area j to derive the expected average response time of the system. The upward and downward transition rates are key inputs to the HQM and form the equations of the equation system in (6). Two important characteristics of the SQM as debuted by [26] are the spatial distribution of the demand as well as the assumption and calculation of districting levels. The later refers to the degree of coverage of each demand area that is provided by servers with downstream preferences. Since the model in this paper considers a large number of servers and demand areas, computational efforts for a complete backup are prohibitive and only third-level districting is used. The term d-th level districting refers to the partitioning of the study area in sub-areas according to the n-th nearest servers. For every d > 1 this means, that whenever the d − 1 nearest server is unavailable, the d nearest server responds. For every level of districting, the demand must be covered. The upward transition rate between vertex a and b of the hypercube can be calculated as follows:
þ
PM m¼2
P
P kab ¼ k1kk þ l1 2N:bl1 ¼1 k2l1 k m1 m2 2 km m 1 Q l1 k \ kl1 lm1 \ kl1 lm2 \ . . . \ kl1 l2
l1 ...;lm 2N:
bli ¼1
i¼1
ð11Þ
326
F. Blank
The equation above denotes that, given the system is in state a, server unit k responds to any demand in its area of responsibility D1kk or any other demand area Dm lk for server l, as well as the m − 1 nearest responsible servers (denoted by l) are 1 m unavailable. k1kk and km lk describe the demands of the demand area Dkk or Dlk respectively. The upward transition rate from state a to state b therefore consists of all demands from the demand areas for which server k is the primary server and from the demand areas for which server k is a lower tier server in the case of the first m − 1 servers of the preference list being unavailable. For a practical example of the partitioning of the study area into the sub areas the reader is referred to Akdogan et al. [30]. For calculating the downward transition rates, i.e. the service rate, there are several approaches in the literature. Geroliminis et al. [26] initially introduced a server-specific weighted-average approach that builds on the calculation of the demand rates without specifically considering the travel times to the atoms. In real life, travel times from the server to the demand area impact significantly the service rate of the server and the ESS. Akdogan et al. [30] therefore have stated an approach that is independent of the demand of the sub areas, but explicitly considers travel times from the server location to the location of the occurring demand. State a and state b differ at exactly one position of their state spaces. The location of the deployed server is then denoted by rk with wkj being the travel time from location rk to demand area j. ;rk is the incident handling rate and T is the given time period. The incident handling rate per sub area then can be expressed as the number of possible deployments per hour with the denominator of (12) consisting of the incident handling time plus two times the mean travel times. The downward transition rate then is the sum of the incident handling rates of all sub areas ðj 2 Ljk Þ that can cause a transition from state a to state b: lab ¼
X j2Ljk
T T þ2 w kj ;r
ð12Þ
k
3.2
Aggregation Algorithm
The HQM sets up a linear equation system with 2N equations. The computational efforts and the solving time increase significantly with N. [27] tackle this problem with postulating a districting approach that reduces the number of equations to N. This is done by the assumption of symmetrical server locations as well as homogeneous demand and hence identical workloads of the servers. The states with the same number of busy servers are then summarized into one “super-state”. The approach presented in this paper does not compromise the expressiveness of the steady states but aggregates the demand areas dynamically with respect to the server locations as well as the demand areas. Since the exploratory study area considers a large number of demand areas, the computation time for generating the upward and downward transition rates increases significantly with the number of servers. Because the calculations have to be done for each proposed solution, the AA has to adapt dynamically to the server locations and the allocation of the demand areas to the servers. The AA is done in the following steps:
A Hypercube Queuing Model Approach
327
1. Determination of the D-nearest servers for each demand area (D = maximum level of districting) 2. Formation of new super demand areas for demand areas with the same server allocation for i = 1, …, D 3. Aggregation of demand of the original demand areas into demand of the super demand areas Since the preferences of the individual demand areas do not change during the aggregation into super demand areas, the calculation of upward transition rates remains the same. The aggregation of the demand areas requires the calculation of new travel times. Equation (13) builds on (12) and uses a weighted demand approach to calculate the new travel times wkj . The fraction of demand from demand region j of the super demand area As is used to weigh the original travel time from server k to demand area j. The downward transition rate is then calculated as a sum of all demand areas that belong to the respective super demand area. l0ab ¼
X j2As 2Ljk
T T þ2 P P kj j2As ;r k
j2As
kj
wkj
ð13Þ
4 Solution Technique, Study Area and Results The objective function in (1) has no closed-form expression. Therefore, an algorithm is needed to solve the proposed model. In this paper, a genetic algorithm (GA) as in [27] and [30] is applied. The GA tries to mimic evolutionary processes and to find optimal or near-optimal solutions by eliminating bad proposed solutions through survival of the fittest. In order to avoid local minima, mutation techniques are used. 4.1
Study Area
For the design of the exploratory study area, a 500 500 grid with 500 demand areas is considered. ESS often operate in urban environments, but also have responsibilities for more rural areas. Therefore, the demand areas are evenly spread over the study area and one urban agglomeration is considered. About a third of the demand points are located within the area of the urban agglomeration. Due to the combination of less and higher populated parts of the study area, the servers have to be located in a way that minimizes the mean response time for both groups of demand areas. The reference time period is one hour. The demand of the demand areas follows a time homogeneous Poisson distribution with the mean of 2 in the urban agglomeration area and 1 in the rural parts of the study area. ;rk is set to 1. 4.2
Results
The model, as well as the corresponding algorithm, were coded in C++ and run on a Intel Core i7 processor. The genetic algorithm was run 125 generations with a population size of 20 individuals in each.
328
F. Blank
75000 without AA
50000
with AA
25000 0 5
6
7
8
9
10
Number of servers
Fig. 1. Computation time in seconds.
The computation time increases significantly with the number of servers, especially in the nine and ten server case. The use of the proposed AA lowers the computation time to about one fourth in the 5 to 8 server-case and about one third in the 9 and 10 server-case. The number of super atoms rises with the servers that are considered due to the higher number of possible districting combinations that arise with more servers. This mitigates the performance advantage of the AA only to a rather small degree, as seen by the shallower course of the AA graph in Fig. 1.
15 14 13 12 11 10
1
25
50
75
100
125
Number of generations
5 Server 7 Server with AA
5 Server with AA 10 Server
7 Server 10 Server with AA
Fig. 2. Mean average response time in minutes for the 5, 7 and 10 server case.
The mean average response time of the system decreases significantly with the consideration of additional servers. With two (five) additional servers the mean response time of the system decreases by about 11(21)% due to the shorter travel times from the server locations to the demand areas. It can be shown that the marginal benefit of additional servers diminishes. The locations found in the optimization process with the AA are then used to compute the travel time when using the full model. The deviation percentage is not significant and under 2% in all cases (Figs. 2 and 3).
A Hypercube Queuing Model Approach
5 Server -0,12
6 Server 1,85
7 Server 1,47
8 Server 0,45
9 Server 0,1
329
10 Server 0,15
Fig. 3. Percentage of deviation.
5 Conclusion In this paper, an approach to incorporate larger study areas and larger number of servers into a hypercube queuing location model was presented. It could be shown that computation times can be reduced significantly while the steady state space as well as the quality of the solutions found is not compromised. Additional servers can also help to reduce the response time of ESS. It could be further shown that the quality of the solutions found is within 2% of the exact model. The proposed AA allows decision makers to include a larger number of servers or location sites into their analysis and decision process. Since the computational error is proven to be marginal, the use of proposed AA within HQM location models allows for a more accurate analysis and more realism, especially in the analysis of large study areas, like metropolitan areas. Future research in this area could include the inclusion of dedicated waiting lines for incoming demand calls, the consideration of different day times as well as the comparison of different dispatch policies. The use of real data from ESS that have responsibilities for both rural and urban areas could add a further benefit.
References 1. Brotcorne, L., Laporte, G., Semet, F.: Ambulance location and relocation models. Eur. J. Oper. Res. 147(3), 451–463 (2003) 2. Caunhye, A., Nie, X., Pokharel, S.: Optimization models in emergency logistics: a literature review. Socio-Econ. Plan. Sci. 46(1), 4–13 (2012) 3. Farahani, R., Fallah, S., Ruiz, R., Hosseini, S., Asgari, N.: OR models in urban service facility location: a critical review of applications and future developments. Eur. J. Oper. Res. 276(1), 1–27 (2019) 4. Toregas, C., Swain, R., ReVelle, C., Bergman, L.: The location of emergency service facilities. Oper. Res. 19(1), 1363–1373 (1971) 5. Church, R., ReVelle, C.: The maximal covering location problem. Pap. Reg. Sci. Assoc. 32(1), 101–118 (1974) 6. Daskin, M., Stern, E.: A hierarchical objective set covering model for emergency medical service vehicle deployment. Transp. Sci. 15(2), 137–152 (1981) 7. Hogan, K., ReVelle, C.: Concepts and applications of backup coverage. Manage. Sci. 34(11), 1434–1444 (1986) 8. Gendreau, M., Laporte, G., Semet, F.: Solving an ambulance location model by Tabu search. Locat. Sci. 5(2), 75–88 (1997) 9. Hakimi, S.: Optimum locations of switching centers and the absolute centers and medians of a graph. Oper. Res. 12(3), 450–459 (1964) 10. Calvo, A., Marks, H.: Location of health care facilities: an analytical approach. Socio-Econ. Plan. Sci. 7(5), 407–422 (1973)
330
F. Blank
11. Carson, Y., Batta, R.: Locating an ambulance on the Amherst campus of the State University of New York at Buffalo. Interfaces 20(5), 43–49 (1990) 12. Larson, R.: A hypercube queueing model for facility location and redistricting in urban facility service. Comput. Oper. Res. 1(1), 67–95 (1974) 13. Larson, R.: Approximating the performance of urban emergency service systems. Oper. Res. 23(5), 845–868 (1975) 14. Halpern, J.: The accuracy of estimates for the performance criteria in certain emergency service queueing systems. Transp. Sci. 11(3), 223–241 (1977) 15. Chelst, K., Barlach, Z.: Multiple unit dispatches in emergency services: models to estimate system performance. Manage. Sci. 27(12), 1390–1409 (1981) 16. Burwell, T., Jarvis, J., McKnew, M.: Modeling co-located servers and dispatch ties in the hypercube model. Comput. Oper. Res. 20(2), 113–119 (1993) 17. Atkinson, J., Kovalenko, I., Kuznetsov, N., Mykhalevych, K.: A hypercube queueing loss model with customer-dependent service rates. Eur. J. Oper. Res. 191(1), 223–239 (2008) 18. Souza, R., Morabito, R., Chiyoshi, F., Iannoni, A.: Incorporating priorities for waiting customers in the hypercube queuing model with application to an emergency service system in Brazil. Eur. J. Oper. Res. 242(1), 274–285 (2008) 19. Rodrigues, L., Morabito, R., Chiyoshi, F., Iannoni, A., Saydam, C.: Towards hypercube queuing models for dispatch policies in queue and partial backup. Comput. Oper. Res. 84, 92–105 (2008) 20. Galvão, R., Morabito, R.: Emergency service systems: the use of the hypercube queueing model in the solution of probabilistic location problems. Int. Trans. Oper. Res. 15(5), 525– 549 (2008) 21. Goldberg, J.: Operations research models for the deployment of emergency services vehicles. EMS Manage. J. 1(1), 20–39 (2004) 22. Takeda, R., Widmer, J., Morabito, R.: Analysis of ambulance decentralization in urban emergency medical service using the hypercube queueing model. Comput. Oper. Res. 34(3), 727–741 (2007) 23. Batta, R., Dolan, J., Krishnamurthy, N.: The maximal expected covering location problem: revisited. Transp. Sci. 23(3), 277–287 (1989) 24. Saydam, C., Aytug, H.: Solving large-scale maximum expected covering location problems by genetic algorithms: a comparative study. Eur. J. Oper. Res. 141(3), 480–495 (2002) 25. Galvão, R., Chiyoshi, F., Morabito, R.: Towards unified formulations and extensions of two classical probabilistic location problems. Comput. Oper. Res. 32(1), 15–33 (2005) 26. Geroliminis, N., Karlaftis, M., Skabardonis, A.: A spatial queuing model for the emergency vehicle districting and location problem. Transp. Res. Part B 43(7), 798–811 (2009) 27. Geroliminis, N., Kepaptsoglou, K., Karlaftis, M.: A hybrid hypercube – genetic algorithm approach for deploying many emergency response mobile units in an urban network. Eur. J. Oper. Res. 210(2), 287–300 (2011) 28. Boyaci, B., Geroliminis, N.: Extended hypercube models for large scale spatial queuing systems. In: 91st Annual Meeting of the Transportation Research Board (2011) 29. Iannoni, A., Morabito, R., Saydam, C.: Optimizing large-scale emergency medical system operations on highways using the hypercube queuing model. Socio-Econ. Plan. Sci. 45(3), 105–117 (2011) 30. Akdogan, M., Bayindir, Z., Iyigun, C.: Locating emergency vehicles with an approximate queuing model and a meta-heuristic solution approach. Transp. Res. Part C 90, 134–155 (2018)
Dynamic Optimization Model for Planning of Multi-echelon Logistic System Activity Mykhaylo Ya. Postan1, Sergey Dashkovskiy2, and Kateryna Daschkovska3(&) 1
3
Department of Management and Marketing, Odessa National Maritime University, Mechnikov Str. 34, Odessa 65029, Ukraine [email protected] 2 Institute of Mathematics, University of Würzburg, Würzburg, Germany [email protected] Faculty of Business Management and Economics, University of Würzburg, Würzburg, Germany [email protected]
Abstract. In our paper, we consider functioning in dynamics (with discrete time) of logistics network including set of suppliers, set of manufacturers and set of points of destination of finished product. The optimization problem is formulated for determination of supply, production and transportation joint plans. It is assumed that total demand at destinations either is given over the planning horizon or is random with given probability densities. The objective function is total logistic cost along the whole network. Keywords: Logistics network Suppliers Manufacturers Finished product Transportation problem Joint planning Total logistic cost Dynamic optimization
1 Introduction It is well-known that most part of real logistic systems may be described as a network with inventory of material or product at each node. Therefore, it is natural to use for modeling, optimization, and analysis of logistic systems (or supply chains) the results of inventory control and optimization theories. The book (Zipkin 2000), for example, covers many recent developments related to or impacting inventory such as ERP systems, supply chain management, JIT, etc. It covers also a wide spectrum of stochastic inventory models. Now a number of different models are developed in logistics describing many different real situations in industry and business (Bramel and Simchi-Levi 1997; Shapiro 2001; Postan 2006; Brandimarte and Zotteri 2007; Smith and Tan 2013). At the same time the inventories in logistic systems have been controlled by some specific for logistical management rules and methods. The most important among them is the “demand driven principle”, that is, management taking into account the feedback between real volumes of finished product sale and its manufacturing. Note also that main goal of any logistical system functioning is material flows movement along the whole supply chain from one supply chain to another one or © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 331–340, 2020. https://doi.org/10.1007/978-3-030-44783-0_32
332
M. Ya. Postan et al.
to end consumer with the minimum total cost (Christopher 2011). These and some other peculiarities of logistical management do not allow to apply immediately the “ready” models from inventory control theory in logistics practices. The most part of existing applications of inventory control theory for logistic systems modeling and optimization takes into account one type of inventory only, e.g. raw material (Bramel and Simchi-Levi 1997; Shapiro 2001; Postan 2006). In the articles (Morozova et al. 2013; Postan et al. 2014) the multi-echelon logistic systems were under consideration with two types of inventory: raw materials and finished product. The corresponding models were based on generalization of the WagnerWhitin dynamic model from inventory control theory. Besides, in the cited works the transportation problem was included in model for the purpose of joint optimization of supply, production and finished product transportation plans of integrated supply chain. Our paper is devoted to further development of the approach to modeling and optimization of logistical networks initiated in the works (Morozova et al. 2013; Postan et al. 2014).
2 Problem Statement Let us consider logistic network including the S enterprises-suppliers manufacturing different complete set for further manufacturing the finished products by M enterprisesmanufacturers, transportation network, and set of points of finished product delivery. Each enterprise-supplier purchases the raw materials and/or complete set from the vendors. We make the following assumptions and simplifications: * The market of raw materials is unlimited. * All ordering of materials, complete set, and delivering the finished products occurs at the start of each period. * The lead time is zero: that is, an order arrives as soon as it is placed. * The production equipment of all enterprises is absolutely reliable. * The capacities of production lines of all enterprises are limited only by capacities of warehouses’ for storage of raw materials and finished products. The sth enterprise-supplier manufactures the Ls kinds of complete set from the Rs kinds of raw materials and other industrial resources. For manufacturing by the sth enterprise-supplier the lth kind of complete set’s unit it is needed to use the rth kind of ð1Þ raw materials in the amount of asrl ; s ¼ 1; 2; . . .; S; l ¼ 1; 2; . . .; Ls ; r ¼ 1; 2; . . .; Rs . The initial inventory level of the rth kind of raw material at the sth supplier’s Rs P ð1Þ ð1Þ warehouse is qsr . It is assumed that qsr W1s ; where W1s is the warehouse’s r¼1
capacity, s = 1,2,…, S. The capacity of warehouse for storage of produced complete set by the sth supplier is W2s and initial inventory level of complete set is Ls ð2Þ P ð2Þ qsl ; qsl W2s . l¼1
The complete set manufactured by enterprises-suppliers are purchased by the mth enterprise-manufacturer for manufacturing the Km types of finished products.
Dynamic Optimization Model for Planning of Multi-echelon Logistic
333
The capacity of warehouse for storage of the complete set at the mth enterprisemanufacturer is denoted by W3m and initial inventory level of each kind of complete set ð3Þ is qslm . It is naturally to assume that following constraint is valid Ls S X X
ð3Þ
qslm W3m ; m ¼ 1; 2; . . .; M:
s¼1 l¼1 ð2Þ
Let aslmk be the amount of the lth of complete set manufactured by the sth supplier needed for manufacturing the kth type of finished product’s unit in the mth enterprisemanufacturer, s ¼ 1; 2; . . .; S; l ¼ 1; 2; . . .; Ls ; k ¼ 1; 2; . . .; Km ; m ¼ 1; 2; . . .; M. The finished products of the mth enterprise-manufacturer come to the warehouse with the capacity W4m from which they must be delivered at the N points of destination (or finite consumption). The initial inventory level of the kth type of finished product at wareKm ð4Þ P ð4Þ house of the mth enterprise-manufacturer is qmk ; qmk W4m ; m ¼ 1; 2; . . .; M. k¼1
Let dmkn be the total demand for the kth type of finished product produced by the mth manufacturer at the nth destination over the planning horizon T. To avoid the trivial situation, we will assume that following conditions hold true ð4Þ
qmk \
X
dmkn ;
n2Bmk
where Bmk ¼ fnjdmkn [ 0; n ¼ 1; 2; . . .; N g; k ¼ 1; 2; . . .; Km . Note that the values dmkn may be determined as a result of market research and demand forecasting at points of destination of finished product. Let us introduce the control variables and corresponding variables describing the inventory levels fluctuation over the planning horizon T: ð1Þ
* Let xsrt be the amount of the rth kind of material ordered and purchased by the sth enterprise-supplier in period t, for t = 1,2,…,T. ð2Þ * Let xslt be the lth kind of complete set which the sth enterprise plans for output at the end of period t, for t = 1,2,…, T. ð3Þ * Let xslmt be the lth kind of complete set produced by the sth supplier purchased by the mth enterprise-manufacturer at the end of period t, for t = 1,2,…,T. * Let ymkt be the amount of the kth type of finished product planned for output by the mth enterprise-manufacturer at the end of period t, for t = 1,2,…,T. * Let zmknt be the amount of the kth type of finished product produced by the mth enterprise-manufacturer which is planned for delivery from warehouse to the nth destination at the end of period t, for t = 1,2,…,T. ð1Þ * Let Isrt be the inventory level of the rth kind of material at the warehouse of the sth supplier at the end of period t, for t = 1,2,…,T. ð2Þ * Let Islt be the inventory level of the lth kind of complete set manufactured by the sth supplier at the end of period t, for t = 1,2,…,T.
334
M. Ya. Postan et al. ð3Þ
* Let Islmt be the inventory level of the lth kind of complete set produced by the sth supplier at the warehouse of the mth manufacturer at the end of period t, for t = 1,2,…,T. ð4Þ * Let Imkt be the inventory level of the kth type of finished product manufactured by the mth manufacturer at the end of period t, for t = 1,2,…,T. To describe the economic effectiveness of supply, production and delivery plans we need the additional initial parameters, namely: ð1Þ
ð1Þ
* Let csrt be the per unit order cost and Ksrt be the fixed order cost for the rth kind of material ordered by the sth enterprise-supplier in period t, for t = 1,2,…,T. ð1Þ * Let eslt be the per unit production cost of the lth type of complete set produced by the sth supplier in period t, for t = 1,2,…,T. ð2Þ ð2Þ * Let cslmt be the per unit order cost and Kslmt be the fixed order cost for the lth kind of complete set produced by the sth supplier ordered by the mth enterprisemanufacturer in period t, for t = 1,2,…,T. ð1Þ ð2Þ * Let hsrt , hslt be the holding cost per unit of the rth kind of material and the lth kind of produced complete set of the sth enterprise-supplier correspondingly in period t, for t = 1,2,…,T. ð3Þ ð4Þ * Let hslmt ; hmkt be the holding cost per unit of the lth kind of complete set produced by the sth supplier at warehouse of the mth manufacturer and the kth finished product of the mth manufacturer correspondingly in period t, for t = 1,2,…,T. ð2Þ * Let cslt be the per unit production cost of the lth kind of complete set manufactured by the sth supplier in period t, for t = 1,2,…,T. ð3Þ * Let cslmt be the per unit cost for purchasing the lth kind of complete set by the mth manufacturer in period t, for t = 1, 2,…,T. ð2Þ * Let emkt be the per unit production cost of the kth type of finished product produced by the mth manufacturer in period t, for t = 1,2,…,T. ð5Þ * Let cmknt be the cost of transportation of the unit of the kth type of finished product from the mth manufacturer to the nth destination in period t, for t = 1,2,…,T. It is obvious that the following inventory-balanced equations are valid: ð1Þ
ð1Þ
ð1Þ
Isrt ¼ Isr;t1 þ xsrt
Ls X
ð1Þ ð2Þ
aslr xslt ;
s ¼ 1; 2; . . .; S; r ¼ 1; 2; . . .; Rs ;
ð1Þ
l¼1 ð2Þ
ð2Þ
ð2Þ
Islt ¼ Isl;t1 þ xslt
M X
ð3Þ
xslmt ;
s ¼ 1; 2; . . .; S; l ¼ 1; 2; . . .; Ls ;
ð2Þ
m¼1 ð3Þ
ð3Þ
ð3Þ
Islmt ¼ Islm;t1 þ xslmt
Km X
ð2Þ
aslmk ymkt ;
k¼1
1; 2; . . .; S; l ¼ 1; 2; . . .; Ls ; m ¼ 1; 2; . . .; M;
ð3Þ
Dynamic Optimization Model for Planning of Multi-echelon Logistic ð4Þ
X
ð4Þ
Imkt ¼ Imk;t1 þ ymkt
zmknt ;
335
ð4Þ
n2Bmk
k ¼ 1; 2; . . .; Km ; m ¼ 1; 2; . . .; M; t ¼ 1; 2; . . .; T; ð1Þ
ð1Þ
ð2Þ
ð2Þ
ð3Þ
ð3Þ
ð4Þ
ð4Þ
where Isr0 ¼ qsr ; Isl0 ¼ qsl ; Islm0 ¼ qslm ; Ikm0 ¼ qkm . The Eqs. (1)–(4) describe the dynamics of inventory levels in warehouses in each period of time. After solving the difference Eqs. (1)–(4), we obtain t X
ð1Þ
Isrt ¼ qð1Þ sr þ
ð1Þ
xsrj
j¼1
Ls t X X
ð1Þ ð2Þ
aslr xslj ;
ð5Þ
j¼1 l¼1
s ¼ 1; 2; . . .; S; r ¼ 1; 2; . . .; Rs ; ð2Þ
ð2Þ
Islt ¼ qsl þ
t X
ð2Þ
xslj
j¼1
t X M X
ð3Þ
xslmj ;
ð6Þ
j¼1 m¼1
s ¼ 1; 2; . . .; S; l ¼ 1; 2; . . .; Ls ; ð3Þ
ð3Þ
Islmt ¼ qslm þ
t X
ð3Þ
xslmj
j¼1
Km X t X
ð2Þ
aslmk ymkj ;
ð7Þ
k¼1 j¼1
m ¼ 1; 2; . . .; M; s ¼ 1; 2; . . .; S; l ¼ 1; 2; . . .; Ls ; ð4Þ
ð4Þ
Imkt ¼ qmk þ
t X
ymkj
t X X
zmknj ;
ð8Þ
j¼1 n2Bmk
j¼1
k ¼ 1; 2; ::; Km ; t ¼ 1; 2; . . .; T: Since the total inventory levels Rs X
ð1Þ
Isrt ;
Ls X
r¼1
ð2Þ
Islt ;
l¼1
Ls S X X s¼1 l¼1
ð3Þ
Islmt ;
Km X
ð4Þ
Imkt
k¼1
for any t can’t exceed the values W1s , W2s , W3m , W4m correspondingly, from (5)–(8), it follows Rs X r¼1
qð1Þ sr þ
Rs t X X
ð1Þ
xsrj
j¼1 r¼1
Ls X Rs t X X j¼1 l¼1 r¼1
ð9Þ ð1Þ ð2Þ aslr xslj
W1s ; s ¼ 1; 2; . . .; S; t ¼ 1; 2; . . .; T;
336
M. Ya. Postan et al. Ls X
Ls t X X
ð2Þ
qsl þ
ð2Þ
xslj
j¼1 l¼1
l¼1
t X M X
ð10Þ ð3Þ xslmj
W2s ; s ¼ 1; 2; . . .; S; t ¼ 1; 2; . . .; T;
j¼1 m¼1 Ls S X X
ð3Þ
qslm þ
Ls t X S X X
s¼1 l¼1
ð3Þ
xslmj
j¼1 s¼1 l¼1
Ls X Km t X S X X
ð11Þ ð2Þ aslmk ymkj
W3m ; m ¼ 1; 2; . . .; M; t ¼ 1; 2; . . .; T;
j¼1 s¼1 l¼1 k¼1 K X
ð4Þ
qmk þ
Km t X X
ymkj
j¼1 k¼1
k¼1
Km t X X X
zmknj W4m ;
j¼1 n2Bmk k¼1
ð12Þ
m ¼ 1; 2; . . .; M; t ¼ 1; 2; . . .; T: Rs X
qð1Þ sr þ
r¼1
Rs t X X
ð1Þ
xsrj
j¼1 r¼1
Ls X Rs t X X
ð1Þ ð2Þ
aslr xslj W1s ; s ¼ 1; 2; . . .; S; t ¼ 1; 2; . . .; T;
j¼1 l¼1 r¼1
On the other hand, the enterprises-suppliers for complete set manufacturing in period t can use only inventories of materials which are in warehouses in the end of period t − 1, that is Ls X
ð1Þ ð2Þ
ð1Þ
aslr xslt Isr;t1 ;
s ¼ 1; 2; . . .; S; r ¼ 1; 2; . . .; Rs ; t ¼ 1; 2; . . .; T:
ð13Þ
l¼1
Further, the all manufacturers in period t can use only inventories of complete set of the sth supplier, which are at warehouse in the end of period t − 1, therefore M X
ð3Þ
ð2Þ
xslmt Isl;t1 ; s ¼ 1; 2. . .; S; l ¼ 1; 2; . . .; Ls ; t ¼ 1; 2; . . .; T:
ð14Þ
m¼1
Similarly, for the mth enterprise-manufacturer the following restrictions must be fulfilled Km X
ð2Þ
ð3Þ
aslmk ymkt Islm;t1 ; s ¼ 1; 2; . . .; S; l ¼ 1; 2; . . .; Ls ; t ¼ 1; 2; . . .; T;
k¼1
m ¼ 1; 2; . . .; M:
ð15Þ
Dynamic Optimization Model for Planning of Multi-echelon Logistic
337
In period t it can’t be delivered the kth type of finished product at the all destið4Þ nations in amount more than inventory level in period t − 1, that is Imk;t1 . Therefore N X
ð4Þ
zmknt Imk;t1 ; k ¼ 1; 2; . . .; Km ; m ¼ 1; 2; . . .; M; t ¼ 1; 2; . . .; T:
ð16Þ
n¼1
At last, the kth finished product must be delivered at the nth destination in amount dmkn over the planning horizon, i.e. T X
zmknt ¼ dmkn ;
k ¼ 1; 2; . . .; Km ; n ¼ 1; 2; . . .; N; m ¼ 1; 2; . . .; M:
ð17Þ
t¼1
From (13)–(16), taking into account the relations (5)–(8), we obtain the following constraints Ls X t X
ð1Þ ð2Þ
aslr xslj qð1Þ sr þ
l¼1 j¼1
ð1Þ
xsrj ; s ¼ 1; 2; . . .; S; r ¼ 1; 2; . . .; Rs ;
ð18Þ
ð2Þ
ð19Þ
j¼1
t X M X
ð3Þ
ð2Þ
xslmj qsl þ
j¼1 m¼1 Km t X X
t1 X
t1 X j¼1
ð2Þ
ð3Þ
aslmk ymkj qslm þ
j¼1 k¼1 t X X
xslj ; s ¼ 1; 2; . . .; S; l ¼ 1; 2; . . .; Ls ;
t1 X
ð3Þ
xslmj ; s ¼ 1; 2; . . .; S; l ¼ 1; 2; . . .; Ls ;
ð20Þ
j¼1 ð4Þ
zmknj qmk þ
j¼1 n2Bmk
t1 X
ymkj ; m ¼ 1; 2; . . .; M; k ¼ 1; 2; . . .; Km ;
j¼1
ð21Þ
t ¼ 1; 2; . . .; T: The following conditions of non-negativity of control parameters must be added to the constraints (9)–(12), (16)–(19) ð1Þ
ð2Þ
ð3Þ
xsrt ; xslt ; xslmt ; ymkt ; zmknt 0; 8s; r; l; m; k; n; t:
ð22Þ
As an objective function, we choose the total logistic cost for the whole supply chain under consideration over the planning horizon. Taking into account the designations introduced previously, the expression for this total cost takes the form
338
C ¼
M. Ya. Postan et al. Rs T S X t X X X ð1Þ ð1Þ ð1Þ ð1Þ ð1Þ ð1Þ f ½csrt xsrt þ Ksrt dðxsrt Þ þ hsrt ðqð1Þ xsrt sr þ t¼1
s¼1 r¼1
Ls t X X
j¼1
ð1Þ ð2Þ
aslr xslt Þ þ
j¼1 l¼1
þ
Ls S X X
ð1Þ ð2Þ
ð2Þ
ð2Þ
½esl xslt þ hslt ðqsl þ
t X
s¼1 l¼1
Ls M X S X X
ð2Þ
ð3Þ
ð2Þ
xslj
j¼1
ð2Þ
t X M X
ð3Þ
xslmj Þ
j¼1 m¼1
ð3Þ
½cslmt xslmt þ Kslt dðxslmt Þ
m¼1 s¼1 l¼1 ð3Þ
ð3Þ
þ hslmt ðqslm þ
t X
ð3Þ
xslmj
j¼1
þ
Km M X X
ð2Þ
Km t X X
ð2Þ
aslmk ymkj Þ
j¼1 k¼1 ð4Þ
ð4Þ
½emkt ymkt þ hmkt ðqmk þ
t X
m¼1 k¼1
þ
Km X M X X
ymkj
t X X
zmknj Þ
j¼1 n2Bmk
j¼1 ð3Þ
cmknt zmknt g;
m¼1 k¼1 n2Bmk
ð23Þ Where dðxÞ ¼ 1 if x [ 0; dð0Þ ¼ 0. Thus, we can formulate the following optimization problem: it is needed to find out ð1Þ ð2Þ ð3Þ the variables xsrt ; xslt ; xslmt ; ymkt ; zmknt satisfying the constraints (9)–(12), (17)–(23) and minimizing the function (23). This optimization problem may be solved, for example, by dynamic programming algorithm or by the method based on reduction of our optimization model to partly integer linear programming problem.
3 Optimization Model for Random Demand Over Planning Horizon Now we will assume that values dmkn ðxÞ are the continuous mutually independent random variables with the given probability densities umkn ðdÞ. Here we will apply the approach proposed in the work (Williams, 1963). Put umkn ¼
T X
zmknt ;
t¼1
where umkn is total amount of the kth finished product planned for delivery to the nth destination before realization of random demand dmkn ðxÞ. In result of its realization, one of two risks may occur: 1. umkn \dmkn ðxÞ, i.e. demand will not be met; 2. umkn [ dmkn ðxÞ, i.e. there is necessity of storage of the kth finished product’s surplus at the nth destination.
Dynamic Optimization Model for Planning of Multi-echelon Logistic
339
We assume that both risks belong to the destinations, i.e. the all manufactured finished products will be sold. Let pmkn be the penalty for the kth finished product’s deficit at the nth destination, and hmkn be the holding cost for storage per unit of the kth product at the nth destination. Then average total logistic cost over the planning horizon is ¼ Cþ C
Km X M X N X m¼1 k¼1 n¼1
Zumkn fpmkn
ðumkn wÞumkn ðwÞdw 0
Z1 þ hmkn
ð24Þ
ðw umkn Þumkn ðwÞdwg; umkn
is concave in respect to the variables ukn . where C is defined by (23). The function C Taking second derivative of C with respect to ukn and applying the Leibnitz rule, we obtain @2 C ¼ ðpik þ hkn Þukn ðumkn Þ: @u2mkn Since pkn þ hkn [ 0 by definition, the expression on the right-hand side of the last equality is non-positive. Hence, function (24) is concave.
4 Conclusion Our paper contributes to the deepening of the knowledge on the recent application of the inventory control theory in the logistics networks analyses and optimal planning. It focuses on a problem of optimization of logistics practices by simultaneously considering the supply, production and transportation plans of two types of inventories: raw materials and finished products to arrive at a cost total logistics minimum. For this purpose, the approach has been developed to modeling and optimization of logistic network on the basis of inventory control theory. The classical Wagner-Whitin model is generalized for the case of final set of suppliers, manufacturers, and points of destination. Our approach allows to realize the optimal synergism in result of better coordination between all participants of multi-echelon supply chain. The further research in this direction may be focused on the following topics: * taking into account the influence of stochastic fluctuations of demand at destinations in each period if time; * consideration of competition among different supply chains forming the logistic network (Postan et al., 2017).
340
M. Ya. Postan et al.
References Zipkin, P.H.: Foundations of Inventory Management, 1st edn. McGrow-Hill/Irwin, Boston (2000) Bramel, J., Simchi-Levi, D.: The Logic of Logistics: Theory, Algorithms, and Applications for Logistics Management. Springer, New York (1997) Brandimarte, P., Zotteri, G.: Introduction to Distribution Logistics. Wiley, New York (2007) Christopher, M.: Logistics & Supply Chain Management (Creating Value-Adding Networks), 4th edn. Prentice Hall, Harlow (2011) Dachkovski S, Wirth, F., Jagalski, T.: Autonomous control in shop floor logistics: analytic models. In: Chryssolouris G, Mourtzis D (eds.) Manufacturing, Modelling, Management and Control 2004. Elsevier Science Ltd, Amsterdam (2005) Smith, J.M., Tan, B. (eds.): Handbook of Stochastic Models and Analysis of Manufacturing System Operations, vol. XXVIII. Springer, New York (2013) Morozova, I.V., et al.: Dynamic optimization model for planning of integrated logistical system functioning. In: Kreowski, H.-J., et al. (eds.) Dynamics in Logistics. Lecture Notes in Logistics, pp. 291–300. Springer, Heidelberg (2013) Postan, M.Y.: Economic-Mathematical Models of Multimodal Transport. Astroprint, Odessa (2006). (in Russian) Postan, M.Y., et al.: Dynamic model for optimization of production and finished products delivery plans in supply chain. Logistyka 4, 2345–2352 (2014) Shapiro, J.F.: Modeling the Supply Chain. Duxbury/Thomson Learning, Pacific Grove (2001) Williams, A.C.: A stochastic transportation problem. Oper. Res. 11(5), 759–770 (1963) Postan, M.Y., et al.: Method of finding equilibrium solutions for duopoly of supply chains taking into account the innovative activity of enterprises. East. Euro. J. Enterp. Technol. 3(87), 25– 30 (2017). https://doi.org/10.15587/1729-4061.2017.103989
Simulation-Based Sensitivity Analysis of Dynamic Contract Extension Elements in Supplier Development Haniyeh Dastyar1(B) and J¨ urgen Pannek2 1
International Graduate School for Dynamics in Logistics, University of Bremen, Bremen, Germany [email protected] 2 IAV GmbH, Development Center, Rockwell Str. 16, 38518 Gifhorn, Germany [email protected] Abstract. Today’s companies consider supplier performance as an essential factor to their competitive advantage specifically in dynamic marketplace. Therefore, supplier development experiences notable attention in variant industries. In this study, we aim to analysis the effects of an efficient supplier development program on the overall supply chain profit. We conduct a sensitivity analysis concerning supplier development cost parameters. To this end, we apply a model predictive control scheme in order to optimize a supplier development program in a monopolistic setting, consisting of one manufacturer and one supplier. Therefore, we conduct a full factorial simulation experiment varying all relevant parameters. The result shows that implementing supplier development provides significantly higher profits for the supply chain. Additionally, the study shows that, a high learning rate leads to the highest profit increase. Hence, manufacturer can invest on enhancing supplier’s learning rate to gain higher profit of supply chain. Keywords: Supplier development Simulation · Sensitivity analysis
1
· Dynamic contract extension ·
Introduction
The quality of supplied components is essential to a manufacturing’s products and therefore to its success in the market. Therefore, manufacturers are highly dependent to the performance of their suppliers like as reliable delivery, cost and quality [14]. Hence, supplier development (SD) received noticeable attention from manufacturers who aim to gain competitive advantages among their competitors. The purpose of supplier development is to enhance the performance of the supplier and to cope with the supply requirements of manufacturers, e.g., by improving reaction times to market changes, enhancing customer services, developing the quality of products, producing new products, or reducing the costs of production [1,5,14]. Therefore, both parties tempt to achieve competitive advantages through supplier development. c The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 341–350, 2020. https://doi.org/10.1007/978-3-030-44783-0_33
342
H. Dastyar and J. Pannek
Partnership requires a new type of relationship. Such as working together to gain shared goals. According to the principle that both partners are able to gain higher achievements through co-operation than competition [3]. Then manufacturers share knowledge and resources in an opportunism protection frame to increase the interest for both engaged parties [10]. Formal contracts are legal means to guarantee the commitments of each partner. These contracts are utilized as the basic tools for safeguarding, specifically in a turbulent setting [11]. One of major drawbacks of long-term contracts is the need of prediction of all possible situations. Moreover, such contracts impose the obligation to track the performance of the business partner, which becomes more and more difficult, especially in an obscure environment [16]. Worthmann et al. [17] conducted dynamic contract extensions in supplier development, and they defined important parameters which affect dynamic contract extension, such as the prohibitive price, price elasticity, contract period, learning rate, supplier development cost and etc. Given this background, the main purpose of the paper is to investigate the effect of supplier development’s parameters on the profitability of supplier development projects. This is particularly important, as manufacturers and suppliers can only influence a small subset of these parameters and, thus, should focus on the more efficient parameters. We seek to answer the following question: Which parameters have a stronger/weaker influence on supplier development programs and how can we assess if an application will be efficient or not. To answer this question, we use the model provided by [17] and perform a sensitivity analysis by performing a full factorial simulation experiment. The underlying model uses a model predictive control approach, also called the receding horizon approach, to optimize decision making for dynamically extending a contract over various (simulated) time steps. The remainder of this paper is structured as follows: First, related literature is briefly reviewed in Sect. 2. In Sect. 3, the model to determine the optimal switching time is described and parameters are detailed. In Sect. 5 and Sect. 4 the simulation study is introduced, and the effects of parameter variations are evaluated and discussed. In the final Sect. 6 conclusions are drawn, and possible future work is outlined.
2
Related Literature
Many researchers have paid considerable attention to supplier development [4,6,9,12–15,17]. In various industries, supplier development has been utilized [14]. In the automotive industry, Toyota started preparing on-site support, to include suppliers into the Toyota Production System [13]. Boeing, Chrysler, Daimler, Dell, Ford, General Motors, Honda, Nissan, Siemens, and Volkswagen followed this collaborative procedure to develop suppliers performance and/or capabilities [12]. In the supplier development context, engaged parties have to decide between formal contracts on the one hand, which represent a formalization as protection against opportunism, and relational contracting, which relies
Simulation-Based Sensitivity Analysis
343
more on a relational mechanism to enhance successful exchange. Relationshipspecific investments conduce to more satisfactory results. Moreover, when the continuation of the relationship is highly expected, the tendency to participate in supplier development activities is higher [9]. According to [15], supplier development is more effective in mature stages in comparison with primary phases of relationship life cycles. Dyer and Singh [4] added that adequate protection mechanisms might affect both costs and the inclination of companies to invest relationship-specific resources for supplier development. Application of mathematical models in general and control theory, in particular, is increasing in decision-making within supply chains. Among them, model predictive control (MPC), also termed receding (rolling) horizon control, is of particular interest as it allows to deal with nonlinear constrained multi-input multi-output systems as well as key performance criteria, see [6] for details. The method itself is a well-established strategy to deal with uncertainties in supply chains, see, e.g., [7]. In [17], for the first time MPC is used in supplier development to mitigate possible contractual drawbacks by dynamical extending the contract.
3
Model Description
In this paper, we follow the setting from [17] and consider a monopolistic supply chain, which consists of one manufacturer M and one supplier S. We assume that M is dominating the market and M determines the quantity of final products, to maximize profit. To this end, M increases/decreases market demands by decreasing/increasing the product sale price. Here, we consider a linear price distribution function p(d) = a − bd, where d indicates the quantity of production and p indicates the sale price, cf., e.g., [2,8,11,17]. a > 0 and b > 0 represent the willingness to pay (prohibitive price) and the price elasticity of the production. In this study, we assume that the willingness to pay is the maximum of production price that costumer can afford. Due to profit maximization, the production quantity d chosen by M is determined by the zero of the first derivative of the profit given by d · (p(d) − cM − cSC ), where cM and cSC denote the unit production costs of M and the supply costs per unit of S. Therefore, we obtain p(d) − cM − cSC − bd = 0 and thereby the optimal production quantity and optimal sale price d =
a − cM − cSC , 2b
p(d ) =
a + cM + cSC 2
In order to include fixed revenues per unit r of the supplier S similar to [2,8,11,17], we define cSC := r+cS , where cS represents the unit production costs of the supplier S. Now, we maximize the overall profit, which is a summation of the manufacturer’s profit J M and the supplier’s profit J S : J(cS ) = J M (cS ) + J S (cS ) =
a − cM (r + cS ) (a − cM − cSC )2 + r 4b 2b
According to the learning concept, greater experience or development in human resource capabilities, production process or machines, enable the supplier to
344
H. Dastyar and J. Pannek
improve its capability. Learning causes less effort, time and in turn also less production costs per unit. In order to formulate the problem more realistic, we take the supplier learning rate into account. We suppose that the manufacturer wants to increase the overall profit of the supply chain by decreasing the unit production costs cS of the supplier via supplier development projects. To this end, we introduce a time-dependent function x : R → R as the ability of supplier ˙ = u(t), x(0) = S to influence its costs via learning, that is cS = c0 xm , withx(t) x0 = 1 where c0 > 0 represents the base unit production cost of the supplier S, and m < 0 characterizes its learning rate. Last, x defines the cumulative number of realized supplier development projects, which can be changed by the manufacturer M via u ∈ [0, ω]. The bound ω > 0 represents a resources availability limitation, which may be due to time, manpower, or budget. A variety of former studies proposed similar models of cost reduction through learning, cf., e.g., [2,8,11,18]. To account for the respective development projects, we integrate supplier development costs cSD u(t) into the profit function and obtain T JT (u; x0 ) =
m 2
(a − cM − c0 x(t) ) − r2 − cSD u(t)dt 4b
0
where cSD ≥ 0 represents the costs of a supplier development action. To have a more realistic setting, we do not consider the problem parameters to be deterministic but some to be stochastic. While parameters such as the learning rate m are stochastic and time-varying, parameters like the contract period T are most likely deterministic and fixed. In either case, the problem reveals a linear quadratic structure, which allows computing the optimal solution ω, if t < t u (t) = 0 else explicitly using Pontryagin’s maximum principle, cf. [8], where t characterizes the optimal switching time of collaboration given by mc0 (x(t) + ωt )m−1 (a − cM − c0 (x(t) + ωt )m ) =
2bcSD t − T
(1)
The latter can be interpreted as follows: Up to t every investment in supplier development results in an increased profit, while expenditures spent after t do not amortize during the contract period and are, thus, not economically reasonable within the considered setting. Note that expenditures spent after t may still result in increased profit if a longer contract period is considered.
4
Experimental Setup
In this paper, we apply a sensitivity analysis, in which varying values for all parameters are considered in a fully factorial manor. Therefore, each parameter
Simulation-Based Sensitivity Analysis
345
is assigned a value range and a step width. For example, a parameter with a range of 10–20 and step width of 5 would result in three values 10, 15, and 20. To assess the impact of all parameters, we simulate each combination of values and record the corresponding performance indicators. We assume the expected value of stochastic parameters according to the ranges given in Table 1, such as the learning rate m, supplier development cost per project cSD , profit margin r, variable cost per unit for supplier c0 , price elasticity b and willingness to pay a. Using these discretized input values, we compute the corresponding output values t and JT (u; x0 ). Table 1. List of parameters [17] Symbol Description
5
Value
Step
T
Number of contract periods/prediction steps 5(months)
ω
Resource availability
m
Learning rate
[−0.4, −0.1]
0.1
a
Willingness to pay
[150, 250]
10
b
Price elasticity
[0.05, 0.2]
0.05
c0
Variable cost per unit (S)
[85, 115]
5
cM
Variable cost per unit (M)
[50, 90]
10
cSD
Supplier development cost per unit
[8000, 12000] 500
r
Supplier profit margin
[10, 20]
1
5
Numerical Results
Simulating the full factorial set of 498.960 experiments, we obtain the results given in Figs. 1, 2 and 3. These figures show the impact of changing values of the given variable on the x-axis on the mean switching time (left) and the mean overall profit (right). For easier comparison, the mean profit is provided for scenarios with supplier development projects (line marked by asterisks) and for scenarios without supplier development (line marked by diamonds). For all figures, they-value always denotes the mean value over all experiments, which uses the provided x-value for the corresponding variable. Due to a large number of experiments, it can be assumed that the mean value correctly represents the influence of each variable on the overall solution. The results strongly support the idea of supplier development, as all derived figures show a significant increase in the overall supply chain profit when investing in supplier development. The parameters are given in Fig. 3, i.e., the learning rate, the supplier’s profit margin, and the supplier development projects’ cost mostly refer to the supplier development and, thus only have a marginal impact on the profit without supplier development. In contrast, the parameters are given in Figs. 1 and 2, i.e., the willingness to pay, the price elasticity, and the supplier’s and manufacturer’s production
H. Dastyar and J. Pannek 40
15
30
10
Profit
Switching Point
346
20 10
104
5
150
200
0
250
150
Prohibitive price (a)
40
3
30
2
20
0.05
0.1
0.15
250
With SD
Profit
Switching Point
Mean optimal contract length
10
200
Prohibitive price (a)
0.2
Price elasticity (b) Mean optimal contract length
10
Without SD
5
1 0
0.05
0.1
0.15
0.2
Price elasticity (b) With SD
Without SD
Fig. 1. Effect of varying values for the willingness to pay (top), the price elasticity (bottom) on the mean value for the optimal switching time and the mean value of the overall profit
cost have a strong impact in both cases. In general, it can be stated that the shape of both scenarios is comparable (apart from the learning rate) however, the mean profit of scenarios with supplier development are significantly higher. For single parameters, most show a monotone trend for both the mean profit and the mean optimal switching point. Exceptions are the willingness to pay and the price elasticity given in Fig. 1. For the willingness to pay, profit shows a monotone and strong increase, which is expected as the willingness to pay denotes the maximum price a manufacturer can take for its product. Regarding the switching point, the contract length strongly decreases until a willingness to pay of approximately 190 is reached. After this point, the optimal contract length increases slowly again. The strong decrease, in the beginning, can be explained by the low willingness to pay. As this value needs to cover all manufacturing costs, a low price requires high investments in supplier development to achieve optimal profit. After the willingness to pay of approximately 190 is reached, more money can be spent on supplier development as the overall costs are covered. From this point on, longer projects can be funded to achieve higher benefits
Simulation-Based Sensitivity Analysis 8
104
6
20
Profit
Switching Point
22
18
4 2
16 80
90
100
110
0
120
Cost per unit (supplier) (c0)
80
90
20
3
Profit
Switching Point
4
10
110
With SD
25
15
100
120
Cost per unit (supplier) (c0)
Mean optimal contract length
5
347
10
Without SD
4
2 1
50
60
70
80
90
Cost per unit (manufacturer) (cM) Mean optimal contract length
0
50
60
70
80
90
Cost per unit (manufacturer) (cM) With SD
Without SD
Fig. 2. Effect of varying values for the supplier cost per unit (top), and the manufacturers price per unit (bottom) on the mean value for the optimal switching time and the mean value of the overall profit
from supplier development, while still retaining a high overall profit. This result shows that supplier development is a useful tool to mitigate low willingness to pays by reducing the manufacturing cost. Even if only a low-profit margin is achievable, high investments still result in an increased profit compared to the scenario where no supplier development is performed. For the profit, it can be observed that a very low elasticity results in the highest profit and accordingly the longest contracting periods. With an increasing elasticity, the profit decreases and tends towards the point where it matches the scenario without supplier development. In terms of the contract length, an increasing length can mitigate this effect for higher elasticities up to a certain degree. The values shown in Fig. 2 depict the cost per unit of product for suppliers and manufacturers. As supplier development directly influences the supplier’s cost per unit, higher values also lead to longer contract horizons. Nevertheless, increasing costs reduce the overall profit, whereas the scenario with supplier development can achieve significantly higher profits. For the manufacturer’s
348
H. Dastyar and J. Pannek 10
20
Profit
Switching Point
25
104
5
15 10
-0.4
-0.3
-0.2
-0.1
0
-0.4
-0.3
Lerning rate (m)
With SD
19
8
18.5
6
Profit
Switching Point
Mean optimal contract length
18
4
10
15
0
20
10
15
20
Profit margin (r)
Mean optimal contract length
With SD
19
8
18.5
6
Profit
Switching Point
Without SD
104
Profit margin (r)
18
10
Without SD
4
4 2
17.5 17
-0.1
2
17.5 17
-0.2
Lerning rate (m)
0.8
0.9
1
1.1
1.2
Project cost (cSD) 104 Mean optimal contract length
0
0.8
0.9
1
1.1
1.2
Project cost (cSD) 104 With SD
Without SD
Fig. 3. Effect of varying values for the learning rate (top), the supplier profit margin (middle) and the cost for supplier development projects (bottom) on the mean value for the optimal switching time and the mean value of the overall profit
production cost, the same trend can be observed: Higher costs lead to longer contracting periods to reduce the overall manufacturing costs. In terms of the profit, it can be seen that the profit for the supplier development scenario decreases stronger than for the higher values of the supplier’s cost per unit. Similar to the price elasticity, the profit for increasing values of cM converges towards the scenario without supplier-development, indicating less efficiency with increasing manufacturing costs.
Simulation-Based Sensitivity Analysis
349
As for the parameters shown in Fig. 3, the profit for the scenario without supplier development remains more or less stationary, regardless the setting. For the learning rate this is to be expected. Also, the increase in project cost shows a slight decrease in supplier development profit. However, in some cases, the profit of supplier development can be increased based on higher investment in supplier, which means higher project cost may lead to higher payoff for manufacturers who invest in the supplier. For the supplier’s profit margin it can be assumed that smaller changes within the provided range don’t have a huge overall impact, whereas a second study would be useful to assess the impact of larger variations. For the project costs, it can be seen that increasing costs generally lead to significantly lower contracting periods and slightly decreasing profits. For the learning rate, it can be seen, that a high learning rate (low value) leads to a considerable increase in profit. Consequently, it can be useful to invest in better learning rates by specific development programs, e.g., by providing trained staff, infrastructure or extended training for the suppliers.
6
Conclusion
In supplier development, one typically uses long-term contracts to entail certain risks. We considered a receding horizon control scheme based on much shorter but repeatedly prolonged contracts, which leads to enhance the supplier development process. Due to shorter contract periods, an added value is generated as both the manufacturer and the supplier gain flexibility. In this paper, by using a full factorial sensitivity analysis, we investigate the impact of willingness to pay, price elasticity, variable cost per unit for the supplier and manufacturer, supplier learning rate, supplier development cost and profit margin for the supplier on the optimal switching time and the average supply chain profit. The results show that supplier development is noticeably profitable in general. Furthermore, greater willingness to pays and learning rates lead to greater profit through supplier development. In contrast, higher price elasticity and cost per unit for manufacturer cause lower profits. The results for the supplier profit margin depict no connection between this parameter, the profit and the optimal supplier development period. Since the willingness to pay and price elasticity are not under the manufacturer’s control, and can be defined with market condition the results of this study highly suggest concentrating on enhancements of the supplier’s learning rate through offering training, developing the suppliers producing processes and etc. In our future work, we plan to enrich the model by less stringent assumptions like, e.g., a linear price distribution. Also, the dynamic conditions of the market can be considered, while we apply the cost function of manufacture and supplier and cost parameters change over time and based on the market condition. Another interesting direction for future research is to expand our study to a network perspective, in which the supply chain consists of more than a single manufacturer and a single supplier.
350
H. Dastyar and J. Pannek
Acknowledgement. Ms. Dastyar’s work was supported by the Friedrich-NeumannStiftung f¨ ur die Freiheit under grant no. ST8224/P612.
References 1. Bai, C., Sarkis, J.: Green supplier development: analytical evaluation using rough set theory. J. Clean. Prod. 18(12), 1200–1210 (2010) 2. Bernstein, F., K¨ ok, A.: Dynamic cost reduction through process improvement in assembly networks. Manage. Sci. 55(4), 552–567 (2009) 3. Dale, B.G., Burnes, B., Reid, I., Bamford, D.: Managing Quality: An Essential Guide and Resource Gateway, 6th edn. Wiley, London (2016) 4. Dyer, J., Singh, H.: The relational view: cooperative strategy and sources of interorganizational competitive advantage. Acad. Manage. Rev. 23(4), 660–679 (1998) 5. Govindan, K., Kannan, D., Haq, A.N.: Analyzing supplier development criteria for an automobile industry. Ind. Manage. Data Syst. 110(1), 43–62 (2010) 6. Gr¨ une, L., Pannek, J.: Nonlinear Model Predictive Control: Theory and Algorithms. Springer, Switzerland (2017) 7. Ivanov, D., Sokolov, B.: Control and system-theoretic identification of the supply chain dynamics domain for planning, analysis and adaptation of performance under uncertainty. Eur. J. Oper. Res. 224(2), 313–323 (2013) 8. Kim, B.: Coordinating an innovation in supply chain management. Eur. J. Oper. Res. 123(3), 568–584 (2000) 9. Krause, D.R.: The antecedents of buying firms’ efforts to improve suppliers. J. Oper. Manage. 17(2), 205–224 (1999) 10. Krause, D.R., Hand¨ ueld, R., Tyler, B.B.: The relationships between supplier development, commitment, social capital accumulation and performance improvement. J. Oper. Manage. 2(25), 528–545 (2007) 11. Li, H., Wang, Y., Yin, R., Kull, T.J., Choi, T.Y.: Target pricing: demand-side versus supply-side approaches. Int. J. Prod. Econ. 136(1), 172–184 (2012) 12. Routroy, S., Pradhan, S.: Evaluating the critical success factors of supplier development: a case study. Benchmark Int. J. 20(3), 322–341 (2013) 13. Sako, M.: Prices, Quality and Trust: Inter-Firm Relations in Britain & Japan. Cambridge University Press, Cambridge (1990) 14. Talluri, S., Narasimhan, R., Chung, W.: Manufacturer cooperation in supplier development under risk. Eur. J. Oper. Res. 207(1), 165–173 (2010) 15. Wagner, S.: Supplier development and the relationship life-cycle. Int. J. Prod. Econ. 129(2), 277–283 (2011) 16. Williamson, O.: Transaction-cost economics: the governance of contractual relations. J. Law Econ. 22(2), 233–261 (1979) 17. Worthmann, K., Proch, M., Braun, P., Schl¨ uchtremann, J., Pannek, J.: Towards dynamic contract extension in supplier development. Logist. Res 9, 1–12 (2016) 18. Yelle, L.: The learning curve: historical review and comprehensive survey. Decis, Sci. 10(2), 302–328 (1979)
Searching for Production Robustness Through Simulation-Based Scheduling Optimization Guilherme Ernani Vieira(&) and Enzo Morosini Frazzon Federal University of Santa Catarina, Florianopolis, SC 88040-900, Brazil {g.vieira,enzo.frazzon}@ufsc.br
Abstract. This paper proposes a new way to consider the dynamics of production execution through discrete event simulation. The proposed method models and simulates a production schedule using spreadsheets supplying input information for a discrete event simulation model that includes randomness (perturbations or time uncertainties) to processing and setup times. This is a new method that allows one to preview, for instance, how robust (resilient) a given schedule really is in midst of real production environment, where resources fail, suppliers delay deliveries, products need reprocessing etc. The proposed approach allows one to more accurately estimate performance of a given schedule execution subject to undesired and unexpected events because it models times using probability distributions instead of deterministic ones, often used by production planners (schedulers) and/or scheduling software tools. This method is very different from traditional mathematical optimization and simulation models, since it simulates the schedule itself, not using dispatching rules nor arrival rates. A three-machine production schedule illustrates the proposed approach. Under the assumptions considered, a 5% increase in total processing in time will probably occur. This waste (loss) was not “seen” during the time the production planner created the schedule (using deterministic setup and processing times). Keywords: Production scheduling Modeling Simulation Control and monitoring of manufacturing processes Robustness analysis
1 Introduction When a production planner (or scheduler) works on the creation of a production plan (or schedule), he usually considers setups and operations times as given and deterministic, ignoring the fact that time uncertainties (randomness) are intrinsic to everyone, everywhere, anytime. Therefore, in reality, it will be quite probable that the actual schedule executed will not be exactly what he had planned using deterministic times. Resources (machines, production lines and operators) will fail, suppliers will delay deliveries, products will need reprocessing and many other unplanned events will take place, requiring adjustments in order to minimize disruptions and changes to what was initially prepared or expected. However, how can one create a good production schedule that takes into consideration the fact that disruptions and changes (randomness) will happen during © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 351–362, 2020. https://doi.org/10.1007/978-3-030-44783-0_34
352
G. E. Vieira and E. M. Frazzon
execution? Can one foresee what the performance will be when unplanned uncertainties occur? This paper presents the idea of simulating a production schedule subject to random variations to processing and setup times in the schedule. This way, one can make better decisions by seeing what may actually happen to a given (theoretical) plan (schedule). Therefore, this method allows one to test feasibility, robustness, completion times, occupation, idleness and any other performance indicator the production planner/scheduler uses, considering more realistic and probable unexpected scenarios, using discrete event simulation (DES) integrated to scheduling data read from spreadsheets. This paper is organized as follows: Besides this introductory section, next section briefly describes some recent works in the literature using DES in production scheduling. Section 3 presents the proposed analysis method. Section 4 brings an example of a simple production schedule and shows how to implement the proposed method. Last section shows some final comments about the proposed development.
2 Some Works Using DES on Production Scheduling The literature brings an enormous amount of works related to production scheduling; however, it seems to be in its infancy regarding works involving schedule simulation. This paper briefly shows some few recent papers (from 2014 on) using simulation to address time uncertainties to production schedules. Vieira et al. (2017) are proposing the use of discrete-event simulation to evaluate production scheduling robustness. (The three machines production schedule used in this paper was taken from this work.) Gyulai et al. (2017) proposed a simulation-based optimization method that utilizes lower level shop floor data to calculate robust production plans for final assembly lines of a flexible considering time changes disturbances. Kück et al. (2016a and b) are working on a data-driven simulation-based optimisation approach for adaptive scheduling and control of dynamic manufacturing systems. Lin and Chen (2015) proposed a simulation optimization approach for hybrid flow shop scheduling problem in semiconductor back-end manufacturing. Pulido et al. (2017) present a case study regarding and analysis of the robustness of production scheduling in aeronautical manufacturing using simulation. For Yang et al. (2016), it is necessary to model and simulate production planning and controls (PPC) with information dynamics in order to analyze the risks caused by uncertainties. A new approach to simulate PPC systems was proposed, aimed at visualizing production processes and comparing key performance indicators as well as optimizing PPC parameters under different uncertainties in order to deal with risks. Using discrete-event simulation, the relationship between robustness and redundancy was analyzed for large-scale manufacturing system configurations. It was found that both a redundancy indicator derived from a classical manufacturing background and elementary flux modes, derived from systems biology, and are significantly correlated with performance robustness in manufacturing systems (2016).
Searching for Production Robustness
353
Pegden (2018) explains in a simple way, ideas very similar to the ones presented in this paper, stating “in some cases hours of computation are expended to generate a schedule; however the basic assumption of deterministic times makes this schedule unrealistic and optimistic from the start. By ignoring variations, APS (Advanced Planning and Scheduling) tools generate schedules that promise more than can be delivered”. Lastly (but far from being an exhaustive review), Leiva (2016) proposed a simulation-based production planning optimization technique integrating metaheuristics, simulation and exact techniques to address the uncertainty and complexity of manufacturing systems. Many other works propose to study schedule robustness in real-world scenarios (where disruptions, delays and undesired events happen during production), most of them, however, do not considered simulation as a tool to evaluate scheduling performance.
3 Main Steps Behind the Proposed Method The proposed method can simulate most of the production schedules found in industries and in the literature. It uses an integrated approach based on spreadsheets and discrete event simulation software to simulate execution of the production plan/schedule created using deterministic times. It is, therefore, a new way to simulate production schedules, not using dispatching rules, arrival rates nor mathematical models. This approach uses MS Excel with the schedule information and ARENA to run the simulation experiments (a newer version, using ProModel, is currently under development). During simulation execution, data regarding sequencing, production and setup times are read from the spreadsheets. The idea is as follows: Input data, or information describing the production schedule itself, is modeled in Excel using different spreadsheets: the first table is used to provide a few general configuration information: basically the number of scheduling resources (machines tools, production lines or operators, for instance) e the corresponding number of operations assigned (scheduled) to it. The other tables contain the sequence of setup and operations times scheduled at each machine. Each resource will require its scheduling data in a separate spreadsheet (in the same file). For the simulation execution, the DES model follows the steps described next: 1. Read information regarding the scheduling resource and the number of operations scheduled to it. 2. For each machine (for simplification purposes, from this point on, a scheduling resource will simply be a “machine”), simultaneously run the following sub-steps: (a) If there is an operation scheduled and not yet executed: when machine becomes available, read expected setup time for operation j from spreadsheet and execute it. Execution (simulated) times will be drawn from a triangular probability distribution, having as mode (most probable) value the deterministic time used in scheduling (“t), minimum value as (t − 5% t = 0.95t) and maximum value
354
G. E. Vieira and E. M. Frazzon
(b) (c) (d) (e) (f)
equal to (t + 15% t = 1.15t). (This idea came from the fact that it is more usual to have a delay than having an early finish.) Otherwise (i.e. there is no operation left to be executed), define completion time for the machine. If this operation has a previous operation, execution must wait until it finishes. Read expected processing time for the operation from spreadsheet and execute it. Update number of operations executed to the machine. If there is another operation scheduled for this machine, go back to step (a).
Once there is no other operation to be executed, calculate performance indicator(s). Currently in the first part of this research, maximum completion time (or makespan) is used as the (only) performance indicator.
4 Applying the Proposed Method to a Three Machine Schedule Example 4.1
Input Data: The Production Schedule
The scenario used to illustrate the proposed method considers a production schedule created for six production orders (PO1…PO6), each requiring two or three operations at three resources: machines M1, M2 and M3 along with corresponding needed setup times (Sij: Setup for POi operation j). Table 1 shows the sequences of machines on which the different product types have to be processed. The required operations are summarized as follows Gyulai et al. (2017): • • • • • •
PO1:[S11+O11](M1) and [S12+O12](M2); PO2:[S21+O21](M2), [S22+O22](M1) and [S23+O23](M3); PO3:[S31+O31](M3), [S32+O32](M2) and [S33+O33](M1); PO4:[S41+O41](M1) and [S42+O42](M3); PO5:[S51+O51](M2) and [S52+O52](M3); PO6:[S61+O61](M1) and [S62+O62](M2).
Table 1. Machine number and operation sequence for each production order. Machine Mk for operation Oij of production order POi i=1 i=2 i=3 i=4 i=5 i=6
j=1
j=2
j=3
M1 M2 M3 M1 M2 M1
M2 M1 M2 M3 M3 M2
– M3 M1 – – –
Searching for Production Robustness
355
Tables 2 and 3 show the setup and processing times for each operation at each machine and Fig. 1 shows a Gantt chart for the whole schedule. According to the computed schedule (using deterministic setup and processing times), the expected total expected production time (Cmax) for this scenario is 513 h. Table 2. Setup times (in hours) for the considered scenario. Setup time for setup Sij of production order POi i=1 i=2 i=3 i=4 i=5 I=6
j=1
j=2
j=3
1 1.5 1.5 1.5 1 1
1.5 1.5 1 2 2 2
– 1 1 – – –
Table 3. Processing times (in hours) for the considered scenario. Processing time for operation Oij of production order POi i=1 i=2 i=3 i=4 i=5 I=6
j=1
j=2
j=3
100 120 129 120 110 100
115 100 85 101 65 75
– 61 85 – – –
Fig. 1. A Gantt chart of the computed optimal production schedule for the scenario tested.
356
4.2
G. E. Vieira and E. M. Frazzon
Input Data: The Production Schedule in Excel
As mentioned in the previous section, input data, or information describing the production schedule itself, is modeled in Excel using different spreadsheets. For the three machines scenario used here to explain the proposed method, four spreadsheets are need and represented in Tables 4, 5, 6 and 7. First table describes the number of operations scheduled and the other three tables the sequence of operations scheduled to each machine, with setup and processing times.
Table 4. Configuration and sequence scheduled for each machine for the considered scenario. Machine # # of operations scheduled 1 5 2 5 3 4
Table 5. M1 scheduled setups and operations. Setup code 11 14 22 16 33
Setup duration 1 1.5 1.5 1 1
Oper. code 11 14 22 16 33
Oper. duration 100 120 100 100 85
Next Seq. Oper. code 21 24 32 26 NULL
Prev. Seq. Oper code NULL NULL 12 NULL 23
Table 6. M2 scheduled setups and operations. Setup code 12 21 23 15 26
Setup duration 1.5 1.5 1 1 2
Oper. code 12 21 23 15 26
Oper. duration 120 115 86 110 75
Next Seq. Oper. code 22 NULL 33 25 NULL
Prev. Seq. Oper code NULL 11 13 NULL 16
Table 7. M3 scheduled setups and operations. Setup Code 13 24 32 25
Setup duration 1.5 2 1 2
Oper. code 13 24 32 25
Oper. duration 129 101 61 65
Next Seq. Oper. code 23 NULL NULL NULL
Prev. Seq. Oper code NULL 14 22 15
Searching for Production Robustness
4.3
357
The ARENA Simulation Model
The total simulation model was decomposed into three hierarchical levels in ARENA. Figure 2 shows the overall simulation model decomposed in six sub-models at the first hierarchical level. A close look at Figs. 3, 4 and 5 allows the reader to identify all hierarchical levels, their blocks and submodules.
Fig. 2. First hierarchical modeling level.
Fig. 3. “Start schedule execution” submodule.
4.4
Simulation Results and Analysis
The total (maximum) production completion time (or makespan) was used as key performance indicator (KPI) to compare the theoretical schedule performance (estimated from a static schedule created with deterministic times) with the performance obtained by simulating the proposed schedule under setup and processing time uncertainties (variations or randomness). As said before, since usually delays are much more prone to happen than early finishing, variations to setup and processing times were modeled as a triangular probability distribution: TRIA(0.95t, t, 1.15t), where t is the deterministic time used by the scheduler. Therefore, during simulation, the actual time to be used will be a random number drawn from this distribution. Again, this will represent delays that occur in real world more realistically into the simulation model.
Fig. 4. “Execute setup and operation at Machine 1” submodule.
358 G. E. Vieira and E. M. Frazzon
Fig. 5. “Calculate KPIs and end schedule execution” submodule.
Searching for Production Robustness 359
360
G. E. Vieira and E. M. Frazzon
Besides these variations, two scenarios were tested: (a) Without machine failures. (b) With machine failures: There is a 10% chance of machine failure during processing. In this case, when it fails, it will be down for maintenance/repair for 2 to 6 h (uniform probability distribution). Based on one hundred simulation runs, the confidence interval for Cmax (considering a 95% confidence level) for both scenarios were: (a) Not considering machine failures: Cmax = 534.89 ± 1.78 h. This corresponds to 4.3% increase in relation to the 513 h from the theoretical production schedule, which representing about 22 h of delay. (See Fig. 6 taken from ARENA’s report.)
Fig. 6. Completion times and Cmax for scenario without machine failure.
(b) With machine failures: Cmax = 537.32 ± 1.85 h. A 4.7% increase, representing a 24 h delay. (See Fig. 7 taken from ARENA’s report.)
Fig. 7. Completion times and Cmax for scenario without machine failure.
5 Final Considerations Despite the fact that this production scheduling scenario was very simple, i.e., only three resources (“machines”) with 14 jobs (each PO operation can be seen as a job), the proposed simulation-based method showed significant difference between the theoretical Cmax (estimated from a schedule created using deterministic setup and processing times) and the mean value of Cmax operating under time uncertainties. In both situations, a delay of approximately 24 h was found, which is very interesting, considering the simplicity of the production schedule used (a 5% increase in production time can represent a lot of waste and money). Imagine in a real manufacturing scenario, with hundreds of orders (jobs), dozens resources, and many other unexpected events
Searching for Production Robustness
361
(e.g., material shortages due to suppliers´ delays or quality issues, operators absenteeism, weather conditions, transportation problems etc.)? One will easily see that the performance, promised by advanced planning and scheduling software - or simply created using spreadsheets, still quite common nowadays - will quite probably not be accomplished in the shop floor. Therefore, the proposed simulation method allows one to foresee what will “in fact” happen during schedule execution and hence, allowing the decision maker to adjust production and/or contracts prior to randomness takes place. The well-known truth is that problems will always occur, but being able to see them beforehand, gives business an edge that can help it succeed in this fierce competitive environment companies live in. Acknowledgment. This work is funded by CAPES (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior) – Brazil – under reference number 99999.006033/2015-06, in the scope of the BRAGECRIM Program.
References Gyulai, D., Pfeiffer, A., Monostori, L.: Robust production planning and control for multi-stage systems with flexible final assembly lines. Int. J. Prod. Res. 55(13), 3657–3673 (2017). https://doi.org/10.1080/00207543.2016.1198506 Kück, M., Ehm, J., Freitag, M., Frazzon, E.M., Pimentel, R.: A data-driven simulation-based optimisation approach for adaptive scheduling and control of dynamic manufacturing systems. In: Wulfsberg, J.P., Fette, M., Montag, T. (eds.) Advanced Materials Research, pp. 449–456. Trans Tech Publications, Pfaffikon, Switzerland (2016a) Kück, M., Ehm, J., Hildebrandt, T., Freitag, M.: Potential of data-driven simulation-based optimization for adaptive scheduling and control of dynamic manufacturing systems. In: Roeder, T.M.K., Frazier, P.L., Szechtman, R., Zhou, E., Huschka, T., Chick, S.E. (eds.) Proceedings of the 2016 Winter Simulation Conference, pp. 2820–2831. Institute of Electrical and Electronics Engineers, Inc., Piscataway (2016b) Leiva, J.E.D.: Simulation-Based Optimization for Production Planning: Integrating MetaHeuristics, Simulation and Exact Techniques to Address the Uncertainty and Complexity of Manufacturing Systems. Ph.D. Thesis. University of Manchester (2016) Lin, J.T., Chen, C.M.: Simulation optimization approach for hybrid flow shop scheduling problem in semiconductor back-end manufacturing. Simul. Model. Pract. Theory 51, 100–114 (2015) Meyer, M.: Redundancy Investments in Manufacturing Systems: The role of redundancies for manufacturing system robustness. Doctor of Philosophy Thesis. Jacobs University (2016) Pegden, C.D.: Risk-based Planning and Scheduling: Why Variation Matters. Not published (2018). https://www.simio.com/resources/white-papers/RPS/Risk-based-Planning-andScheduling-Why-Variation-Matters.pdf Pulido, R., Borreguero-Sanchidrián, T., García-Sánchez, A., Ortega-Mier, M.: Analysis of the robustness of production scheduling in aeronautical manufacturing using simulation: a case study. In: Ríos, J., Bernard, A., Bouras, A., Foufou, S. (eds.) PLM 2017. IFIP Advances in Information and Communication Technology, vol. 517. Springer, Cham (2017)
362
G. E. Vieira and E. M. Frazzon
Vieira, G.E., Kück, M., Frazzon, E., Freitag, M.: Evaluating the robustness of production schedules using discrete-event simulation. IFAC PapersOnLine 50(1), 7953–7958 (2017). Elsevier Yang, S., Arndt, T., Lanza, G.: A flexible simulation support for production planning and control in small and medium enterprises. In: 9th International Conference on Digital Enterprise Technology - DET 2016 – Intelligent Manufacturing in the Knowledge Economy Era. Procedia CIRP, vol. 56, pp. 389–394 (2016)
A Multiagent System for Truck Dispatching in Open-pit Mines Gabriel Icarte1,2(B) , Paulina Berrios2 , Ra´ ul Castillo2 , and Otthein Herzog3,4,5 1
2
International Graduate School for Dynamics in Logistics (IGS), University of Bremen, Bremen, Germany [email protected] Faculty of Engineering and Architecture, Arturo Prat University, Iquique, Chile {gicarte,pberrios,raucasti}@unap.cl 3 Tongji University, Shanghai, People’s Republic of China 4 Jacobs University, Bremen, Germany 5 Center for Computing and Communication Technologies (TZI), University of Bremen, Bremen, Germany [email protected] Abstract. Material handling is an important logistic process for openpit mines. In this process, shovels extract materials and load trucks that transport these materials to different destinations. To support this process, different centralized dispatching solutions have been implemented based on mathematical programming, heuristic processes or simulation modelling. Weaknesses in these methods can be observed in addressing the dynamics of a mine and by not providing a precise dispatching solution. In this paper, we present a solution based on Multiagent Systems (MAS) where the equipment items are represented by intelligent agents that interact with each other to meet the production goals at a minimum cost. The results obtained by applying the MAS in a simulated open-pit mine with actual data show more specific solutions than the current centralized solutions in a practical calculation time frame. In addition, the MAS decreases the truck costs on average by 18%.
Keywords: Truck dispatching Scheduling
1
· Open-pit mine · Multiagent systems ·
Introduction
Material handling is an important logistic process for open-pit mining since it can amount to up to 50% of the operational cost [1]. In this process, trucks and shovels work together to extract and to transport all the material required by the operational plan at minimum cost. In order to reach this objective, dispatching trucks efficiently becomes an important task. However, it is a hard task due to the number of the variables and the dynamics of the environment where the equipment items operate. c The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 363–373, 2020. https://doi.org/10.1007/978-3-030-44783-0_35
364
G. Icarte et al.
The current systems that support the dispatching of trucks follow a centralized approach that commonly uses an allocation model that assigns trucks to trips between loading and unloading points [7]. This solution does not provide a precise operation sequence of the equipment, therefore it cannot secure an efficient use of the equipment. To improve the efficiency of the material handling process, we developed a multiagent system (MAS). The paper demonstrates the applicability of the MAS. To do this, first, we compare the solutions provided by the MAS against a mathematical model and second, we use actual data from a Chilean open-pit mine to compare the actual transported material and the material that could have been transported following the solution proposed by the MAS. The remainder of this paper is structured as follows: Sect. 2 presents some background of truck dispatching in open-pit mines. A mixed integer linear programming (MILP) scheduling model is described in Sect. 3. Section 4 presents the distributed approach based on MAS. Section 5 presents the results and discusses the evaluation of the MAS approach in a case study. Finally, conclusions and outlook are presented in Sect. 6.
2
Problem Definition
In the open-pit mine material handling process shovels extract materials and load trucks that transport these materials to different destinations at the mine. If the extracted material is waste, it is transported to a waste dump, and if it is ore, it is transported to a crusher or a stockpile. Figure 1 shows all operations that a truck must perform to transport materials from a loading point to an unloading point. This is called the truck cycle. This cycle is performed and repeated by each truck until the shift ends.
Fig. 1. The truck cycle.
At first sight, truck dispatching in open-pit mines seems to be a kind of vehicle routing problem (VRP). However, although there are some similarities,
MAS for Truck Dispatching
365
there also are differences: there is no start and end node (depo), the trucks must go to a pickup node (shovel) and then to a delivery node (crusher or stockpile) and repeat this sequence during the entire shift. This implies that a truck can visit a node more than once. The travel times between nodes are short and the number of nodes is too lower than the number of trucks. This produces waiting time at nodes. These differences, make difficult to apply a pure VRP model. Different centralized systems have been implemented to support the dispatching of trucks based on mathematical programming, heuristic processes or simulation. The strengths of these methods are their maturity and their well-known implementation. However, the weaknesses can be observed in addressing the dynamics of a mine [2], by not providing a precise solution [7], the use of estimated information [6], and the time needed to calculate a dispatching solution when the model is too complex. A common strategy applied in some centralized systems is based on the multistage approach [1]. This approach uses a guideline that is computed in the upper stage. Then, this guideline is used by the lower stage as a reference to make real-time dispatching decisions. Despite the use of these systems, the trucks and shovels do not operate efficiently since queues of trucks are built-up in front of shovels and crushers, as well as idle time of shovels. Therefore the problem is how to improve the efficiency in the material handling process. Alternatively, a more specific solution that would allow the equipment items to operate more efficiently would be to set up schedules for each equipment item with all the operations that it must perform, pointing out the start times, end times, etc.
3
Formalization
To address the problem, a mathematical model is formulated based on the work of Patterson [7], which uses a MILP with the objective of minimizing the energy consumption of the shovels and trucks taking into account the targets of the production plan in an open-pit coal mine. The model uses a sequence of loading ‘slots’ per shovel to organize the operations of trucks and shovels. In our model, trucks can be assigned to any shovel. The shovels are assigned to one pit and the material extracted by a shovel must be transported to a destination throughout the shift. Shovels can load one truck at once. At a crusher, one truck can unload at once, whereas in a waste dump or a stock pile several trucks can unload simultaneously. The notation of sets, indices, parameters and decision variables used in the model is shown in Table 1. The objective function (1) is to minimize the cost (in terms of time) that is taken by the shovels and the trucks to perform the operations. Restriction (2) ensures that at most one truck is assigned to each time slot l on each shovel. Restriction (3) ensures that no more than one truck r is loaded on the shovel at once. Restriction (4) ensure that at crushers, no more than one truck can unload at a time. The restriction (5) ensures that an unloading time (μs,l ), starts after the loading starts (λs,l ) plus the time it takes to perform the loading Cs and
366
G. Icarte et al. Table 1. Formulation notation.
Set
Index Description
S
s
R
r
Trucks
Ls
l
Slot time of the shovel s
J
j
Destinations
Shovels
Parameters Cs
Loading time of shovel s
Cj
Unloading time at the destination j
Cjs
Travel time from shovel s to the destination j
Csr
Travel time of truck r to shovel s (only at the beginning of the shift)
Csj
Travel time from the destination j to next shovel s
Ar
Truck capacity
δs
The target of extracted material by shovel s
M
Sufficiently large positive number
Decision variables Xr,s,l
1 if the truck r loads at shovel s in the time slot l, otherwise 0
λs,l
Loading start time of shovel s in time slot l
μs,l
Unloading start time of material extracted by shovel s in time slot l
λseq r,s,l,s l
1 if truck r was loaded by shovel s in time slot l before being loaded in shovel s and slot time l . Otherwise 0
the travel time to the destination Cjs . The restriction (6) ensures that the next loading of a truck in l must be after the truck ends the unloading in l. (Csj + Cs + Cjs + Cj ) (1) M in ∀s,l
Xr,s,l ≤ 1 ∀l, s
(2)
∀r
λs,l+1 − λs,l ≥ Cs ∀l, s
(3)
μs,l+1 − μs,l ≥ Cj ∀s, l
(4)
μs,l ≥ λs,l + Cs + Cjs ∀l, s
(5)
λs ,l ≥ μs,l + Csj + Cj − M (2 − λseq r,s,l,s l − Xr,s,l ) ∀r, l, s, s l seq λr,s,l,s l ∀r, l, s Xr,s,l =
(6) (7)
s ,l
Xr,s,l = r,l
λseq r,s l ,s,l ∀ r, l, s
(8)
s ,l
Xr,s,l Ar ≥ δs ∀s
(9)
MAS for Truck Dispatching
λs,1 ≥ Csr − M (1 − λseq r,0,0,s,1 ) ∀r, s seq λr,0,0,s,l = 1 ∀r
367
(10) (11)
s,l
λseq r,s,l,0,0 = 1 ∀r
(12)
s,l
Restrictions (7) and (8) ensure that the sequence of each truck has one predecessor and a successor. Restriction (9) ensures that the proposed loading targets in the production plan for each shovel are met. Restriction (10) ensures that a truck travel time to its first loading point must be considered at the beginning of the shift. Restriction (11) ensures that all trucks start in a dummy pit. This pit represents the initial place where a truck is at the beginning of the shift. Restriction (12) ensures that all trucks also end the shift in a dummy pit.
4
An Alternative Solution Approach: Multiagent System
A Multiagent System (MAS) is a system collection of agents that are intelligent software programs representing an entity from the real world and/or provide a certain service [4]. The agents act autonomously and make decisions to reach the objectives of their represented entities using their specific data, communication mechanisms and sharing their knowledge. A problem can be divided into smaller problems that the agents can solve optimally due to the smaller complexity of the problem. 4.1
Scheduling MAS Architecture
The objective of the implemented MAS is to accomplish the goals of the production plan at minimal cost. Applying this approach allows us to model truck dispatching in a way that is closer to reality and to avoid the weaknesses of centralized systems. The agents implemented in the MAS include the following ones: – Truck agent: This agent represents a truck of the real world. Its objective is to create a schedule of the operations of the truck at minimal cost. The main specific data used are capacity, loaded velocity, empty velocity, spotting time and unloading time. In addition, the agent uses the layout of the mine. The agent can play the role of a participant in a negotiation process. – Shovel agent: This agent represents a shovel of the real world. Its objective is to create a schedule of the operations of the equipment that it represents considering its target in the production plan. The main specific data are capacity, dig velocity, load velocity and the destination of extracted material. The agent can play the role of an initiator in a negotiation process. – UnloadingPoint agent: This agent represents a crusher, stockpile or waste dump of the real world. Its objective is to create a schedule of the operations of the equipment that it represents. The main specific data is the number of trucks unloading simultaneously.
368
4.2
G. Icarte et al.
Interaction in the Scheduling MAS
In order to create the schedules, the agents must interact with each other using the Contract Net Protocol (CNP) [9] which is a well-known negotiation mechanism for task sharing with near-optimal solutions. In this context, the CNP works as follows: a Shovel agent starts a negotiation process sending a call for proposals (CFP) to the Trucks agents pointing out the time when the shovel is available to load a truck and the idle time from the last loading. When a Truck agent receives the CFP it must evaluate it. The agent checks its schedule and asks the UnloadingPoint agent for information about the prospective waiting time. With this information, the Truck agent calculates the arrival time and the cost to perform all the operations and decides to send a proposal or to refuse the CFP. If it sends a proposal, the Truck agent waits for the answer. The Shovel agent receives the proposals and after receiving all proposals or if the deadline is expired, looks for the best proposal. Then, the Shovel agent sends an acceptance message to the Truck agent that offered the best proposal and sends a rejection message to the other Truck agents. The Truck agent that receives the acceptance of its proposal adds a new assignment to its schedule. The Shovel agent adds it to its schedule. If the Shovel agent does not receive proposals, the negotiation is finished. As the agents work in parallel, several CNP negotiations are done concurrently. As a consequence, a Truck agent may receive several CFPs. If the Truck agent sends a proposal answering one of this CFPs, it must wait for the answer from the Shovel agent, and therefore, the other received CFPs are refused. This situation can generate that the Truck agent refuses a CFP that is a better option than the CFP answered previously. This problem is also called “the eager bidder problem” [8]. To avoid this problem, a confirmation stage was included in the CNP that works as follows: when the Shovel agent finalizes the evaluation of the proposals, it sends a confirmation message to the Truck agent with the best proposal. The Truck agent that receives the confirmation message, could refuse the confirmation (in the case that it has received a better CFP), otherwise it can accept the confirmation. If the Truck agent refuses the confirmation of the Shovel agent, the Shovel agent sends a confirmation message to the next best proposal received. In this way, the Truck agent could decommit a previous proposal sent. If the Shovel agent receives only rejections from the Truck agents, the negotiation process is ended. Figure 2 depicts the interaction between the agents using the CNP with confirmation stage. Table 2 shows a schedule example for a truck created by the MAS using this protocol. 4.3
Decision Making
The decision making process among agents is one of the most important characteristics of a MAS. A bad design of the decision making could generate bad results or lets the agents take more time for their decisions affecting the performance of the MAS.
MAS for Truck Dispatching
369
Table 2. Example of schedule created for a truck. Assignment Destination
Trip start time
Arrival time
Spotting start time
Loading or unloading start time
Assignment end time
0
Shovel.01
00:47:01
01:20:23
01:20:23
01:21:36
01:23:12
1
WasteDump.02 01:23:12
01:32:33
01:32:33
01:32:33
01:33:23
2
Shovel.04
02:18:47
02:18:47
02:20:00
02:21:12
3
WasteDump.03 02:21:12
02:26:38
02:26:38
02:26:38
02:27:28
4
Shovel.04
02:27:28
02:31:37
02:31:37
02:32:50
02:34:02
5
WasteDump.03 02:34:02
02:39:28
02:39:55
02:39:55
02:40:45
02:10:39
Fig. 2. The interaction between the agents using the CNP with the confirmation stage.
The Shovel agents receive proposals from the Truck agents. These proposals mention the time that a truck could start the loading at the shovel, and the time that it takes to perform all the operations. After receiving all the proposals (or if the deadline is expired) the Shovel agent evaluates all the proposals using a utility function. This function promotes those proposals that propose to start the loading on time and with the least time to perform all operations. In this way, the Shovel agent selects the proposal that minimizes its idle time and offers the least truck cost. The Truck agent must decide, after receiving a CFP, if the offer can be performed by the truck. It must determine if the loading time offered by the
370
G. Icarte et al.
shovel fits the schedule of the truck. If there is no time slot, the Truck agent rejects the CFP. If there is a time slot, the Truck agent must calculate the total time that it takes the truck to perform all the operations and it determines if the offer is suitable for the time slot. In this process, the Truck agent applies the Djikstra algorithm [3] to find the shortest paths (from the last unloading point to the shovel and from the shovel to the destination of the material extracted by that shovel) and calculates the travel times. If the offer suits the time slot, the Truck agent sends a proposal, otherwise rejects the CFP. Another decision of the Truck agents is on the confirmation stage. After receiving a confirmation message from the Shovel agent, the Truck agent must decide to confirm the message or reject it. If it is taking part in another negotiation process with a potential better award (in this case a negotiation with a lower cost for the truck) the Truck agent will reject the confirmation message of the Shovel agent, otherwise it will accept it. If the Shovel agent completes a negotiation process without a winner, it starts a new negotiation process, but increases the loading time offered by one minute this time. However, it could happen that the negotiation ends again without a winner, and the Shovel would start another negotiation process adding another minute to the offer. This situation would generate idle time in the schedule of the shovel. To avoid this, the Truck agents consider the shovel idle time from the last loading. If the shovel idle time from the last loading is less than one minute and the Truck agent is taking part in another negotiation with a potential better award, the agent rejects the confirmation message. Nevertheless, if the shovel idle time is higher or equal to one minute, the Truck agent must confirm it. In this way, the Truck agents prefer to achieve the goals of the production plan instead of decreasing their own cost.
5
Results and Discussion
Two experiments were done to validate the approach. The purpose of the first experiment was to compare the time that takes the MAS to generate schedules against the time that takes an exact solver with the implementation of the mathematical model presented in Sect. 3. The purpose of the second experiment was to compare the truck cost obtained by the MAS against actual data. The experiments use actual data from an open-pit copper mine in Chile. In that mine, the equipment items operate in shifts of 12 hours and the material handling is done with a heterogeneous fleet of trucks and shovels. The specific data of the agents were taken from the actual real-world data. The actual data were generated by DISPATCH (TM), which is a centralized system based on dynamic programming [5]. The implemented MAS was deployed and executed in PlaSMA [10], which is a simulation platform for MAS. The implementation of the mathematical model was done in CPLEX. In the first experiment, several simulations run with different parameters. All instances use the same data, i.e., the same mine layout, and trucks and shovels with the same characteristics. The differences are the number of equipment items
MAS for Truck Dispatching
371
and the length of the shift. Table 3 shows different instances (H: length of the time horizon of the shift; the number of shovels; the number of trucks) and the performance results of the MAS and CPLEX in term of calculation time for the schedules. Table 3. Times needed for the MAS and CPLEX to generate the schedules. Instance H Shovels Trucks MAS time (min) CPLEX time (min) 1
1 1
10
0,04
0,06
2
3 3
25
0,43
5,06
3
6 5
40
2,45
>180
4
9 7
60
4,52
–
5
12 9
85
16,74
–
In the case of the bigger instances, the MAS provides the schedules in a practical time frame of about 16 min on a standard PC. This is due to the characteristics of the MAS technology, that lends itself to a mainly distributed organization, parallel processing, and a lower computational power requirement. The results obtained from the mathematical model implemented in CPLEX show that is not possible to get solutions for the bigger instances in practical time. This is because of the increase of variables generates a large number of combinations that the solver must evaluate. In the second experiment, 5 real-world shifts were simulated. The MAS generated schedules for the shovels to extract the same amount of material extracted in the actual data. Table 4 shows a comparison between the actual transported material and the material that could have been transported following the schedules proposed by the MAS. Table 4. Comparison of the production target cost of MAS schedule vs actual data. Id Shift (hours)
Shovels Trucks Actual material transported (tons)
Actual Simulated travel time material (hours) transported (tons)
Simulated travel time (hours)
Delta Delta material travel transported time
1 12
11
99
350.117
821,9
351.659
597,44
+0,44%
−27,31%
2 12
12
96
350.005
796,36
351.895
668,37
+0,54%
−16,07%
3 12
11
98
404.921
849,97
405.903
713,33
+0,24%
−16,08%
4 12
12
94
409.555
813,01
411.345
676,99
+0,44%
−16,79%
5 12
12
93
386.973
783,49
389.404
656,58
+0,63%
−16,2%
The solution provided by the MAS achieves the targets of the production plan at, on average, decreased cost by 18% even with marginally bigger production goals. One of the reasons for these savings is that the MAS travel times of a truck are smaller than the travel times in the actual data since the agents in the
372
G. Icarte et al.
MAS use the shortest path for their travels, whereas the truck operators in the real world decide by themselves which path to follow. Another reason is the use of specific data to allow for more adapted calculations of the operation times of every equipment, and in this way, the agents can create more appropriate and efficient schedules.
6
Conclusions
A Multiagent System for truck dispatching in open-pit mines has been presented. Experimental results show that the MAS provides more precise solutions than a centralized system within a practical computation time frame. In addition, the generated schedules by the MAS are more efficient since they decrease the truck cost on average by 18% meeting even marginally bigger production goals. Future investigations will address two aspects: on the one hand, the dynamic of the material handling process, e.g., dealing with a major change in the mine such as equipment failures or changes in the mine layout. In this case, the affected agents will have to react appropriately, interacting with each other to update their schedules. On the other hand, the MAS will be compared against other methods that provide solutions in practical frame time such as metaheuristics algorithms.
References 1. Alarie, S., Gamache, M.: Overview of solution strategies used in truck dispatching systems for open pit mines. Int. J. Min. Reclam. Environ. 16(1), 59–76 (2002). https://doi.org/10.1076/ijsm.16.1.59.3408. http://www.tandfonline.com/doi/abs/ 10.1076/ijsm.16.1.59.3408 2. Bastos, G.S., Souza, L.E., Ramos, F.T., Ribeiro, C.H.C.: A single-dependent agent approach for stochastic time-dependent truck dispatching in open-pit mining. In: 14th International IEEE Conference on Intelligent Transportation Systems (ITSC), pp. 1057–1062 (2011). https://doi.org/10.1109/ITSC.2011.6082902 3. Djikstra, E.W.: A note on two problems in connexion with graphs. Numer. Math. 1(1), 269–271 (1959) 4. Gath, M.: Optimizing transport logistics processes with multiagent-based planning and control. Ph.D. thesis (2015) 5. Munirathinam, M., Yingling, J.C.: A review of computer-based truck dispatching strategies for surface mining operations. Int. J. Surf. Min. Reclam. Environ. 8(1), 1–15 (1994). https://doi.org/10.1080/09208119408964750 6. Newman, A.M., Rubio, E., Caro, R., Weintraub, A., Eurek, K.: A review of operations research in mine planning. Interfaces 40(3), 222–245 (2010). https://doi.org/ 10.1287/inte.1090.0492 7. Patterson, S.R., Kozan, E., Hyland, P.: Energy efficient scheduling of open-pit coal mine trucks. Eur. J. Oper. Res. 262(2), 759–770 (2017). https://doi.org/10.1016/ j.ejor.2017.03.081 8. Schillo, M., Kray, C., Fischer, K.: The eager bidder problem: a fundamental problem of DAI and selected solutions. In: Proceedings of the 1st international Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS 2002), January, pp. 599–606 (2002). https://doi.org/10.1145/544862.544886
MAS for Truck Dispatching
373
9. Smith, R.G.: The contract net protocol: high-level communication and control in a distributed problem solver. IEEE Trans. Comput. C–29(12), 1104–1113 (1980). https://doi.org/10.1109/TC.1980.1675516 10. Warden, T., Porzel, R., Gehrke, J.D., Herzog, O., Langer, H., Malaka, R.: Towards ontology-based multiagent simulations: plasma approach (2007)
Drone Delivery Using Public Transport: An Agent-Based Modelling and Simulation Approach Raheen Khalid and Stanislav M. Chankov(&) Jacobs University Bremen, Campus Ring 1, 28759 Bremen, Germany [email protected]
Abstract. Drone delivery is considered as one possible solution to the last-mile delivery problems arising from the growth of e-commerce and customer expectations. Besides, urban public transport allows accessibility and connectivity among various locations within a city. Merging these two transportation modes, we propose a new futuristic delivery concept called Drone Delivery using Public Transport (DDPT). DDPT allows drones to deliver packages by riding existing public transport vehicles and aims to resolve the problems of last mile delivery. The purpose of this paper is to introduce the new DDPT concept and investigate its potential by comparing it with the traditional mode of truck delivery. Accordingly, we develop four agent-based simulation models and conduct an initial simulation study with 36 cases by varying the number of packages and their inter-arrival times. The comparison analysis is based on four key performance indicators: (1) Delivery Service Level, (2) Average Delivery Time, (3) CO2 Emissions and (4) Utilization. The results indicate that the DDPT concept is more efficient and environmentally friendly than traditional delivery. Keywords: Drone delivery based simulation
Public transport Last-mile delivery Agent-
1 Introduction The rapid adaption of internet over the past years has played a crucial role in altering methods of consumer consumption by establishing trust towards the online stores [1]. This has led to an exponential rise of the e-commerce market which in turn gave birth to last mile delivery challenges [2]. Last mile delivery is considered the most expensive, wasteful and environmental degrading part of the supply chain [3, 4]. Bringing competitive edge for retailers, many companies are placing efforts in finding solutions to the hurdles faced by last-mile delivery. Drone delivery is one such solution, especially for small and light packages [5]. Drones are unmanned aerial vehicles (UAVs) – electric or hybrid means of transportation that fly above the ground [6, 7]. For last mile delivery, the three main reasons for the adoption of drones are cost, speed and convenience [8]. This has driven Logistic companies, such as Alibaba [9], DHL [10], Google [11] and Amazon [12], to start investing in and testing drone delivery. Amazon predicts that delivery using drones will help to reduce the last mile costs by 80% and reduce delivery time © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 374–383, 2020. https://doi.org/10.1007/978-3-030-44783-0_36
Drone Delivery Using Public Transport: An Agent-Based Modelling
375
significantly [13]. Similar to Amazon, DHL has shown that drones can reduce average delivery time from 40 min to just 8 min along with 80% reduction in cost per delivery when compared with normal truck delivery [10]. However, drones alone face challenges in performing last-mile deliveries due to their battery life constraints [14] and social concerns with regards to numerous drones in the sky [8]. Previous work has focused on combining drones with trucks [15] (TSP-D) and autonomous mobility [16] (DDAM) for delivery. With TSP-D solution to many problems, such as carbon emissions, traffic congestion in last mile delivery, are still unclear. Whereas, DDAM still needs clarity for its feasibility and viability. An alternative delivery method combining drone delivery with public transport is suggested for future research by [16]. This method relies on future assumptions that a drone can carry substantial weight and can charge itself on a charging pad installed on the roof of public transport, but appears to be a possible solution to the last-mile delivery challenges. Thus, the purpose of this paper is to further develop the Drone Delivery using Public Transport (DDPT) concept and investigate its potential by comparing it with the traditional mode of truck delivery. Accordingly, we develop agent-based models to represent both DDPT and traditional truck last-mile deliveries in the city of Bremen, Germany. Subsequently, we conduct an initial simulation study and compare the two delivery modes on four key performance indicators (1) Delivery Service Level, (2) Average Delivery Time, (3) CO2 Emissions and (4) Utilization. The paper proceeds as follows. Section two presents the DDPT concept, while section three explains the methodology used in this study. We present and discuss our results in section four. Lastly, section five provides a brief summary of the investigation, its limitations and outlook for further research.
2 Drone Delivery Using Public Transport (DDPT) Concept Similar to the DDAM concept, the DDPT concepts limits the flying time of the drones by utilizing existing resources. While DDAM suggests that drones could ride on top of autonomous vehicles, DDPT proposes that the drones could use public transport vehicles, such as trams, buses and trains, as an intermediary transportation method for delivery. In essence, the drone picks a parcel from a packet shop, flies to the nearest public transport stop, hops on the public transport, rides it to the public transport stop nearest to the parcel’s destination and flies the remaining distance to the destination.
1
2
3
5
7
6
4 Packet Shop
Drone with package
Bus Stop Near Packet Shop
Bus Stop Near Delivery House
Drone delivering Delivery House package
Fig. 1. Concept illustration of DDPT with one Bus connection.
376
R. Khalid and S. M. Chankov
Figure 1 shows a single transfer process in DDPT with bus consisting of seven steps: Step 1: Drone is loaded with packages at the Packet Shop. Step 2: Drone searches for the nearest Bus stop. Step 3: When the bus with the direction towards the Delivery House arrives, the drone flies onto the roof of the Bus. Step 4: The bus drives in its original route while the drone stays idle. Step 5: When the Bus has reached the stop nearest to the Delivery House, the Drone disconnects from the Bus. Step 6: The drone flies towards the Delivery House. Step 7: Delivery of Package is done by drone at the Delivery House. The process remains the same for train and tram. In addition, the drones can also interchange in the various combinations of public transport depending on the end location of delivery house and the connectivity of the public transport. In that case, the drone gets off at a changeover stop and then hops onto another public transport that goes in the direction of the delivery house. Figure 2 shows an example for a transfer from a bus to a tram.
1
3
2
4
Packet Shop
Drone with package
Bus Stop Near Packet Shop
7
5
8
9
6
Changeover Stop
Tram Stop Near Delivery House
Drone delivering Delivery House package
Fig. 2. Concept illustration of DDPT with a Bus-Tram changeover.
3 Methodology 3.1
Simulation Setup
Agent-based simulation is a common approach for modeling package delivery [17]. In order to compare the DDPT concept to traditional truck delivery, two separate agentbased models are created to represent the two delivery methods respectively. Moreover, two different delivery types are modelled: (1) intra-city delivery and (2) delivery from an online retailer. In the intra-city scenario, it is assumed that a package is dropped in a packet shop in the city and needs to be delivered to a customer in the same city, while in the delivery from an online retailed, the package needs to be delivered from the distribution center of the online retailer (located out of the city center) to a customer inside the city. The models are developed using AnyLogic Software to model package deliveries in the western part of the city of Bremen, Germany. The locations of five DHL offices are used as Packet Shops (PS) and Amazon’s regional distribution center in Winsen is considered to represent the online retailer. For the DDPT model, the routes and schedule of the local public transport are adopted. The average speeds of the public
Drone Delivery Using Public Transport: An Agent-Based Modelling
377
transports are set the same as stated by transport service providers: for bus 40 km/h, for tram 40 km/h and for train 120 km/h. The average speed of drone is set to 80 km/h [18]. For the truck delivery model, trucks perform milk runs to deliver packages running at an average speed of 40 km/h (aligned with the urban speed limit in Germany). 3.2
DDPT Models
The intra-city DDPT delivery model includes the following agents: Delivery House, Tram, Bus, Public Transport Stop and Packet Shop. Figure 3 shows the general model process flow. The packages are divided in the five Packet Shops according to the population density of their corresponding regions. Some packages need to be delivered to delivery houses from the same region as the Packet Shop where they were generated (intra-region), while other times they need to be delivered to the region of another Packet Shop (inter-region). On start-up two drones leave each Packet Shop, one carrying inter-region packages and the other – intra-region. 84% of packages weigh less than 4–5 kg [19], whereas a drone can carry up to 300 kg [20, 21] and thus it is assumed that a drone can carry 10 packages at once. Customized functions are created within the model to navigate the drone agent by calculating shortest distances between public transport stops, its current location and packages’ destination locations. Customized functions are also created for the drone to hop onto the correct public transport in the right direction and find out the changeover stop; this is achieved by comparing collections of stops and the stop nearest to the delivery house.
Fig. 3. Model process flow – DDPT: intra-city delivery.
378
R. Khalid and S. M. Chankov
The model for DDPT delivery from an online retailer is similar to the DDPT intracity model with small modifications. The Packet Shop agents are replaced by one distribution center agent in Winsen. The packages are created at the distribution center and are assigned to one of the five Bremen regions depending on their delivery destination. There is one drone agent doing deliveries to each region. There are additional train agents, which are used by the drones to travel from Winsen central station to Bremen central station. Once, in Bremen central station, the drones continue using the bus and tram network. 3.3
Truck Delivery Models
The traditional truck delivery model includes Delivery House, Packet Shop and Truck agents for both intra-city and online-retailer deliveries. Similar to the DDPT intra-city model, the packages are divided into inter-region and intra-region packages based on population density. Inter-region deliveries can be performed on the same day, while inter-region deliveries can only be conducted on the next day, as we assume that it takes one day to deliver the packages to their corresponding packet shop. Each packet shop is assigned one truck performing a milk run inside the packet shop’s region. In the online-retailer truck delivery model, we assume that the delivery from the distribution center to the packet shop nearest to the package destination takes one day. The remaining delivery is the same as in the intra-city delivery model. 3.4
Key Performance Indicators (KPIs)
For comparison between DDPT and traditional truck delivery, four KPIs, focusing on operational and environmental aspects, are monitored in the simulation models: (1) Delivery Service Level, (2) Average Delivery Time, (3) CO2 Emissions and (4) Utilization. Delivery service level is an important factor for delivery and is calculated by obtaining the percentage of total packages delivered from the total packages available. Delivery time is the time between customer’s order generation and package delivery. The CO2 emissions are used as a measure for the carbon footprint of the two delivery methods. They are calculated based on the distance travelled and the rate of CO2 emissions for each vehicle: 162 g/km for trucks [22], 0.004 g/km for drones [23], 46 g/km for trams [24], 92 g/km for bus [24] and 9.7 g/km for trains [25]. The utilization of the drone or truck is calculated as the ratio of the time the drone or truck spends delivering packages to the total time of the drone or truck in the system.
4 Results and Discussion 4.1
Experiment Generation
The aim of the simulation is to compare DDPT and traditional truck delivery, which is done by changing two parameters. A total of nine simulation scenarios are created for each of the four simulation models by varying the number of packages and their interarrival times. The total number of packages is varied with the parameter values of: 70,
Drone Delivery Using Public Transport: An Agent-Based Modelling
379
140 and 210. In addition, three different package inter-arrival times are used: no interval, random intervals and fixed intervals. No Interval represents the case when all packages are available for delivery at the start of the simulation run. Random and fixed intervals represent the cases when a new package is generated after a random or fixed interval respectively. Each experiment is iterated for 10 times. Based on the public transport schedule in Bremen and the fact that drones can operate longer times independently, the DDPT simulation time is set to 8:00–20:00, whereas for truck delivery the standard working hours of 8:00–17:00 are used [26]. A screenshot presenting an overview of the simulation animation can be found in Fig. 4.
(a) DDPT Intra-city Delivery
(b) Truck Intra-city Delivery
Fig. 4. Simulation experiment overview.
4.2
Experiment Results
Figure 5 shows the results for the four KPIs obtained from the experiments by varying the two parameters: number of packages and inter-arrival times. The service level for drones is higher than for trucks across both the intra-city and online retailer delivery (Fig. 5(a)). In the intra-city delivery, DDPT has the advantage of being able to deliver inter-region packages directly on the same day, while the truck delivery needs to wait for the inter-region packages first to be delivered to their corresponding Packet Shop, thus leading to lower service level for the truck delivery. For online-retailer delivery, all the packages that need to be delivered by truck take one day to reach from the distribution center to the Packet Shops in the city, which lowers the service level for truck delivery excessively. On the other hand, the DDPT drones, already located at the premises of the retailer, can leave the distribution center immediately, take the train to Bremen and deliver packages directly. By combining drone delivery with public transport, DDPT can fulfil more deliveries faster. The delivery time for trucks is a lot higher than for drones across both the intra-city and online retailer delivery (Fig. 5(b)). This stems from the one-day delay in package delivery of inter-region packages and package delivery from retailer to Packet Shops. Drones are not impeded by the delay barriers which is the primary reason for the delivery time being much less.
380
R. Khalid and S. M. Chankov
(a) Delivery Service Level
(b) Average Delivery Time
(c) CO2 Emissions
(d) Utilization Fig. 5. KPIs simulation results.
Drone Delivery Using Public Transport: An Agent-Based Modelling
381
CO2 emissions in the truck delivery scenarios are greatly higher than emissions in the DDPT scenarios across both delivery types (Fig. 5(c)). The results for utilization are different than the results for the other KPIs as the Truck Utilization is slightly higher than the drone utilization for both delivery types (Fig. 5(d)). For all the experiments we see that as the number of packages increases, the Utilization, delivery time and CO2 emissions increases, while the delivery service decreases. When comparing the time parameters of each models, we see that the No Interval scenario has the highest delivery service level with the least delivery time. The better efficiency of no intervals is because all the packages arrive at the start of the day and the drones and trucks all leave together which reduces the overall utilization as less time is spent delivering. After observing the experiment results, it can be implied that DDPT is better than truck delivery for package delivery. DDPT is much faster, has higher service levels, and emits 2–4 times less CO2 emissions than truck delivery. One slight drawback of the DDPT concept is the fact that the utilization of drone is less than trucks. However, when looking at the values of average delivery time and service levels, it can be ignored.
5 Conclusion An agent-based simulation study was conducted to investigate the viability of the DDPT concept when compared to traditional truck delivery. Our findings suggest that DDPT is viable and more efficient than truck delivery for intra-city deliveries as well as deliveries from online retailers. Delivery companies can achieve greater customer satisfaction through DDPT due to lower average delivery time and greater service levels. Delivery companies can also significantly reduce CO2 emissions by adopting the concept of DDPT, reducing the deliveries’ environmental impact. Thus, the main contribution of this paper is the development and evaluation of an innovative and novel concept that can help companies handle the last mile delivery challenges. Even though, the simulation analysis and design are created based on realistic values and considerations, there are still certain limitations during the analysis and the design phase of the simulation. The DDPT models are generated based on futuristic assumptions (1) a drone can carry substantial weight and can charge itself on a charging pad installed on the roof of public transport, (2) legal regulations would allow drones to fly from a public transport stop to the final package destination and (3) advances in drone technology would make its operating costs comparable to the costs of traditional truck delivery. Delays, traffic problems in public transportation or actual customer order density have not been considered. Lastly, only 10 iterations are run per simulation scenario due to the limited time and computing capacity. Future research can further study the DDPT concept by comparing it with other existing delivery concepts including real-world traffic flows and package destination distributions. To bring the DDPT concept to life, a pilot run can be conducted utilizing the public transportation network of a particular city such as Bremen where approximately 500–600
382
R. Khalid and S. M. Chankov
Drones can easily replace the conventional truck type of delivery. This prototype can be used to evaluate the installation of drone charging pads on public transportation, the connection of drone with the public transport, the way the customers receive packages from the drones, the pay load and weight capacity of the drones and the placement of drone at the public transportation stops. This would contribute to a better understanding of the challenges and issues that the DDPT concept can face in real life.
References 1. Gefen, D.: E-commerce: the role of familiarity and trust. Omega 28(6), 725–737 (2000) 2. Madlberger, M., Sester, A.: The last mile in an electronic commerce business model-service expectations of Austrian online shoppers (2005). https://pdfs.semanticscholar.org/ea69/ 972a1a4608226005cf3261645eec1c013aa3.pdf [Accessed 4 Jul. 2019] 3. Gevaers, R., Van de Voorde, E., Vanelslander, T.: Cost modelling and simulation of lastmile characteristics in an innovative B2C supply chain environment with implications on urban areas and cities. Procedia Soc. Behav. Sci. 125, 398–411 (2014) 4. Ülkü, M.A.: Dare to care: shipment consolidation reduces not only costs, but also environmental damage. Int. J. Prod. Econ. 139(2), 438–446 (2012) 5. Xu, J.: Design perspectives on delivery drones (2017). https://pdfs.semanticscholar.org/ 5d42/ccdc6b5afd0ff8407a389789e1055de84fef.pdf. Accessed 7 Jul 2019 6. Culus, J., Schellekens, Y., Smeets, Y.: A drone’s eye view (2018). https://www.pwc.be/en/ documents/20180518-drone-study.pdf 7. Maarten, O.: Using cargo drones in last-mile delivery. Deloitte Nederland (2019). https:// www2.deloitte.com/nl/nl/pages/consumer-industrial-products/articles/using-cargo-drones-inlast-mile-delivery.html 8. Brar, S., Rabbat, R., Raithatha, V., Runcie, G., Yu, A.: Drones for Deliveries (2015). https:// scet.berkeley.edu/wp-content/uploads/ConnCarProjectReport-1.pdf 9. Reagan, J.: Alibaba Ramps Up the Drone Delivery Game in China. Dronelife (2015). https:// dronelife.com/2015/02/05/alibaba-ramps-drone-delivery-game-china/ 10. Deutsche Post DHL Group: DHL launches its first regular fully-automated and intelligent urban drone delivery service, 16 May 2019 (2019). https://www.dpdhl.com/en/mediarelations/press-releases/2019/dhl-launches-its-first-regular-fully-automated-and-intelligenturban-drone-delivery-service.html 11. X, The Moonshot Factory: Transforming the way goods are transported (2012). https://x. company/projects/wing/ 12. Holland, A.: Amazon’s Drone Patents (2017). https://dronecenter.bard.edu/files/2017/09/ CSD-Amazons-Drone-Patents-1.pdf 13. Amazon.com: Amazon.com: Prime Air (2019). https://www.amazon.com/Amazon-PrimeAir/b?ie=UTF8&node=8037720011 14. Lavars, N.: SkySense pad starts charging your drone the moment it lands. Gizmag (2014). https://newatlas.com/skysense-pad-charging-drone-lands/34592/ 15. Agatz, N., Bouman, P., Schmidt, M.: Optimization approaches for the traveling salesman problem with drone. Transp. Sci. 52(4), 965–981 (2018) 16. Yoo, H.D., Chankov, S.M.: Drone-delivery using autonomous mobility: an innovative approach to future last-mile delivery problems. In: 2018 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM) (2018)
Drone Delivery Using Public Transport: An Agent-Based Modelling
383
17. Cheng, P., Chankov, S.M.: Crowdsourced delivery for last-mile distribution: an agent-based modelling and simulation approach. In: 2017 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM) (2017) 18. Slyvester, G.: E-Agriculture in action: Drones for Agriculture. Food and Agriculture Organization for the United Nations (2018). http://www.fao.org/3/I8494EN/i8494en.pdf 19. Austin, R.: IPC cross border e commerce shopper survey 2017 (2017). https://www.ipc.be/ */media/documents/public/markets/2016/ipc-cross-border-e-commerce-shoppersurvey2017.pdf 20. DD Counter Measures: 10 Facts on How Much Weight a Small Drone Can Carry. DD Counter Measures (2018). https://www.ddcountermeasures.com/how-much-weight-a-smalldrone-can-carry/ 21. Suprapto, B.Y., Putro, B.K.: Optimized neural network-direct inverse control for attitude control of heavy-lift hexacopter. J. Telecommun. Electron. Comput. Eng. 9(2–5), 103–107 (2017) 22. European vehicle market statistics (2018). http://eupocketbook.org/wp-content/uploads/ 2019/04/ICCT_Pocketbook_080419.pdf 23. Park, J., Kim, S., Suh, K.: A comparative analysis of the environmental benefits of dronebased delivery services in urban and rural areas. Sustainability 10(3), 888 (2018). https:// pdfs.semanticscholar.org/1ba9/46503ea2325b413969b27f04da02d80e4214.pdf 24. Munich Transport Corporation (MVG) Sustainability Report. MVG (2014). https://www. mvg.de/dam/mvg/ueber/nachhaltigkeit/mvg-nachhaltigkeitsbericht-eng.pdf?fbclid= IwAR2cAx6CYxS0HXs4d3CB5HFWNG2YAVxaezDP0y18unDij_FyFYJedq2rJGg 25. Sims, R., et. al.: 8 Transport Coordinating Lead Authors: Lead Authors: Review Editors: Chapter Science Assistant (2014). https://www.ipcc.ch/site/assets/uploads/2018/02/ipcc_ wg3_ar5_chapter8.pdf 26. Teter, B.: How Do Standard Courier Services Work? Eurosender (2016). Accessed 5 Aug 2019
Part V: Intelligent Production and Logistics Systems
Perspectives on the Application of Internet of Things in Logistics Ícaro Romolo Sousa Agostino(&), Charles Ristow, Enzo Morosini Frazzon, and Carlos Manuel Taboada Rodriguez Industrial and Systems Engineering Department, Federal University of Santa Catarina, Florianópolis, SC 88040-900, Brazil [email protected], [email protected], [email protected], [email protected]
Abstract. Platforms based on IoT (Internet of Things) technologies can connect sensors and devices along the supply chains of production and logistics systems, as well as end-users of products, enabling efficient and customized solutions. This paper aims to present perspectives for the application of IoT technologies in logistics, covering theoretical and practical aspects. A Systematic Literature Review was carried out to identify the main characteristics of the research area, providing an updated bibliographic portfolio of studies related to the theme and grouping the theoretical studies and the practical perspectives analyzed. As results, the bibliometric analysis showed the continuous growth of the research area and the most important scientific journals that publish content related to IoT technologies in logistics. In the content analysis, the perspectives are grouped into: (i) conceptual propositions and requirements, (ii) new methods and models to support decision making, (iii) development of technology-based approaches and (iv) empirical studies. As a conclusion, the article presents the description of directions for future researches. Keywords: Logistic
Internet of Things Systematic literature review
1 Introduction Logistics systems involve the management of flows of assets, services, finance and information, focusing on supply chain performance, commonly presenting complex and dynamic characteristics. Technological advances such as Internet of Things (IoT), have increased the availability and volume of data in production and logistics systems (Tao et al. 2018). IoT technologies provide such capabilities extending the interconnection between several devices linked to industrial and logistics processes, allowing the intermittent collection of data, guiding synchronous decision making by considering several aspects of the current state of the system (Qu et al. 2016; Tao et al. 2018). Thoben et al. (2017) argue that one of the main aspects of the fourth industrial revolution is linked to the incorporation of IoT technologies in production and logistics,
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 387–397, 2020. https://doi.org/10.1007/978-3-030-44783-0_37
388
Í. R. S. Agostino et al.
allowing horizontal and vertical integration of various processes. Such approaches based on real-time information can provide high-quality platforms for decision-makers (Heger et al. 2017), and there is a growing need to develop feasibility studies for the implementation of IoT platforms in production and logistics systems (Leusin et al. 2018; Qu et al. 2016). This study presents as an original contribution, a portfolio of updated papers focusing on IoT and its applications in logistics. This paper addresses the theme, providing an overview of previous researches, as well as opportunities and practical theoretical gaps of research to be explored. In this way, a Systematic Review of Literature (RSL) was conducted guided by the following research questions: (i) what is the temporal evolution and main journals that link research involving IoT in logistics? (ii) What are the important themes for research involving IoT in logistics? (iii) What are the research perspectives and what is the current state of development of the area?
2 Methodology To analyze and evaluate the current literature on IoT in Logistic Systems, a systematic literature review (RSL) was conducted (Moher et al. 2016). The search process considered the Web of Science (WoS) and Scopus bases, being recognized as the largest repositories of scientific documents (Guerrero-Bote and Moya-Anegón 2012). Table 1 shows the final versions of the search strings used, as well as the number of results in terms of the number of publications. Only publications of journals and in English were considered. Table 1. Search strings Database WoS Scopus Total
Search string TS = ((“Internet of Things” OR “IoT”) AND (“Logistic*”)) TITLE-ABS-KEY-AUTH((“Internet of Things” OR “IoT”) AND (“Logistic*”))
Result 224 339 563
The search was conducted in April 2019, 563 articles were found, 224 of which were from the Web of Science database and 339 from Scopus. The terms used for the first construct were “Internet of Things” and “IoT” and for the second “logistic*” and the search was applied to the titles, abstracts and keywords of the articles. The research protocol was then built according to the process model presented in Fig. 1.
Perspectives on the Application of Internet of Things in Logistics
389
Fig. 1. Methodological research flow process
The classification taxonomy consists of five macro-stages: (i) identification of the research problem; (ii) the search in the bases of journals; (iii) removal of duplicate articles; (iv) removal of publications outside the time interval between 2014 and 2019 and removal of articles that did not obtain at least 3 citations (in Scopus database). This criteria allowed select only recent and relevant papers; (v) reading the titles and abstracts to identify the alignment with the research questions. The final selection totaled 33 articles published in journals with thematic alignment. The bibliometric and content analysis were performed considering this group. The software R 3.5.2 was used as a computational resource. All analyses were supported by the package “Bibliometrix” 2.1.2 (Aria and Cuccurullo 2017).
3 Results 3.1
Bibliometric Analysis
Figure 2 illustrates the temporal evolution of the publications in the selected portfolio. Considering the temporal cut of the last five years, it is possible to highlight the growth of the area in the last years, with a positive linear trend during this period. This analysis allows highlighting the growing interest in IoT-based applications in logistics considering the selection criteria adopted.
Fig. 2. Publications temporal evolution
Figure 3(a) shows the top ten journals with the highest concentration of publications in the analyzed group. The journal “Industrial Manufacturing & Data Systems” has the highest number of publications. In Fig. 3(b) are presented the top ten journals
390
Í. R. S. Agostino et al.
most frequently cited by the papers of the analyzed group. In this case, it is observed great prominence to the “International Journal of Production Economics” among other journals that link studies of operations management, logistics and technology.
Fig. 3. (a) Journals of analyzed group (b) Most cited journals in the analyzed group
To identify the most recurrent terms it was used a co-occurrence network of the keywords of the papers by multidimensional scaling (Huang et al. 2005) using edge betweenness centrality clustering algorithm (Prell 2012), as shown in Fig. 4. It can be seen that the selected group of publications is aligned with the theme to be investigated. The term “internet of things” appearing as central and “radio frequency identification” (RFID) indicates the direction of research for the use of this technology in conjunction with logistics.
Fig. 4. Keywords co-occurrence network
Figure 5 illustrates a thematic map of the area, relating the density and centrality of the terms most evidenced in the titles of the analyzed publications (for more details see Cobo et al. 2011). The themes located in the upper-right quadrant represent motor themes, being well developed and important areas for the structuring of a search field.
Perspectives on the Application of Internet of Things in Logistics
391
The terms “Internet” and “radio frequency identification” appear in this classification. The themes in the upper-left quadrant represent specialized themes, appearing the term “wireless sensor networks”.
Fig. 5. Thematic map
In the lower-left quadrant are the emerging themes, appearing “architecture”, “management” and “things”. Finally, in the lower-right quadrant are the base themes, which are considered important for the development of the field, being more generic and transversal. The terms “logistics” and “internet of things” are in this category. 3.2
Content Analysis
This section presents as result a portfolio of studies classified by perspectives and the evolution over time. The bibliographic portfolio classification considered four perspectives: (i) theoretical and/or conceptual studies, (ii) proposition of new methods and models for decision-making, (iii) development of applied technological approaches and (iv) empirical studies. Table 2 presents the classified by perspective. Theoretical and/or conceptual studies (i) presented about 15% of the total of portfolio publications. The studies in this classification mostly deal with literature reviews and the proposition of concepts. The highest concentration of studies occurred in the perspective (ii), about 48%, dealing with the proposition of methods, models, and approaches to manage and/or optimize logistics using IoT. The perspective (iii) considered studies that dealt with the analysis and development of technologies involving software and hardware, enabling sensitive and responsive capabilities of IoT-based devices and systems. This classification concentrated about 15% of the publications. Studies dealing with applications with empirical evidence represented about 22% of the analyzed portfolio. Predominantly it was observed studies that relate characteristics of industry 4.0 with logistics, exploring concepts of cyber-physical systems with interconnection along supply chains, focusing on greater control and efficiency, as well as approximation and integration of production and logistics.
392
Í. R. S. Agostino et al. Table 2. Studies classification by perspective
Classification Theoretical or Conceptual Proposing a method or model
Applied technological approach Empirical studies
Author (year) Tu et al. (2018a), Qu et al. (2017), Maslarić et al. (2016), Guo et al. (2015), Zou et al. (2014) Accorsi et al. (2018), Zhang et al. (2018), Lee et al. (2018), Guo et al. (2017), Yan (2017), Zhang et al. (2017), Cho and Kim (2017), Zhong et al. (2016), Kang et al. (2016), Thürer et al. (2016), Yang et al. (2016), Qiu et al. (2015), Kim et al. (2015), Kong et al. (2015), Chen et al. (2014a), He and Chu (2014) Verdouw et al. (2018), Chuang et al. (2017), Li et al. (2015), Chandra and Lee (2014), Chen et al. (2014b) Tu (2018), Hopkins and Hawking (2018), Tu et al. (2018b), Trappey et al. (2017), Qu et al. (2016), Papert et al. (2016), Hu et al. (2016)
The evolution of publications by proposed perspectives is presented in Fig. 6.
Fig. 6. Perspectives for digital twin in cyber-physical systems over time
It is noticed a recent direction for proposing new methods and models together with applied studies, indicating the maturation of the area with a focus on evaluating the theory developed in real applications of IoT technologies in logistics contexts. In the following subsections, each perspective was analyzed to highlight the recent advances in the area, as well as trends observed in the studies from a scientific and practical point of view. Theoretical and Conceptual Studies The theoretical and conceptual aspects of the incorporation of IoT technologies in logistics are discussed by several authors. Zou et al. (2014) discussed concepts related to the application of RFID in food logistics systems and highlighted the technology as promising to enable interconnectivity and systemic intelligence in logistics. Guo and Qu (2015) conceptually analyzed the influence of IoT and cloud computing in the context of “big data” in intelligent logistics systems, the authors discuss three system architectures to integrate these tools. Maslarić et al. (2016) present an analysis of the impact of Industry 4.0 on logistics, pointing out the concept of “Physical Internet” as a response to changes in the industry supported by IoT technologies. Qu et al. (2017)
Perspectives on the Application of Internet of Things in Logistics
393
analyzed the application of systems dynamics models in the context of IoT adoption in logistics and production systems, providing an implementation and analysis roadmap for integration of industrial and transportation information. Tu et al. (2018a) propose a theoretical framework for IoT implementation in logistics production and supply chains. In a general way, conceptually the area presents maturation regarding the adoption of IoT in logistics, the authors in this perspective present convergent directions from the theoretical point of view with recommendations focused on practice. New Methods and Models for Decision Making The perspective “proposition of new methods and models for decision making” considering IoT in the logistics context presented a great variety of approaches considering several aspects of logistics and their interface with production. In the studies of Guo et al. (2017) and Zhang et al. (2018) the integration production and logistics from IoT technologies is addressed. In this same direction Chen et al. (2014a), Kang et al. (2016), Kim et al. (2015), Qiu et al. (2015) and Yang et al. (2016) discuss the proposition of approaches considering the integration of logistics services with the use of IoT technologies, highlighting as a focus the capacity for real-time monitoring of operations. Kong et al. (2015) and Lee et al. (2018) developed approaches focused on the management of warehouses and distribution systems. In both studies it is proposed the application of IoT for real-time control with adaptive capacity. Zhong et al. (2016), Cho and Kim (2017) and Yan (2017) address the incorporation of RFID devices, proposing this technology in a real-time management model. Another context observed in the studies dealt with the proposition of approaches aimed at perishable food products. Accorsi et al. (2018), Yan (2017) and Zhang et al. (2017) presented proposals for IoT incorporation in this context. Applications in specific contexts were proposed by He and Chu (2014) who addressed the use of IoT for information management focused on maritime logistics. As well as Thürer et al. (2016) proposed an approach for reverse logistics of solid waste considering the incorporation of IoT. In a general way, in this perspective, the studies were based on simulation methods to evidence the proposed methods and models, in some cases using real data to demonstrate performance gains, but not showing a direct and empirical application. Development of Applied Technological Approaches In this perspective, studies focused on technological development, addressing both computational development (e.g. software and systems) and device development. The studies of Li et al. (2015) proposed real-time tracking systems, specifying sensing devices and proposing computational tools. In the same direction Chuang et al. (2017) developed a study for real-time monitoring using wireless and Bluetooth technologies, the authors perform a prototyping of the proposed approach demonstrating the applicability. Chandra and Lee (2014) developed a data transmission architecture for monitoring temperature and humidity in food transport.
394
Í. R. S. Agostino et al.
The study developed by Verdouw et al. (2018) also focused on food logistics, developing an architecture for information management in the supply chain supported by IoT technologies. In a more generic context, Chen et al. (2014) developed a system for cloud computing using RFID devices to control and monitor the supply chain. In this perspective, it was possible to evidence a maturation of the research area, in which it was identified the transition of the theoretical concepts and proposition of models towards the technological development of the area. Empirical Studies Few application studies are reported in the literature in the analyzed bibliographic portfolio. Tu (2018) conducted a survey type survey in Taiwan considering the adoption of IoT in the context of production and logistics, the author used a system of structural equations to analyze the information raised indicating that factors such as cost, external pressures and the perception of benefits influence the adoption of IoT. In this direction Hu et al. (2016) used exploratory factor analysis to analyze in China the perception of customers, concluding that IoT technologies will improve the companycustomer relationship. Case studies involving IoT and logistics were developed by Hopkins and Hawking (2018) that focused on the influence of IoT adoption on safety and operational costs in truck transportation. Tu et al. (2018b) conducted an experimental study to implement IoT technology in a complex industrial system, involving transportation and logistics, finding results aimed at greater flexibility and control. Qu et al. (2016) conducted a case study focusing on production and logistics synchronization demonstrating effectiveness related to IoT adoption in this context. Papert et al. (2016) analyzed the pharmaceutical sector considering multiple case studies in Germany, indicating the adoption of IoT as a solution for the management of complex logistics chains. The study by Trappey et al. (2017) deals with the evaluation of the development and adoption of IoT technologies through the analysis of patent registration.
4 Conclusion This paper presented an exploratory systematic review covering the application of IoT technology in logistics. The perspectives identified allow evidencing the maturity of IoT adoption in logistics with a recent direction for proposing new methods and models and applied studies with a focus on evaluating the theory developed in real applications. Some limitations in this research were noted, as follows: (1) only two databases were consulted, (2) the inclusion criteria were limited to the English language, only publications of journals during the time interval between 2014 and 2019 and with at least 3 citations (3) studies applied for specific areas were excluded. Despite of this, the systematic literature review follows a proper methodology to answer the addressed research questions and therefore the study’s purpose was achieved. Obtained results serve as a general overview of the state-of-the-art in the application of IoT technologies in logistics.
Perspectives on the Application of Internet of Things in Logistics
395
As future research directions, three main directions stand out: (a) Evidence of a generic model for IoT adoption in logistics: the various approaches found present convergence in most aspects. However, it was observed a need for evidencing the alternatives proposed by the authors, to consolidate and integrate technical and operational aspects related to IoT in the logistics context; (b) Integration production and logistics: Recurring applications of IoT in the sense of integration production and logistics can provide high quality platforms for joint decisions focused on the comprehensive control of operations in the industrial context; (c) Application of real-time data monitoring effective control of material flows and information through real-time data collection with a focus on transportation and handling. Acknowledgements. This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - under reference number 99999.006033/2015-06, in the scope of the BRAGECRIM program.
References Accorsi, R., Cholette, S., Manzini, R., Tufano, A.: A hierarchical data architecture for sustainable food supply chain management and planning. J. Cleaner Prod. 203, 1039–1054 (2018) Aria, M., Cuccurullo, C.: Bibliometrix: an R-tool for comprehensive science mapping analysis. J. Informetrics 11, 959–975 (2017) Chandra, A., Lee, R.: A method of WSN and sensor cloud system to monitor cold chain logistics as part of the IoT technology. Int. J. Multimed. Ubiquit. Eng. 9, 145–152 (2014) Chen, L., Chen, Y., Hsu, C.: A new approach to integrate Internet-of-Things and software-as-aservice model for logistic systems: a case study. Sensors 14, 6144–6164 (2014a) Chen, Y., Zhao, S., Zhai, Y.: Construction of intelligent logistics system by RFID of internet of things based on cloud computing. J. Chem. Pharm. Res. 6, 1676–1679 (2014b) Cho, S., Kim, J.: Smart logistics model on Internet of Things environment. Adv. Sci. Lett. 23, 1599–1602 (2017) Chuang, H., Lee, H., Chang, J., Weng, C., Shaikh, O., Huang, L.: Real-time monitoring via patch-type piezoelectric force sensors for Internet of Things based logistics. IEEE Sens. J. 17, 2498–2506 (2017) Cobo, M.J., López-Herrera, A.G., Herrera-Viedma, E., Herrera, F.: An approach for detecting, quantifying, and visualizing the evolution of a research field: a practical application to the fuzzy sets theory field. J. Informetr. 5, 146–166 (2011) Guerrero-Bote, P., Moya-Anegón, F.: A further step forward in measuring journals’ scientific prestige: the SJR2 indicator. J. Inf. 6, 674–688 (2012) Guo, Y., Qu, J.: Study on intelligent logistics management information system based on IOT and cloud computation in big data era. Open Cybern. Syst. J. 9, 934–941 (2015) Guo, Z., Zhang, Y., Zhao, X., Song, X.: A timed colored petri net simulation-based self-adaptive collaboration method for production-logistics systems. Appl. Sci. 7, 235 (2017) He, W., Chu, X.: A novel navigation information management system for food maritime logistics based on Internet of Things. Adv. J. Food Sci. Technol. 6, 280–283 (2014) Heger, J., Grundstein, S., Freitag, M.: Online-scheduling using past and real-time data. An assessment by discrete event simulation using exponential smoothing. CIRP J. Manuf. Sci. Technol. 19, 158–163 (2017)
396
Í. R. S. Agostino et al.
Hopkins, J., Hawking, P.: Big Data Analytics and IoT in logistics: a case study. Int. J. Logistics Manag. 29, 575–591 (2018) Hu, M., Huang, F., Hou, H., Chen, Y., Bulysheva, L.: Customized logistics service and online shoppers’ satisfaction: an empirical study. Internet Res. 26, 484–497 (2016) Huang, J., Tzeng, G., Ong, C.: Multidimensional data in multidimensional scaling using the analytic network process. Pattern Recogn. Lett. 26, 755–767 (2005) Kang, Y.-S., Park, I.-H., Youm, S.: Performance prediction of a MongoDB-based traceability system in smart factory supply chains. Sensors 16, 2126 (2016) Kim, J.-S., Lee, H.-J., Oh, R.-D.: Smart Integrated Multiple Tracking System development for IOT based Target-oriented Logistics Location and Resource Service. Int. J. Smart Home 9, 195–204 (2015) Kong, X.T., Fang, J., Luo, H., Huang, G.Q.: Cloud-enabled real-time platform for adaptive planning and control in auction logistics center. Comput. Ind. Eng. 84, 79–90 (2015) Lee, C.K., Lv, Y., Ng, K.K., Ho, W., Choy, K.L.: Design and application of Internet of Thingsbased warehouse management system for smart logistics. Int. J. Prod. Res. 56, 2753–2768 (2018) Leusin, M., Frazzon, E., Maldonado, M.U., Kück, M., Freitag, M.: Solving the job-shop scheduling problem in the industry 4.0 era. Technologies 6, 107 (2018) Li, Y.N., Peng, Y.L., Zhang, L., Wei, J.F., Li, D.: Quality monitoring traceability platform of agriculture products cold chain logistics based on the Internet of Things. Chem. Eng. Trans. 46, 517–522 (2015) Maslarić, M., Nikoličić, S., Mirčetić, D.: Logistics response to the industry 4.0: the physical internet. Open Eng. 6, 511–517 (2016) Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G., PRIMA Group.: Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. Int. J. Surg. 8, 336–341 (2016) Papert, M., Rimpler, P., Pflaum, A.: Enhancing supply chain visibility in a pharmaceutical supply chain. Int. J. Phys. Distrib. Logistics Manag. 46, 859–884 (2016) Prell, C.: Social Network Analysis: History. Theory and Methodology. Sage, Thousand Oaks (2012) Qiu, X., Luo, H., Xu, G., Zhong, R., Huang, G.Q.: Physical assets and service sharing for IoTenabled Supply Hub in Industrial Park (SHIP). Int. J. Prod. Econ. 159, 4–15 (2015) Qu, T., Lei, S.P., Wang, Z.Z., Nie, D.X., Chen, X., Huang, G.Q.: IoT-based real-time production logistics synchronization system under smart cloud manufacturing. Int. J. Adv. Manuf. Technol. 84, 147–164 (2016) Qu, T., Thürer, M., Wang, J., Wang, Z., Fu, H., Li, C., Huang, G.Q.: System dynamics analysis for an Internet-of-Things-enabled production logistics system. Int. J. Prod. Res. 55, 2622– 2649 (2017) Tao, F., Qi, Q., Liu, A., Kusiak, A.: Data-driven smart manufacturing. J. Manuf. Syst. 48, 157– 169 (2018) Thoben, K.-D., Wiesner, S., Wuest, T.: “Industrie 4.0” and smart manufacturing-a review of research issues and application examples. Int. J. Autom. Technol. 11, 4–16 (2017) Thürer, M., Pan, Y.H., Qu, T., Luo, H., Li, C.D., Huang, G.Q.: Internet of Things (IoT) driven Kanban system for reverse logistics: solid waste collection. J. Intell. Manuf. 30, 2621–2630 (2016) Trappey, A.J., Trappey, C.V., Fan, C.-Y., Hsu, A.P., Li, X.-K., Lee, I.J.: IoT patent roadmap for smart logistic service provision in the context of Industry 4.0. J. Chin. Inst. Eng. 40, 593–602 (2017) Tu, M., Lim, M.K., Yang, M.-F.: IoT-based production logistics and supply chain system – Part 1. Ind. Manag. Data Syst. 118, 65–95 (2018a)
Perspectives on the Application of Internet of Things in Logistics
397
Tu, M., Lim, M.K., Yang, M.-F.: IoT-based production logistics and supply chain system – part 2. Ind. Manag. Data Syst. 118, 96–125 (2018b) Tu, M.: An exploratory study of Internet of Things (IoT) adoption intention in logistics and supply chain management. Int. J. Logistics Manag. 29, 131–151 (2018) Verdouw, C.N., Robbemond, R.M., Verwaart, T., Wolfert, J., Beulens, A.J.: A reference architecture for IoT-based logistic information systems in agri-food supply chains. Enterp. Inf. Syst. 12, 755–779 (2018) Yan, R.: Optimization approach for increasing revenue of perishable product supply chain with the Internet of Things. Ind. Manag. Data Syst. 117, 729–741 (2017) Yang, R., Li, B., Hu, Y.: An experimental study for intelligent logistics: a middleware approach. Chin. J. Electron. 25, 561–569 (2016) Zhang, Y., Guo, Z., Lv, J., Liu, Y.: A framework for smart production-logistics systems based on CPS and industrial IoT. IEEE Trans. Ind. Inf. 14, 4019–4032 (2018) Zhang, Y., Zhao, L., Qian, C.: Modeling of an IoT-enabled supply chain for perishable food with two-echelon supply hubs. Ind. Manag. Data Syst. 117, 1890–1905 (2017) Zhong, R.Y., Lan, S., Xu, C., Dai, Q., Huang, G.Q.: Visualization of RFID-enabled shopfloor logistics Big Data in Cloud Manufacturing. Int. J. Adv. Manuf. Technol. 84, 5–16 (2016) Zou, Z., Chen, Q., Uysal, I., Zheng, L.: Radio frequency identification enabled wireless sensing for intelligent food logistics. Phil. Trans. Roy. Soc. A: Math. Phys. Eng. Sci. 372, 20130313 (2014)
AIDA 4.0: Architectures of Industry 4.0 Demonstrated Through Application Scenarios in Business Game Julia Feldt (geb. Wagner)1,2(&) and Henning Kontny1 1
Fakultät Wirtschaft und Soziales, University of Applied Sciences in Hamburg, Alexanderstraße 1, 20099 Hamburg, Germany {j.wagner,henning.kontny}@haw-hamburg.de 2 University of the West of Scotland, Technology Avenue, Glasgow G72 0LH, UK
Abstract. Industry 4.0 is a very promising trend which could allow companies to achieve new levels of efficiency such as highly automated and adaptive supply chains, innovative products and even expansion to new markets. At the same time, due to numerous available scenarios of digitalization within a supply chain, it is very challenging for practitioners to choose one of them and to evaluate the adopted strategy in comparison to other possibilities. In order to provide support for practitioners and find an application-oriented way to teach Industry 4.0, a business game AIDA 4.0 was modeled. It is based on a typical process structure of a producing company, whereas proposed digitalization architectures reflect the current state of research. The AIDA 4.0 allow one group of participants to develop a scenario for their company based on one of the proposed architectures then compare their results with another group, who followed a different strategy. Keywords: Business game
Digitalization Industry 4.0
1 Introduction In last year’s companies are claiming about increasing global competition [1] at the same time looking for ways to optimize their supply chains [2, 3] in order to meet growing customer expectations. Scientists also underline the importance of supply chain for the commercial success of a company [4, 5] focusing on the necessity to adapt to changing demands to use the chances of the potential growth [6]. Additionally to the “traditional” key performance indicators such as costs and level of service, new indicators such as the ability to react to the unforeseen events as well as adaptability become increasingly important [7, 8]. Flexibility and robustness can only be achieved by autonomous highly-digitalized systems, which gain information in real-time modus and are able to take own decisions along with ability to communicate with each other [9]. Although solutions enabling the digitalization of supply chain processes, also often referred to as Industry 4.0 or Internet of Things, are highly in trend and got substantial interest from management, many companies are struggling to decide which solution © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 398–408, 2020. https://doi.org/10.1007/978-3-030-44783-0_38
AIDA 4.0: Architectures of Industry 4.0
399
will bring them highest advantages and which digitalization strategy should they choose [10–12]. At the same time researchers, Kotzab [13] and Zijm [14] emphasize that digitalization should be accompanied with the qualification of employees operating in logistics on all levels in order to adapt them to new technologies. For the above reasons digitalization of supply chains is a highly topical subject for practitioners and for scientists. Therefore, this paper pursues the aim to provide the tool which would allow managers and students better understand different scenarios of digitalization along the supply chain at a cross-functional level using the method of the business game.
2 Theoretical Background 2.1
Impact of Business Games on Learning
Practitioners are acknowledging the benefits of business games for a very long time and have successfully implemented a wide variety of games for stuff education in logistics, some examples of which are provided below: • “The Lean Leap Logistics Game” [15], which aims to encourage collaboration as well as demonstrating the “Bullwhip Effect”. • “Agents playing the Beer Distribution Game” [15]. A version of Beer Distribution Game in which results of human players are compared with the results of decisions made by a program and in the end providing proof for the efficiency of autonomous IT decisions. • “Stackelberg-game-based modeling” from Yue and You [16] demonstrates a decentralized perspective on optimization of supply chains, presented based on a case study on forestry and biofuel supply chain. • “ArKoH- Harbor Game” [17], describes an online game in which different scenarios of harbor development such as a Smart Port or a Stagnation are played and compared with each other from a perspective of different roles (i.e. harbor worker). • Mobile Learning App “MARTINA” [18], which represents a mobile application for the creation of qualification games for logistics personnel in different areas, i.e. Cargo Securing Game. The main target of the app is the development of universally applicable gamification app. Serious (Business) games are mostly implemented for adult education in order to increase engagement while teaching business-relevant subjects. The main reasons for the serious games are: safety (i.e. simulating dangerous situations in a virtual environment), cost-effectiveness (compared to real-world experience) as well as longerlasting learning effect [19]. Chin, Dukes, and Gamson [20] have performed an analysis of 40 years in simulation and gaming and came to the conclusion that games can be effectively used for learning of serious topics and are often preferred by students over other teaching techniques. For above reasons the business game as a teaching method was chosen to discuss and teach such a complicated subject as digitalization of supply chain processes.
400
2.2
J. Feldt (geb. Wagner) and H. Kontny
Struggles to Find an Optimal Supply Chain Strategy in Times of Digitalization
The McKinsey Survey of over 300 industrial companies from the USA, Japan and Germany [21] revealed that most of them struggle to define a convertible strategy for the Industry 4.0 (here will be often referred to as digitalization), despite their fear for being completely displaced from market as it happened to Motorola and Nokia after development of smartphones or to Kodak in times of digital photography [22]. Digitalization of the processes within the supply chain will lead to the creation of autonomous elements which are connected with each other and are known in current literature as Cyber-Physical Systems (CPS), whereas the operation system with CPS is often referred to as Cyber-Physical Production System (CPPS) [23, 24]. According to this approach, any area, or a process or even a product of an enterprise can become “smart”, creating a nearly unlimited number of scenarios for possible development and making any strategic decision very difficult. For that reason presented research is concentrating on only three possible architectures (s. Sect. 4). This would allow the participants of the business game AIDA 4.0 to develop an understanding of how digitalization of an enterprise area can influence the information flow without losing their focus. 2.3
Related Studies and Research Question
The subject of digitalization or Industry 4.0 is meanwhile strongly represented in diverse studies as well as in literature. Studies, which are focused on supply chain and even describe the development trends often ignore the system changes, which is caused by digitalization [25–27]. Other studies recognize the need of cross-linked and adaptive supply chain management systems, nevertheless, they do not deliver a comparison to other development scenarios [5, 7, 28, 29]. To close the above gap, we started the development of possible scenarios for the supply chain digitalization on the example of a bike producing company, which led us to the following research questions: RQ #1: Which scenarios of process and system integration along the supply chain will be possible within the next years? RQ #2: How each scenario will influence the information flow and which impact will it have on the supply chain? Three architectures with suitable information flow/scheduling approaches for a nonexisting company will be described in Sect. 4, thus providing the answer to the first research question. Whereas the presented Figs. 1, 2 and 3 visualize the impact of each scenario on the supply chain, hence answering the second question.
AIDA 4.0: Architectures of Industry 4.0
401
3 A Fictional Company “Bike Manufacturer” and AIDA 4.0 Game To understand potential benefits from the digitalization of processes along the supply chain, it is helpful to start with an understanding of high-level processes of a fictional company “Bike Manufacturer”. Although the company is not real, the processes are build according to the information which was gained from projects, meetings and workshops with diverse companies within initiatives, which support technological development in the industry such as “Mittelstandszentrum 4.0” and “Logistics initiative” in last three years. “Bike Manufacturer” orders raw materials and spare parts from diverse suppliers, which enter the warehouse and after quality control in goods receipt area will be stored in a central warehouse. From the central warehouse, raw materials can be moved toward production in order to manufacture components for bicycles. Afterward bought spare parts and manufactured components will be assembled in one of the three assembly areas into finished goods which are diverse types of bicycles. As any other producing company, the most important target of “Bike Manufacturer” is to fulfill customer orders while gaining profit from sales. The “Bike Manufacturer” is a mid-size company with fragmented IT systems. According to O’Leary, around 90% of IT systems are not fully automated [30], forcing companies to use Excel sheets and Access databanks to close the gaps [31]. Within the AIDA 4.0 game each team will represent one “Bike Manufacturer” and play against other teams 1 to 10 rounds (or less, depending on available time) with changing conditions in each round (i.e. broken machine or sick assembly workers) which should trigger new round of planning processes. In order to steer their company, players will be able to take decisions such as: investments in IT systems (all start as a company with conventional centralized supply chain as described in Sect. 4.1), planning decisions for purchasing, production and assembly (if the area, is not fully automated) etc. The results will be measured based on key performance indicators, such as level of inventory, lost sales, profitability and others. The main target of a game is to provide the players with the insights on differences between the possible levels of digitalization and demonstrate how they can change the operations as well as overall performance of a company.
4 Strategies for Digitalization and Operations Scheduling Instead of focusing on very detailed possible scenarios of digitalization such as 3D printing and plentiful other existing trends, the high-level logic, as proposed by ScholzReiter and Sowade [32] was chosen as most appropriate. Both researchers argued, that although self-steering and automation can be implemented in different ways, it can be divided into three types:
402
J. Feldt (geb. Wagner) and H. Kontny
• Conventional production system (as already existing at most companies, described in Sect. 4.1) • Partially-digitalized architecture (for more information s. Sect. 4.2) • Fully digitalized architecture (as described in Sect. 4.3) They provide a comprehensible and well-structured way to compare highlyautomated supply chain with the supply chain with some automated smart object as well as analyze strengths and weaknesses from both scenarios in comparison to the current situation (first scenario). Additionally to the architecture types three types of operations scheduling were chosen to describe the information flow in each scenario [33, 34]: • Predictive scheduling (which fits a classic production system, since it tries to predict events (i.e. customer order) instead of reacting to them) • Reactive scheduling (correspond to fully digitalized architecture, since the smart objects can react to events in real-time modus) • Hybrid dynamic scheduling (match the logic of partially-digitalized architecture). Both approaches, architectures and scheduling techniques allow evaluating the impact of digitalization from the intercompany perspective, in contrary to narrow consideration of the impact on one area ignoring others. 4.1
Conventional Centralized Supply Chain with Predictive Scheduling
At the start of the day X the ERP system will show the customer orders till previous day, X-1 (s. Fig. 1). The purchasing department will run an availability check based on customer orders as well as planned demand for the next weeks and place purchasing orders by suppliers in order to get the needed spare parts and raw materials on time. Production will start the fulfillment of the respective production volumes per machine, which will be calculated in PPS (Production Planning System) based on demand planning data for the next one to six weeks. Assembly will start the fulfillment of the assembly orders based on the same logic as production. In case of disturbance, human input will be required for any decision-taking steps. Produced/assembled orders will be reported at the end of the day and available the next day after the goods are shifted to a central warehouse. Information on material movements is available in near-real-time modus (excluding time needed for physical movement of goods). All entities do not take any autonomous decisions and concentrate on order fulfillment, ERP and PPS can take standardized decisions but are unable to react to disturbances in real-time.
AIDA 4.0: Architectures of Industry 4.0
403
Fig. 1. Conventional centralized approach to supply chain
4.2
The Architecture of Partially Digitalized Supply Chain
Due to some level of automation of information flow, customer orders are shown in the system in real-time modus. Still, purchasing orders will not be places in real-time modus, since they are usually placed for a longer period of time. In-House Production, Assembly and Logistics are enabled to take decisions within their area for smaller entities (such as Machine or Assembly Cells), which significantly adds flexibility to the process, since they can share information, “communicate”, with each other helping to resolve arising issues in near-real-time modus. For example, the Production entity can automatically generate a transfer order for spare parts needed for the production. Still, the data such as the capacity of a machine or assembly cell are calculated based on a plan data instead of real data. In the case of deviances from the plan such as note of illness of assembly worker (unexpected event), the solution will take time and could require human support for the decision-making process. Feedback on produced/assembled goods cannot be carried out in real-time since the respective entities (machine or assembly cell) are not digitalized. This way the overall process still has some “black boxes”.
404
J. Feldt (geb. Wagner) and H. Kontny
Fig. 2. Partially digitalized supply chain
4.3
The Architecture of Fully Digitalized Supply Chain – Reactive Scheduling
All information flows (except that from suppliers which cannot be influenced by internal IT solution) match the material movements in near real-time modus. All entities which are responsible for operations are enabled to take respective decisions and communicate results or issues to other relevant entities. The overall process is selfadaptive, fast and transparent.
AIDA 4.0: Architectures of Industry 4.0
405
Fig. 3. Fully digitalized reactive supply chain
5 Conclusions Presented architectures can be used to support practitioners by providing them with a structured way to evaluate different scenarios of company development. Additionally, they can be used for projects on digitalization for mapping of the current status of the IT systems and information flows as well as to visualize and analyze current weaknesses. Moreover, fully digitalized or partially digitalized scenarios of AIDA 4.0 can be used at companies for workshops and discussions about a preferred digitalization strategy, allowing to analyze weaknesses and strengths of both scenarios in comparison to the status quo. It can help the employees to develop a deeper understanding of their own processes and information flows. Besides its educational facet, the AIDA 4.0 can be used as a facilitation tool at team-building events to promote collaboration between different departments. For students and researchers, AIDA 4.0 provides a gamified tool for learning on current topics of Industry 4.0 with a strong link to the real-world supply chains. The game demonstrates the effects of different digitalization strategies on the whole supply chain, thus teaching students from the start how to assess managerial decisions from the intercompany perspective, instead of focusing themselves on one department.
406
J. Feldt (geb. Wagner) and H. Kontny
Next step in game development is to complete a digital solution (first steps are already modeled in OTD-NET [35]), which would allow teams to play 10 rounds and help to quantify the described effects of different architectures on company KPIs such as level of inventory vs. lost sales. This paper should serve as an inspiration for other researchers to create solutions for real-world issues using gamification techniques and making the content of the research as applicable for the practitioners as possible.
References 1. Bauernhansl, T., ten Hompel, M., Vogel-Heuser, B. (eds.): Industrie 4.0 in Produktion. Automatisierung und Logistik. Springer, Wiesbaden (2014) 2. Ollesch, J., Hesenius, M., Gruhn, V., Alias, C.: Real-time event processing for smart logistics networks. In: Proff, H., Fojcik, T.M. (eds.) Mobilität und digitale Transformation, pp. 517–532. Springer, Wiesbaden (2018). https://doi.org/10.1007/978-3-658-20779-3_32 3. Ponte, B., Sierra, E., de la Fuente, D., Lozano, J.: Exploring the interaction of inventory policies across the supply chain: an agent-based approach. Comput. Oper. Res. 78, 335–348 (2017). https://doi.org/10.1016/j.cor.2016.09.020 4. Wellbrock, W.: Innovative Supply-Chain-Management-Konzepte. Springer, Wiesbaden (2015). https://doi.org/10.1007/978-3-658-09181-1 5. Yuvaraj, S., Sangeetha, M.: Smart supply chain management using Internet of Things (IoT) and low power wireless communication systems. In: International Conference on Wireless Communications, Signal Processing and Networking (WiSPNET), pp. 555–558. IEEE (2016) 6. Bogaschewsky, R., Müller, H.: Industrie 4.0: Wie verändern sich die IT-Systeme im Einkauf und SCM? 7. Monostori, J.: Supply chains robustness: challenges and opportunities. Procedia CIRP 67, 110–115 (2018). https://doi.org/10.1016/j.procir.2017.12.185 8. Fisel, J., Duffie, N., Moser, E., Lanza, G.: Changeability - a frequency perspective. Procedia CIRP 79, 186–191 (2019). https://doi.org/10.1016/j.procir.2019.02.043 9. Wagner, J., Kontny, H.: Use case of self-organizing adaptive supply chain. Epubli (2017). https://doi.org/10.15480/882.1471 10. Tjahjono, B., Esplugues, C., Ares, E., Pelaez, G.: What does industry 4.0 mean to supply chain? Procedia Manuf. 13, 1175–1182 (2017). https://doi.org/10.1016/j.promfg.2017.09. 191 11. Kolberg, D., Berger, C., Pirvu, B.-C., Franke, M., Michniewicz, J.: CyProF – insights from a framework for designing cyber-physical systems in production environments. Procedia CIRP 57, 32–37 (2016). https://doi.org/10.1016/j.procir.2016.11.007 12. Jede, A., Teuteberg, F.: Towards cloud-based supply chain processes: designing a reference model and elements of a research agenda. Int. J. Logist. Manag. 27, 438–462 (2016). https:// doi.org/10.1108/IJLM-09-2014-0139 13. Kotzab, H., Teller, C., Bourlakis, M., Wünsche, S.: Key competences of logistics and SCM professionals – the lifelong learning perspective. Supply Chain Manag. Int. J. 23, 50–64 (2018). https://doi.org/10.1108/SCM-02-2017-0079 14. Zijm, H., Klumpp, M.: Future logistics: what to expect, how to adapt. In: Freitag, M., Kotzab, H., Pannek, J. (eds.) Dynamics in Logistics, pp. 365–379. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-45117-6_32
AIDA 4.0: Architectures of Industry 4.0
407
15. Holweg, M., Bicheno, J.: Supply chain simulation–a tool for education, enhancement and endeavour. Int. J. Prod. Econ. 78, 163–175 (2002). https://doi.org/10.1016/S0925-5273(00) 00171-7 16. Yue, D., You, F.: Stackelberg-game-based modeling and optimization for supply chain design and operations: a mixed integer bilevel programming framework. Comput. Chem. Eng. 102, 81–95 (2017). https://doi.org/10.1016/j.compchemeng.2016.07.026 17. Duin, H., Gorldt, C., Thoben, K.-D.: Serious Gaming als Instrument zur Kompetenzentwicklung für Hafenfachkräfte. In: Bullinger-Hoffmann, A.C. (ed.) Zukunftstechnologien und Kompetenzbedarfe, pp. 145–162. Springer, Heidelberg (2019). https://doi.org/10.1007/ 978-3-662-54952-0_9 18. Klumpp, M., Neukirchen, T., Koop, W.: Mobile learning and human-artificial cooperation in logistics. In: Proff, H. (ed.) Mobilität in Zeiten der Veränderung, pp. 371–382. Springer, Wiesbaden (2019). https://doi.org/10.1007/978-3-658-26107-8_28 19. Bellotti, F., Kapralos, B., Lee, K., Moreno-Ger, P., Berta, R.: Assessment in and of serious games: an overview. Adv. Hum.-Comput. Interact. 2013, 1–11 (2013). https://doi.org/10. 1155/2013/136864 20. Chin, J., Dukes, R., Gamson, W.: Assessment in simulation and gaming: a review of the last 40 years. Simul. Gaming 40, 553–568 (2009). https://doi.org/10.1177/1046878109332955 21. McKinsey Digital: Industry 4.0 after the initial hype. Where manufacturers are finding value and how they can best capture it (2016) 22. Roth, A. (ed.): Einführung und Umsetzung von Industrie 4.0. Springer, Heidelberg (2016) 23. Monostori, L., Kádár, B., Bauernhansl, T., Kondoh, S., Kumara, S., Reinhart, G., Sauer, O., Schuh, G., Sihn, W., Ueda, K.: Cyber-physical systems in manufacturing. CIRP Ann. 65, 621–641 (2016). https://doi.org/10.1016/j.cirp.2016.06.005 24. Thoben, K.-D., Wiesner, S., Wuest, T.: “Industrie 4.0” and smart manufacturing – a review of research issues and application examples. Int. J. Autom. Technol. 11, 4–16 (2017). https:// doi.org/10.20965/ijat.2017.p0004 25. Verdouw, C.N., Wolfert, J., Beulens, A.J.M., Rialland, A.: Virtualization of food supply chains with the Internet of Things. J. Food Eng. 176, 128–136 (2016). https://doi.org/10. 1016/j.jfoodeng.2015.11.009 26. Rezapour, S., Farahani, R.Z., Pourakbar, M.: Resilient supply chain network design under competition: a case study. Eur. J. Oper. Res. 259, 1017–1035 (2017). https://doi.org/10. 1016/j.ejor.2016.11.041 27. Mortazavi, A., Arshadi Khamseh, A., Azimi, P.: Designing of an intelligent self-adaptive model for supply chain ordering management system. Eng. Appl. Artif. Intell. 37, 207–220 (2015). https://doi.org/10.1016/j.engappai.2014.09.004 28. Biedermann, L., Kotzab, H.: Erfolgsfaktoren zur zukünftigen Gestaltung resilienter Supply Chains – Konzeption eines Bezugsrahmens. In: Bode, C., Bogaschewsky, R., Eßig, M., Lasch, R., Stölzle, W. (eds.) Supply Management Research, pp. 235–254. Springer, Wiesbaden (2019). https://doi.org/10.1007/978-3-658-23818-6_11 29. Basnet, C., Seuring, S.: Demand-oriented supply chain strategies – a review of the literature, p. 17 (2016) 30. O’Leary, D.E.: Supporting decisions in real-time enterprises: autonomic supply chain systems. Inf. Syst. E-Bus. Manag. 6, 239–255 (2008). https://doi.org/10.1007/s10257-0080086-0 31. Rentrop, C., Zimmermann, S.: Shadow IT: management and control of unofficial IT. In: ICDS, pp. 98–102 (2012) 32. Scholz-Reiter, B., Sowade, S.: Der Beitrag der Selbststeuerung zur Wandlungsfähigkeit von Produktionssystemen (2010). www.sfb637.uni-bremen.de/pubdb/repository/SFB637-B2-10005-IIC.pdf
408
J. Feldt (geb. Wagner) and H. Kontny
33. Foerster, F., Mueller, D., Scholz, D., Michalik, A., Kiebler, L.: Integration of condition based maintenance orders into the decision-making of autonomous control methods. Procedia CIRP 81, 216–221 (2019). https://doi.org/10.1016/j.procir.2019.03.038 34. Ouelhadj, D., Petrovic, S.: A survey of dynamic scheduling in manufacturing systems. J. Sched. 12, 417–431 (2009). https://doi.org/10.1007/s10951-008-0090-8 35. Liebler, K., Beissert, U., Motta, M., Wagenitz, A.: Introduction OTD-NET and LAS: orderto-delivery network simulation and decision support systems in complex production and logistics networks. In: 2013 Winter Simulations Conference (WSC), pp. 439–451. IEEE, Washington, DC (2013). https://doi.org/10.1109/WSC.2013.6721440
Machine Learning in Production Scheduling: An Overview of the Academic Literature Satie L. Takeda-Berger1(&) , Enzo Morosini Frazzon1 Eike Broda2 , and Michael Freitag3
2
,
1 Industrial and Systems Engineering Department, Federal University of Santa Catarina, Florianópolis, Brazil [email protected], [email protected] Faculty of Production Engineering, University of Bremen, Bremen, Germany [email protected] 3 BIBA - Bremer Institut für Produktion und Logistik GmbH at the University of Bremen, Bremen, Germany [email protected]
Abstract. Production scheduling is an important tool for a manufacturing system, where it can have a significant impact on the productivity of a production process. In this sense, the application of machine learning can be very fruitful in this field, since it is an enabling computer programs to automatically make intelligent decisions based on data to improve performance at the manufacturing system. Therefore, this paper aims to explore the use of machine learning in production scheduling under the Industry 4.0 context. A systematic literature review was conducted to identify the main machine learning techniques currently employed to improve production scheduling. As a result, bibliometric analysis evidenced the continuous growth of this research area and identified the main machine learning techniques applied. Finally, the gaps leading to further research are highlighted. Keywords: Machine learning Production scheduling Manufacturing system
1 Introduction In the modern world, manufacturing companies are increasingly gaining access to vast amounts of data from several sources (Flath and Stein 2018). Extracting knowledge from this data holds great potential to improve and add value to production processes (Mulrennan et al. 2018). This provides the foundation for the fourth industrial revolution, better known as “Industry 4.0”. In the production context, Industry 4.0 is defined as the intelligent flow of the workpieces machine-by-machine in a factory, based on real-time communication between machines (Leyh et al. 2017). In this sense, Peruzzini et al. (2017) commented that Industry 4.0 intends to make manufacturing intelligent and adaptive using flexible and collaborative systems to solve problems and make the best decisions. Rüßmann et al. (2015) describe nine groups of technologies to enable the realization of Industry 4.0, which are: (i) Big Data and Analytics; (ii) Autonomous Robots; (iii) Simulation; © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 409–419, 2020. https://doi.org/10.1007/978-3-030-44783-0_39
410
S. L. Takeda-Berger et al.
(iv) Horizontal and Vertical System Integration; (v) Industrial Internet of Things; (vi) Cybersecurity; (vii) Cloud; (viii) Additive Manufacturing and (ix) Augmented Reality. However, this paper focuses on Big Data Analytics (BDA), and more specifically on Machine Learning (ML) techniques applied in production scheduling. Production scheduling aims to accomplish the optimal sequence of tasks, optimally allocating limited resources to processing tasks over time (Li and Ierapetritou 2008). Several different approaches to address the problems of production scheduling have been found in the literature (Li and Ierapetritou 2008). However, increasingly dynamic market conditions have spurred the development of new and modified methods for production scheduling, increasing the complexity of today’s manufacturing environment (Baldea and Harjunkoski 2014). Thus, to overcome some of today’s significant challenges of complex manufacturing systems, machine learning techniques have been used. These data-driven approaches can find complex and non-linear patterns in data of different types and sources. Such data is transformed into relevant information that can support decision-makers or can be used automatically to improve the system (Wuest et al. 2016). In this context, this paper aims to conduct a systematic literature review on the use of machine learning in production scheduling under the Industry 4.0 context. To this end, the research seeks to answer the following question: What are the main machine learning techniques currently employed to perform production scheduling? The review was conducted on the scientific bases Web of Science and Scopus through a bibliometric analysis to identify the main applied ML techniques and content analysis to identify future perspectives. This paper is structured as follows: In Sect. 2, the systematic literature review methodology is described. The third section presents and analyzes the results of the literature review. Finally, in Sect. 4, the final considerations of the research are exposed.
2 Research Methodology In this section, the process and methodology that was followed to conduct this study are presented. The systematic literature review was based on the guidelines established by Moher et al. (2009) and Tranfield et al. (2003). The research methodology was divided into three steps: (i) search and papers collection, (ii) papers screening, and (iii) results analysis. In the first step, the search was performed in the Scopus and Web of Science databases. These databases are considered the largest repositories of scientific documents (Guerrero-Bote and Moya-Anegón 2012). For the search the keyword combinations were defined, according to Table 1. The idea is to search for papers combining machine learning and scheduling for production approaches. Additional keywords have been tested, but they added many non-theme-related papers. Thus, the keywords used resulted in more appropriate documents, potentially including all papers for analysis. As delimitation of the search, only papers written in English and published in and after 2011 have been reviewed. This year delimitation was defined since the objective is to obtain an overview of the use of machine learning in production planning in the era of the fourth industrial revolution (Alcácer and Cruz-Machado 2019; Kagermann et al. 2011).
Machine Learning in Production Scheduling
411
Table 1. Keywords search. Databases Scopus Web of Science Total
Keywords TITLE-ABS-KEY ((“machine learning”) AND (“scheduling”) AND (“production”)) TS = ((“machine learning”) AND (“scheduling”) AND (“production”))
Results 82 47 129
According to Table 1, searching the databases resulted in 129 papers found. For step (ii), papers screening, first the duplicates were removed and afterwards criteria were established to select only papers relevant to the research, as follows: (a) only articles or conference papers; (b) analysis of abstracts to identify theme alignment; and (c) full text papers that can be accessed by CAPES Portal de Periódicos. After the completion of step (ii), the final portfolio was 31 papers. It should be noted that the research was conducted in June 2019. The research protocol was built according to the process model presented in Fig. 1, the main purpose of this study was to select only papers that are related to the adoption of machine learning techniques in the context of production scheduling.
Fig. 1. Process flow to the systematic literature review.
Thus, with the final portfolio, it was possible to perform the third step, which will be presented in the next section. In the third step (iii), results analysis, a bibliometric analysis was performed to identify the main characteristics of the research area and content analysis to answer the research question of the study.
3 Bibliometric Analysis of the Portfolio For the bibliometric analysis, the Bibliometrix tool was utilized, developed for performing comprehensive science mapping analysis. It was programmed in R, is opensource, and provides a wide variety of statistical and graphical techniques (Aria and
412
S. L. Takeda-Berger et al.
Cuccurullo 2017). The bibliometric analysis considered the portfolio of 31 papers to be evaluated according to the dimensions: publications temporal evolution, papers per journal, most cited papers, and highlighted keywords. Figure 2 shows the year of publication of each of the papers that compose the portfolio. This filter is useful to compare the papers in different time slices tracing its historical evolution. As the portfolio is relatively small, these numbers just give an idea of the development, but in this case, it can be seen that there were few publications between 2011 and 2013, but from 2015 the number of publications has grown constantly, except for a small break in 2017. Between 2017 and 2018, publications increased by 300%. This analysis highlights the growing interest in applications of machine learning techniques in the production scheduling environment. 15 10 5 0 Papers
2011 2012 2013 2015 2016 2017 2018 3
2
1
4
5
4
12
Fig. 2. Temporal evolution of publications.
Figure 3 shows the ten journals with the highest concentration of publications in the analyzed portfolio. The International Journal of Production Research and Procedia CIRP contain the largest amount of published papers, followed by IFAC Papers Online and the Computers & Industrial Engineering journal. Figure 4 shows the top ten journals where papers, which have been cited by papers of the analyzed portfolio, have been published according to the Scopus database. In the International Journal of Production Research most of the citing papers have been published, followed by several other journals that link studies of operations management, engineering, and technology. Int. J. of Production Research Procedia CIRP IFAC-PapersOnline Comp. and Ind. Engineering Expert Systems With Applications Comp. and Chemical Engineering Euro. J. of Op. Research Alexandria Engineering Journal Adv. in Intel. Systems and Comp. Journal of Simulation
3 3 2 2 1 1 1 1 1 1 0
1
2
Fig. 3. Papers per journal.
3
4
Machine Learning in Production Scheduling Int. J. of Production Research Comp. and Chemical Engineering Euro. J. of Operational Research Nature Machine Learning IEEE Trans. on Semi. Man. Aiche J. Procedia CIRP Int. J. of Production Economics Comp. and Ind. Engineering
413
66 16 12 10 10 10 10 9 9 8 0
10
20
30
40
50
60
70
Fig. 4. Journals, in which papers have been published that have been cited by the 31 papers
Figure 5 shows a thematic map considering the most evidenced terms in the keywords of the publications, relating the density and centrality of the terms from four perspectives, according to Cobo et al. (2011): (1) motor, (2) specialized, (3) emerging and (4) basic themes. The density can be read as a measure of the theme’s development, and the centrality can be read as the importance of the theme in the entire research field (for more details see Cobo et al. 2011). Each bubble represents a network cluster, i.e., the bubble name is the keywords, belonging in the cluster, with the higher occurrence value, the bubble size is proportional to the cluster keywords occurrences, and the bubble position is set according to the cluster centrality and density (Cobo et al. 2011).
Fig. 5. Thematic mapping of the area.
The upper-right quadrant represents motor themes, i.e., they are well developed and important terms for the structuring of a research field, in this case “decision making”, “production system” and “data mining” appear in this classification. Commonly, the motor-themes present strong centrality and high density. Thus, the placement of terms in this quadrant implies that they are related externally to concepts applicable to other themes that are closely related. The upper-left quadrant has themes with well-developed
414
S. L. Takeda-Berger et al.
internal ties, in which case the term “production scheduling” appeared in this classification. However, in this quadrant the themes are very specialized and a little bit more peripheral in character. The terms “optimization”, “machine learning techniques”, and “manufacturing execution system” were classified in the lower-left quadrant, which presents themes are both weakly developed and marginal. The themes of this quadrant usually have low density and low centrality, i.e., are considered emerging, which need for further studies for development in these fields. Finally, in the lower-right quadrant are the basic themes, considered important for the development of the field, but not yet well developed. In this quadrant is the term “learning system”, which more broadly considers aspects of studies that have adopted learning techniques for manufacturing systems. Due to bubble size it is possible to observe that this term had a lot of occurrences in the papers analyzed, highlighting its importance in research in this field. Figure 6 shows a multi-dimensional scaling keywords co-occurrence network (Huang et al. 2005) using the walktrap clustering algorithm (Pons and Latapy 2005), generated automatically by Bibliometrix. With this map, it is possible to identify the most correlated clusters of terms and the intercession between the themes investigated in this research. In this sense, the analysis may occur as follows. The position of terms may indicate terms with more occurrence in articles (centrality) and terms with less occurrence (periphery). Just as the box size also indicates the most frequently occurring terms (larger boxes) and the least frequently occurring terms (smaller boxes). In addition, the line size is proportional to the co-relation of the terms, i.e., how stronger is the line, more related these words were in the papers.
Fig. 6. Keywords co-occurrence network.
Three main clusters were formed in green, red, and blue. The green one presents the most frequent terms, due to the centrality of the words and the size of the boxes. Additionally, the words of this cluster are strongly correlated with each other, as they presented thick lines, such as “learning system”, “artificial intelligence”, and “scheduling”. Furthermore, the words of this cluster also generate a vast connection to the red and blue clusters.
Machine Learning in Production Scheduling
415
In summary, an analysis of the formed clusters can be performed. The red cluster includes the terms most related to the techniques adopted in the portfolio papers, such as neural networks, decision trees, machine learning. The blue cluster relates to application strategies, such as scheduling problems, production systems, decision making. Finally, the green cluster can be classified with terms related to innovation and industry 4.0, such as learning systems, artificial intelligence, big data, learning algorithms.
4 Main Machine Learning Techniques Applied in Production Scheduling In order to reply to the main question of this paper, Fig. 7 shows the compiled numbers of uses of each machine learning technique family found in the selected papers of the portfolio. The formation of family groups was based on Marsland (2015) and Géron (2019). It is important to mention that, in the case of papers applying several technique families, only the best performing technique, according to the results presented by the authors of the papers, was counted. Artificial Neural Networks Regression Tree-based Models Genetic Algorithms K-nearst Neighbor Bayesian Models Clustering Support Vector Machines Principal Component Analysis Association Rule
8 7 5 3 2 2 1 1 1 1 0
1
2
3
4
5
6
7
8
9
Fig. 7. Machine learning techniques families and their number of uses.
This analysis shows that four techniques are most applied. These are Artificial Neural Networks, Regression, Tree-based models, and Genetic Algorithms. Artificial Neural Networks (ANNs) technique has widely been used in production scheduling (Heger et al. 2016). This technique can provide answers to inputs on an online system which seeks optimizations and acts faster than traditional heuristics which can take, hours, days or weeks to provide desirable results (Gomes et al. 2017). Lee et al. (2018) proposed an architecture framework to implement the cyber-physical production systems cooperating with other manufacturing information systems for quality prediction and operation control in metal-casting processes. The authors used Decision Tree, the Random Forest Model, the ANNs model, and the Support Vector Machine model. Among them, the ANNs model showed the highest accuracy. The second technique, Regression, also had a wide application in the portfolio’s papers. This technique has a good prediction performance in comparison to other techniques. Additionally, the Regression method is used for forecasting and finding the
416
S. L. Takeda-Berger et al.
causal relationship between variables (Heger et al. 2013). Gaussier et al. (2015) studied how predictions of the running times may help in obtaining a better schedule. For this purpose, the authors used Regression method and cost functions that are used to learn the prediction model. The results showed an average gain of 28% compared to the classical EASY policy algorithm. The third most applied technique were Tree-based models. These techniques are the most commonly known method when building decision models (Bergmann et al. 2017). Lubosch et al. (2018) proposed an algorithm that combines Monte Carlo Tree Search and Decision Tree to improve the production system’s performance. This flexible scheduling algorithm can easily adapt to different types of problems and situations while still finding near-optimal schedules. The results show that a combination of these two techniques is a promising way to handle complex industrial scheduling problems. Finally, Genetic Algorithms (GA) was the fourth most technique applied in the portfolio’s papers. The GA could be used as a feature selection method to search for an optimal feature subset from a large number of candidates (Shapiro 2011). Ma et al. (2016) proposed a learning-based scheduling framework for semiconductor manufacturing system. For this, the hybrid algorithm based on GA, Simulated Annealing (SA) and Extreme Learning Machine (ELM) was developed. The result indicates that the proposed method satisfies the requirement of real-time scheduling well and gets better performance compared to using the ELM method only. In summary, the papers in the portfolio show that the adoption of ML techniques has been attracted much attention in recent years. Especially in production scheduling since it’s one of the most important activities in a manufacturing company. Figure 8 shows the evolution of the four main ML techniques mentioned in the portfolio and described previously. This graphic gives an overview of the evolution of these methods for this specific, small selection of papers.
Fig. 8. Temporal evolution of the main ML techniques.
Among the four techniques, it is possible to notice that there was a growth mainly in the use of the technique of ANNs. Such fact may be justified due to currently a huge quantity of data available to train neural networks, and ANNs frequently outperform other ML techniques on very large and complex problems (Géron 2019). In addition, there was also an increase in the use of Regression methods. Regression is a useful and
Machine Learning in Production Scheduling
417
widely used statistical learning method since many fancy statistical learning approaches can be seen as generalizations or extensions of Regression methods (James et al. 2013). In the end, there are a wide variety of aspects that need to be considered when developing production scheduling models. Due to the nature of the manufacturing systems, there is no best technique for all scenarios, settings, and objective functions. However, manufacturing processes have been updated to follow the trends pointed out by Industry 4.0 and, more studies applying ML techniques can increase knowledge in this field of research.
5 Conclusion and Further Research This research paper conducted a systematic literature review to explore the use of machine learning in production scheduling in the Industry 4.0 context. This research realized a bibliometric analysis and identified the main machine learning techniques currently employed to improve production scheduling. The analysis of 31 papers shows that the number of publications has grown constantly since 2015; this shows a great interest in the application of machine learning techniques in the production scheduling environment. Additionally, emerging themes have been identified. Terms like “machine learning techniques” and “learning system” are considered a recent field to deserve more studies and exploration, especially for manufacturing systems environments. In summary, many different approaches to improve scheduling in dynamic stochastic environments have been studied. However, it is important to develop more studies applying these techniques to provide more results and knowledge about this emerging field in the industry 4.0 context. For the bibliometric analysis only 31 papers have been used which have been published from 2011 on. For further research it might be interesting to also consider older paper and approaches. This would also lead to more significant results regarding the total usages of the different methods. But this paper focusses on the research since the establishment of the term Industry 4.0 and therefore gives a good overview of the main machine learning techniques applied in production scheduling. So, it answers the paper’s research question and therefore the study’s purpose was achieved. For future research, manufacturing is an area where the application of machine learning can be very fruitful. The inclusion of these techniques for predictive and reactive production scheduling could be a good way to face the disruptions during the production execution. Moreover, the exploration of other machine learning techniques that have not been studied extensively in the analyzed portfolio’s papers is an interesting field, capable of providing more insight into the behavior and the possible advantages of such techniques by applying them at production scheduling problems. Acknowledgements. This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001. It is also funded by the German Research Foundation (DFG) under reference number FR 3658/1-2 and by CAPES under reference number 99999.006033/2015-06, in the scope of the BRAGECRIM program.
418
S. L. Takeda-Berger et al.
References Alcácer, V., Cruz-Machado, V.: Scanning the industry 4.0: a literature review on technologies for manufacturing systems. Eng. Sci. Technol. Int. J. 22(3), 899–919 (2019) Aria, M., Cuccurullo, C.: bibliometrix: an R-tool for comprehensive science mapping analysis. J. Inf. 11(4), 959–975 (2017) Baldea, M., Harjunkoski, I.: Integrated production scheduling and process control: a systematic review. Comput. Chem. Eng. 71, 377–390 (2014) Bergmann, S., Feldkamp, N., Strassburger, S.: Emulation of control strategies through machine learning in manufacturing simulations. J. Simul. 11(1), 38–50 (2017) Cobo, M.J., López-Herrera, A.G., Herrera-Viedma, E., Herrera, F.: An approach for detecting, quantifying, and visualizing the evolution of a research field: a practical application to the fuzzy sets theory field. J. Inf. 5(1), 146–166 (2011) Flath, C.M., Stein, N.: Towards a data science toolbox for industrial analytics applications. Comput. Ind. 94, 16–25 (2018) Gaussier, E., Glesser, D., Reis, V., Trystram, D.: Improving backfilling by using machine learning to predict running times. In: International Conference for High Performance Computing, Networking, Storage and Analysis, SC 2015, 15–20 November 2015 (2015) Géron, A.: Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems. O’Reilly Media, Boston (2019) Gomes, M., Silva, F., Ferraz, F., Silva, A., Analide, C., Novais, P.: Developing an ambient intelligent-based decision support system for production and control planning. In: Madureira, A.M., Abraham, A., Gamboa, D., Novais, P. (eds.) Intelligent Systems Design and Applications, pp. 984–994. Springer, Cham (2017) Guerrero-Bote, V.P., Moya-Anegón, F.: A further step forward in measuring journals’ scientific prestige: the SJR2 indicator. J. Inf. 6(4), 674–688 (2012) Heger, J., Branke, J., Hildebrandt, T., Scholz-Reiter, B.: Dynamic adjustment of dispatching rule parameters in flow shops with sequence-dependent set-up times. Int. J. Prod. Res. 54(22), 6812–6824 (2016) Heger, J., Hildebrandt, T., Scholz-Reiter, B.: Dispatching rule selection with Gaussian processes. Cent. Eur. J. Oper. Res. 23(1), 235–249 (2013) Huang, J.J., Tzeng, G.H., Ong, C.S.: Multidimensional data in multidimensional scaling using the analytic network process. Pattern Recognit. Lett. 26(6), 755–767 (2005) James, G., Witten, D., Hastie, T., Tibshirani, R.: An Introduction to Statistical Learning. Springer, Cham (2013) Kagermann, H., Lukas, W.D., Wahlster, W.: Industrie 4.0 – Mitdem Internet er Dinge auf dem Wegzur 4. Industriellen Revolution (2011). https://www.vdi-nachrichten.com/TechnikGesellschaft/Industrie-40-Mit-Internet-Dinge-Weg-4-industriellen-Revolution. Accessed 31 July 2019 Lee, J., Noh, S.D., Kim, H.J., Kang, Y.S.: Implementation of cyber-physical production systems for quality prediction and operation control in metal casting. Sensors 18(5), 1428–1444 (2018) Leyh, C., Martin, S., Schäffer, T.: Industry 4.0 and lean production—a matching relationship? An analysis of selected industry 4.0 models. In: 2017 Federated Conference on Computer Science and Information Systems (FedCSIS), pp. 989–993 (2017) Li, Z., Ierapetritou, M.: Process scheduling under uncertainty: review and challenges. Comput. Chem. Eng. 32(4–5), 715–727 (2008)
Machine Learning in Production Scheduling
419
Lubosch, M., Kunath, M., Winkler, H.: Industrial scheduling with Monte Carlo tree search and machine learning. In: Kjellberg, T., Wang, L., Ji, W., Wang, X.V. (eds.) 51st CIRP Conference on Manufacturing Systems, CIRP CMS 2018, vol. 72, pp. 1283–1287 (2018) Ma, Y., Qiao, F., Lu, J.: Learning-based dynamic scheduling of semiconductor manufacturing system. In: 2016 IEEE International Conference on Automation Science and Engineering, CASE 2016, November 2016, pp. 1394–1399 (2016) Marsland, S.: Machine Learning: An Algorithmic Perspective. CRC Press, Boca Raton (2015) Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G.: Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Ann. Intern. Med. 151(4), 264–269 (2009) Mulrennan, K., Donovan, J., Tormey, D., Macpherson, R.: A data science approach to modelling a manufacturing facility’s electrical energy profile from plant production data. In: 2018 IEEE 5th International Conference on Data Science and Advanced Analytics (DSAA), pp. 387–391 (2018) Peruzzini, M., Grandi, F., Pellicciari, M.: Benchmarking of tools for user experience analysis in industry 4.0. Procedia Manuf. 11, 806–813 (2017) Pons, P., Latapy, M.: Computing communities in large networks using random walks. In: International Symposium on Computer and Information Sciences, pp. 284–293 (2005) Rüßmann, M., Lorenz, M., Gerbert, P., Waldner, M., Justus, J., Engel, P., Harnisch, M.: Industry 4.0: the future of productivity and growth in manufacturing industries, vol. 9, no. 1, pp. 54– 89. Boston Consulting Group (2015) Shapiro, J.: Genetic algorithms in machine learning. In: Advanced Course on Artificial Intelligence, pp. 146–168 (2011) Tranfield, D., Denyer, D., Smart, P.: Towards a methodology for developing evidence-informed management knowledge by means of systematic review. Br. J. Manag. 14(3), 207–222 (2003) Wuest, T., Weimer, D., Irgens, C., Thoben, K.: Machine learning in manufacturing: advantages, challenges, and applications. Prod. Manuf. Res. 4(1), 23–45 (2016)
Software-Defined Mobile Supply Chains Uwe Aßmann1, Udo Buscher2, Sven Engesser3, Jörn Schönberger4(&), and Leon Urbas5 1
Faculty of Computer Science, Technische Universität Dresden, 01062 Dresden, Germany [email protected] 2 Faculty of Business and Economics, Technische Universität Dresden, 01062 Dresden, Germany [email protected] 3 Faculty of Arts, Humanities and Social Science, Technische Universität Dresden, 01062 Dresden, Germany [email protected] 4 “Friedrich List” Faculty of Transport and Traffic Sciences, Technische Universität Dresden, 01062 Dresden, Germany [email protected] 5 Faculty of Electrical and Computer Engineering, Technische Universität Dresden, 01062 Dresden, Germany [email protected]
Abstract. Supply chain designs and processes are the outcome of a bunch of different hierarchically organized planning problems. The hierarchy induces different decision update cycles for the individual decisions. This classification is mainly based on the assumption that the re-location of facilities causes disruptive efforts and should be avoided as long as possible. Reactions on demand variations (often called agility activities) are therefore mainly addressed by intensified transportation efforts causing noise and substance emissions. To overcome this undesired situation the concept of “software-defined mobile supply chains” is proposed. The idea is to mobilize so far immobile production assets using smart mobile production units supported by a software-concept that ensures the adaptation of the mobile production units to a varied demand situation. Keywords: Supply chain planning
Cyber-physical systems Agility
1 Introduction Modern value creation systems (called “Supply Chains” or “Supply Networks”) contribute to greenhouse gas (GHG) emissions in the different value creation steps starting from procurement, over production, distribution, purchasing as well as consumption of finished goods. Production as well as transportation operations seem to be the most crucial contributors to GHG emissions of supply consortia. The design as well as the processes in a supply chain are the outcome of several well-balanced individual decisions. They are aligned to fulfill given demand with © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 420–430, 2020. https://doi.org/10.1007/978-3-030-44783-0_40
Software-Defined Mobile Supply Chains
421
maximal efficiency to achieve one or several planning objectives. Thus, any demand fluctuation and variation endangers the performance of this orchestrated interaction of different units and assets. A quick, cheap and reliable adaptation of the design and process decisions is necessary to recover from such a suboptimal situation [1]. This paper introduces a new adaptation approach to reduce negative impacts resulting from demand fluctuation, which is an often-observed kind of dynamics in logistics. The main proposed innovation is the mobilization of (so far immobile) stationary production assets (like factories or warehouses or terminals). Instead of limiting the demand-oriented supply chain setup adaptation in response to a demand variation to the revision of the transportation network as made in the supply chain planning matrix [2], we propose to enable a quick and efficient relocation of production sites. Such a relocation is possible if small units, which can be moved easily, form the production system. Furthermore, these units can be combined in a modular (or “Lego-like”) fashion to quickly realize and setup an immediately needed special purpose production facility. We call a supply chain “mobile” if used production sites are prepared to be relocated quickly and efficiently at reasonable costs. Doing so, the production system should adapt quickly to a volatile environment. In order to ensure a smooth and reliable interaction and orchestration of a bunch of mobile but modular assets it is necessary to establish and connect a specific information and communication layer with a mobile supply chain. This layer is used to realize information exchange and broadcasting among and towards all assets in a mobile supply chain. However, it is also used to specify information interfaces to mobile assets, which might be integrated for a certain period into the mobile supply chain consortium. While the mobile assets represent the “hardware” ingredients of a mobile supply chain, the aforementioned layer is interpreted as the “software” that glues the hardware components together. A Software-Defined Mobile Supply Chain (SD-MSC) uses mobile assets whose interactions are made available and are defined by an adequate software layer. This manuscript introduces into the concept of SD-MSCs. The central research question addressed here is “What is an SD-MSC and why do we need both hardware and software concepts to exploit its abilities to improve value creation processes and systems?” We start with the positioning of the idea of a Software-Defined Mobile Supply Chain within the existing literature in Sect. 2. Next, we describe and analyze supply chain scenarios in which mobile production facilities are already deployed (Sect. 3). Section 4 addresses the transformation of a “traditional” supply chain into an SD-MSC if we assume that mobile production assets are available, that can be relocated quickly. Section 5 proposes research areas that particularly address the challenges of developing a transferrable management regime for SD-SCMs.
2 Literature-Based Positioning of the Idea of SD-MSCs The integration of economic, technical as well as environmental objectives in the management of supply networks within the last 20–30 years has led to the concept of the closed loop supply chain [3]. Here, the closure of the material flow to a cycle flow of materials is proposed to reduce the negative impacts of supply chain operations. In
422
U. Aßmann et al.
the focus of the associated research manufacturing and production, related concepts are found. Transport operations, which have also a significant contribution to GHG emissions, have received less interest. In supply chain management there is the paradigm that production facilities are immobile assets. They form focal points in value creation systems [2]. The idea behind is to generate significant economies of scale by condensing large production volumes at few network nodes. However, this concept induces significant challenges to the distribution of semi-finished or finished goods. First, global transportation still increases and causes huge volumes of emissions [4]. Second, the growth of large nodes in supply nets is limited by a lack of expansion opportunities [5]. Third, system inherent challenges like the empty container repositioning caused by imbalanced material flows in the global context cannot be avoided but cause costs [6]. Technological innovations like 3D-printing, sintering and cutter systems are expected to influence supply chain wide material flows. In addition, informationprocessing technologies like cloud-based sensor and actuator systems, flexible and programmable production systems with remote control have penetrated the supply chain application environment. Together, they are initiating an ongoing revision and reengineering of the design of supply chains and their control paradigms [9]. Recent developments and innovations are summarized under the terms Internet of Things as well as Industry 4.0 and Logistics 4.0 [8]. They finally postulate a new design concept for both the material flow as well as the information flow system of a supply chain. While the old centralized design concept of a supply chain seems to be unable to fulfill individual demand of far away customers the new design concept particularly addresses this challenge. However, it requires replacing centralized production at least partially by a decentralized manufacturing close to the demand site [7]. In this context, SD-SCMs represent a promising supply chain control concept. It founds on the idea that so far immobile production sites are substituted by a network of mobile production resources. This network adapts itself dynamically to fluctuating and variable or volatile demand as well as to the availability of the mobile resources. Here, mobility is interpreted as information mobility in the sense of ubiquity [10] as well as physical mobility both appearing in different appearance intensities. While ubiquity enables the responsibility for supply chain decision (“moving competences”) physical mobility finally refers to the physical movement of production facilities. Mixtures of mobility between these two extreme scenarios can be setup continuously. An individual “mobility degree” can be installed for each individual scenario. Beside the system internal aspects of the mobilization of production facilities (discussed in Sect. 3), also mega-trends can be observed that will foster the dissemination of SD-SCMs. First, the maturity of technology that supports the further formation of the “Internet of Things” will support ideas similar to the SD-MSC [8]. Second, the reactivity of a supply chain becomes more and more important as a competitive advantage of a supply chain consortium [1]. Production must come closer to demand sites but demand is moving and to preserve short fulfillment times, production must follow the demand in time and space. SD-MSCs contribute to the dissemination of “ad-hoc” supply chains which form spontaneously and that exist only for a short time period [11]. Finally, SD-MSCs have to be investigated with respect to
Software-Defined Mobile Supply Chains
423
environmental aspects. In particular, the ability that production facilities preserves close distances to markets is promising with respect to contribute to the reduction of transportation and save associated emissions [12].
3 Mobile Production Units A mobile production unit is every facility that is (a) mobile and (b) that is able to contribute to the manufacturing of products and services. Here, “mobile” refers to the property of a production unit that its dismantling, its movement to another location and its re-construction and setup is reasonable since the construction of another (second) units at another location would be impossible or too costly. The physical extent of such a mobile production unit varies from small tools up to complete factories and everything in between (like so-called “container factories”, [16]). In the literature, some applications are reported in which mobile supply chain production assets are already in use. They are moved in order to maintain or increase the supply chain process performance. The individual reasons for the deployment of mobile production facilities are different. The following examples have been selected to demonstrate that different types of applications and different ideas as well as objectives are involved into the idea to mobilize production units in a mobile supply chain. 3.1
Oil Industry
Floating Production Storage and Offloading Units (FPSO) are vessels that host production and/or warehousing capacities for the processing of crude oil produced on the high shores. After an oil field is exhausted an FPSO can be moved (or moves itself) to another oil field. Here, FPSOs are deployed as “mobile assets” since production capacities are needed at a certain place for a certain time (period). The installation of a transportation link (by vessel or pipeline) to a central immobile production facility is often impossible or too expensive. 3.2
High Sea Fishery
The success of industrial fishery depends on several factors. Among them, the availability of fish in a certain fishery region mainly determines the quantity of caught fish. From period to period the availability of fish varies. In addition, the fishery regions in which a fishery boat fleet is allowed to operate is often far away from the intended markets. It is therefore not possible to install and maintain a factory to process the fresh fish. In this context, fishery production vessels combine production, warehousing as well as transportation capabilities. The immediate onsite processing of fresh fish as well as the ability to freeze the fresh fish, make intermediate visits to ports as well as intermediate transportation of small batches unnecessary.
424
U. Aßmann et al.
3.3
Perishable Goods
Supply chains setup for the processing and distribution of perishable goods are faced with several special requirements. Often, the maximal time until the goods are distributed are crucial. In other scenarios, the processing time of a product is used to modify it, like for example a green banana that matures during the several week vessel trip from America to Europe. Here, the vessel moves to the source of the raw material (the green banana) and the maturing of the banana happens during the vessels trip. This “mobile asset” is deployed to ensure that the final product reach the market in the right quality. 3.4
Blood Supply Chains
Supply chain processes designed for the treatment of scarce products must offer a very high level of reliability and security, like cash-in-transit. Human blood is a very scarce but sensitive “raw material” that is required to save lives. Human blood donators are the only source of the raw material but the available quantity as well as the demanded quantities of donated blood are uncertain. In addition, processing times of full blood are very short. As a consequence blood donation facilities are often mobile and move to the location of blood supply instead of letting donators travel to the facilities to reduce the donator’s travel efforts and therefore, motivate more persons to donate their blood [13]. 3.5
Advantages of Mobile Production Assets
In the high sea oil, production scenarios larger investments can be avoided in regions where raw material supply and/or demand is high only for a short period. Furthermore, mobile production assets are used to ensure that perishable goods reach the market in the requested quality as seen in the fishery scenario but also in the perishable goods example. In addition, the gathering of larger quantities of raw materials and the fulfillment of larger demand requires mobile production facilities here. Finally, mobile production units are expected to contribute significantly to the reduction of transportation efforts (in the blood collection example) and the accompanying side effects like noise and harmful substance emission (in the fishery and the perishable food applications).
4 From Mobile Assets to Software-Defined SCMs 4.1
Impacts on Supply Chain Planning and Decision Making
At first glance, it seems that the ability to realize short-term re-location decisions mainly depends on the ability to move physically the assets. However, such a relocation decision requires preparations, i.e. the solving of relocation decision tasks. A (re-) location decision task is connected to or has impacts on other decision (tasks) within a supply chain setup. A commonly accepted scheme to classify the different supply chain related decision tasks is the supply chain planning matrix (SCPM) [2].
Software-Defined Mobile Supply Chains
425
The SCPM represents a hierarchical planning concept by sorting supply chain planning tasks according to increasing re-planning frequency. Decisions for decision tasks with low re-planning flexibility are preserved and protected for a long time and are used as invariant input parameters to those planning tasks with a higher re-planning flexibility. According to the SCMP-concept, location decisions are long-term decisions with a very low re-planning flexibility. These decision tasks are called “strategic”. In addition, one might also consider distribution planning tasks in the context of location planning. Distribution planning decides about the installation of transportation links. They have a higher re-planning frequency and fall into the category of “tactical” decision tasks (left part Fig. 1). If we increase the frequency of (re-) location decisions, it is necessary to consider other decision tasks with higher re-planning frequencies because the outcome of (re-)location decisions cannot be used as input to those “operative tasks” anymore since (re-)location is now an operative decision task (right part of Fig. 1). This means that capacity as well as transportation capacity determination tasks must now be solved in parallel to re-location tasks, which tremendously increases the complexity of operative decision tasks. Very complicated, complex and interwoven decision problems that must be solved when controlling the mobile assets in a SD-MSC.
Fig. 1. Decision tasks associated with re-location tasks in SD-MSCs
4.2
The Meaning of “Software Definition”
The environment in which the re-locations of movable assets are realized vary from local areas (like one production site) to the global perspective (entering new markets). The idea of a SD-SCM comprises not only the movement of individual assets from one original to another new place in reaction to a change in the SCM-environment. Instead, groups of several independent assets should interact in order to adapt the SCM to the new challenge. To realize such an adaptation it is necessary that mobile assets can receive and submit information about own abilities and desires but about those opportunities of the other assets. Furthermore, a control regime is needed that finally determines how the mobile assets are orchestrated. Such a regime can be controlled centrally or (at least partially) in a decentralized manner [7]. The degree of autonomous decisions of individual assets must be adjustable to fit to a specific situation. Finally,
426
U. Aßmann et al.
it is necessary that different entities can be moved. Here, an entity can be the asset itself (“physical movement”) or allowances/responsibilities can be assigned to or removed from assets if reasonable (“privilege movement”). As a prerequisite for the connectivity it is necessary that each mobile asset in a supply chain can be integrated into a powerful and reliable communication infrastructure using wireless lan, other mobile communication networks (based on G4 or G5) or Bluetooth (e.g. for truck platoons). The asset’s hardware must be equipped with innovative software that is able to detect, process, and submit all information. Since different types of hardware (from different contributors) must be integrated into a specific SCM setup without long-term manual but quick autonomous reconfiguration, clearly specified interfaces and protocols are needed. The concept of multi-agent systems to represent the group of assets in a supply chain has been already investigated in the context of management of logistics systems. Technological progress in the last decade w.r.t. cloud-based information sharing now offers additional opportunities to realize a cooperation of individual agents. Instead of bilateral information exchange between agents, more general information exchange concepts are possible. However, even a cloud-based information exchange require a structure. Furthermore, rules for storing, processing, deleting and updating information stored in the cloud must be stated in order to avoid an information overload. Altogether, the physical production assets in a supply chain that can move or can be moved, enriched by the communication network that links them with a cloud or similar data exchange system and a system-wide decision making concept realizes the specific cyber-physical-system [14] which we name “Software-Defined Mobile Supply Chain”. Finally, the resulting complex decision problems related to the high frequency relocation outlined in Subsect. 4.1 must be solved automatically. For that, specific planning/decision support tools are needed. Therefore, additional software is needed to processes the available planning information with the goal to make the best process decisions for the integrated decision tasks. The delegation of decision tasks must be defined. In summary, it is necessary to develop accompanying software for mobile production assets that enables the ad-hoc orchestration of mobile production assets, establishes and deploys a sufficient communication and support the solving of the new integrated decision tasks.
5 Outline of a Research Agenda for SD-MSCs The dissemination of the idea of SD-MSCs into an implementable and realistic concept requires a research agenda that prepares and accompanies the mitigation of current concepts of physical goods distributions into SD-MSCs. With the goal to structure the necessary research into an agenda three research areas are defined (cf. Fig. 2):
Software-Defined Mobile Supply Chains
427
Fig. 2. Research landscape for SD-MSCs
• Research area “architecture”: hard-& software-systems for the representation and the control of SD-MSCs • Research area “optimization”: planning and deployment of material flows in SDMSCs under the considerations of costs, revenues, robustness, stability, reliability, • Research area “information modelling”: software-oriented SD-MSC-modelling, connectivity to existing systems and interfaces. In order to ensure that a comprehensive and integrated concept for a SD-MSC will be developed it is necessary to investigate each research area in depth but, even more challenging, it is required to address the intersections of the three areas. Five explicit research topics (1) – (5) are identified and investigated during the period from 2019 until 2022 by the cluster on “Software-Defined Mobile Supply Chains” of the Boysen TU Dresden Research Training Group “Mobility in Transition”. This group is an interdisciplinary research and doctoral study group [15]. Project 1 - Integrated Production and Distribution Planning in Software-Defined Mobile Supply Chains: This project addresses the question of how production and distribution are coordinated when there is the possibility of relocating production facilities at comparatively short notice. Supply chain objectives cannot be achieved without considering the interdependencies between production, plant relocation and logistics costs. Particular attention must be paid to the temporal decomposition of the planning period in order to capture adequately the various cost effects. The problems represented in mathematical models are solved with the help of mathematical programming and metaheuristics in order to analyze under which conditions mobile production plants prove to be particularly advantageous. Project 2 - Decentralized Decision Making in SD-MSCs: Spatial transformations of value-creating resources are the major ingredient of the SD-MSC-concept. The goal is to increase the efficiency of these resources. In this line, it becomes necessary to enrich the already complex bunch of supply chain related time- and factual decision problems by spatial decision tasks. The deployment of mobile resources (vehicles) is challenging. The corresponding resource allocation tasks are summarized under the term vehicle routing and scheduling problems (VRSP). Solving VRSPs requires decisions that are
428
U. Aßmann et al.
made reactively, dynamically and under consideration of locally available information in order to exploit the emergence of the SD-MSC. With the objective to enable the efficient and effective solving of highly complex resource deployment decision problems in SD-MSCs this project investigates the conceptualization, configuration and prototypical evaluation of decentralized decision making approaches especially for VRSP applications in SD-MSCs. Project 3 - Confidentiality in SD-MSCs: This project develops methods and algorithms that make it possible to evaluate and maintain a required degree of confidentiality in mobile cyber-physical production systems. We argue that a privacy-based approach that is a task quality oriented anonymization of the data to be exchanged between plant operators and service providers is superior to any encryption technique, because we can optimize the benefits and risks of the amount of shared information. For the systematic processing of research questions methods of information modeling are used and security analysis applied to mobile cyber-physical production system. Project 4 - Fog Computing in Software-Defined Mobile Supply Chains: To control mobile plants, a recently developed cloud computing approach, called fog computing, is very interesting to employ, because fogs are sensor-actor-based near-clouds that control cyber-physical systems as complex machines, robots, as well as assembly lines, and can be used as well for mobile production plants. A fog keeps all data private and moves with the mobile production plant. In addition, a fog maintains a world model (“digital twin”) of its physical world, i.e., the mobile production plant. In this project, the fog should use an initial robot (embryo) to construct an assembly line of robots. This means that the fog of the mobile production plant must know how to build up a plant (specified a target world model) and how to assemble it. Project 5 - Trust in Mobile Value Creation Chains with Informational Character: Trust is highly relevant in supply chain management because it increases commitment, performance and efficiency. This project analyses the role of trust in Software-Defined Mobile Supply Chains (SD-MSC). It focuses in particular on the influence of information exchange and information quality on trust, as well as on the interplay of confidence and confidentiality. The project develops a conceptual model of trust, which will be empirically tested through surveys among employees and social science experiments. Based on the findings, recommendations for optimizing the information exchange and implementing confidence-building measures in SD-MSC will be formulated.
6 Conclusions and Future Work We have outlined the concept of software-defined mobile supply chains as an approach to transform a traditionally designed and controlled supply chain into an agile value creating system, which adapts to demand variation by a quick as well as reactive relocation of mobile production assets. Although we have seen that mobile production facilities are already in use it is necessary to develop a new information as well as resource deployment regime, since new challenges must be managed.
Software-Defined Mobile Supply Chains
429
As a first step towards the realization and dissemination of SD-MSCS we have identified three research areas that must be addressed in future research. Five explicitly stated research tasks have been formulated that are positioned in the intersections of at least two of these areas. Acknowledgement. The authors thank the Friedrich und Elisabeth Boysen-Stiftung for the financial support given for this work during the third Boysen-TU Dresden-Graduiertenkolleg “Mobility in Transition”.
References 1. Sharma, N., Sahay, B., Shankar, R., Sarma, P.: Supply chain agility: review, classification and synthesis. Int. J. Logist. Res. Appl. 20, 1–28 (2017) 2. Rohde, J., Meyr, H., Wagner, M.: Die supply chain planning matrix. PPS Manag. 5, 10–15 (2000) 3. Difrancesco, R.M., Huchzermeier, A.: Closed-loop supply chains: a guide to theory and practice. Int. J. Logist. Res. Appl. 19(5), 443–464 (2016) 4. Rauch, E., Dallasega, P., Matt, D.T.: Sustainable production in emerging markets through Distributed Manufacturing Systems (DMS). J. Clean. Prod. 135, 127–138 (2016) 5. Gertz, C.; Wagner, T.: Konfliktfelder von wachsenden Logistikknoten. In: Schrenk, M., Popovich, V.V., Engelke, D., Elisei, P. (eds.) Proceedings of the 13th International Conference on Urban Planning and Regional Development in the Information Society, Vienna, pp. 305–314 (2008) 6. Xie, Y., Liang, X., Ma, L., Yan, H.: Empty container management and coordination in intermodal transport. Eur. J. Oper. Res. 257, 223–232 (2017) 7. Hülsmann, M., Scholz-Reiter, B., Windt, K. (eds.): Autonomous Cooperation and Control in Logistics. Springer, Heidelberg (2011) 8. Barreto, L., Amaral, A., Pereira, T.: Industry 4.0 implications in logistics: an overview. Procedia Manuf. 13, 1245–1252 (2017) 9. Srai, J.S., Kumar, M., Graham, G., Phillips, W., Tooze, J., Ford, S., Beecher, P., Raj, B., Gregory, M., Tiwari, M.K., Ravi, B., Neely, A., Shankar, R., Charnley, F., Tiwari, A.: Distributed manufacturing: scope, challenges and opportunities. Int. J. Prod. Res. 54(23), 6917–6935 10. Estrada, E.: Information mobility in complex networks. Phys. Rev. E 80(2), 026104 (2009) 11. Franke, J., Widera, A., Charoy, F., Hellingrath, B., Ulmer, C.: Reference process models and systems for inter-organizational Ad-Hoc coordination - supply chain management in humanitarian operations. In: 8th International Conference on Information Systems for Crisis Response and Management (2011) 12. Demir, E., Bektaş, T., Laporte, G.: A review of recent research on green road freight transportation. Eur. J. Oper. Res. 237(3), 775–793 (2014) 13. Schiefer, G.: Motive des Blutspendens: Tiefenpsychologische Untersuchung mit Gestaltungsoptionen für das Marketing von Nonprofit-Organisationen des Blutspendewesens, Deutscher Universitätsverlag (2006) 14. Püschel, G., Piechnick, C., Aßmann, U.: Generative und simulative Softwaretests für selbstadaptive, cyber-physikalische Systeme. In: Aßmann, U., Demuth, B., Spitta, T., Püschel, G., Kaiser, R. (eds) Software Engineering & Management, vol. 239 of LNI, p. 135. GI (2015)
430
U. Aßmann et al.
15. https://tu-dresden.de/bu/wirtschaft/bu/forschung/forschungsprojekte/projekte/3-boysen-tudresden-graduiertenkolleg-mobilitaet-im-wandel-herausforderungen-und-loesungen-fuertechnik-umwelt-und-gesellschaft?set_language=en 16. de Wet, P., Oosthuizen, G.A., Burger, M.D., Oberholzer, J.F.: Sustainably manufacturing a bamboo bicycle in a container factory. Procedia Manuf, 7, 234–239 (2017)
Clustering for Monitoring Logistical Processes in General Cargo Warehouses Andreas Neubert(B) PKE Deutschland GmbH, Logistics Competence Center, Egellsstraße 21, 13507 Berlin, Germany [email protected]
Abstract. This paper analyses and evaluates the monitoring methods of logistical processes in place in general cargo warehouses. For this purpose, existing monitoring processes are examined and evaluated. Based on the results, the paper proposes a method for monitoring logistical processes. The paper also looks at various clustering algorithms. The proposed method is applied with data obtained in a general cargo warehouse. The results are described and discussed with a view to their application in warehouses. Keywords: Monitoring · Time series databases Tracking and tracing · General cargo warehouse
1
· Clustering ·
Introduction
Due to the different properties of general cargo [1,2], there has, to date, only been a low level of automation in general cargo warehouses. General cargo is therefore subject to manual handling in the warehouses, with pallets either being unloaded from trucks to the storage area by forklift trucks or hand pallet trucks in the goods receipt area or, in the goods issue area, loaded from the storage area onto the trucks. This manual handling, again and again, gives rise to errors in the logistical process, which manifest themselves as misloading, incorrect storage or damage to the stored goods or equipment in the warehouses. If these errors were identified in advance, savings could be made. Workflows in general cargo warehouses are planned with the help of warehouse management systems. If a package is to be moved in the warehouse, the forklift driver receives a paperless transport order via WLAN on his terminal or scanner (MDE). The forklift driver then carries out the transport order and, after having executed the order, enters the fact that the order has been completed in the warehouse management system. However, it is not clear whether the data entries in the system have actually been carried out or whether they were carried out correctly. For instance, the forklift driver may have forgotten c The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 431–440, 2020. https://doi.org/10.1007/978-3-030-44783-0_41
432
A. Neubert
to enter data after having unloaded a pallet at a bin location. It is also possible that he entered data even though he unloaded the package at the wrong storage location, resulting in incorrect storage. This results in long search times for the package for subsequent loading since the package is not in the storage bin that has been recorded in the warehouse management system. No information is available on the actual status in the cargo warehouse because there is no feedback about the actual situation in the warehouse management system. This calls for an analysis of logistical operations in general cargo warehouses in order to be able to detect and counteract logistical errors. The paper examines the following research question: Q1: How and to what extent can clustering methods be used to monitor logistical processes in a general cargo warehouse? For this purpose, existing monitoring processes in general cargo warehouses are being examined and evaluated. On the basis of the results, the paper proposes a method for monitoring logistical processes. In the process, different clustering algorithms are also examined. The proposed method is applied with data from a general cargo warehouse. The results are described and analyzed.
2 2.1
Monitoring of Logistical Workflows in General Cargo Warehouses Monitoring Processes in General Cargo Warehouses
Different areas, methods, and techniques are in place in general cargo warehouses to control logistical processes. These are described in greater detail below. Identification Point (I-Point). The recording of goods receipt and the combination of loading aids and stored goods to form a storage unit takes place at an identification point [3]. The dimensions and weights of the loading units are recorded here [4]. The storage locations are also determined at the identification point [4]. Measuring Point. There are so-called measuring points within the warehouse with a view to quality assurance. For example, temperature measurements must be carried out at the interfaces of a cold chain for perishable goods in order to preserve evidence [5]. In addition to the condition of the goods, the place and time are also known. Warehouse Inspection. In a general cargo warehouse, the goods are inspected to ensure that they are properly stored. It is checked, for example, whether liquids are leaking from the packages. In the case of warehouses with shelving systems, a regular shelving inspection is carried out pursuant to DIN EN 15635 to ensure occupational safety and health and accident prevention. The standard describes the application and maintenance of stationary steel shelving systems [6]. The warehouse stock levels can also be recorded. Inventory. According to Bichler, inventory is the recording of physical material stock (warehouse stock levels) by means of measurements, weighing and
Clustering for Monitoring Logistical Processes
433
figures [3]. The target stock is managed in the warehouse management system. The target stock is brought into line with the actual physical status [3]. Deviations can ensue from errors during the inventory, theft or booking errors during goods receipt or goods issue [3]. Smart Pallets. The research institutes Fraunhofer IML and Fraunhofer IFF have presented different concepts setting out how the pallet, besides being a loading aid, can also provide information. The Fraunhofer IML and the companies Deutsche Telekom and EPAL have 500 so-called smart pallets in practical use. A low-cost tracker is integrated into an EPAL pallet. The tracker registers shocks, positions, tilt angle, acceleration and temperature of the pallet and transmits the data via Narrow-Band IoT (NBIoT) [7]. Deutsche Telekom expects a device that is capable of narrow band to cost less than $5 per module [8]. Together with the companies Cabka IPS, Telent and metraTec, Fraunhofer IFF developed an IoT pallet that provides data on the location, movements, and condition of the load and the stresses and strains to which it is exposed. The data of the plastic pallets are transmitted via an Low Power Wide Area Network (LPWAN). Critical deliveries are thus better secured and controlled [9]. Tracking System. A shipment tracking system is integrated into a general cargo warehouse in order to search for the location of a package, document existing damage and enhance security. Video technology is used to monitor every corner in the warehouse, positioning technology to locate the barcode scanners and software technology to connect all the data with each other. The data is used for investigations and searches if a problem is subsequently noted in the logistical process. Tracking systems do not report damage in real time. Evaluation of Monitoring Procedures. The whereabouts of the package, plus the specific quality of the package, are known at the I-point or measuring points at a certain point in time. However, the route the package takes across the warehouse afterwards is unknown. Warehouse inspections and inventories are time-consuming and tie up warehouse staff. If tags or trackers were attached to the pallets in the pallet exchange system, these location tags would also leave the warehouse. The tags would have to be attached at goods receipt and removed again at goods issue because the acquisition costs of the tags would otherwise be lost. This requires additional manual work, which is a hindrance to the logistical process. If a complaint has been made, the searches in the tracking system are carried out retrospectively in the past. The tracking system does not report problems in real time. In summary, it can be stated that the existing monitoring procedures, in part, monitor the location and quality of the package at certain points in the warehouse. However, there is no real-time monitoring of the packages on their way through the warehouse. Knowledge of the route through the warehouse is important, however, in order to document any damage that may occur. Even the putting down of the package at an unplanned bin location can only be detected if the route through the warehouse is known. Given the lack of information on
434
A. Neubert
the packages’ path in the warehouse, the existing monitoring procedures are therefore not sufficient.
3 3.1
Method Monitoring of Logistical Processes
Ten Hompel defines monitoring as the technical recording and processing of system and process states [4]. An IT monitoring system monitors the IT infrastructure and processes using metrics. If metrics are defined and supplied for logistical processes, the metrics can be collected and evaluated by an IT monitoring system. The monitoring of logistical processes can thus be realized by using a monitoring system for IT. According to Kramm, the requirements for a monitoring system for IT include surveillance of the status of the system, alerting, diagnosis, quality measurement and monitoring of standardized configurations [10]. The monitoring system is expected to provide conclusions or concrete statements on the reliability of systems and components, in addition to the detection of errors [10]. To make this possible, the data must be saved for later evaluation. Here the storage of the data in a database is the obvious solution. Current data from the warehouse is needed for real-time monitoring in order to be able to evaluate the data. A shipment tracking system supplies the current data of the logistical processes in the warehouses and is to be used as the basis for real-time monitoring. It thus supports real-time monitoring by providing the current data and metrics that arrive continuously at the system over time. The result is a sequence of data values in chronological order, the so-called time series. 3.2
Database for Storing Time Series
Different types of databases can be used to store time series in a database. In Relational Database Management Systems (RDBMS), data is stored in twodimensional tables. The tables consist of a fixed number of columns with a fixed data type [11]. Examples of RDBMS include Microsoft SQL Server, MySql, Oracle, PostgreSQL. An advantage is the structured storage of data. A disadvantage lies in the fact that the scheme must be clear right from the very beginning. The duration for inserts or selects is extended above a certain size. Using a Round Robin Database (RRD), data is written by means of the round robin procedure. Once the available memory space has been used up, old data will be overwritten. An example is the RRDtool. The required storage space is fixed from the outset and does not change. This constitutes an advantage. Disadvantages of the RRDtool are as follows: (a) it can only store numeric values, (b) it stores data at constant time intervals, and (c) the average calculation poses problems for the RRDtool. A Time Series Database (TSDB) is a database optimized for time series management [12]. Amazon Timestream, InfluxDB, OpenTSDB, Prometheus are examples of TSDB. An advantage is the storage of numerical data and character strings for any points in time. A disadvantage is that Join is only possible via the time column.
Clustering for Monitoring Logistical Processes
3.3
435
Clustering
Everitt defines clustering as follows: “Given a number of objects or individuals, each of which is described by a set of numerical measures, devise a classification scheme for grouping the objects into a number of classes such that objects within classes are similar in some respect and unlike those from other classes.” [13] There are different clustering algorithms mentioned in the scientific literature. K-Means [14], Mean shift [15], Expectation Maximization (EM) Clustering with Gaussian Mixed Models (GMM) [16], Hierarchical Clustering [17], and DensityBased Spatial Clustering of Applications with Noise (DBSCAN) [18] are examples of clustering algorithms. Since the number of clusters in the warehouse is unknown, algorithms in which the number of clusters must be defined in advance are not suitable. Consequently, K-Means and EM clustering are not suitable. The Mean-Shift algorithm is not selected because it is not easy to define the radius. The Hierachical Clustering algorithm has the disadvantage of the long computing time. It is also difficult to choose a meaningful hierarchical level. This leaves the DBSCAN algorithm, which is also suitable for outlier detection in addition to clustering. Each point in the dataset is evaluated to determine whether it can be assigned to a cluster. A point that cannot be assigned to a cluster is marked as “noise”. The noise points are the outliers at the end of the algorithm. 3.4
Experimental Structure and Procedure
Research Data. The data used come from the Visual Location Management System (VLMS) [19] shipment tracking system (see Sect. 2.1), which is integrated into a general cargo warehouse. The tracking and tracing system obtains and archives tracking positions of the scanners, logistical data such as scanned barcodes and video data. The localization positions and the scanned barcodes are used for the examination. The scanned barcodes are then specified in the payload. The data was stored in a TSDB (see Sect. 3.2). The TICKstack [20] of Influx Data was used as TSDB. Clustering Algorithm. The DBSCAN algorithm (see Sect. 3.3) was used for clustering. The DBSCAN algorithm uses MinPts and Epsilon as parameters. Epsilon is the value for the distance used in order to determine the neighbouring points of the examined data point. The MinPts value indicates the minimum number of neighbouring points that must be reached in order for the examined point to become the first point of the cluster. The MinPts ranges between 3 and 10. Epsilon is determined for every number of MinPts. The number of the resulting cluster is documented. The optimum pair of MinPts and Epsilon is selected from that table. Ester et al. proposed a procedure to establish Epsilon that uses the calculation of kNNdist [18]. Kassambara [21] described how to use the data mining tool R [22] in order to determine Epsilon by means of the
436
A. Neubert
kNNdistplot diagram of the package dbscan. R is used in this paper to apply the DBSCAN algorithm in line with Kassambra. The package influxdbr was employed to access the InfluxDB database. The data points were filtered in such a way that only those positions in the general cargo warehouse where a barcode was actually scanned were used for the DBSCAN algorithm. Expenses. The used DBSCAN algorithm supplies 4 lists as a result, with one of the lists describing the assignment of the data points to the respective cluster. The collected data and the calculated clusters were graphically visualized. The diagrams were created using the function plot of R and the function fviz cluster of package factoextra.
4
Results
The results of two consecutive hours are described below. 4.1
Experiment 1: DBSCAN, raw data 2, 00:00–01:00 a.m.
Within the time period of 00:00–01:00 a.m., 495 data points were archived for which a bar code was scanned. Figure 1 shows the distribution of data points in the general cargo warehouse. Determining MinPts and Epsilon. Table 1 shows the sequence of MinPts from 3 to 10. The value of Epsilon was identified for each of the MinPts. The tables shows the resulting number of clusters. The higher the MinPts the lower is the number of the cluster. Here, the selection of MinPts = 8 with Epsilon = 5.3 is a reliable method in order to obtain 7 clusters as a result. Nonetheless, the reduction of MinPts or the magnification of MinPts results in 7 clusters. Figure 2 sets out the result of the DBSCAN algorithm with MinPts = 8 and Epsilon = 5.3. The X and Y axes are normalized. The DBSCAN algorithm detects outliers. The outliers are shown as black dots in Fig. 2. The dots do not belong to any cluster. Table 1. Experiment 1
Table 2. Experiment 2
MinPts Epsilon Cluster
MinPts Epsilon Cluster
3 4 5 6 7 8 9 10
3 4 5 6 7 8 9 10
3.8 4.0 4.0 4.9 5.0 5.3 5.6 5.9
16 13 11 8 7 7 7 8
2.75 2.90 3.05 3.10 3.30 3.60 3.90 4.10
6 5 5 5 5 5 5 5
Clustering for Monitoring Logistical Processes
437
50 y [m]
40 30 20 10 0
20
40
60 80 x [m]
100 120
Fig. 1. Plot raw data 2, 00:00–01:00 a.m., barcode scans
y [norm.]
1 0 -1 -2 -2
-1
0 x [norm.]
1
2
Fig. 2. DBSCAN raw data 2, 00:00–01:00 a.m., eps = 5.3, MinPts = 8
4.2
Experiment 2: DBSCAN, raw data 2, 01:00–02:00 a.m.
Within the time period of 01:00–02:00 a.m., 271 data points were archived for which a bar code was scanned. Figure 3 shows the distribution of data points in the general cargo warehouse. Determining MinPts and Epsilon. Table 2 shows the sequence of MinPts from 3 to 10. Here, the selection of MinPts = 8 with Epsilon = 3.6 is a reliable method in order to obtain 5 clusters as a result. Figure 4 shows the result of the DBSCAN algorithm with MinPts = 8 and Epsilon = 3.6.
5
Discussion
The study investigated whether clustering can support monitoring of logistical processes in general cargo warehouse. For this purpose, the DBSCAN clustering algorithm was applied to logistical data. It turned out that clustering provides
438
A. Neubert
50
y [m]
40 30 20 10 40
60 x [m]
100
80
Fig. 3. Plot raw data 2, 01:00–02:00 a.m., barcode scans
y [norm.]
2 1 0 -1 -2
-1
0 x [norm.]
1
Fig. 4. DBSCAN raw data 2, 01:00–02:00 a.m., eps = 3.6, MinPts = 8
the necessary storage locations for the unloading and loading of trucks. This knowledge can be used for monitoring, since barcode scans can be compared with the locations for recurring orders. If the locations of the scans show deviations from the storage bins detected by clustering, this deviation from the logistical process can be reported to the forklift driver. The manual determination of the parameters Epsilon and MinPts of the DBSCAN algorithm is a weakness. Further research has to be done for an automatic determination of Epsilon and MinPts in order to integrate the DBSCAN algorithm into the warehouse. Another weakness of the study is the assumption that the loading and unloading processes should be repetitive. If there is no change in the storage bins in the warehouse the handling processes will indeed be repeated. The storage bins in a warehouse are not replanned on a daily, but rather on a monthly or quarterly basis. The storage bins remain in place for at least one month. The assumption of recurring processes can therefore be made. The research question Q1 can be answered as follows: The DBSCAN clustering algorithm can be used to support both a consignment tracking system and a
Clustering for Monitoring Logistical Processes
439
real-time monitoring system in a general cargo warehouse to monitor logistical processes. If packages have been put in the wrong storage bin, this can be remedied by detecting and reporting scans outside the usual storage locations. Long search times for packages at goods issue would thus be avoided. The DBSCAN clustering algorithm can also be used to detect and delete outliers in the data records so as to prevent a distortion of further calculations.
6
Conclusions
The article analyses and evaluates the methods currently used to monitor logistical processes in a general cargo warehouse. Since these methods fail to continuously monitor the actual status, this article presents a new method for monitoring logistical processes, which was derived from a monitoring system for IT infrastructures. The required time series and their storage in a TSDB were described. The data mining method DBSCAN for clustering was applied to data of a tracking system and documented. The outcome is that DBSCAN can be used to identify slots where transport orders are executed. The cluster formation and the slots in question change on an hourly basis. If recurring route planning is carried out in warehouses over time, the detected clusters can be used for a real-time monitoring function that generates alerts in the event of incorrect stacking or incorrect loading. If the warehouse clerk or forklift driver scans a package outside a cluster area, a warning can be displayed to the clerk or forklift driver on the MDE device. In addition, the information about cluster assignment can be used as an attribute for other algorithms, such as classification. Further research work to be conducted includes the automatic determination of the parameters Epsilon and MinPts of the DBSCAN algorithm.
References 1. Martin, H.: Transport- und Lagerlogistik: Systematik, Planung, Einsatz und Wirtschaftlichkeit, E-Book, pp. 59–60. Springer, Wiesbaden (2016) 2. Klaus, P., Krieger, W., Krupp, M.: Gabler Lexikon Logistik: Management logistischer Netzwerke und Fl¨ usse., p. 545. Gabler Verlag, Wiesbaden (2012) 3. Bichler, K., Krohn, R., Philippi, P., Schneidereit, F.: Kompakt-Lexikon Logistik: 2.250 Begriffe nachschlagen, verstehen, anwenden. Springer, Wiesbaden (2017) 4. Ten Hompel, M., Heidenblut, V.: Taschenlexikon Logistik: Abk¨ urzungen, Definitionen und Erl¨ auterungen der wichtigsten Begriffe aus Materialfluss und Logistik. Springer, Heidelberg (2011) 5. Dantzer, H.: Techniken der Qualit¨ atssicherung im Lagerwesen und G¨ uterversand: Klimaschutz – Korrosions- und Feuchtigkeitsschutz – K¨ uhl- und Tiefk¨ uhl-Logistik, 1. Auflage, p. 104. expert (1995) 6. DIN EN 15635:2009-08: Steel static storage systems - Application and maintenance of storage equipment (2009)
440
A. Neubert
7. von Janczewski, B.: Datengold der Logistik: Neuer 5G-kompatibler Tracker macht Paletten intelligent, 35. Deutscher Logistik-Kongress, Fraunhofer-Institut f¨ ur Materialfluss und Logistik, Presse und Medien, Pressemitteilung/15.10.2018 (2018). https://www.iml.fraunhofer.de/de/presse medien/pressemitteilungen/ Live-Palettentracking DLK.html 8. NARROWBAND IoT - Bahnbrechend f¨ ur das Internet der Dinge, Whitepaper, Deutsche Telekom AG, Oktober 2017 (2017) 9. Maresch, R.: Digitale Logistik spart Kosten - Fraunhofer IFF auf Deutschem Logistik Kongress, Fraunhofer-Institut f¨ ur Fabrikbetrieb und -automatisierung IFF, Fraunhofer-Institut f¨ ur Fabrikbetrieb und -automatisierung IFF, Presse und ¨ Offentlichkeitsarbeit, 16.10.2018 (2018). https://idw-online.de/en/news?print=1& id=704094 10. Kramm, Th: Monitoring mit Zabbix, Das Praxishandbuch, Grundlagen, Skalierung, Tuning, Erweiterung, pp. 6–7. dpunkt.verlag, Heidelberg (2016) 11. Freiknecht, J., Papp, S.: Big Data in der Praxis: L¨ osungen mit Hadoop, Spark, HBase und Hive: Daten speichern, aufbereiten, visualisieren, Carl Hanser Verlag M¨ unchen (2018) 12. Time Series DBMS - DB-Engines Encyclopedia. https://db-engines.com/en/ article/Time+Series+DBMS 13. Everitt, B.: Cluster Analysis, 2nd edn, p. 1. Heinemann Educational Publishers, London (1980) 14. MacQueen, J.B.: Some methods for classification and analysis of multivariate observations. In: Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability, pp. 281–297. University of California Press, Berkeley (1967) 15. Wu, K.-L., Yang, M.-S.: Mean shift-based clustering. Pattern Recognit. 40(11), 3035–3052 (2007) 16. Gupta, M.R., Chen, Y.: Theory and use of the EM algorithm. Found. Trends(R) Sig. Process. 4(3), 223–296 (2011) 17. Gordon, A.D.: A review of hierarchical classification. J. R. Stat. Soc. Ser. A Gen. 150(2), 119–137 (1987). Wiley for the Royal Statistical Society 18. Ester, M., Kriegel, H.-P., Sander, J., Xu, X.: A density-based algorithm for discovering clusters in large spatial databases with noise. In: Proceedings of 2nd International Conference on Knowledge Discovery and Data Mining (KDD 1996), pp. 226–231 (1996) 19. Brochure VLS, Die Software, p. 2. https://v-l-s.com/wp-content/uploads/2015/ 10/vls software broschuere-deu.pdf 20. InfluxData: Open Source Time Series Platform—InfluxData. https://www. influxdata.com/time-series-platform/ 21. Kassambara, A.: DBSCAN: density-based clustering essentials - Datanovia. https://www.datanovia.com/en/lessons/dbscan-density-based-clusteringessentials/ 22. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria (2019). https://www.R-project. org/
Using RFID to Monitor the Curing of Aramid Fiber Reinforced Polymers Marius Veigt1(&), Marco Cen2, Elisabeth Hardi3, Walter Lang2, and Michael Freitag1,4 1
4
BIBA - Bremer Institut für Produktion und Logistik GmbH at the University of Bremen, Bremen, Germany {vei,fre}@biba.uni-bremen.de 2 IMSAS - Institut für Mikrosensoren, -aktoren und -systeme at the University of Bremen, Bremen, Germany {MCen,WLang}@imsas.uni-bremen.de 3 Faserinstitut Bremen e.V., Bremen, Germany [email protected] Faculty of Production Engineering, University of Bremen, Bremen, Germany
Abstract. In the production of fiber-reinforced composites, an increase of resource efficiency is necessary to ensure the future competitiveness of this production technology in Germany. Therefore, wireless sensor methods are currently being researched to monitor the curing of fiber-reinforced composites online and in situ during the production process. At the same time, radio frequency identification (RFID) technology is being used to identify fiberreinforced composite components in production and logistics. In this paper, we present results to combine these two approaches to monitor the cure of aramid and carbon fiber-reinforced composites in situ and online using RFID technology. By experimenting, we have determined a relationship between the received signal strength indicator (RSSI) of the RFID transponder integrated into an aramid fiber-reinforced composite component and the curing of this component. As comparison method for the verification of the curing processes, we used the dielectric analysis as a well-known method for cure monitoring. The results show that it is possible to monitor the process of aramid fiber-reinforced composites through the RSSI value. However, the so-called curing transponder does not work with carbon fiber-reinforced composites. To complement the research, we proposed a surface modification method for RFID tags, based on a preliminary study on the surface activation of polymer films using oxygen plasma. Keywords: Fiber reinforced polymers activation
RFID Cure monitoring Surface
1 Introduction 1.1
Motivation
The integration of fibers, e.g. glass or carbon fibers, into polymers leads to very advantageous material properties. In particular, a high tensile and compressive strength © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 441–450, 2020. https://doi.org/10.1007/978-3-030-44783-0_42
442
M. Veigt et al.
as well as a high rigidity combined with low weight, which are reasons for the high potential of these materials for lightweight constructions [1]. It is essential to cure the components in a targeted manner to achieve these material properties. Therefore, cure monitoring is of central importance. Due to this, many research activities are currently ongoing [2–5]. However, there is still no wireless measurement method that can be used for cure monitoring as wells as in the subsequent processes of the product life cycle. 1.2
State of the Art
For cure monitoring, the dielectric analysis (DEA) is widely used. The method can detect all transition phases and critical points during curing [6]. However, a disadvantage of this method is that the used sensors affect the mechanical properties of the composites negatively [7]. Due to this, the sensors will be usually removed after the cure monitoring, and they have no further benefit during the product life cycle [8, 9]. Furthermore, the sensors are wired. Cabling causes installation effort, and it is also a potential weak point in the vacuum build-up, allowing air to penetrate the composite, which has another negative effect on the material properties [6]. In order to circumvent this vulnerability, wireless methods are used. Especially ultrasound has gained acceptance. However, this method requires a high degree of implementation. Due to disturbances such as temperature, pressure, and variability of the sensors, the measuring instruments must be calibrated elaborately [10]. Therefore, this measurement method is relatively susceptible and must be set up by specialists. This causes additional costs without the equipment can be reused in the further product life of the composites [8, 9]. An alternative are fiber Bragg sensors [8], which are used to monitor most phases during curing and can remain in the component for use in structural health monitoring applications [9]. However, these sensors are considered to be fragile and susceptible to mechanical stress, so the probability of damaging a sensor is high. This results in increased costs and inefficiency [7, 8]. Even if further developments remedy these weaknesses, this method is still criticized for the fact that these sensors, as well as the sensors used in DEA, have to be wired tightly, which also leads to effort and potential weak points in the vacuum build-up [6]. Wireless and chipless sensors, which can be integrated into the component, are currently being researched. During the cure of a composite, these sensors shift their response frequency at 25 GHz resulting from the dielectric permittivity change of the fiber composites [3]. However, these sensors also have to be removed from the components after curing, and thus, they are useless in the further product life of the component. On the other hand, the contact area between fiber-reinforced polymer and a hosted sensor is known as interphase and plays an essential role during the transfer of mechanical loads [11]. A range of different treatments, including wet chemistry, UV, and plasma among others [12] can be used to modify the surface of the sensor and improve its integration within the fiber composite. Plasma treatments can easily graft molecules on polymer surfaces, which can increase their adhesion to other materials [13, 14] by chemical reactions of such molecules [15, 16].
Using RFID to Monitor the Curing of Aramid Fiber Reinforced Polymers
443
It can be concluded that a cure monitoring method is being sought that allows the sensors to remain in the composite without adversely affecting the mechanical properties of the host material and provides a benefit in the further product life cycle. The sensors should also be able to be read out wirelessly and should be as low cost as possible. 1.3
Research Approach
The possibility to integrate RFID transponders into fiber composite without affecting the component properties was developed by Zettler et al. [17] and Gray [18]. The use of RFID transponders offers several potential benefits along the product life cycle of the composites, e. g. in production planning and control [19, 20], Track & Trace within the supply chain [19, 21, 22] and for component authentication as an intervention against product piracy [23]. In addition, the RFID transponders are readout wirelessly, and a wide range of transponders can be purchased at a low price. Veigt [5, 24] and Hardi [25] have shown that RFID transponders can be used for cure monitoring of glass fiber reinforced plastics. They exploit the changes in the dielectric permittivity of the fiber composites to measure an effect on the transmission behavior of the RFID transponders integrated into the composites. They discovered a correlation between the signal strength and the curing progress of the glass fiber reinforced plastics. However, until now, there is no investigation if RFID transponders are suitable to monitor the curing of composites with carbon or aramid fibers. In the following, experiments will be presented to investigate if this effect can be observed at carbon and aramid fiber reinforced plastics as well. The dielectric properties of aramid fibers are similar to those of glass fibers. Therefore, the first thesis to be examined is: Into aramid fiber reinforced plastics integrated RFID transponder can be used to monitor the curing process of the composite. The dielectric properties of carbon fibers are contrary to those of glass fibers. It is known that carbon blocks the radio waves of RFID [26]. Hence, in case of carbon fibers, the RFID transponder will be placed on top of the composite. The second thesis to be examined is: The contract of the RFID transponder with the resin is sufficient to measure the RSSI change during the curing process of the carbon composite. One of the goals during the integration of sensors into fiber-reinforced composites is to obtain a proper bonding between them to minimize the downgrading of the mechanical properties of the host composite. A potential method to achieve this is by surface modification of the embedded sensors. For this purpose, oxygen plasma can be used to graft hydroxyl groups on fibers and polymers [13, 14], which can lead to a chemical bond with the epoxy resin of the target fiber-reinforced composite. Considering the latter, oxygen plasma treatments were performed on polymer films in order to activate their surface. Those samples were characterized to measure changes in their surface wettability. The films were made from Polyimide, which is material commonly used as a substrate for electronic devices [7, 27]. The general idea of the research approach is presented in the scheme of Fig. 1.
444
M. Veigt et al.
Wireless communication
RFID Transponder
Bonding to composite Fiber composite Fig. 1. Schema of the proposed approach for integration of cure monitoring sensor.
2 Measurement Background 2.1
Dielectric Analysis (DEA)
A widely used method to characterize the polymeric resins during the curing process is the DEA. Free ions and dipoles are existing within the resin and cause the dielectric behavior of it. The free ions can move in the uncured resin and thus dominate the dielectric response. Ion migration is reduced to nearly zero as soon as the resin is cured. In the cured state, the dielectric response is dominated by the alignment of the dipoles, which, in contrast to the migration of ions, is a frequency-dependent phenomenon [28]. Thus, the complex electrical impedance can be used to monitor the curing of thermoset resins or adhesives, as explained by Mijovic et al. in [29]. The frequency-independent resistivity or direct current resistivity (also called ion viscosity) correlates with the cure state throughout cure. Interdigitated electrodes on a substrate in contact with the material under test can be used as a sensor for studying the dielectric properties and for observation of cure state of the material under test. 2.2
Radio Frequency Identification (RFID)
An RFID system consists of two components. The transponder is attached to an object to be identified. It usually consists of an antenna and an electronic microchip. The reader (and writer) consists of a high-frequency module (transmitter and receiver), a control unit, and one or more antennas. The reader provides energy without wire contact to the transponders [30]. A minimum field strength arriving at the transponder is necessary to activate the transponder and supply it with sufficient energy for the operation. This arriving field strength E depends on the transmission power of the reader PEIRP, the magnetic permeability µ and the electric permittivity e of the transmission medium as well as the distance r between the reader antenna and the transponder antenna, see Eq. (1) [30]: rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffi PEIRP l e E¼ 4pr 2
ð1Þ
Using RFID to Monitor the Curing of Aramid Fiber Reinforced Polymers
445
The strength of the received signal, also called backscatter signal, can be measured using the received signal strength indication (RSSI). It is just an indicator because interferences, such as other radio waves and reflections, affect the measurement. Nevertheless, the RSSI is used in locating application [31, 32]. RFID systems exist in many different variants. An essential distinguishing feature is the operating frequency. For low frequency (LF) systems, the frequency lies between 30 kHz and 300 kHz, for high frequency (HF) systems between 3 MHz and 30 MHz, for ultra-high frequency (UHF) systems between 300 MHz and 3 GHz and for microwave above 3 GHz. In the case of application of production planning and control as well of logistics, the UHF system has gained wide acceptance. The European frequency band for UHF systems is regulated from 865 MHz to 868 MHz [30].
3 Material 3.1
Integration of RFID Transponders Within Composites
For the experiment, a hand laminate with 16 layers of carbon fiber fabrics and another hand laminate with 16 layers of aramid fiber fabrics was build-up. The EPIKOTE resin L20 and curing agent H531 were used to cure the composites at room temperature. Three RFID transponders were centrally integrated into the aramid composite (between layers 8 and 9) and another three transponders on top of the carbon composite, see Fig. 2. The RFID transponder PROtag 3 mini from tagItron was used. It is an UHF Gen 2-compliant transponder so that it is used as a standardized transponder, e.g., in logistics. It is designed for communications from 860 MHz to 960 MHz, and it is available at reasonable prices ( 0 and the number of process stations mps > 0 defines the size of the output layer. The number of hidden layers nh > 0 and their sizes mh > 0 are set by the user. The lambda and the first hidden layer as well as each hidden and its succeeding layer are densely connected. The input-fingerprint vector consists of received signal strength indicator (RSSI)-values [17]. The higher an RSSI, the more power is present in a received signal. In our case, the RSSI ranges from −99 to −44 dBm. The microcontroller listens to the channels 1, 6, 13 and, for each monitored access point, writes the average RSSI of up to 30 beacon frames into the fingerprint. If no beacon frames are received from an access point the average RSSI is set to −99 dBm. The preprocessing function of the input in the lambda layer is defined as λ(x) := λF · (x + λS ) ∀ x ∈ R with parameters λF , λS ∈ R.
(1)
In the hidden layers rectifier and in the output layer softmax are the activation functions, r(x) := rectifier(x)
= max(0, x)
s(x, i) := softmax(x, i) =
exp(xi ) j=1,...,n exp(xj )
∀ x ∈ R,
(2)
∀ x ∈ Rn , i ∈ {1, . . . , n},
(3)
so that the i-th entry of the output vector defines the probability that the fingerprint was taken in station i. The input Σ to the first hidden and, in the case of nh = 1 hidden layer, Σ to the output layer satisfy mh map Σ = bk + wk, λ(f ) , (4) =1 k=1 mps mh = bm + Σ w m,k rectifier(Σk ) , (5) k=1
m=1
with weights w·,· , w ·,· ∈ R and bias b· , b· ∈ R. Let X define the union of all weights and bias and let xmin , xmax ∈ X denote their minimum and maximum. To save memory, a quantization technique maps = {0, 1, . . . , n every floating point x ∈ X to an integer x (x) ∈ X } satisfying 2cw xmax − xmin ∈ 0, ¯ · xs −x| , where xs := , x (x) = arg minx¯∈X |xmin + x n n (6) where the euclidean norm of weight and bias vectors is by constraint smaller , the smaller the or equal to parameter cw > 0. The smaller cw and greater n approximation error. The required memory capacity is further reduced by converting the quantized values, that consist of one- to four-digit base 10-numbers for n = 832 −1 = 6888, to two-digit base 83-numbers and encoding them by the ASCII-characters 40 to 122, where character 92 (‘\’) is replaced by character 37. Consequently, separators between weights and bias become unnecessary.
484
M. Redeker and A. Wagenitz
Table 2. The number of station-wise distinct training and validation fingerprints and an analysis of the strengths of the received signals in these fingerprints.
The categorical cross-entropy loss function is minimized by supervised learning stochastic gradient descent (SGD) with a constant learning rate 0.01. This iterative process adapts the network’s weights and bias, which are initialized by samples randomly drawn from a uniform distribution within [−0.05, 0.05] or [0, 0.05]. The training data is shuffled before the single iteration over the entire training data set and per gradient update 32 training data are used. The overall network-string forwarded from the central unit to the products by MQTT messaging protocol reads “λF $λS $xmin $xs ” plus, for every layer following the lambda layer, “!activation-function-code$input-dim$outputdim$weights-string$bias-string” with separators “$” and “!”. The products decode the string, update the positioning network and use it to determine their position and to monitor their individual production schedule.
3
A Test in the GME’s Production Process
A test was conducted between 10 April and 08 May 2019 in the GME’s production process. The manufacturing stations S1 , . . . , S6 were equipped with three microcontrollers positioned at one, three and five sixths of their length and two microcontrollers covered the final station Sf at one and three fourth of its length. These twenty stationary microcontrollers took fingerprints during working times, i.e. usually between 07:00 and 19:00 from Monday to Friday, and forwarded them to the central unit for the training and validation of neural positioning networks. Each fingerprint consisted of the receiving strengths of map = 9 monitored access points of the shop floor wireless LAN-infrastructure. During working hours on 08 May four additional microcontrollers were moved through the production process, took fingerprints and predicted their position using one trained positioning network. Their fingerprints are also used as test data by the central unit. The central unit uses the station-wise distinct fingerprints from the stationary microcontrollers to train and validate neural positioning networks. In each case, 80% randomly selected data sets are used for training and the remainder for validation. Table 3 shows the positioning accuracy and the prediction probability for the validation data in dependence on the choice of the learning and
An Edge Neural Cyber-Physical Production System
485
Table 3. Categorical accuracy – the percentage of correct predictions – and mean prediction probability – the mean of the maxima of the output vectors – for the validation data for 100 networks per parameter set each with randomly initialized weights and bias: mean and standard deviation at the end of the networks’ training as well as an indicator rating their mean learning, which equals the mean of either categorical accuracy or mean prediction probability determined after 1, 2, . . . , 100% of training: the steeper the mean learning curve the closer to 100 is the indicator. Highlighted in gray are the parameters of the Benchmark network. In each parameter category, the parameters of the Benchmark network and the other networks equal except for the category’s parameter.
486
M. Redeker and A. Wagenitz
network parameters. The Benchmark network is trained and validated with the 46391 distinct data sets from the period 24 April to 7 May (see Table 2), it is optimized by stochastic gradient descent, its λ-function is defined as λ(x) := 1(x + 88.74), where −88.74 is the training data’s mean RSSI, it contains one hidden layer with 15 nodes and with the rectifier activation function, bias are not used, no weight is negative and the euclidean norm of the weight vectors does not exceed one. The Benchmark network’s parameters were chosen because of the resulting high positioning accuracy and prediction probability for the validation data, the high execution speed and memory economy in the positioning phase, due to rectifier activation in the hidden layer, the small number of hidden nodes and no bias; because of the high quantization weight approximation, due to the fact that the minimum and maximum weight are greater or equal to zero and less or equal to one; and, finally, because of the steep learning curves. For training, validation and test data the maximum relative output error caused by the weight
Fig. 3. Mean learning curves of the categorical accuracy and of the mean prediction probability for the training (solid darkgray), validation (dashed black) and test data for 100 Benchmark networks: values determined after 1, 2, . . . , 100% of training. The horizontal axis represents the percentage of training, the vertical axis either categorical accuracy (a) or mean prediction probability (b).
Fig. 4. Mean learning curves of the categorical accuracy for the training (solid darkgray), validation (dashed black) and test data for 100 networks with training and validation data from the working time in the period 24 April to 08 May. The axis are defined as in Fig. 3(a).
An Edge Neural Cyber-Physical Production System
487
Fig. 5. Positioning predictions for the test data from the four additional microcontrollers on 08 May.
approximation is 0.0022. The number of nodes in its one hidden layer could have been smaller, but in order to be prepared for a possible installation of additional access points or more complex data, 15 hidden nodes seemed to be reasonable. Figure 3 shows that the learning curves of the positioning accuracy and prediction probability for training and validation data are very similar. Contrarily, after twenty percent of training these curves are constantly above the curves for the test data, which was generated by the four additional microcontrollers on 08 May. Since this difference is also present in the learning curves when training and validation data from 08 May is added, as depicted in Fig. 4, the absence of proper training data is most likely not the cause. It seems either be caused by the fact that, firstly, training and validation data on the one hand and test data on the other are generated by different, not handpicked microcontrollers or, secondly, the microcontrollers generating training and test data are stationary while the ones generating test data are moved from station to station. Before the begin of work on 08 May the central unit initialized 100 Benchmark networks with random weights and trained and validated each of them with training and validation data from 24 April to 07 May, where the training of one network executed by two cores running at 2.30 GHz takes only five seconds. It selected the one Benchmark network with the maximum sum of the learning indicators for categorical accuracy and prediction probability for the validation data and forwarded it to the four additional microcontrollers, which passed on this day during working hours through the production process in the sequence S1 → S2 → S3 → S4 → S5 → S6 → Sf . These microcontrollers repeatedly took fingerprints and identified 94.59% of their positions correctly, see Fig. 5. Their average prediction probability was 90.97%.
488
4
M. Redeker and A. Wagenitz
Conclusion
This article presented an edge neural cyber-physical production monitoring system in a German medium-sized enterprise’s production process, that was not fit for providing data of the production progress. For retrofitting, each product is equipped with a battery-powered, cost-efficient and not handpicked wireless LAN-microcontroller that monitors its product’s production progress autonomously by taking fingerprints of the strengths of received signals of the existing, unmodified shop floor wireless LAN-infrastructure, by inputting these fingerprints into a neural positioning network and predicting the station in which its product is momentarily being processed, and by comparing predicted to planned stations of its product’s individual production schedule. In the event a microcontroller detects a significant plan-deviation, it sends a warning message to its product’s agent in an agent-based planning system and recommends an adaptation of the production schedule. In combination the monitoring system, the corresponding agent in the planning system and the product’s individual production schedule form a digital twin of the product. While its production schedule is passive data, its monitoring system and planning agent are active routines. The system creates transparency about the observance of planned schedules, which enables the planning system to adapt schedules at all times in full knowledge of the current production status. It functions on its own with regular bidirectional communication of the agents with an ERP-system or an MES. Its integration into one of these systems might ease these interactions. On the other hand this might complicate the interactions of agent and monitoring system. Furthermore, the decentralized evaluation of the sensors’ data minimizes the necessary data exchange and prevents unnecessary adaptations of the schedule. The results of the presented test, which was performed during working hours, prove, as the process stations were identified correctly in 95% of cases, the system’s functionality. To further improve the positioning precision, a microcontroller conducts a plausibility check in case of a predicted station-deviation or if a prediction probability is below 50%. For example, it is implausible that a step of a planned schedule is skipped. If a check fails, the prediction is repeated once and then accepted without checking the plausibility again. Additional test results indicate that the minimum distance between individual process stations should be at least five to six meters. This relatively weak precision is the result of the approach of not making high quality demands on the wireless LAN-infrastructure and the microcontrollers used. In case of smaller distances UWB-positioning techniques can be added to the system. In its current state, however, it is sufficiently precise for the enterprise’s production process. The positioning system is very memory efficient using a neural network that consists of nine input and lambda, fifteen hidden and seven output nodes and whose structure and parameters can be encoded in an ASCII-string of about 500 digits. Beyond that, it quickly converts fingerprints into process station predictions – needed are only about 250 additions and multiplications each as well as 22 evaluations of the rectifier and exponential function.
An Edge Neural Cyber-Physical Production System
489
The knowledge about logistic processes can be encoded in microcontrollers. Microcontrollers can use neural networks for predictions. These qualities allow the products’ microcontrollers to autonomously determine the individual process stations in which their products are being processed and, whenever they deem it necessary, to recommend the adaptation of the planned production schedule. Acknowledgment. We would like to thank the German Bundesministerium f¨ ur Wirtschaft und Energie (BMWi) for financial support under grant 16KN069721 of the ZIM-project InPro - Agent.Pro/Methode und Hardware zur Lokalisierung des Werkst¨ ucks und des Agenten, Dr. Andreas Baar and his team from innos Sperlich GmbH for their support within the Netzwerk f¨ ur intelligente Produktionstechnologien (InPro), our project partners A&T Solution GmbH and Helmut-SchmidtUniversit¨ at/Universit¨ at der Bundeswehr Hamburg as well as Jan Fischer and our colleagues from HAW Hamburg’s Business Innovation Lab.
References 1. Chen, B., Wan, J., Shu, L., Li, P., Mukherjee, M., Yin, B.: Smart factory of industry 4.0: key technologies, application case, and challenges. IEEE Access 6, 6505–6519 (2017). https://doi.org/10.1109/ACCESS.2017.2783682 2. Zuehlke, D.: Smart factory - towards a factory-of-things. Ann. Rev. Control 34(1), 129–138 (2010). https://doi.org/10.1016/j.arcontrol.2010.02.008 3. Hofmann, E., R¨ usch, M.: Industry 4.0 and the current status as well as future prospects on logistics. Comput. Ind. 89, 23–34 (2017). https://doi.org/10.1016/j. compind.2017.04.002 4. Zhang, Y., Qian, C., Lv, J., Liu, Y.: Agent and cyber-physical system based selforganizing and self-adaptive intelligent shopfloor. IEEE Trans. Ind. Inform. 13(2), 737–747 (2017). https://doi.org/10.1109/TII.2016.2618892 5. Mourtzis, D., Vlachou, E.: A cloud-based cyber-physical system for adaptive shopfloor scheduling and condition-based maintenance. J. Manuf. Syst. 47, 179–198 (2018). https://doi.org/10.1016/j.jmsy.2018.05.008 6. Bracht, U., Geckler, D., Wenzel, S.: Digitale Fabrik. Springer, Heidelberg (2018). https://doi.org/10.1007/978-3-662-55783-9 7. Vogel-Heuser, B., Bauernhansl, T., ten Hompel, M.: Handbuch Industrie 4.0 Bd.1 Produktion. Springer, Heidelberg (2017). https://doi.org/10.1007/978-3-66245279-0 8. (Reza) Zekavat, S.A., Buehrer, R.M.: Handbook of Position Location: Theory, Practice, and Advances. Wiley-IEEE Press (2019) 9. Negri, E., Fumagalli, L., Macchi, M.: A review of the roles of digital twin in CPSbased production systems. Procedia Manuf. 11, 939–948 (2017). https://doi.org/ 10.1016/j.promfg.2017.07.198 10. Rosen, R., von Wichert, G., Lo, G., Bettenhausen, K.D.: About the importance of autonomy and digital twins for the future of manufacturing. IFAC-PapersOnLine 48(3), 567–572 (2015). https://doi.org/10.1016/j.ifacol.2015.06.141 11. WEMOS Electronics. Wemos D1 mini Pro. https://wiki.wemos.cc/products:d1:d1 mini pro. Accessed 12 June 2019 12. Espressif Systems. ESP8266. https://www.espressif.com/en/products/hardware/ esp8266ex/overview. Accessed 12 June 2019
490
M. Redeker and A. Wagenitz
13. ISO/IEC 20922:2016: Information technology – Message Queuing Telemetry Transport (MQTT) v3.1.1. https://www.iso.org/standard/69466.html. Accessed 19 July 2018 14. Goodfellow, I., Bengio, Y., Courville, A.: Deep Learning. MIT Press, Cambridge (2016). http://www.deeplearningbook.org 15. Chollet, F., et al.: Keras (2015). https://keras.io 16. Abadi, M., et al.: TensorFlow: large-scale machine learning on heterogeneous systems (2015). Software tensorflow.org 17. Sauter, M.: From GSM to LTE: An Introduction to Mobile Networks and Mobile Broadband. Wiley, Hoboken (2011)
Part VI: Human-Machine Interaction
Augmented Reality in the Packing Process: A Model for Analyzing Economic Efficiency Tim Woltering, Andre Sardoux Klasen, and Carsten Feldmann(&) University of Applied Sciences Münster, Münster, Germany {tim.woltering,andre.sardoux-klasen, carsten.feldmann}@fh-muenster.de
Abstract. The use of augmented reality (AR) in outbound logistics is associated with potentially strong stimuli for cost savings and throughput time. Nevertheless, the benefits of AR compared to conventional methods require a holistic analysis for investment decision making. Until now, research has only assessed case-study-related potentials and selected aspects of the technology. This paper answers the following research questions: How can the economic efficiency of AR in the packing process be quantified by utilizing a holistic model of value drivers? How can AR be technically implemented for packing processes in outbound logistics? What economic profit results from the use of AR technology in a case company’s packing process? The presented model enables the investment decision to be supported based on economic value added (EVA), thereby providing an assessment of value drivers in packing systems. Cost drivers are identified on the basis of the Supply Chain Operations Reference (SCOR) process model. The technical and economic validation of the model was carried out by means of an empirical study: Expert interviews were conducted for validating the model elements. Data collection by a prototype at a mechanical-engineering company was used to calculate the value contribution. The mapping of cause-effect relationships within the framework of EVA driver trees has proven itself in both the expert interviews and the prototype validation. The field experiment at the case company demonstrated a positive value contribution of AR, in particular regarding employee productivity, length and variance of throughput time, quality aspects, volume utilization, and quantity of packing material used. Keywords: Augmented reality AR Packing Packaging Economic value added EVA Value contribution Cost analysis Cost drivers
1 Introduction The packing process according to the Supply Chain Operations Reference (SCOR) model is defined as activities such as sorting/combining products, packing/kitting the products (pasting labels and barcodes, etc.), and delivering the products to the shipping area for loading [1]. Increasing customer needs for individuality and quality are reflected both in the products and in the packing requirements. Manifold demands such as a large number of different components and frequent non-cubic shapes (e.g., in the mechanical engineering industry) create a complexity that is difficult for employees to © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 493–503, 2020. https://doi.org/10.1007/978-3-030-44783-0_46
494
T. Woltering et al.
master efficiently in terms of time, quality, and cost pressure [2]. Augmented reality (AR) is a group of technologies allowing for the situational visualization of information which has the potential to reduce complexity for the employee [2]. With AR, reality is interactively enriched or extended by the inclusion of additional information [3]. It is especially suitable for use in industrial environments because it enhances the visual perception of users by displaying additional computer-generated information in the field of vision [4, 5]. Examples include information about the pick location for warehouse operators and assembly instructions for employees at the production line [5, 6]. Within the packing process, the use of AR manifold assistance can be imagined, e.g. for the instruction of an optimal packing scheme [6], indications of areas requiring special protection, selection of packing materials or display of packing instructions. A packing system consists of the packaging material itself and is supplemented by the packing process: packing material (such as a covering or stuffing) is utilized to enclose or protect goods for shipping or storage [7]. This article focuses on the process of packing goods in outbound logistics. The objective of a supply chain is to maximize the overall value generated: i.e., the difference between what the value of the final product is to the customer and the costs of supply-chain activities for fulfilling the customer order [8]. Therefore, the aim of this research project is to establish a model for calculating the economic value added (EVA) to determine the value contribution for investment decision making. The model was validated within a case company by means of an AR test installation. Before the validation, the EVA model was set up in the form of value-driver trees based on extensive desk research and expert interviews. Literature analysis was based on the approach of vom Brocke [9] (cf., Fig. 1 and Appendix 1).
Step
Result
I. Definition of type and scope of the analysis
Six databases were included
II. Concept design for the investigation
Cross table with the search words was created
III. Literature research
4232 items have been identified
IV. Literature analysis
44 relevant papers selected
V. Identification of the research gap
Results of literature review
Fig. 1. Process of literature analysis
A total of 44 publications were selected on the basis of the criteria currency, relevance, authority, accuracy, and purpose. The present studies are characterized by a wide heterogeneity in terms of the applied analyses, the empirical data base, and the presentation of the results, so the findings are not strictly comparable. The practiceoriented research focuses primarily on decentralized order picking, assembly, maintenance and repair [10–14]. A large number of articles examine the topic mainly from a theoretical perspective of potential applications in a production logistics environment [15–18]. The packing process is not addressed sufficiently [2, 19]. In the literature, no comprehensive and coherent statements are offered about the direction and impact of value drivers on costs. None of the articles identified present a model for investment decision making and evaluation of the advantages. Mättig et al. prove the efficiency of
Augmented Reality in the Packing Process
495
AR in the packing process under laboratory conditions [2], but with neither reference to a natural working environment nor with consideration of the actual economic value contribution. Thus, an EVA-based analysis of AR in packing processes is needed to understand and prioritize the value drivers for a calculation scheme for investment decision making. To ensure value orientation, the profitability analysis was based on the concept of EVA, which is widely accepted as a financial metric for measuring value [20]. EVA is a measure of economic (not accounting) profit and is defined as the difference between net operating profits after taxes (NOPAT) and capital charge, which depends upon total invested capital and a weighted average cost of capital (WACC) [20, 21]. Against this background, the following research questions are addressed: RQ1: How can the economic efficiency of AR in the packing process be quantified by utilizing a holistic model of value drivers? RQ2: How can AR be technically implemented for packing processes in outbound logistics? RQ3: What economic profit results from the use of AR technology in the case company’s packing process?
2 Methodology First, an overall procedure model was created, which is based on the modeling process according to Adam [22] as presented in Fig. 2. Feedback Capture of the real initial problem: symptoms
Formulation of the problem
Analysis of relevant characteristics and relationships
Reproduction in the structureholding model
Verification of the model
Absence of systematization
Root cause analysis and problem
Characteristics of task
Model design
Verification of applicability
No comprehensive understanding of the value contribution of Augmented Reality in the packing process
No reference model for comprehensive identification of value drivers
Linking the SCOR model to the EVA
Reference model for systematic analysis of the value contribution
Verification of the model on the basis of a proof-ofconcept
Feedback
Fig. 2. Overall modeling process
In a second step, a structural model was established to comprehensively identify the value drivers in the packing process. To identify the operational costs and asset impacts associated with implementing AR, a successive approach is applied. First, cost drivers are identified per process activity. A cost driver is any factor which causes a change in the cost of an activity, reflecting any linkages or interrelationships that affect it [23]. Business process models have been proven to represent knowledge and can be utilized as a basis for aggregating different types of information [24]. Prior to identifying cause-
496
T. Woltering et al.
effect relationships, all relevant supply-chain processes have been investigated and modeled in the overall research project. In this paper, however, only the influences on the sub-process “material packing” within the SCOR process sD1 are presented (see Fig. 3). After modeling the process, the direction of influence is determined per cost driver (increase versus decrease of the reference parameter) using scenario analysis [25]. The cause-effect relationships identified between value drivers and costs were initially established based on desk research and then validated in a second step by ten expert interviews in the German manufacturing industry. The interviews were carried out as semi-structured guided interviews with interview partners from different companies. The answers were transcribed, and a qualitative content analysis was performed afterwards following the approach of Mayring [26].
sD1 Deliver
Process Order
Plan Transport
Pick Product
Ship Product
Pack Product Degree of planning reliability
Degree of volume utilization
Degree of process control
Degree of complexity
Degree of transparency
Total lead time
Invoice
Fig. 3. Subprocess sD1 deliver
When discussing the cost drivers with the interviewed experts, the process activities depicted in Fig. 3 were accredited with the major cost impacts in an AR scenario. A higher degree of volume utilization of the package (such as a covering carton box, container) is achieved by providing information regarding the geometry and quantity (packing pattern), weight, and points of the packages which require special protection. The aforementioned information also optimizes the consumption of packaging materials. These are the box size (transport costs) and packing material, such as fillers (material costs). AR situationally visualizes relevant information by displaying additional computer-generated information in the field of vision, thereby reducing complexity for the employee. This could potentially lead to shorter throughput times in the packing process, mainly driven by a higher level of transparency, process control, and planning reliability. Overall, the experts assume that packing costs decrease when AR is used. In a third step, the structural model encompassing the value-driver trees needed to be linked to the EVA as the target figure for economic profit. As for supply-chain management (SCM), the highest objective of the EVA concept–maximizing the value added respectively business success–should be achieved by means of value drivers. An investment in AR technologies is to be assessed based on the same approach. The supply-chain value drivers are identified according to the EVA concept and then further systematically decomposed per SCM target area: costs (target: low operational costs),
Augmented Reality in the Packing Process
497
assets (target: low investments in assets), and sales volume (target: high turnover). These objectives contribute to increasing EVA: Higher sales volume and lower operational costs result in higher profits. Reduced working capital results in lower capital costs. Costs are broken down according to SCM processes–source, make, deliver, return–as described in the SCOR model [1]. The means of achieving higher sales revenue in SCM is high customer satisfaction facilitated by a high logistics service level, determined by delivery reliability and lead time [8, 27]. The approach of Feldmann/Pumpe was employed to evaluate the economic efficiency of the model based on the EVA [28]. In the last step, an AR prototype was developed by using the software engine Unity from Unity Technologies and the Vuforia plugin from PTC Inc. The validation of the assumed cause-effect relationships took place at the case company, a German manufacturing engineering company, to calculate the influence on the EVA. With regards to the AR hardware, a head-mounted device (“AR glasses”) would have been desirable under the selection criteria of mobility and flexibility. Against the background of employee acceptance, AR glasses were rejected and a solution based on a tablet PC was established instead (see Fig. 4). For the test scenario, the pack pattern was established manually. The pattern was then integrated into the prototype. As a tracking solution, a marker in the form of a DIN A4 sheet at the bottom of the box was chosen to ensure the correct positioning of the 3D objects. Other tracking approaches are conceivable such as tracking the box as a marker or AR without marker. An interface to other IT systems was not developed. In the AR camera view, the warehouse operator can observe the goods to be packaged and the optimum positioning in the cardboard box. Both the goods to be boxed and packing materials are displayed step by step as 3D objects in the optimal position. The objects are placed with consideration of space saving, transport safety, and packing time.
Fig. 4. Prototype in action
498
T. Woltering et al.
A total of four groups were formed, each of which performed a case number of n = 10. Two of these groups were composed of inexperienced employees in the packing process (Experimental Design 1), the other two groups with experienced employees (Experimental Design 2). Two scenarios for the test setup were prepared. Scenario 1 was set up without AR support (control group), Scenario 2 with AR support (experimental group, see Table 1). This resulted in a control group and experimental group four both scenarios. The experiment was carried out in the logistics center of the case company. The materials for the packing task were made available to the test persons at a packing area, which consisted of a workbench, a lift truck and a rack filled with cardboard boxes and packing materials. For scenarios 1 and 2, a total of eight goods of different geometries and weights had to be packed. In Scenario 1, a suitable cardboard box first had to be selected out of a selection of different sizes. The test groups were advised to pack as space-saving as possible and to assure the safety of the goods. For the safety of the goods, the test subjects could use as much packing material as needed. In Scenario 2, the tablet was mounted on a tripod, and the tablet’s camera was aligned with an AR marker mounted in the box. With the help of the touch display, the employee was guided through the packing process by visualizing the process stepwise. Looking at the tablet’s display, the 3D objects virtually overlaid the box (see Fig. 4). The throughput time per cycle was measured with a stopwatch. The arithmetic mean values of the time measurement were tested for significance using a t-test and a significance level of a = 5%. Quality and box-volume utilization were evaluated by comparison with an optimum packing scheme determined beforehand. A quality deficit was defined by the aspect of inadequate transport safety. The results are presented in Table 1 and show a significant difference (Scenario 1: p = 0.21%; Scenario 2: p = 3.63%) in terms of throughput time, volume utilization, and quality for both groups, inexperienced and experienced, in test Scenario 2. The reduction of variance is also noticeable. Especially inexperienced employees were able to increase their volume utilization and quality due to the assistance of the AR system. Against the background of the frequent employment of semi-skilled temporary workers with a high fluctuation in this process area, this is a relevant finding for practitioners.
Table 1. Results of the experiment: volume utilization Group
Scenario 1 1. Inexperienced; without AR Experience level Low AR support No Experimental design 1 Volume utilization 50% Mean (minutes) 4.06 p value 0.21% Variance 0.53 Quality 40%
Scenario 2 2. Inexperienced; 3. Experienced; with AR without AR Low High Yes No 2 100% 70% 3.13 3.16 3.63% 0.22 0.84 100% 60%
4. Experienced; with AR High Yes 100% 2.57 0.05 100%
Augmented Reality in the Packing Process
499
A significant reduction could also be demonstrated in the throughput time. The non-representative results indicate that AR technology enables newly hired and temporary employees to become productive more quickly, and it also requires a lower level of experience knowledge. This would accommodate trends in logistics such as high fluctuation or a high proportion of temporary workers. The experiment confirms the hypotheses presented in the model. The results were used to calculate the actual value contribution of the case company, based on the EVA. Table 2 shows the cost effect triggered by the described causes. The savings of process costs were calculated based on the measured cycle times in the test setup. For this purpose, the percentage reduction of the throughput time was interpolated to all packing processes. Table 2. Cost drivers Cost drivers (in €) Process costs Packing material Packing aids Write-offs Maintenance Balance
2019 – −18,857.36 −46,957.05 29,880.00 3,000.00 −32,934.41
2020 2021 2022 −59,847.60 −119,695.21 −179,542.81 −18,857.36 −18,857.36 −18,857.36 −46,957.05 −46,957.05 −46,957.05 38,173.00 38,466.00 36,879.00 3,000.00 3,000.00 3,000.00 −84,489.01 −144,043.61 −205,478.22
Table 3 shows the change in assets in the balance sheet. Due to the existing highperformance level of the case company’s packing processes and barriers in the data collection of the sales drivers, these were excluded from the calculation. With regard to the calculation of an optimal packing scheme, the assumption of an automated software solution was made such as PUZZLE. Both the license costs for the software to determine the optimal packing scheme and the development costs for the app presented were considered as asset drivers. However, due to the faster processing time and higher packing quality, positive effects on sales are assumed. Taking into account the cost and asset effects discussed, an AR-induced value contribution of €175,508 (discounted from 2018) could be demonstrated for the case company (see Table 4). Table 3. Asset driver Asset driver (in €) 2019 Redesign of workplaces 3,000.00 Software development 100,000.00 Licenses 40,000.00 Tablets 2,640.00 Balance 145,640.00
2020 2021 – – 40,000.00 € – – – 880.00 880.00 40,880.00 880.00 €
2022 – – – 880.00 € 880.00 €
500
T. Woltering et al. Table 4. EVA calculation
EVA calculation (in €) 2018 NOPAT 21,330.067 NOA 194,187.537 WACC 5.60% EVA 10,455.565 EVA changes – Discounted – Sum 175,508
2019 21,363.001 194,360.472 5.60% 10,478.815 23.250 22.017
2020 21,414.556 194,252.179 5.60% 10,536.434 57.619 51.669
2021 21,474.110 194,211.886 5.60% 10,598.245 61.811 52.489
2022 21,535.545 194,213.473 5.60% 10,659.590 61.345 49.332
3 Conclusion and Future Research This paper aims to establish a model to identify the contribution of an investment in AR in the packing process to increase company value. Transparency regarding the causeeffect relations should help to assess an investment in AR technology. The model presented is a systematically structured framework of value drivers and resulting effects on costs. Taking a dynamic EVA perspective, an exemplary in-depth decomposition was provided for the packing process, as this area has been widely neglected in current research. The model supports investment decisions for practitioners, i.e., by comparing the employment of AR versus a conventional scenario without AR. Utilizing the concept of EVA for calculating the economic efficiency, it provides a consistent framework for analyzing the effects of AR oriented to value creation (RQ1: How can the economic efficiency of AR in the packing process be quantified by utilizing a holistic model of value drivers?). Moreover, the same framework can be utilized to optimize the investment across the life cycle by integrating the value drivers into operational controlling and periodically measuring the economic value added. The implementation of the prototype at the case company has successfully proven the technical suitability of AR in the packaging process (RQ2: How can AR be technically implemented for packing processes in outbound logistics?). Within the course of the field experiment at the case company, a positive value contribution could be demonstrated (RQ3: What economic profit results from the use of AR technology in the case company’s packing process?). However, some limitations of the methodology should be mentioned. The results of the expert interviews are not representative and can be related only to the manufacturing sector. Furthermore, of the 20,000 different materials within the case company’s portfolio and the resulting potential combination, only one combination was selected for the experiment, which limits the transfer of the findings to the whole portfolio. The monetary assessment must be considered on a case-by-case basis, as the WACC level, process cost rates, packing material costs, and other applicable model conditions cannot be generalized. Prior to applying the model, a decision needs to be made regarding a specific AR technology. The model has to be interpreted in view of the specific context of the company and product portfolio analyzed. The findings of the case company indicate the best suitability for a low mix/high volume portfolio of materials to be packed. The strengths of influences between value drivers and costs depend on the
Augmented Reality in the Packing Process
501
company’s context as well. This leads to the fact that transferability to other companies is given only to a very limited extent. For future research, applications of the methodology must be carried out for further iterations and improvements. Further case studies are desirable which focus on a specific AR technology or industry and which use the presented framework for systematic analysis. In addition, other target areas can be integrated into the model, such as ecological aspects.
Appendix 1 Databases Web of Science N = 870 De Gruyter N = 323
Further Sources of Information
EBSCOhost N = 182 SCOPUS N = 1410
ECONbiz N = 226 WISO N = 1115
Monographs, Studies, newspaper arƟcles, informaƟon from research insƟtuƟons and companies N = 63
Identified Articles N = 4232
Examining Titel and Abstract N = 37 Forward and Backward Search N=7 Full Text N = 44
List of Keywords Context Augmented Reality
Context Logistics
Erweiterte Realität
augmented reality
Logistikkosten
Augmented Reality
AR
Supply-Chain-Kosten
supply chain costs
Mixed Reality
mixed reality
Transportkosten
transport costs
Logistik
logistics
Digitaler Assistent
human-computer interaction human-machine interaction digital assistant
AND
Mensch-MaschineInteraktion Mensch-ComputerInteraktion
logistics costs
Supply-Chain
supply chain
Wirtschaflichkeit
economic efficiency
Wirtschaflichkeitsanalyse
economic viability
Economic Value Added
economic value added
Geschäftswertbeitrag
profitability analysis
Verpackungskosten
cost-effectiveness
Verpackungsprozess
economic analysis
Verpackungslogistik
investment analysis packaging proof of concept
502
T. Woltering et al.
References 1. Supply Chain Council Inc. Supply Chain Operations Reference Model (SCOR), 11th edn. (2012) 2. Mättig, B., Lorimer, I., Kirks, T., Jost, J.: Untersuchung des Einsatzes von Augmented Reality im Verpackungsprozess unter Berücksichtigung spezifischer Anforderungen an die Informationsdarstellung sowie die ergonomische Einbindung des Menschen in den Prozess. Wissenschaftliche Gesellschaft für Technische Logistik, pp. 97–106 (2016) 3. Azuma, R., Baillot, Y., Behringer, R., Feiner, S., Julier, S., MacIntyre, B.: Recent advances in augmented reality. IEEE Comput. Graph. Appl. 21, 34–47 (2001) 4. Schmidt, L., Wiedenmaier, S., Oehme, O., Luczak, H.: Benutzerzentrierte Gestaltung von Augmented Reality in der Produktion. In: Stary, C. (Hrsg.) Mensch & Computer 2005: Kunde und Wissenschaft–Grenzüberschreitungen der interaktiven ART. Oldenbourg Verlag, München (2005) 5. Alt, T.: Augmented Reality in der Produktion. Herbert Utz Verlag, München (2003) 6. Mättig, B., Jost, J., Kirks, T.: Erweiterte Horizonte–Ein technischer Blick in die Zukunft der Arbeit. In: Wischmann, S., Hartmann, E.A. (Hrsg.) Zukunft der Arbeit–Eine praxisnahe Betrachtung. Springer Vieweg, Berlin (2018) 7. Lange, V.: Verpackungs- und Verladetechnik. In: Arnold, D., Kuhn, A., Furmans, K., Isermann, H., Tempelmeier, H. (Hrsg.) Handbuch Logistik. Springer Verlag, Heidelberg (2008) 8. Chopra, S., Meindl, P.: Supply-Chain Management, 6th edn. Pearson, Boston (2015) 9. vom Brocke, J., Simons, A., Niehaves, B., Riemer, K., Plattfaut, R., Cleven, A.: Reconstructing the giant: on the importance of rigour in documenting the literature search process. In: European Conference on Information Systems, vol. 17, pp. 2206–2217 (2009) 10. Schwerdtfeger, B., Klinker, G.: Supporting order picking with augmented reality. In: 7th IEEE/ACM International Symposium on Mixed and Augmented Reality (2008) 11. Kohn, V., Harborth, D.: Augmented reality–a game changing technology for manufacturing processes? In: 26th European Conference on Information Systems, pp. 1–18 (2018) 12. Ong, S.K., Yuan, M.L., Nee, A.Y.C.: Augmented reality applications in manufacturing: a survey. Int. J. Prod. Res. 46, 2707–2742 (2008) 13. Webel, S., Bockholt, U., Engelke, T., Peveri, M., Olbrich, M., Preusche, C.: Augmented reality training for assembly and maintenance skills. In: BIO Web of Conferences, vol. 1 (2011) 14. Schmidt, L., Wiedenmaier, S., Oehme, O., Luczak, H.: Benutzerzentrierte Gestaltung von Augmented Reality in der Produktion. In: Stary, C. (Hrsg.) Mensch & Computer 2005: Kunde und Wissenschaft–Grenzüberschreitungen der interaktiven ART. Oldenbourg Verlag, München (2011) 15. Kückelhaus, M.: Eleven reasons to consider augmented reality in logistics. Focus, pp. 14–17 (2015) 16. Hammerschmid, S.: Chances for virtual and augmented reality along the value chain. In: Communications in Computer and Information Science, vol. 748, pp. 352–359 (2017) 17. Michel, R.: Insight into smart glasses. Modern Materials Handling, pp. 44–50 (2018) 18. Lang, A., DastagirKota, M.S.S., Weigert, D., Behrendt, F.: Mixed reality in production and logistics: discussing the application potentials of Microsoft HoloLens. Procedia Comput. Sci. 149, 118–129 (2019) 19. Stoltz, M.-H., Giannikas, V., McFarlane, D., Strachan, J., Um, J., Srinivasan, R.: Augmented reality in warehouse operations: opportunities and barriers. IFAC-Papers OnLine 50(1), 12979–12984 (2017)
Augmented Reality in the Packing Process
503
20. Young, D.S., O’Byrne, S.F.: EVA and Value-Based Management. McGraw Hill, New York (2001) 21. Ehrbar, A.: EVA. Wiley, New York (1998) 22. Adam, D.: Planung und Entscheidung: Modelle, Ziele, Methoden. Verlag Gabler, Wiesbaden (1996) 23. Porter, M.E.: Competitive Advantage. Free Press, New York (1998) 24. Lindemann, C., Jahnke, U., Moi, M., Koch, R.: Analyzing product lifecycle costs for a better understanding of cost drivers in additive manufacturing. In: 23th Annual International Solid Freeform Fabrication Symposium - An Additive Manufacturing Conference (2012) 25. Fischer, T.M., Möller, K., Schultze, W.: Controlling: Grundlagen, Instrumente und Entwicklungsperspektiven. Schäffer-Poeschel Verlag für Wirtschaft, Stuttgart (2015) 26. Mayring, P.: Einführung in die qualitative Sozialforschung. Beltz Verlag, Weinheim (2002) 27. Schnetzler, M.J., Sennheiser, A., Schonsleben, P.: A decomposition-based approach for the development of a supply chain strategy. Int. J. Prod. Econ. 105, 21–42 (2006) 28. Feldmann, C., Pumpe, A.: A holistic decision framework for 3D printing investments in global supply chains. Transp. Res. Procedia 25, 677–694 (2016)
Assessment of Cognitive Strain in Digital Logistics Work: Background, Analysis and Implications Matthias Klumpp1,2(&), Vera Hagemann3, and Martina Schaper3
2 3
1 Georg-August-Universität Göttingen, Wilhelmsplatz 1, 37073 Göttingen, Germany [email protected] FOM Hochschule Essen, Leimkugelstr. 6, 45141 Essen, Germany Universität Bremen, Bibliothekstraße 1, 28359 Bremen, Germany {vhagemann,martina.schaper}@uni-bremen.de
Abstract. Digital logistics processes are not only a technological challenge. In addition, the question of human cognitive strain and human-computer interaction are important success factors for digital logistics work concepts. After a background introduction, conceptual and an analytical frameworks for identifying cognitive strain in digital logistics work are presented and applied to specific logistics activities. Characteristic features are the speed and density of work tasks, the requirement for teamwork and inter-organizational cooperation as well as the important role of motivation in the “people business” of logistics. Further research is warranted regarding the specific characteristics of digital logistics work and the impact on cognitive workload of human workers.
1 Introduction Digital concepts are changing logistics structures and supply chain management approaches. A significant body of research is dealing with these concepts and implications (Montreuil 2011; Ballot et al. 2014; Fawcett and Waller 2014; Gunsekaran and Ngai 2014; Crainic and Montreuil 2016; Orlikowski 2016; Wieland et al. 2016; Phan et al. 2017; Sternberg and Norrman 2017; Pilati and Regattieri 2018; Sendlhofer and Lernborg 2018; Borangiu et al. 2019; Klumpp et al. 2019; Klumpp and Zijm 2019). However, a research gap consists in the field of cognitive ergonomics in logistics work: In digital processes, workers are facing new challenges regarding their job demands as for example responsibilities are shifting from operational questions towards supervisory tasks as more and more processes are automated e.g. in autonomous material handling or other AGV systems. Therefore, an analysis regarding assessments of cognitive ergonomics, cognitive strain and possible mitigations is an important research task with many logistics business practice implications. In a concurrent example with order picking in warehouse settings, pick-by-voice systems are putting a high level of strain on workers as the cognitive load in processing spoken commands – also depending on language capabilities – is high and at the same time, manual processes have to be performed. Therefore, future technologies will also © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 504–515, 2020. https://doi.org/10.1007/978-3-030-44783-0_47
Assessment of Cognitive Strain in Digital Logistics Work
505
migrate towards different communication channels (“pick-by-light” etc.) as the cognitive capacity of warehouse personnel is the bottleneck factor herein. The specific contribution of this publication is the application of existing theory elements regarding cognitive ergonomics for logistics processes especially within digitalization projects. Furthermore, a concept development contribution is presented in outlining required elements of an elaborated cognitive ergonomics analysis concept in logistics as basis for further derived action options. The text is structured as follows: In Sect. 2, the conceptual elements from existing theory regarding cognitive ergonomics are presented and applied towards logistics contexts. Section 3 is presenting a concept development regarding an analytical framework for logistics and Sect. 4 is discussing implications for this regarding the typical environment of logistics and transportation. Section 5 is providing a conclusion and outlook for further research on this topic of cognitive ergonomics in digital logistics processes.
2 Conceptual Framework Cognitive Ergonomics Within the design of work places, and what is more important today, the design of work processes, cognitive stress and strain of employees are important to consider for successful work and the healthiness and employability of workers. The assessment of cognitive stress and strain at the workplace is also prescribed by labor protection law (ArbSchG §§ 4 & 5). The foundation for risk assessment tools is the DIN EN ISO 10.075. Cognitive stress refers to all external and objective factors influencing the human, for example the temperature, work equipment or upset customers. Cognitive strain or demands mean the consequences of these factors within the human, for example an increase in heart rate, blood pressure or hormone release, fatigue, errors, aggressive behavior or an increase in consumption of alcohol or nicotine (Kaufmann et al. 1982; Rohmert and Rutenfranz 1975). Fatigue can result from a long-lasting execution of a task and hard work and is a reversible mitigation of productivity of a single organ (e.g. eyes or arms) or the whole organism. Important for the reduction of fatigue and an increase in productivity are phases of recreation during work (Ulich 2001). Regarding manual labour as well as easy and repetitive cognitive tasks, more short breaks during work lead to a higher recreational value than a few but long-lasting breaks (Ulich 2001). Based on digitized changes in work leading to more monitoring tasks instead of an active physical participation within the production process, the work is becoming more and more monotonous for employees (Ruiner and Wilkesmann 2016). Monotony is a state of lowered psychophysical arousal (Bartenwerfer 1970) and occurs in lowstimulus environments. It further results from long-lasting executions of repetitive and uniform activities. Monotony especially occurs when the worker has to narrow his attention to a single uniform activity and is not able to execute other relieving secondary activities (Ulich 2001). An essential distinguishing feature between fatigue and monotony is that monotony disappears immediately after a change of activities, whereas fatigue does not disappear or decreases very slowly. Thus, in order to prevent fatigue and monotony in modern digitized work places phases of recreation are important as well as, for example, secondary tasks integrated into the accountability of employees monitoring processes (see Fig. 1).
506
M. Klumpp et al.
Fig. 1. Cognitive stress and strain implication chain.
The development of perceived stress results from a real or perceived imbalance between situational demands and requirements on the one hand and perceived resources for coping with them on the other hand. Stress means a subjective and intensive unpleasant state of pressure. People are afraid of being faced with high aversive, subjectively perceived temporal nearby and long-lasting situations, which cannot be controlled and their avoidance appears to be important at the same time (Schaper 2011, p. 477). An established classification of stress factors in work distinguishes demands of the work task (e.g. information overload or unexpected disturbances), demands from work role (e.g. responsibility or missing support from others), demands from the environment (e.g. noise or light), demands from social environment (e.g. climate or organizational change) and demands from the person itself (e.g. fear of tasks or missing qualification) (Richter and Hacker 1998). High-level work demands mainly create experiences of mental strain in employees when they have only small job decision latitude (Karasek 1979). For example, imagine the situation where the customer is able to intervene into the work process of a driver delivering parcels; because the customer can indicate at every time that he or she wants to give back parcels to the driver. Thus, the driver receives constantly new information about his or her route, without considering his current workload, and gets a feeling of being controlled externally without any decision latitude. From research, we already know that the healthiness of the employees is significantly influenced by work that is characterized by task identity, task significance and autonomy (Brousseau 1976). Tasks characterized by high demands and low decision latitudes provoke an increase in stress experience and can lead to psychosomatic problems and cardiovascular diseases in the long term (Schaper 2011). Tasks characterized by low demands and low decision latitudes also lead to negative consequences, in this case to the experience of monotony. However, it is also known that a high decision latitude in work is not always good. High decision latitudes can evoke uncertainty of action and then often lead to burnout symptoms. Intervening variables related to the person preventing these negative consequences of high decision latitudes are the self-regulatory capacities and self-efficacy of the employees as well as a match of qualifications to the job (Schaper 2011).
Assessment of Cognitive Strain in Digital Logistics Work
507
Table 1. Stress factors (Richter and Hacker 1998). Work tasks Work role Environment Social environment Information overload Responsibility Noise Climate Unexpected disturbances Missing support Light Organizational change
In order to detect demands in work that lead to high cognitive strain in employees, and to prevent negative consequences as described above and in Table 1, risk assessment tools are helpful. These instruments are filled in by the employees and assess their experienced cognitive strain. From the general outline, the following application areas for logistics can be derived, also referring to required qualifications and skills in the workforce (Lin and Chang 2018): Warehouse processes, especially picking and packing; driving work processes and professions; production logistics tasks and processes; value added services in contract logistics.
3 Analytical Framework The health report 2018 of the Techniker Krankenkasse (TK) (2019) reports health statistics of all its policyholders. According to that, for all employed people who are insured with the TK, 6.13 million cases of illness were reported in 2018, resulting in 82 million sick days. Mental illness is concerning in particular: Compared to the year 2000, mental illnesses increased over 190% until 2018 (Techniker Krankenkasse 2019). The average duration per sick leave is 15.49 days. The absenteeism for mental illnesses, at 25.6 days, is significantly longer (Badura et al. 2016). These numbers can be transferred to the logistics sector. The quality of work in logistics is ranked in the lower midfield by the DGB Index Gute Arbeit (2016). For example, 41% of those employed in transport and warehousing state that the workload has increased as a result of digitization and 58% find the increasing surveillance and performance monitoring by digital means particularly demanding (DGB Index Gute Arbeit, 2016). Therefore, it is necessary to address psychological risks with the help of psychological diagnostics in order to protect the health and employability of employees. Companies now face the challenge of protecting and promoting the health of their employees. This requires detailed information about activities that are expected to bare high cognitive stress. Information on the consequences of cognitive stress, i.e. cognitive strain, is also needed in order to implement effective work design measures. A large variety of instruments and methods for measuring cognitive stress and strain, as well as processes for assessing cognitive endangerment, are available for this purpose and can be used by companies. Lists of the procedures can be found, for example, at (Gemeinsame Deutsche Arbeitsschutzstrategie (GDA) 2017) or Kauffeld (2019). In the following section, two recent instruments to assess work related psychological risk factors are presented. The first one is mentioned as one of 28 instruments used by the partners of the Psyche work programme of the Joint German Occupational Health and Safety Strategy (Arbeitsgruppe Psyche der Gemeinsamen Deutschen Arbeitsschutzstrategie, GDA-Psyche 2017). The second instrument was developed,
508
M. Klumpp et al.
refined and applied in the field over the last ten years (Metz and Rothe 2017). The first one is the Kompaktverfahren Psychische Belastung (KPB) edited by the Institut für angewandte Arbeitswissenschaft (ifaa) and updated in 2017. The KPB procedure and assessment instrument focuses on measuring cognitive stress and objective data such as absenteeism, fluctuation, accidents and conflicts at work, and others. Questions regarding New Work (i.e. flexibility, mobility and permanent availability) are also included. That makes the KPB a suitable tool to assess cognitive stress for small and medium sized businesses. However, cognitive strain is not included in the assessment. Information gained with the KPB can only be used to restructure critical working environments, but not to derive measures for personnel development, since no information about the experiences of the employees is collected. Furthermore, the influence of major trends such as digitization, automation and participation in change processes is not yet taken into account. Another recent procedure is the Screening Psychischer Arbeitsbelastung (SPA) by Metz and Rother (2017). The SPA is composed of four different parts. The first part can be employed by experts to collect information about the working conditions, i.e. cognitive stress. The other three parts are questionnaires filled in by the employees to gain information about cognitive strain and its consequences in terms of psychosomatic and other symptoms (e.g. back pain, insomnia, restlessness, pensive tendencies). Organizational and social resources are also covered by the SPA asking about aligning personal qualifications with work tasks, learning opportunities, the support of colleagues and the social climate. As a whole, the SPA combines different perspectives to measure cognitive stress, cognitive strain, and psychological and physiological consequences, which makes the SPA a valid screening instrument. However, the application of the SPA procedure is partly limited to experts and because of the four-part structure, the measurement is time-consuming. Our aim is to make it easier for companies to access and carry out psychological risk assessments. Thus, new or revised measuring instruments are needed which take up and incorporate current developments and changes in the world of work, such as digitization of workplaces, increasing work density, new forms of work, value changes (virtual) teamwork and the introduction of new technologies. These tools should also be applicable by employees and not only by experts and efficient to fill in. In order to address this issue, the DIAMANT project funded by the German Federal Ministry of Labour and Social Affairs developed a risk assessment tool to assess cognitive stress and strain in digital work context. This tool is based on a mixedmethod approach and will be used as a quantitative screening instrument in alignment with the international standard for ergonomic principles relating to mental workload (DIN EN ISO 10075-3:2004) after completion. In the following, the basics and the procedure to develop this tool are explained in more detail. The development of the risk assessment tool was based on a comprehensive desk research for existing literature and measurement methods. With the help of the information received, guidelines for problem-centered interviews were drawn up for employees and managers, thus taking the qualitative aspect into account. 34 interviews were conducted transcribed and evaluated by means of content analysis. New findings from the results were then transferred to the questionnaire.
Assessment of Cognitive Strain in Digital Logistics Work
509
On the basis of the interview data and desk research findings the risk assessment tool was developed containing different scales. One scale measures the “perceived support through technology”. There, technology is conceived as a resource at the workplace that positively supports workers in their work activities. The scale contains four items that can be answered on a five-point Likert-scale using the values never rarely - occasionally - often - always. The four items of the scale “perceived support through technology” are: (1) The technology makes it easier for me to organize my work tasks. (2) Through technology I can improve the quality of my work. (3) Through technology I can reduce mistakes in my work. (4) Through technology I can achieve my work goals. The interviews also vividly illustrate the working conditions in logistics, especially the limited scope of action and highly structured tasks. In past risk assessments only these working conditions, reflecting cognitive stress, were considered. Therefore, it is important to measure cognitive strain and its consequences for people working in such an environment. The measurement of cognitive strain shows whether and in what way there is a need for action. For these reasons, it is important to ask not only about the risk factors but also about the personal attitude of the employees in order to be able to intervene based on a balance of cognitive strain. In addition to the interview results, existing measuring instruments and other literature were used to develop the risk assessment tool. Questions on various factors in work situations were adopted, modified and created. These five factors relate to work tasks and activities, work environment, organizational climate, work process and organization as well as personal attitude (Gemeinsame Deutsche Arbeitsschutzstrategie (GDA) 2012). This will be exemplified further in Sect. 4. The difference to existing measurement instruments lies in the specific inclusion of technical factors in order to pay special attention to the digitization of work processes. The scope of included aspects in the survey is also a decisive factor for the new instrument. By investigating the various factors related to digitization of work, a broad spectrum of potential risk factors is identified. This combination of risk factors and digitalization makes it possible to obtain a comprehensive survey of cognitive stress and strain at work. A further gain of this new instrument is the ability to precisely measure cognitive strain, which is collected individually through personal attitudes. An example for this aspect is the acceptance of technology, which has a great influence on how employees deal with the introduction of new technology and how they cope with it, as well as the resulting cognitive strain when rejecting technical innovations. Thus, it is possible with this instrument to provide a stress and strain balance of the employees of the company and to deal with it more specifically. Because different jobs, work tasks and personalities need individual measures to be effective.
4 Implications for Digital Logistics Work In general, logistics work tasks are changing from specific work and skill application towards supervision of automated systems, e.g. with autonomous driving for trucks on streets or forklifts in production and retail facilities. Therefore, cognitive stress and strain – especially cognitive readiness for the application of supervision tasks – is
510
M. Klumpp et al.
necessary. In addition, specific work areas as well as the organizational climate in most logistics work environments can be outlined according to the category structure presented in Sect. 3: As Table 2 is outlining, the specifically addressed subcategories for cognitive stress and strain analysis are highly relevant for logistics work processes due to the general work settings found in most cases. Table 2. Risk assessment categories within logistics processes. Work tasks & activity There is a crucial role of logistics in most production and value chain environments: high importance of exact and compliant work
Work environment Work environments in logistics are often hampered by bad light or temperature (too cold, too hot) situations, therefore an evaluation also for the cognitive side of work is highly relevant
Organizational climate The organizational culture in most logistics organizations is rudimentary due to high fluctuations of personnel, broad mix of backgrounds and qualifications as well as migration background impacts (language)
Workflow & organization Teamwork is highly relevant for logistics jobs along the supply chain and a participation of workers is required but often neglected. This could be changed by technology application but also would put further effort on the cognitive workload balance of workers
Personal attitude In many cases, highly motivated people are found in logistics. Nevertheless, in many occasions, motivation and attitude can be improved as for example acceptance between blueand white-collar workers is limited in many cases
Applied to a typical picking job routing, this can be exemplified as follows: Work tasks are very crucial in terms of the accurate and timely delivery and compliance with all commands (given e.g. by voice). The work environment for pickers is often demanding regarding physical ergonomic as well as regarding cognitive demands e.g. in finding and sorting items. Regarding organizational climate, pickers are often dissatisfied with the very diverse and changing setup of teams and changing technical work requirements due to new ordering and optimization systems in their work area. Workflow requirements are very prominent in logistics, e.g. with the subsequent internal and external transport for pickers after they finish an order (e.g. also packaging requirements). This exerts also a specific cognitive workload due to time, speed and deadline requirements. Finally, personal attitude plays an important role also for pickers as logistics workers: A personal sense of accuracy, speed and flow systems is required in order to perform well in this area. This again is highly connected to cognitive workload and performance. The following Table 3 is further outlining specific implications for digital work tasks and processes in logistics.
Assessment of Cognitive Strain in Digital Logistics Work
511
Table 3. Implications within digital logistics processes. Work tasks & activity Digital work tasks have to be evaluated upfront regarding cognitive stress and strain in order to prevent too exhaustive situations for all workers
Work environment Physical work environment resources have to be evaluated beforehand, too. Especially with high cognitive stress and strain environments the physical environment is of even more importance than may be supposed intuitively. Physical and cognitive environments act in a multiplicative order in this sense
Organizational climate Cultural and language requirements and situations have to be adapted early on in order to prevent misperceptions and anger
Workflow & organization Teamwork has to be supports by e.g. digital communication devices (Smartphones, Social Media) in order to enable logistics workers to cope with digital work tasks, i.e. seek help from coworkers also electronically
Personal attitude Personal motivation and attitudes have to be supported and enhanced e.g. by training for digital devices and systems. This could include peer-to-peer training by first teaching “multiplicators” and then the rest of the worker groups
5 Conclusion and Outlook The chapter has presented the urgency and background for the increasing importance of cognitive workload analysis in digital logistics developments and workplaces. Furthermore, a comprehensive analysis concept was developed and the specific application to logistics work processes discussed. The importance of the topic does warrant further research and there are essential interrelations with other research fields as presented subsequently: Current developments in the assessment of cognitive strain are also trying to determine it not only by subjective information but also with objective data. Therefore, the approach cognitive neuroergonomics originated (Reiser et al. 2018). This approach is trying to assess cognitive strain with neurophysiological measurements, such as electroencephalography (EEG) or functional near-infrared spectroscopy (fNIRS). Applying the EEG theta wave band is of interest, for example, as it correlates with working memory processes and increases with more and more task demands (Reiser et al. 2018). Such an approach for the assessment of cognitive strain at work might also be of interest for logistics, as the employees’ working memory is becoming more and more the bottleneck in work processes, and not the physical motion any more, as described in the example in the introduction regarding pick-by-voice systems. These measurements can be applied to quantify cognitive strain objectively and to complement
512
M. Klumpp et al.
subjective data, resulting from interviews or questionnaires, for example. First studies in this field of cognitive neuroergonomics have shown that cognitive strain can be assessed in logistics as well as in mechanically production tasks (Mijovic et al. 2017; Wascher et al. 2014). Today, EEG and fNIRS can be used in a mobile and noninvasive way and they are able to assess, for example, working memory processes or prefrontal cortex activity (Reiser et al. 2018). EEG can be applied, for example, for assessing the employees’ mental workload as a mental state, so that man-machine interfaces can be monitored in real-time and the index for workload can be used for a reduction of stress. Certainly, these methods carry many problems when applying them in real work settings. The fNIRS is able to assess the level of cognitive strain very well for easy and moderate tasks, but it has problems to do so when tasks are becoming more and more difficult and the mental workload increases (Reiser et al. 2018). Furthermore, these measurement tools are not really user friendly and it seems that it is not possible for employees to carry these instruments at their bodies all shifts long. Thus, the acceptance by individuals is questionable. It is also of importance, that these measurements are able to discriminate mental workload from emotional processes (Mandrick et al. 2016). For scientific validated data, researches even recommend to apply both measurements at once in order to compensate the weaknesses of each of them (Reiser et al. 2018). This again makes it more difficult to assess cognitive strain of employees with these measurements in real-work settings. These developments can be compared to the increasingly interesting developments of motion-capture video analysis of ergonomic strain for example in production and intralogistics (Bortolini et al. 2019). An increased level of technology-based analysis is in line with the innovation pathway of digital logistics in general. For a final outlook regarding the importance of cognitive strain analysis in digital logistics work, the typical work of truck drivers especially in urban logistics can be described in the digitally enhanced work situations of today and tomorrow: Increasingly due to Cyber-Physical Systems (CPS) and dynamic routing concepts, online orders from customers (shipment of orders, return of parcels from customers), drivers for parcel delivery services are confronted with minute-by-minute changes in their routing schedule. It can be assumed that a maximum cognitive workload exists – individually for each person – until which such changes are productive. However, above such a theoretical maximum level of changes e.g. by hour, the individual might face cognitive overload and not be able to perform driving and delivery tasks any more. This highlights that further research is needed in order to analyze such thresholds and design appropriate answers to this.
References Arbeitsgruppe Psyche der Gemeinsamen Deutschen Arbeitsschutzstrategie (GDA-Psyche). Instrumente und Verfahren zur Gefährdungsbeurteilung psychischer Belastung. Sicher ist sicher, 68(4) (2017). https://www.gda-psyche.de/SharedDocs/Publikationen/DE/instrumenteund-verfahren-zur-gefaehrdungsbeurteilung-psychischer-belastung.pdf?__blob= publicationFile&v=2. Accessed 24 July 2019
Assessment of Cognitive Strain in Digital Logistics Work
513
Badura, B., Ducki, A., Schröder, H., Klose, J., Meyer, M.: Fehlzeiten-Report 2016. Unternehmenskultur und Gesundheit – Herausforderungen und Chancen. Springer, Heidelberg (2016) Ballot, E., Montreuil, B., Meller, R.: The physical Internet – the network of logistics networks. Programme de recherche et d’innovation dans les transports terrestres (PREDIT). La Documentation française, Paris (2014) Bortolini, M., Faccio, M., Gamberi, M., Pilati, F.: Motion Analysis System (MAS) for production and ergonomics assessment in the manufacturing processes. Comput. Ind. Eng., in press. (2019) Bartenwerfer, H.: Psychische Beanspruchung und Ermüdung. In: Mayer, A., Herwig, B. (eds.) Handbuch der Psychologie (Bd. 9: Betriebspsychologie, S.), pp. 168–209, Hogrefe Göttingen (1970) Borangiu, T., et al.: Digital transformation of manufacturing through cloud services and resource virtualization. Comput. Ind. 108, 150–162 (2019) Brousseau, K.R.: Effects of Job Experience on Personality. A Theoretical and Empirical Investigation. Yale University Press, New Haven (1976) Crainic, T.G., Montreuil, B.: Physical Internet enabled hyperconnected city logistics. Transp. Res. Procedia 12, 383–398 (2016) DGB-Index Gute Arbeit: DGB-Index Gute Arbeit. Der Report 2016. Wie die Beschäftigten die Arbeitsbedingungen in Deutschland beurteilen. Mit dem Themenschwerpunkt: Die Digitalisierung der Arbeitswelt – Eine Zwischenbilanz aus der Sicht der Beschäftigten (2016). https://index-gute-arbeit.dgb.de/++co++8915554e-a0fd-11e6-8e36-525400e5a74a. Accessed 17 July 2019 DIN EN ISO 10075: Ergonomische Grundlagen bezüglich psychischer Arbeitsbelastung. Teil 3: Grundsätze und Anforderungen an Verfahren zur Messung und Erfassung psychischer Arbeitsbelastung (DIN EN ISO 10075-3) (2004) Fawcett, S.E., Waller, M.A.: Supply chain game changers– mega, nano, and virtual trends – and forces that impede supply chain design. J. Bus. Logist. 35(3), 157–164 (2014) Gemeinsame Deutsche Arbeitsschutzstrategie (GDA): Leitlinie, Beratung und Überwachung bei psychischer Belastung am Arbeitsplatz (2012). http://www.gda-portal.de/de/Betreuung/ Leitlinie-PsychBelastung.html Gemeinsame Deutsche Arbeitsschutzstrategie (GDA): Arbeitsprogramm Psyche. Instrumente und Verfahren zur Gefährdungsbeurteilung psychischer Belastung. In: Erich Schmidt Verlag (eds.), sicher ist sicher, pp. 194–196. Erich Schmidt, Berlin (2017) Gunsekaran, A., Ngai, E.W.T.: Expert systems and artificial intelligence in the 21st century logistics and supply chain management. Expert Syst. Appl. 41(1), 1–4 (2014) Institut für angewandte Arbeitswissenschaft e. V. (ifaa) (Hrsg.): KPB – Kompaktverfahren Psychische Belastung. Werkzeuge zur Durchführung der Gefährdungsbeurteilung. Springer, Berlin (2017). https://doi.org/10.1007/978-3-662-54898-1_1 Karasek, R.A.: Job demands, job decision latitude, and mental strain: implications for job redesign. Adm. Sci. Q. 24(2), 285–308 (1979) Kauffeld, S.: Arbeits-, Organisations- und Personalpsychologie für Bachelor. Springer, Heidelberg (2019) Kaufmann, I, Pornschlegel, H., Udris, I.: Arbeitsbelastung und Beanspruchung. In: Zimmermann, L. (ed.) Humane Arbeit für Arbeitnehmer (Bd. 5: Belastungen und Stress bei der Arbeit), pp. 13–48. Rowohlt, Reinbeck (1982) Klumpp, M., Zijm, W.M.H.: Logistics innovation and social sustainability: how to prevent an artificial divide in human-computer interaction. J. Bus. Logist. (2019). https://doi.org/10. 1111/jbl.12198
514
M. Klumpp et al.
Klumpp, M., et al.: Production logistics and human-computer interaction - state-of-the-art, challenges and requirements for the future. Int. J. Adv. Manufact. Technol. (2019). https://doi. org/10.1007/s00170-019-03785-0 Lin, C.-C., Chang, C.-H.: Evaluating skill requirement for logistics operation practitioners: based on the perceptions of logistics service providers and academics in Taiwan. Asian J. Shipp. Logist. 34(4), 328–336 (2018) Mandrick, K., et al.: Why a comprehensive understanding of mental workload through the measurement of neurovascular coupling is a key issue for neuroergonomics? Front. Hum. Neurosci. 10 (2016). https://doi.org/10.3389/fnhum.2016.00250 Metz, A.-M., Rothe, H.-J.: Screening Psychischer Arbeitsbelastung. Ein Verfahren zur Gefährdungsbeurteilung. Springer, Wiesbaden (2017). https://doi.org/10.1007/978-3-65812572-1 Mijovic, P., et al.: Neuroergonomics method for measuring the influence of mental workload modulation on cognitive state of manual assembly worker. In: Longo, L., Leva, M. (eds.) Human Mental Workload: Models and Applications, H-WORKLOAD 2017, pp. 213–224. Communications in Computer and Information Science (2017) Montreuil, B.: Towards a physical Internet: meeting the global logistics sustainability grand challenge. Logist. Res. 3(2), 71–87 (2011) Orlikowski, W.: Digital work: a research agenda. In: Czarniawska, B. (ed.) A Research Agenda for Management and Organization Studies, pp. 88–96. Edward Elgar Publishing, Northampton (2016) Petermann: Psychologische diagnostik. In: Wirtz, M.A. (ed.) Dorsch – Lexikon der Psychologie (2019). https://portal.hogrefe.com/dorsch/gebiet/psychologische-diagnostik/. Accessed 17 July 2019 Phan, S., et al.: Physical Internet and interconnected logistics services: research and applications. Int. J. Prod. Res. 55(9), 2603–2609 (2017) Pilati, F., Regattieri, A.: The impact of digital technologies and artificial intelligence on production systems in today industry 4.0 environment. Netw. Ind. Q. 20(2), 16–20 (2018) Wirtz, M.A.: Psychometrie. In: Wirtz, M.A. (ed.) Dorsch – Lexikon der Psychologie (2019). https://portal.hogrefe.com/dorsch/psychometrie-1/. Accessed 12 July 2019 Richter, P., Hacker, W.: Belastung und Beanspruchung. Stress, Ermüdung und Burnout im Arbeitsleben. Asanger, Heidelberg (1998) Reiser, J., Pacharra, M., Wascher, E.: Neurophysiologische Methoden zur Erfassung von Arbeitsbeanspruchung – Ein Überblick und aktuelle Entwicklungen. Wirtschaftspsychologie, 20(1), 23–31 (2018) Rohmert, W., Rutenfranz, J.: Arbeitswissenschaftliche Beurteilung der Belastung und Beanspruchung an unterschiedlichen industriellen Arbeitsplätzen. Bundesminister für Arbeit und Sozialordnung, Referat Öffentlichkeitsarbeit, Bonn (1975) Ruiner, C., Wilkesmann, M.: Arbeits- und Industriesoziologie. Soziologie im 21. Jahrhundert. UTB, Paderborn (2016) Schaper, N.: Wirkungen der Arbeit. In: Nerdinger, F.W., Blickle, G., Schaper, N. (eds.) Arbeitsund Organisationspsychologie, pp. 475–495. Springer, Heidelberg (2011) Sendlhofer, T., Lernborg, C.M.: Labour rights training 2.0: the digitalisation of knowledge for workers in global supply chains. J. Clean. Prod. 179, 616–630 (2018) Sternberg, H., Norrman, A.: The physical Internet – review, analysis and future research agenda. Int. J. Phys. Distrib. Logist. Manag. 47(8), 736–762 (2017) Techniker Krankenkasse (TK, Ed.): Gesundheitsreport 2019 – Pflegefall Pflegebranche? So geht’s Deutschlands Pflegekräften (2019). Accessed 17 July 2019. https://www.tk.de/ resource/blob/2066542/2690efe8e801ae831e65fd251cc77223/gesundheitsreport-2019-data. pdf
Assessment of Cognitive Strain in Digital Logistics Work
515
Ulich, E.: Arbeitspsychologie. vdf Hochschulverlag, Zürich (2001) Wascher, E., Heppner, H., Hoffmann, S.: Towards the measurement of event-related EEG activity in real-life working environments. Int. J. Psychophysiol. 91(1), 3–9 (2014) Wieland, A., Handfield, R.B., Durach, C.F.: Mapping the landscape of future research themes in supply chain management. J. Bus. Logist. 37(3), 205–212 (2016)
Human Factor in Forecasting and Behavioral Inventory Decisions: A System Dynamics Perspective Kavith Balachandra1 , H. Niles Perera1(&) and Amila Thibbotuwawa1,2 1
,
Department of Transport and Logistics Management, Faculty of Engineering, University of Moratuwa, Katubedda 10400, Sri Lanka [email protected], [email protected] 2 Department of Materials and Production, The Faculty of Engineering and Science, Aalborg University, 9100 Aalborg, Denmark [email protected]
Abstract. Inventory decisions and demand forecasts are at the heart of supply chain management. Effective integration of these two activities is key to organizational success. Despite this, literature haven’t extensively explored the interrelation between judgmental inputs into forecasts and inventory decisions. This paper draws insights by connecting these two vital activities, applying system dynamics theory at a multinational company dealing in the heavy industries sector. Empirical evidence suggests that despite the growth of various statistical and computational advancements in managing inventory and forecasts, practitioners still frequently adjust the final output. The literature outlines that some practitioners are privy to information that might not be exposed to the system/model which are likely to increase supply chain performance when used effectively. The outcome of this research aids in avoiding double-counting of contextual information and improving forecast accuracy, eventually leading to better supply chain performance. Keywords: System Dynamics Judgmental forecasting Behavioral supply chain management Behavioral inventory decisions Case study
1 Introduction Supply chains (SC) are a combination of processes linked together [1–3]. Inventory management and demand management are significant components of supply chains. Smooth integration among these processes leads to organizational success but it is challenging. The heavy equipment consists of products having high complexity. These possess a high number of assemblies, sub-assemblies and component configurations. As a result, this industry has numerous stock keeping units (SKU), necessitating inventory forecasting to be automated [4]. Fully automated inventory (product) forecasting models used in practice, focus on quantitative models without capturing contextual knowledge [5]. Thus, decision-makers often adjust automated model results based on competitor effects, organization process changes, product promotions and © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 516–526, 2020. https://doi.org/10.1007/978-3-030-44783-0_48
Human Factor in Forecasting and Behavioral Inventory Decisions
517
manufacturing constraints. Similarly, demand drivers that are difficult to measure may be missing from the normative model. Adjustments to the forecasting model affects supply chain decision-making in areas such as inventory control, purchasing, production planning, cash flow planning, etc. The extent of interrelation between judgmental forecasting and inventory decision is not deeply understood [5]. Moreover, driving causes and effects are generally not well recognized. This research aims to find knowledge that helps academia and the industry to partially understand the interrelation between judgmental forecasting and inventory decisions. For this purpose, we study supply chain dynamics at a leading multinational heavy equipment manufacturer. The company has cross-functional inventory and forecasting processes. We apply System Dynamics (SD) theories to ascertain the interrelationship.
2 Literature Review 2.1
Demand Forecasting
The goal of a supply chain is to achieve high customer satisfaction by delivering quality service at an affordable cost [6]. Demand forecasting provides the initial step for a wide variety of processes in a firm including production planning, inventory decisions and pricing [7]. It consists of a process of prediction, projection or estimating expected demand of a particular product over a considered time frame [8] using time series techniques [9]. Generalized statistical forecasting models have difficulty in addressing volatile market environments due to seasonality, trend, economic conditions and special events like promotions, new product launches and market dynamics [10, 11]. Due to the difficulty of addressing aggregate market dynamics through a model, empirical evidence suggests that managers eventually adjust the output of the statistically driven demand forecasts [4] since the forecaster aims to improve the forecasting accuracy through their expert knowledge and contextual information. 2.2
Judgmental Forecasting
Demand forecasting and supply chain operations rely heavily on human decisionmaking. Judgmental forecasting takes many forms. But in general, it refers to the human influence on a final forecast. This could well be in the form of even the selection of the appropriate forecasting model [12]. Evaluation of survey-based studies reveal that judgmental forecasting could not be neglected even in well-established forecasting support systems (FSS) [5]. Judgmental forecasting and adjustments evolve in different stages in the forecasting process. Examples are candidate model defining, model selections, model parameterization, production forecasting and forecast adjustments [13]. The variants of judgmental forecasting are pure judgmental forecasting, combining forecast (the average result of several forecasts) and judgmentally adjusting the statistically derived forecast generated by a system/software [5, 14]. Survey studies reveal that the application of the above methods vary in the industry [15]. In this research we focus on the combined forecast and judgmentally adjusted statistical forecasts instead of focusing on pure judgmental forecasts since this aligns with the case company’s practices [5].
518
K. Balachandra et al.
Literature argued that judgmental forecasts are either equally precise or better than statistical forecasts [15]. Further, research illustrates that human intervention could improve forecasting accuracy when the right information drives this intervention [16]. However, managers often adjust statistically derived forecasts without having valid reasons. For instance, due to the lack of transparency of forecasting software packages (black box effect), occasionally humans tend to see patterns where there is actually no pattern. Moreover, organizational cultures have an effect on adjustments [17]. 2.3
Inventory
Demand forecasting is the initial trigger of inventory management [10]. Inventory is a significant part of working capital. Therefore, efficient inventory management is important for the sustainability and growth of a business [18]. Keeping inventory allows a firm to reduce delays and avoid stockouts to keep customers satisfied and maintain market-share [19]. Secondly, inventory is critical to smoothen the production process by balancing the production schedule. This is to distribute fixed costs associated with a production run optimally across as many products as possible [19]. 2.4
Inventory Decisions
According to literature, inventory decisions can be divided into strategic, tactical and operational level inventory decisions. In the context of inventory management, demand forecasting is a critical input for stock control models or operational level inventory decisions such as when to order, how much to order, reorder points, order up to levels, replenishment quantities and safety stocks [20]. This is triggered by personalities and motivations of those in the position to intervene with or circumvent standard inventory decisions [21]. 2.5
Effects on Forecasting Adjustment to Inventory Decisions
Most companies nowadays use inventory models through either specific inventory control software or more general ERP software [22]. However, inventory models generally rely on complete certainty of the future demand distribution while aggregate inventory cost drives performance measurement. Despite judgmental adjustment being widely practiced in the industry to enhance inventory performance, its impacts on inventory decisions have not been academically scrutinized extensively [23]. 2.6
System Dynamics
SD can be used to solve complex systems problems with a combination of quantitative and qualitative methods depending on feedback control theory and simulation technology [24]. Supply chains use SD modeling for inventory decision and policy development, demand amplification, supply chain design and integration, and international supply chain management [25]. As per SD, models judge objective systems to study and plan future action corresponding decision making of the object system. Table 1 outlines recent publications
Human Factor in Forecasting and Behavioral Inventory Decisions
519
on supply chain management with a focus on SD [25]. SD relies on information feedback and control with appropriate structuralist and functionalist systems methodologies.
Table 1. Recent industry-focused system dynamics literature Author I. P. Tama1*, Z. Akbar1, and A. Eunike1
Paulo Fernando Pinto Barcellos, Margareth Rodrigues de Carvalho Borella
Year Journal 2017 International Conference on Industrial and System Engineering 2015 International Journal of Humanities and Social Science
Ghada Elkady, 2014 International Jonathan Moizer, Journal of and Shaofeng Liu Innovation, Management and Technology 2013 International J. Mulaa*, F. Journal of CampuzanoProduction Bolarinb, M. Research Díaz-Madroñeroa and K.M. Carpioa Yang Feng 2012 International Conference on Solid-State Devices and Materials Science
Industry Vegetable product supply chain
Major findings Coordination in the supply chain increases the total supply chain profit.
Traditional four-tier Green Supply Chain & Reverse Logistics Grocery Retail- yogurt Supply Chain
Modeling framework [27] for developing customized supply chain models.
Automobile supply chain
Traditional four-tier supply chain
Reference [26]
Proposes a conceptual [28] modeling framework on small retailers’ supply chain collaboration and a decision support tool. Proposes a model to [29] improve the operational transport and procurement planning process of the SC considered. Bullwhip effect can [24] be decreased through information sharing.
3 Methodology 3.1
Research Question and Model
We aim to identify the internal influences in judgmental adjustments and finding out its effects on inventory decisions. SD enables the identification of relationships between inter-system behaviors through computer simulation at the focal firm. We focus on modeling the judgmental adjustments and inventory decision-making using SD to represent reality. The steps therein are explained below [29, 30];
520
K. Balachandra et al.
• Defining the problem: In this stage, the problem needs to be clearly verified whether the problem can be addressed using system language or not. • Conceptualization: Defining different elements in the system and find relationship influences or develop a causal loop diagram. • Formalization: Developing stock and flow diagram with model equations. • Performance: Simulating the model to define the causes. • Evaluation: Evaluation of model validity and quality. • Exploitation: Analyzing alternative policies that can be applied to the system. The supply chain model was built through model-building workshops with SC experts in the focal company while relying heavily on SD literature [22, 29, 30]. The focal company has decentralized supply chain decision-making into different echelons such as customer service, demand planning, production planning, inventory optimization and warehousing. Defining Boundaries and Model Construction Once the system domain is defined, we were able to ascertain system elements and define boundaries and gaps in the study. Inspired by prior literature (i.e.: Table 1), we use systems dynamics to investigate and model complex dynamic problems [31]. The model is derived from literature and data gathered from industry decision makers. We also used applied statistical approaches to parameterize the model attributes from the focal company data. Conceptualization The conceptualization of a problem is the most vital activity in the development of a SD model [31]. In this instance, we anchor the concept based on practices at the focal company and literature. Vensim, a well- recognized SD simulator, was used to ascertain the relationships outlined in Figs. 1 and 2. Feedback loops are also illustrated here (i.e. R: reinforcing loop, B: balancing loop) [30, 31].
+ Forecasting Qty
R
+ Adjustment ( for order confirmation process)
+ Sales
+ R
+ +
Demand
Confirmed Orders
Allocation
Stock
+
+ B
Customer Order
+
+ B Customer Satisfaction -
+ Order book
B
+
+ Planning Qty
- Delivery Lead Time
+ B
Order confirmation Lead time
Customer Complain
+
+ Adjustments + B Agreement & Restrictions
+
+ Back Orders
Fig. 1. Causal loop diagram for customer order management - order flow.
Human Factor in Forecasting and Behavioral Inventory Decisions
521
Formalization Causal loop representation allows an easy understanding of the status quo. The next step will be the model’s dynamic performance evaluation using stock and flow diagrams (phase: Formalization). First, we consider the flows between elements and formulation among two main flows: 1. Order flow: Order acceptance, confirmation and order fulfillment. 2. Goods flow: Production, inventory levels and shipments.
Compound Requirement
+
+ Coverage Plan Incentive-Production based
+
R
+
+ + Production Qty + +
R
B
+
Srap Rate + Adjustment fill up
Over Production B
Production Weight (per day) -
+ Adjustment
Productivity B + Unit Cost
+ Planning Qty B
Competition for + resources -
B
+ Working Days (Month)
+ Mold Change
Machine Breakdown -
B -
Adjustment-fill down
+
Plant Capacity
B
Preventive Maintaince
Plant Performance
-
Fig. 2. Causal loop diagram for capacity management - goods flow.
Fig. 3. Stock and flow diagram of supply chain model before judgmental adjustment (literaturebased).
Interactions among order flow and goods flow is based on literature [30]. We find that many causes influence order flow and goods flow with respect to levels, rates, auxiliaries and constants.
522
K. Balachandra et al.
Levels:
Backorder Level Desired forecast level Inventory level Rates: Order intake rate Order fulfillment rate Forecast adjustment rate Capacity rates Production start rate Shipment rates Auxiliaries: Customer order rate Accepted percentage % Accepted perceived by planners Pressure of acceptance to forecast adjustment Accepted order rates Maximum capacity Capacity utilization Capacity utilization perceived by planners Pressure of capacity utilization to forecast adjustments Delivery delay Delivery delay perceived by planners Pressure of delay dispatch to forecast adjustment
Order Intake Rate - Order Fulfillment Rate Forecast Adjustment Rate - Capacity Rate Production Start Rate - Shipment Rate Accepted Order Rate Customer Service Level Fraction of Order Dispatch Order Backlog Pressure to Forecast Adjustment + Pressure of Capacity Utilization to Forecast Adjustments + Pressure of Delay Dispatch to Forecast Adjustment Desired Forecast Working Days Capacity Rate - (Inventory Level/Working Days) Fraction of Shipment Rate Inventory Level Customer demand Market share MIN (Max. Capacity-Accepted Order Rate, Customer Order Rate) Accepted percentage Sensitivity to Planner to Accepted percentage Accepted Perceived by Planners Target Accepted percentage MIN (Capacity Rate, Max. Capacity) Daily Capacity Working Days Capacity Rate/Max. Capacity MAX (Capacity Utilization, Sensitivity Planner to Capacity Utilization-Capacity ceiling) (Max. Capacity Capacity Utilization perceived by Planners)-Capacity Rate Order Backlog Fraction of Backorders Delivery Delay Sensitivity to Planner to Delivery Delay Delivery Delay perceived by Planners Target Delivery Delay
4 Simulation and Analysis of the Model After the formulation process of the stock and flow diagrams, we simulate those diagrams (supply chain model structures before and after judgmental adjustment -i.e. Figs. 3 and 4) using Vensim 7.2. The next step of SD is performance. This simulation output facilitates building a quantitative graphical relationship between model
Human Factor in Forecasting and Behavioral Inventory Decisions
523
attributes using input data. Simulation results prove that there are three main internal causes for judgmental adjustments in the focal industry when moving into hyper stable situations [30]. These are due to factors such as order acceptance, back ordering and capacity utilization [24] .
Customer Demand
Fraction of Order Dispatch
Market share
Daily Capacity
Order Intake Rate
Order Backlog
Customer Order Rate
Working Days
Max. Capacity
Accepted Order Rate
Accepted %
Production Start Rate
Order Fulfillment Rate
Sensitivity to Planner to Accepted %
Accepted Percived by Planners
Fraction of Backorders
Delivery Delay
Delivery Delay Percived by Planners
Sensitivity to Planner to Delivery Delay
B
B Inventory Level Pressure of Acceptence to Forecast Adjustment Fraction of Shipment Rate
Shipment Rate
Desired Forecast Capacity Rate
Forecast Adjustment Rate
Capacity Utilization
Sensitivity Planner to Capacity Utilization-Capacity ceiling
Presuure of Delay Dispatch to Forecast Ajdustment
Target Delivery Delay
Target Accepted %
B
Capacity Utilization Percived by Planners
Pressure of Capacity Utilization to Forecast Adjustments
Fig. 4. Stock and flow diagram of supply chain model structure after judgmental adjustments.
Fig. 5. Simulation result of judgmental adjustment’s effects on the inventory level.
524
K. Balachandra et al.
Fig. 6. Simulation result of pure statistical Fig. 7. Simulation result of judgmental adjustforecast on inventory level based on excess ment’s effects on inventory level based on and shortage production capacities. product perishability.
5 Evaluation Simulation results show the significant relationship between inventory decisions and judgmental adjustments in the focal industry. Findings illustrate that inventory level only piles up for a shorter period when practicing judgmental adjustments. It will gradually stabilize with customer response time. In pure statistical forecasts, customer service level significantly depends on manufacturing capacity, if the manufacturing rate is lower than the customer order rate that causes poor customer service level (i.e. Fig. 6.). Due to such effect judgmental intervention need to be applied to avoid stock accumulation.
6 Conclusion This paper explains the interrelationship between judgmental forecasting and inventory decisions. The study identified the main root causes for internal environment driven factors for judgmental adjustments through analysis of the system with and without judgmental intervention. This suggests the value add of judgmental adjustments in this industry where there’s an information gap. However, erroneous adjustment in one place may lead to negative impacts on the whole supply chain. Comparing the simulation results, we find that the sensitivity level of decisionmakers affects the size of the adjustment. This study suggests that human factor (judgmental adjustment) is key for hyper stable demand forecasting and inventory decisions. Findings also indicate the salience of recruiting the most appropriate employees for specific decision-making roles. The study checks for the effect of product perishability. For both functional and perishable product inventories, inventory level piles up and stabilize over time. However perishable products inventory level stabilizes at a higher level. Future research is encouraged to explore more applied research using SD to understand the effect of human intervention on forecasting and inventory decisions. Findings of SD-based research can trigger future research using other methodologies.
Human Factor in Forecasting and Behavioral Inventory Decisions
525
Artificial intelligence (AI) based decision support systems (DSS) are likely to revolutionize supply chain decision-making in the future. Demand forecasting and inventory decision-making at the forefront of this transformation. However, AI-enabled decision support is still at an embryonic stage of development [32]. Neural network is a commonly used AI technique that can be used to formulate a DSS made up of interconnected causal factors. The translation process of the decision-making environment to a neural network is challenging. However, SD is a solid steppingstone for crafting neural networks. Interconnecting SD and neural networks can lead to the development of an improved DSS [33]. The formulated SD model covering behavioral aspects of forecasting and inventory decisions will be a vital initiative for future neural network development [34].
References 1. Perera, H.N., Sugathadasa, P.T.R.S.: Investigating manufacturing supply chain transformations: a Sri Lankan context. In: International Symposium on Logistics 2014 (2014) 2. Perera, H.N., Sudusinghe, J.I.: Longitudinal analysis of supply chain transformation project management. In: Engineering Research Conference (MERCon), pp. 153–158 (2017) 3. Perera, S., Perera, H.N., Kasthurirathna, D.: Value chain approach for modelling resilience of tiered supply chain networks. In: Engineering Research Conference (MERCon), pp. 159– 164 (2017) 4. Syntetos, A.A., Kholidasari, I., Naim, M.M.: The effects of integrating management judgement into OUT levels: in or out of context? Eur. J. Oper. Res. 249(3), 1–11 (2016) 5. Perera, H.N., Hurley, J., Fahimnia, B., Reisi, M.: The human factor in supply chain forecasting: a systematic review. Eur. J. Oper. Res. 274(2), 574–600 (2019) 6. Ramezanian, R., Saidi-Mehrabad, M.: Capacitated production planning problem considering the detailed scheduling constraints in a flow shop environment. Int. J. Manag. Sci. Eng. Manag. 7(4), 293–302 (2012) 7. Jl, S., et al.: Structure of the allosteric regulatory enzyme of purine biosynthesis. Science 264(5164), 1427–1433 (1994) 8. Bon, A.T., Leng, C.Y.: The fundamental on demand forecasting in inventory management. Aust. J. Basic Appl. Sci. 3(4), 3937–3943 (2009) 9. Armstrong, J.S.: Principles of Forecasting: A Handbook for Researchers and Practitioners, 1st edn. Springer, Berlin (2001) 10. Baecke, P., De Baets, S., Vanderheyden, K.: Investigating the added value of integrating human judgement into statistical demand forecasting systems. Int. J. Prod. Econ. 191, 85–96 (2017) 11. Trapero, J.R., Pedregal, D.J., Fildes, R., Kourentzes, N.: Analysis of judgmental adjustments in the presence of promotions. Int. J. Forecast. 29(2), 234–243 (2013) 12. Petropoulos, F., Kourentzes, N., Nikolopoulos, K., Siemsen, E.: Judgmental selection of forecasting models. J. Oper. Manag. 60(May), 34–46 (2018) 13. Petropoulos, F., Wang, X., Disney, S.M.: The inventory performance of forecasting methods: evidence from the M3 competition data. Int. J. Forecast. 35(1), 251–265 (2019) 14. Harvey, D.: Neoliberalism as creative destruction. Ann. Am. Acad. Pol. Soc. Sci. 610(1), 21–44 (2007) 15. Sanders, N.R., Manrodt, K.B.: Forecasting software in practice: use, satisfaction, and performance. Interfaces (Providence) 33(5), 90–93 (2003)
526
K. Balachandra et al.
16. Önkal, D., Gönül, M.S., Lawrence, M.: Judgmental adjustments of previously adjusted forecasts. Decis. Sci. 39(2), 213–238 (2008) 17. Onkal, D., Gönül, M.S.: Judgmental adjustment: a challenge for providers and users of forecasts. Foresight Int. J. Appl. Forecast. (1), 13–17 (2005) 18. Lemke, S.: Inventory optimization in manufacturing organizations. ProQuest Dissertation theses, p. 191 (2015) 19. Disney, S.M., Naim, M.M.: Dynamic simulation modeling for lean logistics. Int. J. Phys. Distrib. Logist. Manag. 27(4), 174–196 (1997) 20. Kapuscinski, R., Zhang, R.Q., Carbonneau, P., Moore, R., Reeves, B.: Inventory decisions in dell’s supply chain. Interfaces 34(3), 191–205 (2004). https://pubsonline.informs.org/doi/ 10.1287/inte.1030.0068 21. Bendoly, E.: System dynamics understanding in projects: information sharing, psychological safety, and performance effects. Prod. Oper. Manag. 23(8), 1352–1369 (2013) 22. Belbaǧ, S., Çimen, M., Taş, A., Tarim, Ş.: A research on corporate enterprise resource planning (ERP) systems used for supermarket supply chain inventory management in Turkey. Eur. J. Sci. Res. 38(3), 486–499 (2009) 23. Prak, D., Teunter, R.: A general method for addressing forecasting uncertainty in inventory models. Int. J. Forecast. 35(1), 224–238 (2018) 24. Feng, Y.: System dynamics modeling for supply chain information sharing. Phys. Procedia 25, 1463–1469 (2012) 25. Angerhofer, B.J., Angelides, M.C.: System dynamics modelling in supply chain management: research review, pp. 342–351 (2002) 26. Tama, I.P., Akbar, Z., Eunike, A.: Implementation of system dynamic simulation method to optimize profit in supply chain network of vegetable product. In: IOP Conference Series Materials Science and Engineering, vol. 337, no. 1 (2018) 27. Singh, S., Sharma, P.C., Barcellos, P.F.P., de Borella, M.R.C.: System dynamics as a tool for green supply chain management: a theoretical ransom. Int. J. Humanit. Sci. 5(4-1), 121–133 (2015) 28. Elkady, G.: A decision support framework to assess grocery retail supply chain collaboration: a system dynamics modelling approach. Int. J. Innov. Manag. Technol. 5(4), 232–238 (2014) 29. Mula, J., Campuzano-bolarin, F., Díaz-madroñero, M., Carpio, K.M.: A system dynamics model for the supply chain procurement transport problem: comparing spreadsheets, fuzzy programming and simulation approaches. Int. J. Prod. Econ. 51(13), 4087–4104 (2013) 30. Sterman, J.D.: Business Dynamics: Systems Thinking and Modeling for a Complex World. McGraw-Hill, New York (2000) 31. Bezemer, J.J., Akkermans, H.A.: Not with a bang, but with a whimper : understanding delays in semiconductor supply chain dynamics. In: Proceedings of the 21st International Conference Systems Dynamic Society, May 2003 32. Mehrjerdi, Y.Z., Bioki, T.A.: System dynamics and artificial neural network integration: a tool to evaluate the level of job satisfaction in services. Int. J. Ind. Eng. Prod. Res. 25(1), 13– 26 (2014) 33. Azadeh, A., Shoushtari, K.D., Saberi, M., Teimoury, E.: An integrated artificial neural network and system dynamics approach in support of the viable system model to enhance industrial intelligence: the case of a large broiler industry. Syst. Res. Behav. Sci. 31, 236– 257 (2014) 34. Kofjac, D., Skraba, A., Kljajic, M.: Neural network modeling by system dynamics methodology - basic concepts. In: EUROCON 2003, pp. 424–428 (2003)
Improving Human-Machine Interaction with a Digital Twin Adaptive Automation in Container Unloading Jasper Wilhelm1(B) , Thies Beinke1 , and Michael Freitag1,2 1
2
BIBA - Bremer Institut f¨ ur Produktion und Logistik GmbH at the University of Bremen, Hochschulring 20, 28359 Bremen, Germany {wil,ben,fre}@biba.uni-bremen.de Faculty of Production Engineering, University of Bremen, Badgasteiner Straße 1, 28359 Bremen, Germany
Abstract. The unloading of containers is a tedious task that a decreasing number of workers is willing to take on. (Semi-)autonomous systems are already available but limited to clearly defined scenarios due to a rigid level of autonomy. This paper focuses on an adaptive concept for a human-centered semi-autonomous unloading process. First, available systems are analyzed regarding their level of autonomy and the integration of the human operator. Following this, a concept integrating the digital twin and adaptive automation is presented. The usage of a digital representation with adapting autonomy allows combining the strength of humans and machines. In the presented system, the operator uses his cognitive advantage to provide specific support when the machine reaches its limits. Keywords: Cyber-physical systems · Digital twin · Autonomous systems · Human-machine interaction · Contract logistics
1
Introduction
Over the past decades, increasing globalization has led to a strong increase in the volume of transported goods [37]. A large share of these goods are shipped in containers, transported to the hinterland by train or truck and manually unloaded at their destination [33]. Due to the often heavy cartons, a harsh working environment, and the monotonous work activity, this arduous task places high physical and psychological demands on the human workers. The decreasing number of workers willing to take up such work, combined with increasing competitive pressure in the logistics sector, creates a need for autonomous systems for the container-unloading task. There is already a large number of unloading systems on the market, but no product has achieved widespread use, particular in contract logistics. Due to the limitation on specific scenarios, fully autonomous systems may often not provide c The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 527–540, 2020. https://doi.org/10.1007/978-3-030-44783-0_49
528
J. Wilhelm et al.
the reliability and availability needed in this sector. This is different for semiautonomous system, in which the task of the human changes from manual work to that of a supervisor of a semi-autonomous system. This approach is pursued within the framework of the IHATEC project “IRiS – Interaktives Robotiksystem zur Entleerung von Seecontainern” (Interactive robotic system for emptying sea containers), in which the operator can support the system during all stages of the unloading-process, following the principle of adaptive automation. The project aims to use the cognitive capabilities of the employees to cope with complex situations and at the same time to let the technical system deal with the physically strenuous activities. This contribution presents a new approach for a human-machine interaction (HMI) for remote control based on a combination of adaptive automation (AA) and the Digital Twin (DT), allowing an operator to control autonomous systems only if needed. The possibilities resulting from such a division of labor, both for HMI in general and for container unloading in particular, are the subject of this article. The remainder of this work is organized as follows. In Sect. 2, the concept of levels of autonomy (LOA), AA and DT are presented. Section 3 reviews currently available container-unloading systems and classifies them according to the LOA presented in Sect. 2. A potential framework for the application of AA in complex environments is presented in Sect. 4, followed by a description of the IRiS project as a specific application in Sect. 5. Section 6 concludes this article and presents both future work and further perspectives.
2
Fundamentals
Automation is largely understood as the allocation of physical and/or cognitive tasks to a system. In case of physical allocation and interaction, [4] distinguishes between four different scenarios (coexistence, synchronized operation, cooperation, and collaboration, see Table 1). In each case, human and machine exist in the same physical area, not separated by a fence. While in a scenario described as coexistence, human and robot are not separated by a fence, they are still not sharing the same workspace but work alongside. In a synchronized operation human and robot share their workspace, but their work is synchronized so only one entity is using the workspace at each time. When cooperating, human and robot work in the same workspace at the same time, but without contact (neither direct, nor due to the work piece). In case of collaboration, human and robot (cobot) work in the same workspace, at the same time, with a direct or indirect form of contact. This taxonomy is well suited for the description of physical collaboration. Since this work focuses on a HMI for remote control, so a more generic explanation is needed.
Improving Human-Machine Interaction with a Digital Twin
529
Table 1. Amount of physical interaction between a human worker and a robot [4] Level of cooperation Workspace Time Contact (1) (2) (3) (4)
2.1
Coexistence Synchronized Cooperation Collaboration
√ √ √
√ √
√
Levels of Autonomy
For the general automation of systems exists a vast amount of literature focusing on the LOA [34]. All analyzed publications differ concerning the processes involved but come to a conclusion, that there exist several different LOA. With an increasing LOA, the machine takes over a higher number of tasks and also more responsibility. The four-stage model by [31] represents a recognized consensus for human information processing (Fig. 1).
Acquisition of information
Analysis of information
Decision making
Implementation of action
Fig. 1. Four-stage model of human information processing [31]
The taxonomy in [16] consists of ten subsequent levels following this scheme but using slightly different wording. Monitoring is here the acquisition of information about the environment. Generation gathers the analysis of this information and the combination with additional knowledge to create a possible action. Selection is the process of selecting a single action out of all possible actions created in the previous step. Implementing is the execution of the selected task, resulting in a change in the environment. The different levels of autonomy are listed in Table 2. For further explanation on the different levels, we refer to the relevant literature. 2.2
Adaptive Automation
With increasing computational performance, the dynamic distribution of decision-making responsibility between humans and computers was first discussed in the 1970s [10,27]. The idea of AA is the optimization of cooperation and an efficient labor allocation between an automated system and its human operator [21]. It aims at a workload regulation, enhanced performance, and a reduction of mundane activities for the human operator by a dynamic management of the LOA. In adaptable systems this shift is initiated by the operator,
530
J. Wilhelm et al. Table 2. LOA taxonomy for human-computer performance [16] Level of autonomy (LOA)
Mon
Gen
Sel
Imp
(1) Manual control H H H H (2) Action support H/S H H H/S (3) Batch processing H/S H H S (4) Shared control H/S H/S H H/S (5) Decision support H/S H/S H S (6) Blended decision making H/S H/S H/S S (7) Rigid system H/S S H S (8) Automated decision making H/S H/S S S (9) Supervisory control H/S S S S (10) Full automation S S S S Mon: Monitoring, Gen: Generation, Sel: Selecting, Imp: Implementing Accountability at: Human (H), System (S)
whereas in adaptive systems, both user and system can induce such a change. The automatic shift of LOA can be triggered by critical events, operator performance, operator physiological assessment, modeling, or a combination of those [31]. One of the first uses of AA was in the aircraft sector intending to provide pilots with relevant information at the right time [26]. Other examples include software analysis [28], air traffic control [20], and driver assistant systems [7] More recently, AA has also been applied in the field of manufacturing and process control [13,35]. All these applications use adaptation as a support strategy for human decision-making, offering the LOA optimal for the human at any time. The decision for a shift in LOA in these examples is made mostly by the system. AA has shown to increase productivity and employee satisfaction [8,18]. 2.3
Digital Twin
There is a multitude of different definitions for the DT, focusing on different aspects, among other things the life cycle of the product, the complexity and detailing of the modeling or simulation, the purpose of the employment or the information exchange [25]. A production-related definition describes the DT as “[. . . ] a virtual representation of a production system that is able to run on different simulation disciplines that is characterized by the synchronization between the virtual and real system [. . . ]” [25]. Synchronization is achieved by sensory data acquisition and networked intelligent devices as well as mathematical models and real-time data analysis. Reference [22] distinguishes between model, shadow, and twin and classifies them depending on automatic or manual data information flow. A model offers no automated data transfer, while in a shadow the virtual
Improving Human-Machine Interaction with a Digital Twin
531
object is automatically updated with respect to changes in the physical world, and in the twin, the virtual object is capable of influencing the physical object. Currently, the functionalities of the DT are used in particular for predicting and optimizing the behavior of production systems. Applications include the virtual representation [1] and predictive maintenance [6] of milling machines and a model synchronization of collaborative robots [2].
3
Container Unloading Systems – State of the Art
Autonomous systems for the automated unloading of containers are a continuous topic of research. In [36] a system for individual handling of items stacked in a container is described. This and other handling technologies were applied in the RobLog project [15,32]. Due to cycle-time limitations of individual handling of items, new systems utilizing a multi-grip were developed [23,24]. Besides the limitation of the individual gripping technique, the performance of the sensory detection and subsequent object recognition is a limiting factor in the application of fully autonomous unloading systems to chaotic packing scenarios. Despite rapid progress in classification tasks, machine vision still faces different problems regarding their industrial implementation. Machine vision relies on a vast number of labels [17] and suffers severe performance losses when objects are obscured or distorted [14]. Several semi-autonomous systems on the market do not face some of the problems of autonomous systems. In these systems, the human operators perform the tasks of sensory processing, perception, and decision making concerning the gripping task. The systems are designed to relieve the user physically while increasing overall performance. Since a human operator controls them, the systems can be used in almost any scenario without compromising efficiency through prone autonomy. In the following, first an overview of automated systems for the containerunloading task for both fully- and semi-autonomous systems is given. Afterwards, the systems are classified regarding their LOA concerning the information processing as described in Fig. 1. The classification is carried out according to the definition in [16]. This assessment serves as a basis for the classification of the LOA according to Table 2. The following list serves solely as an overview of currently available systems. In addition to the solutions presented here, there are other products on the market that do not differ in their functional principle from the systems mentioned. Since the focus of this paper is on container unloading systems, only such systems are evaluated. The classification of systems is gathered in Table 3. Parcel Picker. The Parcel Picker consists of a stationary telescopic belt conveyor and an unloading unit, consisting of an operator’s platform with a gripping tool and a full-width conveyor at ground level [5]. The belt moves the discharge unit, actively steered by the system operator. An operator stands on the machine, pulling the parcels onto the conveyor system using a manually controlled and
532
J. Wilhelm et al.
actuated gripper. The packages are unloaded in bulk and transferred to the downstream conveyor system via the telescopic belt conveyor with integrated plates for directional control. The unloading process is performed by the operator, managing all tasks mentioned in Table 2. The parcel picking process is performed by the operator, using a manually guided arm. The system serves as a movable conveyor belt, supporting the operator during the transport of the parcels, which is not viewed as part of the implementation. Carton Mover. The Carton Mover consists of a mobile platform with an inclinable conveyor system at the end of which an extendable gripper is mounted on a bucket, similar to the Empticon II system developed in the RobLog project [32]. The system is equipped with an operator’s platform from where the operator controls the machine movements [9]. The system is manually controlled via two joysticks. A human operator controls the unloading process, operating bucket, gripper, and conveyor. While keeping full control of the system, the operator authorizes the action implementation, which is then performed by the machine (teleoperation) [9]. C2. The main structure of the C2 is an open linear kinematics mounted on a mobile platform. Interchangeable vacuum grippers are mounted at the end of the kinematics. They grip the objects to be unloaded and pull them onto a small panel, similar to the Parcel Robot developed in the RobLog project [32]. The parcels are placed on a conveyor system and transported to an integrated palletizing unit. The entire system can move into the container to unload cartons. The recently launched autonomous version of this system is covered in a later paragraph [11]. The machine is equipped with laser sensors, measuring all movements of the manipulator, and cameras monitoring the process. A human operator, supported by an automatic height and gripping-point correction, controls the system. The systems support the human in monitoring activities. Both human and system generate possible control options with the human maintaining control over the selection. The system performs the selected option [11]. Ultra. The Ultra mobile robot is an autonomous system developed for usage in small spaces. It consists of an omnidirectional mobile platform, mounted with an arm equipped with a conveyor belt and an end effector, similar to the Empticon II system developed in the RobLog project [32]. Stereo infrared cameras identify individual parcels. Mobile base and arm move synchronously, successively picking up parcels that are conveyed on the arm of the robot [3]. After the initial setup, the system operates autonomously. Multiple cameras with infrared stereo vision identify the scenery, including the cartons for unloading. Security scanners ensure safe operation in autonomous mode [3].
Improving Human-Machine Interaction with a Digital Twin
533
Robotic Truck Unloader and Autonomous C2. The Robotic Truck Unloader (RTU) and the autonomous version of the C2 consist of a jointedarm robot mounted on a mobile platform. A flexible gripper is mounted at the end of the arm, picking cartons and placing them on an attached conveyor belt. The entire platform can move into a container to unload cartons [12,38]. Both systems are fully autonomous. A 3D camera mounted on the gripper (RTU) or a laser scanner (C2) records the content of the container, analyzing the packing scenario. The analyzed conditions after each pick are used to make an autonomous decision about the next parcels to be handled [12,38]. Robotic Unloader. The Robotic Unloader (RU) combines a grip of multiple cartons with a moving floor, as used in the parcel Picker. The system consists of a straddle-arm with a full-width vacuum gripper mounted as the end effector. The machine is mounted on multiple wheels and can move into a container [19]. The system is fully autonomous. Machine learning and a connection to other systems are implemented to increase the performance of the autonomous system. The container can be unloaded by either picking individual cartons, entire rows or by moving the conveyor floor to unload all cartons at the same time. A human supervisor might monitor different machines from a remote location [19]. Of the six analyzed systems, three are fully autonomous with an LOA of 10. They perform all stages of the unloading process without human input, relying on a correct environmental analysis for the reliant processing of cartons. In case of unforeseen situations, these systems must pause the process and wait for a human operator at the point of work. The not fully autonomous systems are of different LOA. While the parcel picker only supports in a later stage of the unloading process, the Carton Mover supports the operator through teleoperation. The C2 supports the operator not only in the implementation but also by offering enhanced information and supported machine control. None of the semi-autonomous systems offer support in the action selection task. The non-autonomous systems can be summarized as handling aids, supporting the human in less computational-intense tasks. The operator performs the action selection and keeps responsibility for the entire process chain.
4
Adaptive Automation in Container-Unloading
It was shown that the existing systems are either fully autonomous solutions or handling aids. Although the semi-autonomous solutions rely on the cognitive capacity of the human operator mainly for the task of action selection, the operator is involved in all steps shown in Fig. 1. Thus the machine is heavily dependent on the operator who has to support the machine – to a large extent – in simple tasks. However, autonomous systems have the disadvantage that if the system intelligence fails due to unforeseen situations, direct human intervention is required. Therefore, people are still necessary in the immediate vicinity of the system or more extended downtime has to be accepted.
534
J. Wilhelm et al. Table 3. Classification of autonomous unloading solutions System
Mon
Gen
Sel
Imp
LOA
Parcel Picker H H H H 1 Carton Mover H H H H/S 2 C2 H/S H/S H S 41 Ultra S S S S 10 RTU S S S S 10 RU S S S S 10 Mon: Monitoring, Gen: Generation, Sel: Selecting, Imp: Implementing Accountability at: Human (H), System (S) 1
The C2 is rated in its semi-autonomous version. For the fully autonomous version see RTU.
By utilizing the process segmentation shown in Fig. 1, a structured collaboration between human and machine can be achieved. With a clearly defined structure of the internal logic and the artificial intelligence of the unloading system, the operator can support the semi-autonomous system precisely where it encounters problems. The system can operate fully autonomously, assuming no unforeseen situations occur. In case of a malfunction, a human operator can remotely supply additional knowledge and information exclusively at the critical task, which the system previously failed. The difference to current systems, in which a supervisor must also intervene, is the provision of clearly defined intervention points. Human-readable interfaces between the individual process steps of the systems information processing (see Fig. 1) allow the operator to understand the system’s behavior at all levels and modify or overwrite individual signals if necessary. The human operator can be supported with all previously gathered and pre-processed information as well as the raw sensor data. This enables maximum support of the human, who conversely supports the system directly at the point of failure. Once the operator supplies the missing information, the system can return into its autonomous LOA, freeing the human from repetitive tasks and resulting in an optimal LOA at every time. For the operator to engage with the different tasks of the signal processing chain, a digital representation of the system, including its internal processes and information, is necessary. The operator needs both direct access to all available data and the means to change the control as needed. The resulting cyber-physical system (CPS) needs a multi-model representation for the different tasks and bidirectional communication between human and CPS. Following the concept of the DT as presented in [22], the physical and digital object of the CPS share all task-relevant information bidirectionally. The digital object should include all information concerning its physical counterpart. Adding the human as an additional stakeholder interacting with the digital object allows for interaction
Improving Human-Machine Interaction with a Digital Twin
535
with the physical world through the information stored in the digital object. The difference between common HMI and an interaction supported by a DT is shown in see Fig. 2.
Physical Object
Legend
Digital Object
Common HMI
Human Operator
HMI using a DT representaƟon
Fig. 2. The role of the digital twin in human-machine interaction (adapted from [22])
5
Application
A system for the semi-autonomous unloading of containers is currently being built as part of the project “IRiS – Interaktives Robotiksystem zur Entleerung von Seecontainern” (Interactive robotic system for emptying sea containers). The control logic is following the previously described approach of AA using a DT representation, constructed of different sub-systems. An individual module represents each task of the four stages of the information processing (acquisition, analysis, decision making, and implementation). These modules are connected to a representation of the physical system (e.g., drive-train, kinematics, grippers), which serves as the basis for calculating the autonomous behavior of the system. The environment is sensed by the machine, which is performing the sensory processing. Four cameras with stereoscopic vision create a colored real-time representation of the surrounding environment, pre-processing the sensed information using image recognition. This information is fed into internal models for information analysis, serving as a basis for both decision making within the DT. The creation of hypotheses for the decision making is currently implemented using heuristics which are chosen by evaluating a cost function. The result of the decision-making process is given to the physical representation of the machine, performing the action implementation task. The implementation is performed by calculating the control signals, using inverse kinematics. The autonomous system is assumed to handle most of the scenarios that may occur with the user offering the necessary help to support the system from a distance. Depending on the kind of failure, the human intervenes at different stages of the processing chain. A representation of available means of interaction is shown in Fig. 3. In addition to Fig. 2, this figure is further segmented into environment, machine, and human operator, with the digital object only being a part of the machine. The machine—especially its digital object—enables the operator to interact with the environment.
J. Wilhelm et al.
State of world
Sensory processing
Environment
Analysis of information
Sensor reading
Model change Model observation Analysis change
Decision making
Logic change
Human operator
536
Decision change
Modification of world
Action Implementation
environment
Teleoperation
machine
Legend form
user
color object
task
physical
digital
human
arrow Common HMI
HMI using a DT
Fig. 3. Proposed integration of the user into the digital twin and its internal decision making process
The pre-processed data enables the worker to quickly grasp the current situation. The user can either monitor the pre-processed sensor data (e.g. image recognition to highlight edges of cartons) or virtually move around in a 3Drepresentation of the current situation. This virtual scenario is updated live, combining current sensor readings with knowledge about the physical system. Additional information is projected into this model, forecasting the results of possible control signals of the user. Figure 4 gives a brief overview of the 3Dmodel as used in the user interface. The intervention paths of the operator are manifold. With the ability to supervise all decisions made by the system, the user can modify or overwrite all results of the DT without interfering with the following steps. These decisions are of short term effect, modifying only the current information or signal. It includes the correction of the model representation, e.g., to improve the automatic detection of cartons, and the revision of the decision-making output, e.g., overwriting the selected cartons to be unloaded next. Additionally, the operator
Improving Human-Machine Interaction with a Digital Twin
537
Fig. 4. Example of guiding information in the virtual representation of the unloading scenario as presented to the user
might bypass the entire process chain of the DT, resulting in ordinary teleoperation. For long-lasting changes in the system, the operator also can change the basis of the internal processes, modifying the artificial intelligence part of the DT. This, however, would exceed the task of supervising. An evaluation of different user interfaces and control concepts for direct modification of the information and signals has been performed. The user interface and different controllers for the model observation and the teleoperation have been evaluated with potential users of the system through a laboratory pre-test. Based on the test results, the user is supported by a 3D-representation of the current situation in the container, augmented with additional information regarding the machine orientation inside the container. Comparing different means of control, a classical game-pad was found to offer the highest success rate and user satisfaction.
6
Results and Outlook
This paper presents a new approach for the integration of a human operator into the DT concept, utilizing the idea of AA. The DT representation of the cyberphysical system offers various interfaces for the human operator to take over control of a single step of the signal processing chain, eliminating small errors during different parts of the semi-autonomous signal processing. This concept allows the design of autonomous systems with a flexible LOA, which offers the optimal path of intervention for each potential failure. To operator can modify the system’s behavior by changing the parameters of autonomous tasks or by modifying the system’s outputs. This concept is currently being integrated into a semi-autonomous machine for the container-unloading task. Following a modular approach, all tasks of
538
J. Wilhelm et al.
the information processing and physical subsystems of the system are modeled individually within the DT, and are connected by standardized interfaces based on the physical system. The models are currently being implemented in Matlab and are connected to the physical system using OPC-UA and gRPC. The user interface is designed in Unity. This model-based approach allows the accuracy and computation time demands in each system to be particularly adapted to the requirements of the simulation. With the implementation of a physical simulation, new strategies for manual operation of the robot utilizing software-in-the-loop approaches can be tested. This could also allow deep learning strategies to improve the artificial intelligence by e.g., supervised learning or reinforcement learning. Simulation-based optimization would open up the possibility of developing new image recognition and unloading strategies, without encountering a potential problem first. With the digital object of the system generating novel scenarios, user and system would train using an artificial problem description. By connecting several systems, the knowledge provided by the user can also be made available in the network, supporting multiple systems without the user directly interacting with them. Acknowledgment. This work is funded by the German Federal Ministry of Transport and Digital Infrastructure (BMVI) as part of the research project 19H17016C.
References 1. Angrish, A., Starly, B., Lee, Y.S., Cohen, P.H.: A flexible data schema and system architecture for the virtualization of manufacturing machines (VMM). J. Manuf. Syst. 45, 236–247 (2017). https://doi.org/10.1016/j.jmsy.2017.10.003 ´ de la Cruz, M., Aguinaga, D.: Sustainable 2. Ardanza, A., Moreno, A., Segura, A., and flexible industrial human machine interfaces to support adaptable applications in the Industry 4.0 paradigm. Int. J. Prod. Res. 57(12), 4045–4059 (2019). https:// doi.org/10.1080/00207543.2019.1572932 3. Bastian Solutions: Robotic Truck Trailer - Case Loader/Unloader. Technical report Indianapolis (2018). https://www.bastiansolutions.com 4. Bauer, W., Bender, M., Braun, M., Rally, P., Scholtz, O.: Lightweight robots in manual assembly – best to start simply! pp. 1–61, Fraunhofer IAO (2016) 5. BEUMER Group GmbH & Co: KG: BEUMER Parcel Picker. Technical report, Beckum (2019). https://www.beumergroup.com 6. Cai, Y., Starly, B., Cohen, P., Lee, Y.S.: Sensor data and information fusion to construct digital-twins virtual machine tools for cyber-physical manufacturing. Procedia Manuf. 10, 1031–1042 (2017). https://doi.org/10.1016/j.promfg.2017.07.094 7. Calefato, C., Montanari, R., Tango, F.: Advanced drivers assistant systems in automation. LNCS (including its subseries Lecture Notes in Artificial Intelligence (LNAI) and Lecture Notes in Bioinformatics (LNBI)), LNCS (PART 2), vol. 4551, pp. 768–777 (2007) 8. Calefato, C., Montanari, R., Tesauri, F.: The adaptive automation design. In: Human Computer Interaction: New Developments. InTech (2008). https://doi.org/ 10.5772/5878
Improving Human-Machine Interaction with a Digital Twin
539
9. Carton Mover. https://www.cartonmover.com 10. Chu, Y.Y., Rouse, W.B.: Adaptive allocation of decisionmaking responsibility between human and computer in multitask situations. IEEE Trans. Syst. Man Cybern. 9(12), 769–778 (1979). https://doi.org/10.1109/TSMC.1979.4310128 11. Copal Handling Systems: C2 Container Unloader. Technical report, Heerenberg (2019). https://www.copalhandlingsystems.com/products/copal-c2/ 12. Copal Handling Systems: Container Unloader C2 mobile. Technical report, Heerenberg (2019). https://www.copalhandlingsystems.com 13. D’Addona, D.M., Bracco, F., Bettoni, A., Nishino, N., Carpanzano, E., Bruzzone, A.A.: Adaptive automation and human factors in manufacturing: an experimental assessment for a cognitive approach. CIRP Ann. 67(1), 455–458 (2018). https:// doi.org/10.1016/j.cirp.2018.04.123 14. Dodge, S., Karam, L.: A study and comparison of human and deep learning recognition performance under visual distortions. In: 26th International Conference on Computer Communications and Networks (ICCCN) (2017). https://doi.org/10. 1109/ICCCN.2017.8038465 15. Echelmeyer, W., Bonini, M., Rohde, M.: From manufacturing to logistics: development of a kinematic for autonomous unloading of containers. Adv. Mater. Res. 903, 245–251 (2014). https://doi.org/10.4028/www.scientific.net/AMR.903.245 16. Endsley, M.R., Kaber, D.B.: Level of automation effects on performance, situation awareness and workload in a dynamic control task. Ergonomics 42(3), 462–492 (1999) 17. Forster, D., Sheikh, A.S., L¨ ucke, J.: Neural simpletrons: learning in the limit of few labels with directed generative networks. Neural Comput. 30(8), 2113–2174 (2018). https://doi.org/10.1162/neco a 01100 18. Gombolay, M., Bair, A., Huang, C., Shah, J.: Computational design of mixedinitiative human-robot teaming that considers human factors: situational awareness, workload, and workflow preferences. Int. J. Robot. Res. 36(5–7), 597–617 (2017). https://doi.org/10.1177/0278364916688255 19. Honeywell Intelligrated: Robotic Unloader, Technical report, Mason (2019). https://www.intelligrated.com 20. Kaber, D.B., Perry, C.M., Segall, N., McClernon, C.K., Prinzel, L.J.: Situation awareness implications of adaptive automation for information processing in an air traffic control-related task. Int. J. Ind. Ergon. 36(5), 447–462 (2006). https://doi. org/10.1016/j.ergon.2006.01.008 21. Kay, M.: Adaptive automation accelerates process development. BioProcess Int. 4(4), 70–76 (2006) 22. Kritzinger, W., Karner, M., Traar, G., Henjes, J., Sihn, W.: Digital Twin in manufacturing: a categorical literature review and classification. IFAC-PapersOnLine 51(11), 1016–1022 (2018). https://doi.org/10.1016/j.ifacol.2018.08.474 23. Mortensen Ernits, R., Kunaschk, S., Rohde, M., Freitag, M.: Autonome Entladung ¨ von Kaffees¨ acken aus Uberseecontainern durch ein innovatives Greifverfahren. Ind. Manag. 31(6), 51–55 (2015) 24. Mortensen Ernits, R., Beinke, T., Freitag, M., Rohde, M.: Automatic unloading of coffee sacks out of sea containers – special pile situations and challenges for gripping. In: International Conference on Material Handling, Constructions and Logistics (MHCL 2019) (2019) 25. Negri, E., Fumagalli, L., Macchi, M.: A review of the roles of digital twin in CPSbased production systems. Procedia Manuf. 11, 939–948 (2017). https://doi.org/ 10.1016/j.promfg.2017.07.198
540
J. Wilhelm et al.
26. Parasuraman, R., Bahri, T., Deaton, J.E., Morrison, J.G., Barnes, M.: Theory and design of adaptive automation in aviation systems. (Progress Rep. No. NAWCADWAR-92033-60), Naval Air Development Center Aircraft Division, Warminster, PA (1992) 27. Rouse, W.B.: Adaptive Allocation of decision making responsibility between supervisor and computer. In: Monitoring Behavior and Supervisory Control, pp. 295– 306. Springer, Boston (1976). https://doi.org/10.1007/978-1-4684-2523-9 24 28. Scerbo, M.W.: Adaptive automation. In: Neuroergonomics, pp. 239–252. Oxford University Press (2006). https://doi.org/10.1093/acprof:oso/9780195177619.003. 0016 29. Schluse, M., Priggemeyer, M., Atorf, L., Rossmann, J.: Experimentable digital twins-streamlining simulation-based systems engineering for industry 4.0. IEEE Trans. Ind. Inf. 14(4), 1722–1731 (2018). https://doi.org/10.1109/TII.2018. 2804917 30. Schmidtler, J., Knott, V., H¨ olzel, C., Bengler, K.: Human centered assistance applications for the working environment of the future. Occup. Ergon. 12(3), 83–95 (2015). https://doi.org/10.3233/OER-150226 31. Sheridan, T.B., Parasuraman, R.: Human-automation interaction. Rev. Hum. Factors Ergon. 1(1), 89–129 (2005). https://doi.org/10.1518/155723405783703082 32. Stoyanov, T., Vaskevicius, N., Mueller, C.A., Fromm, T., Krug, R., Tincani, V., Mojtahedzadeh, R., Kunaschk, S., Mortensen Ernits, R., Canelhas, D.R., Bonilla, M., Schwertfeger, S., Bonini, M., Halfar, H., Pathak, K., Rohde, M., Fantoni, G., Bicchi, A., Birk, A., Lilienthal, A.J., Echelmeyer, W.: No more heavy lifting: robotic solutions to the container unloading problem. IEEE Robot. Autom. Mag. 23(4), 94–106 (2016). https://doi.org/10.1109/MRA.2016.2535098 33. UNCTAD: Quantity of loaded freight in international maritime trade from 1970 to 2017 (in million metric tons loaded). In: Statista. Statista Inc. (2018). https:// www.statista.com/statistics/234698/loaded-freight-in-international-maritimetrade-since-1970/ 34. Vagia, M., Transeth, A.A., Fjerdingen, S.A.: A literature review on the levels of automation during the years. What are the different taxonomies that have been proposed? Appl. Ergon. 53, 190–202 (2016). https://doi.org/10.1016/j.apergo. 2015.09.013 35. Valente, A., Mazzolini, M., Carpanzano, E.: An approach to design and develop reconfigurable control software for highly automated production systems. Int. J. Comput. Integr. Manuf. 28(3), 321–336 (2015). https://doi.org/10.1080/ 0951192X.2014.880810 36. Vaskevicius, N., Mueller, C.A., Bonilla, M., Tincani, V., Stoyanov, T., Fantoni, G., Pathak, K., Lilienthal, A., Bicchi, A., Birk, A.: Object recognition and localization for robust grasping with a dexterous gripper in the context of container unloading. IEEE International Conference on Automation Science and Engineering, pp. 1270– 1277 (2014). https://doi.org/10.1109/CoASE.2014.6899490 37. World Trade Organization: World Trade Statistical Review 2019. World Trade Organization, Geneva (2019). https://www.wto.org/english/res e/statis e/ wts2019 e/wts19 toc e.htm 38. Wynright Corp: Robotic Container Unloader (RTU). Technical report, Arlington (2016). https://robotics.wynright.com
Requirements for an Incentive-Based Assistance System for Manual Assembly Christoph Petzoldt1(&), Dennis Keiser2, Thies Beinke1, and Michael Freitag1,2 BIBA – Bremer Institut für Produktion und Logistik GmbH at the University of Bremen, Bremen, Germany [email protected] Faculty of Production Engineering, University of Bremen, Bremen, Germany 1
2
Abstract. The customization of products and variable market demands result in increasing product varieties and smaller product volumes. As full automation of such processes is not cost-effective yet, the assembly and inspection are often performed at manual workstations. As a consequence, assembly workers have to manage complex assembly processes with a wide diversity of assembly components and varying assembly steps. This increases the need for individual assistance in modern assembly systems. So far, assistance mainly focuses on some process-related aspects of assembly processes, while system acceptance, motivational aspects and individual support needs of the worker are not considered. Therefore, based on theoretical discussions and expert interviews, this paper defines requirements for human-centered assistance systems that combine individual assistance with incentive systems. In a case study, the obtained requirements for incentive-based assistance systems are applied to a modeled assembly process of an extruder for a 3D printer. Finally, general implications and dependencies of the requirements on manual assembly are discussed. Keywords: Assistance systems Manual assembly Requirements definition Operator 4.0 Incentive-based assistance Industry 4.0
1 Introduction The paradigm of fully automated production systems was seen as the target structure for future production systems [1]. The trend of customized products and strong customer orientation leads to increasing varieties and smaller batch sizes, down to batch size one [2, 3] and mass customization [4]. This results in highly complex assembly systems where automation is often not suitable for the requirements of the market [5]. To meet the demand for flexibility and make the complexity of processes manageable, manual or hybrid workstations are used. The cognitive abilities of humans enable these production systems to manage the high degree of complexity. Furthermore, unpredictable events can be solved efficiently and quickly. However, the assembly process has a key impact on the performance of an industrial company [6] as the assembly of manufactured products accounts for more than 50% of total production time and 20% of total production costs [7]. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 541–553, 2020. https://doi.org/10.1007/978-3-030-44783-0_50
542
C. Petzoldt et al.
Therefore, it is necessary to support the employees in manual assembly stations. The fourth industrial revolution (Industry 4.0) represents a great opportunity for workers to become part of the intelligent system [8] as new types of interactions between operators and machines become available [9]. However, the integration of workers into Industry 4.0 systems is a significant challenge [10]. Therefore, the humancentered concept of Operator 4.0 was proposed by [11, 12] to analyze these challenges regarding different support dimensions and to facilitate human-machine interaction and cooperation using novel technologies. Assembly workers can be supported with context-related information using digital assistance systems [13]. The assistance systems support the employees to cope with complexity in highly flexible production structures. Assisting systems such as pick-by-light and control systems using multisensors are state of the art in industrial practice. Other solutions provide information via projection or head-mounted displays during the assembly process [14]. The systems mainly focus on the product and process requirements, while the separate consideration of employee requirements is insufficient. The acceptance by the employees is one of the key factors and the lack of consideration leads to unused potentials of production and assembly systems [15]. Based on the characteristics of assembly systems, typical fields of application can be defined. High flexibility, small to medium lot sizes, wide variety, and average productivity are characteristics of manual assembly, which result in a high need for assistance. Figure 1 illustrates the sensible fields of application by means of the described criteria. Variety large
small
high
Flexibility
Productivity
low
Field of application for assembly assistance systems
low
high low
high
Lot size
Fig. 1. Field of application for assembly assistance systems (adapted from [16]).
For the described field of application, our objective is to develop an assembly assistance system which considers both the production-related and the employee requirements sufficiently. Hence, we determine the relevant requirements in the following sections. In the second section of this paper, we present related work on assistance systems and incentive systems and review state-of-the-art assembly assistance systems. Requirements for incentive systems are known from management science, and requirements for assembly assistance systems are discussed in some publications and
Requirements for an Incentive-Based Assistance System
543
standards as well [17–19]. So far, however, the requirements for systems that combine assembly assistance with incentive-based methods for a human-centered assembly assistance system have not been defined. Therefore, in the next section, the authors consider the practical point of view using expert interviews and summarize the results of theoretically analyzed requirements and expert interviews, which result in requirements for incentive-based assistance systems. Eventually, in the fourth section, we apply the requirements in a case study exemplary for the manual assembly of an extruder for a 3D printer. By modeling the assembly process using process management tools, the impact of incentive and assistance systems is shown and finally, dependencies between different requirements are discussed.
2 Related Work on Assistance Systems and Incentive Systems The term assistance is broadly defined and therefore, a precise definition of assistance systems is difficult [15]. According to [20], assembly assistance systems are technical systems that receive and process information via sensors and other inputs to assist employees in carrying out their assembly tasks. These systems are classified based on types of assistance and are distinguished between physical and information assistance [21]. Energetic assistance aims to reduce the physical strain of employees. Information assistance systems provide employees with the right information at the right time [20]. A more detailed classification is proposed by [22]. Accordingly, assistance systems can be classified by the degree of support, the type of support and its objective. We propose the enhancement of the taxonomy regarding the assistance system location. Due to the practical importance of flexibility at the point of use, this dimension is included. Figure 2 shows the classification of assistance systems in morphological structure and outlines the characteristics of the assistance systems for manual assembly discussed in this paper.
Assistance System Characteristics
Characteristic values Mobile
Location
Degree of support
Type of support
Objective of support
Low Physical (e.g. exoskeleton, manipulator) Compensational
Stationary
Medium
High
Variable
Cognitive (e.g. projection-supported assembly assistance system)
Sensorial (e.g. augmented reality supported picking)
Preserving
Expansion
Depending on application case
Fig. 2. Morphological structure for classification of assistance systems and desired properties of assistance systems for manual assembly (highlighted in gray). The degree of support often depends on the complexity of the assembly product; however, a variable and individualized support is desired. The objective of support highly depends on the application case; for example, exoskeletons can be used to support operators with handicaps (compensational), for exhaustive or ergonomically unfavorable tasks (preserving) or for lifting and handling very heavy components (expansion).
544
C. Petzoldt et al.
According to the classification scheme, all types of assistance systems can be found in industrial production. Particularly assembly, maintenance, and logistic processes are supported by assistance systems. The main objective for the implementation is to increase productivity by avoiding errors and accelerating the work processes. Further goals are the improvement of ergonomics and safety. Besides, assistance systems facilitate the inclusion of workers so that even less-qualified employees are able to fulfil the job requirements. The wide range of applications and characteristics is also shown by a morphology for assembly assistance systems introduced by [18] and a typology for the Operator 4.0 [10]. As highlighted in Fig. 2, we focus on stationary, human-centered assistance systems for manual assembly. Within this field of consideration, state-of-the-art assistance systems were identified on the basis of multidimensional search criteria and analyzed according to their assistance functions. Our research was based on publications, technical data from system providers and trials on exhibitions to investigate systems from both research and industry. As summarized in Table A1 in the appendix, the considered state-of-the-art assistance systems mainly focus on cognitive support of the worker by optical pick-to-light systems highlighting the required assembly components. Some systems additionally use multisensory cameras in order to monitor the assembly progress or correctness and completeness of the assembled product. Partly, also the projection of additional process information related to the current assembly step is implemented. Only very few assistance systems consider the possibility of integrating incentive approaches and individualized support. Also, the ergonomic situation of the worker is analyzed only by a single assistance system which automatically adjusts the height of the assembly station before starting work. However, as inferred from Table A1, a system that extensively combines individualized assembly assistance with incentives and ergonomic aspects does not yet exist. Incentive systems are used in all hierarchical levels of companies. Incentive systems can be classified into mechanical, economic and behavioral incentive systems. Mechanical incentive systems are based on Taylor’s motivation theory using only monetary incentives. Economic systems extend mechanical incentive systems utilizing individualized monetary incentives. In the field of assembly, companies traditionally only apply monetary incentives. However, according to motivational research, the focus on monetary incentives does not lead to a permanent motivation of employees [23]. Especially in relatively simple tasks, performance is highly dependent on motivation [24]. Therefore, in manual assembly, the mere use of monetary incentives may lead to unmotivated employees and a performance drop of the assembly stations. Furthermore, the quality decreases and error rates increase. Therefore, non-monetary incentives are increasingly important for the development of modern incentive systems so that behavioral incentive systems are proposed. An approach for the implementation of behavioral incentives is to use game design elements (e.g. scores, best-lists and team best-lists, achievements and insignia, target values, challenges, attainable levels, progress indicators, and narratives) in non-gaming contexts [25], which is best known as gamification [26]. Gamification is increasingly gaining the attention of academics and practitioners as studies measure an increase in performance and efficiency in production and logistic systems [27, 28].
Requirements for an Incentive-Based Assistance System
545
3 Assistance Systems Requirements in Manual Assembly For the successful development of an effective assembly assistance system, the definition of requirements is important. Therefore, the authors gathered information from industrial developers, standards and technical literature detailed below. By summarizing and analyzing the given information, we define and discuss three main requirement categories for assembly assistance systems that must be considered: technical, process-related and organizational requirements. We extend these with the requirement of supporting employee motivation, which together form the four key requirements for an incentive-based assistance system for manual assembly. Technical Requirements. For assembly assistance systems, the human-machineinteraction (HMI) plays an important role. As assembly assistance systems are sociotechnical systems, the technical requirements are mainly related to this. [18] state that accurate and targeted information provision is a crucial aspect for the design of assembly assistance systems to avoid unnecessary movements and assembly errors. However, to achieve the acceptance of the worker, assistance systems should not provide redundant or superfluous information leading to the requirement of individual assistance. We extract further requirements from an established standard in HMI, which is the ISO 9241 part 110 [19]. Based on this, suitability for the task, self-descriptiveness, controllability, conformity with user expectations, error tolerance, suitability for individualization, and suitability for learning, are requirements for incentive-based assembly assistance systems. To support the acceptance among the operators, all the listed aspects should be considered. Furthermore, ergonomic and safety aspects and regulations are important technical requirements that must be considered in assembly assistance systems. The safety requirements must be taken into account particularly in case of movable parts (e.g., robots for physical support). Process Requirements. Process related requirements for assembly assistance systems primarily derive from the product requirements, as well as the general requirements at present assembly systems. The essential requirement in industry is high quality, aiming at 100% quality [17]. In order to achieve the quality goals in assembly, process reliability is critical and forms a fundamental requirement for the development of assembly assistance systems. Consequently, assembly assistance systems should implement automatic quality control functions for the assembled products. The frequently changing market demand is a challenge for assembly systems. In order to cope with uncertainties in terms of lot sizes and variety, assembly systems need to be flexible [4]. Flexibility is one of the key success factors for the future of manufacturing and is, therefore, a crucial requirement for incentive-based assembly assistance systems.
546
C. Petzoldt et al.
Organizational Requirements. Economic efficiency is defined as a requirement for assistance as well as incentive systems. To accomplish this organizational requirement, the financial benefit of the incentive-based assembly system should be higher than the investment costs. With regard to increasing acceptance, the system advantages regarding economic benefits and improvement of working procedures must be clearly visible for the operators, which we refer to as the requirement of transparency. Finally, an obvious requirement is data safety and security [17]. Assembly assistance systems are usually integrated into the existing data network structures of companies (e.g. ERP-Systems). Hence, the system is a potential target for cyberattacks. When assistance systems collect person-related data, e.g. for the implementation of individualized assistance, IT security is also important from the point of view of the operator. Requirement of Supporting Employee Motivation. As detailed in the second section, the motivation of the worker influences task performance. Therefore, an assembly assistance system should encourage motivation in a targeted and long-term manner. As monetary incentive systems do not provide sustainable motivation [23], we suggest the implementation of behavioral incentive systems as a requirement for incentive-based assembly assistance systems.
4 Requirements Impact Analysis Based on Assembly Process Modelling After the requirements for incentive-based assistance systems have been defined, the implications on assembly processes are discussed in this section for a single operator assembly station with an informational assistance system. To analyze the effects on the process and to ensure a goal-orientated development of the assistance system, a granular process modeling of the current process flow is necessary. According to the VDI guideline 2860 (technology for handling and assembly), the principal activities of assembly processes are joining, handling, fitting, controlling, and auxiliary functions [29]. With these activities and based on the modeling language Business Process Model and Notation (BPMN), a detailed process flow can be created. For our case study, we select a product that is typically assembled at manual workstations, namely an extruder for a 3D printer. Figure 3 shows a picture of the extruder and an excerpt of the modeled assembly process. The workstation is already equipped with an assistance system, which guides the operator through the entire assembly process. On a screen in front of the worker, all work steps including detailed instructions are displayed. After a successfully completed work step, the operator confirms the finished step manually. Subsequently, the assistance system provides new
Requirements for an Incentive-Based Assistance System
547
Suitable for automatic execution by intelligent assistance system
Suitable for further cognitive assistance
Pick engine
Pick Retainer
Check cable
Engine in assembly device
Auxiliary functions
Legend:
Assistance System
Extruder for a 3D printer
Mount Retainer
Fitting
Operator
Controlling Handling
Joining
instructions and new parts can be taken from the assembly part boxes. The picking of assembly components is supported by a pick-by-light system. A separate quality check (e.g., by camera-based assembly inspection) is not included.
Confirmation
Process approval
Confirmation
Picking request
Fig. 3. Potentials of incentive-based assistance systems illustrated by a modeled process flow for the assembly of an extruder for a 3D printer. Process steps highlighted in blue are suitable for automatic execution by an intelligent assistance system. Process steps in yellow can be supported with further cognitive assistance. In addition, continuous incentive systems can be implemented to motivate the employee and thus affect the performance of all manual assembly process steps.
Following the defined requirements, the authors propose to extend the current assembly system with additional hardware and software. Optimization potentials arise in particular due to the manual confirmation after every process step. With appropriate camera technology and intelligent monitoring algorithms, these steps can be executed automatically by an assistance system. Other steps, such as manual controlling steps (e.g. the check of cable shown in Fig. 3) can be performed automatically using cameras and industrial image processing methods as well. This would also improve the process reliability due to assured quality checks. Additionally, we propose to cognitively assist
548
C. Petzoldt et al.
the worker firstly by projecting mounting positions and individual, qualificationdependent further information on the worktop. Secondly, to support the motivation of the worker, we propose to apply gamification on the remaining process steps. To meet the defined technical requirements, intuitive and simple design of the gamification application is needed. Appropriate dimensioning of the game design elements is crucial for the success of gamification applications. The selection of game design elements follows the self-determination theory and the underlying basic needs. A classification and allocation of game design elements and basic needs is introduced by [30]. However, implementation costs should also be considered, as for example, designing a narrative can be more time-consuming than defining performance indicators to determine a score.
Fig. 4. Causal chain explaining the dependencies between the defined requirements for incentive-based assembly assistance systems.
Requirements for an Incentive-Based Assistance System
549
From the modeled process flow of the case study, we infer general propositions for the potentials of incentive-based assistance systems in manual assembly by considering the fundamental assembly process steps. Firstly, control and auxiliary functions can be automatically executed by intelligent assistance systems. Consequently, such process steps can be omitted from the manual process workflow. Secondly, we suggest that the remaining process steps, i.e. handling, fitting and joining, are supported variably, individually and cognitively, with the primary objective of increasing performance. Additionally, this variable degree of support is especially useful for the training phase of new employees and the launch of new product varieties. Thirdly, in addition to supporting or automating single process steps, we propose to provide incentives throughout the entire assembly process. As incentive systems contribute to increasing motivation and performance of the operator, this positively affects all manual process steps. Figure 4 summarizes the defined key requirements and shows their dependencies. The defined technical requirement constitute the basic prerequisite for achieving system acceptance by the operator to generate satisfaction with the work situation. Incentive systems (e.g., gamification methods) can be used supplementary to achieve and maintain the motivation of the operator in the long term. The willingness to perform is directly influenced by the motivation, which underlines the importance of both motivation-increasing incentive systems and the implementation of the technical requirements. Therefore, the fulfillment of the technical requirements (e.g. by implementing of automatic quality control or supporting the worker’s abilities through individual assistance) and the attainment of an adequate motivation (e.g. potentially supported by the implementation of incentive systems) leads to the fulfillment of the process requirements. In turn, high process reliability, product quality as well as the flexible reaction to variable market situations and a high number of product varieties ultimately increases the overall performance of the assembly system. Finally, this results in the fulfillment of the defined organizational requirement, namely economic efficiency.
5 Conclusion and Outlook In this paper, we propose an incentive-based assistance system for manual assembly, which combines individual assistance and incentives. Since the requirements for assistance and incentive systems in assembly have so far only been considered separately, we defined the requirements for incentive-based assembly assistance systems. Thereafter, this paper discussed the impacts of the defined requirements on manual assembly processes based on a case study. A gamification approach has been described as a solution for an incentive system in manual assembly environments to design nonmonetary incentives.
550
C. Petzoldt et al.
In our ongoing research, we focus on the development of a stationary, cognitively and individually supporting, incentive-based assistance system based on the requirements defined above, which considers both the assembly product and the employee at manual assembly stations. Our assistance system uses an intelligent vision system to determine the progress of the assembly process for automatic confirmation of a completed assembly step. Also, we implement an automated quality control system analyzing the quality of the assembly product and its assembly completeness. Furthermore, we derive information about the execution speed and reaction rate of the worker from the vision system. From this data and the complexity of the assembly product, we calculate the qualification level as well as the individual support needs and adapt the assistance system accordingly. Moreover, we evaluate the ergonomic work situation of the operator. Particularly, we analyze the body posture to advice the worker with potentials for ergonomic improvements. In addition, we apply performance incentives using an intelligent gamification approach to motivate the worker with individualized, short- and long-term goals. Hence, with the proposed incentive-based assistance system, we facilitate human-machine cooperation leading to a healthy, smarter, augmented and social operator in relation to the Operator 4.0 typology. For our future research on incentive-based assembly assistance systems, we intend to investigate the characteristic of preventive support with regard to ergonomics during the assembly work processes more closely. By analyzing the ergonomic situation of the operator during work, the height and dimensions of the assembly station can be adjusted. In addition, learning from the execution manner of conducted work over time enables optimization of the assembly station configuration, e.g., with regard to the positioning of component boxes. With this approach, the currently being developed individual, incentive-based assistance system could be extended by automatic physical individualization of manual assembly stations. Acknowledgment. The authors would like to thank the European Regional Development Fund (EFRE) and the Bremer Aufbau-Bank (BAB) for their support within the project AxIoM Gamified AI assistance system for support of manual assembly processes (funding code: FUE0619B).
Requirements for an Incentive-Based Assistance System
551
Appendix See Table A1. Table A1. Overview of state-of-the-art assistance systems for manual assembly. Assistance functionalities are analyzed with particular focus on incentive systems, ergonomic assistance, and individualization of support.
Active-Assist
AS Pro Assembly Pro
Der Schlaue Klaus Der Assistent
ELAM EMU (electrical assembly support) Ergonomic Assembly 4.0 HumanInterfaceMate Laser projectors
motionEAP Nexonar Assembly Scout OAS (OnePiece-FlowAssistance System)
QualityAssist xtend Smart Assembly Trainer
Bosch Rexroth AG Assembly Solutions GmbH LAP GmbH Laser Applikationen Optimum datamanagem ent solutions GmbH Ulixes Robotersyste me GmbH Armbruster Engineering GmbH & Co. KG
MiniTec Smart Solutions GmbH
Individualization of support
Human-MachineInterface
Camera
commercial
yes
PickbyLight
Screen, Projection, Gestures, Speech
-
commercial
yes
-
Screen, Projection
-
commercial
Screen
Camera
commercial
Camera
commercial
Camera (optional)
commercial
Camera
commercial
Camera
research
yes
yes
yes
-
yes
partially
yes
PickbyLight
partially
PickbyLight
yes
yes
PickbyLight
yes
yes
yes
partially
yes
(Gamification)
yes
yes
partially
yes
Gamification [17]
Projection
Screen, Projection
Screen, Projectio, Augment ed Reality Ergonomic workstationadjust ment
-
-
PickbyLight
Projection
Camera
commercial
-
Projection
-
commercial
Camera
research
Hand tracking by motion sensor on gloves
commercial
PickbyLight
Projection, Adaptive projection Augmented feedback (3profiles) Reality
Screen, Hand Position Augmented Tracking Reality
yes
yes
yes
yes
PickbyLight
Screen
Camera detecting markers on watch
research
yes
yes
Hand Position Tracking
Screen
Ultrasonic sensor with data gloves
commercial
yes
PickbyLight
Screen, Projection, Gestures
Camera
research
Sarissa GmbH
Fraunhofer IOSB
Stage
Projection
Arkite B.V.
soft2tec GmbH Fraunhofer IFF; Treston Germany GmbH
Monitoring sensors
PickbyLight
yes
yes
Fraunhofer IPA
Z-Laser GmbH Schnaithmann Maschinenbau GmbH; University of Stuttgart
Picking assistance
Incentive system
Ergonomics analysis
Developer or Producer
Assembly instruction control Automated assembly quality control Automatic assembly step confirmation
Assistance functions Assembly assistance system
yes
yes
Specific training phase program
552
C. Petzoldt et al.
References 1. Zäh, M.F., Beetz, M., Shea, K., Reinhart, G., Bender, K., Lau, C., Ostgathe, M., Vogl, W., Wiesbeck, M., Engelhard, M., Ertelt, C., Rühr, T., Friedrich, M., Herle, S.: The cognitive factory. In: Changeable and Reconfigurable Manufacturing Systems, pp. 355–371. Springer, London (2007) 2. Zhang, Z.: Manufacturing complexity and its measurement based on entropy models. Int. J. Adv. Manuf. Technol. 62, 867–873 (2012). https://doi.org/10.1007/s00170-011-3872-7 3. Stecken, J., Linsinger, M., Sudhoff, M., Kuhlenkötter, B.: Didactic concept for increasing acceptance of consistent data standards using the example of assistance systems in assembly. Procedia Manuf. 31, 277–282 (2019). https://doi.org/10.1016/j.promfg.2019.03.044 4. Koren, Y.: The Global Manufacturing Revolution: Product-Process-Business Integration and Reconfigurable Systems. Wiley, Hoboken (2010) 5. Müller, R., Vette-Steinkamp, M., Hörauf, L., Speicher, C., Bashir, A.: Worker centered cognitive assistance for dynamically created repairing jobs in rework area. Procedia CIRP 72, 141–146 (2018). https://doi.org/10.1016/j.procir.2018.03.137 6. Andolfatto, L., Thiébaut, F., Lartigue, C., Douilly, M.: Quality- and cost-driven assembly technique selection and geometrical tolerance allocation for mechanical structure assembly. J. Manuf. Syst. 33, 103–115 (2014). https://doi.org/10.1016/j.jmsy.2013.03.003 7. ElMaraghy, H., ElMaraghy, W.: Smart adaptable assembly systems. Procedia CIRP 44, 4–13 (2016). https://doi.org/10.1016/J.PROCIR.2016.04.107 8. Peruzzini, M., Grandi, F., Pellicciari, M.: Exploring the potential of Operator 4.0 interface and monitoring. Comput. Ind. Eng. 105600 (2018). https://doi.org/10.1016/j.cie.2018.12.047 9. Lorenz, M., Rüßmann, M., Strack, R., Lueth, K.L., Bolle, M.: Man and machine in Industry 4.0. Bost Consult Gr 18 (2015) 10. Ruppert, T., Jaskó, S., Holczinger, T., Abonyi, J.: Enabling technologies for Operator 4.0: a survey. Appl. Sci. 8 (2018). https://doi.org/10.3390/app8091650 11. Romero, D., Bernus, P., Noran, O., Stahre, J., Fast-Berglund, Å.: The Operator 4.0: human cyber-physical systems & adaptive automation towards human-automation symbiosis work systems. In: IFIP Advances in Information and Communication Technology, pp. 677–686. Springer, Cham (2016) 12. Romero, D., Stahre, J., Wuest, T., Noran, O.S., Bernus, P., Fast-Berglund, Å., Gorecky, D.: Towards an Operator 4.0 typology: a human-centric perspective on the fourth industrial revolution technologies. In: International Conference on Computers & Industrial Engineering (CIE46), Tianjin, China (2016) 13. Keller, T., Bayer, C., Bausch, P., Metternich, J.: Benefit evaluation of digital assistance systems for assembly workstations. Procedia CIRP 81, 441–446 (2019). https://doi.org/10. 1016/j.procir.2019.03.076 14. Lampen, E., Teuber, J., Gaisbauer, F., Bär, T., Pfeiffer, T., Wachsmuth, S.: Combining simulation and augmented reality methods for enhanced worker assistance in manual assembly. Procedia CIRP 81, 588–659 (2019). https://doi.org/10.1016/j.procir.2019.03.160 15. Sochor, R., Kraus, L., Merkel, L., Braunreuther, S., Reinhart, G.: Approach to increase worker acceptance of cognitive assistance systems in manual assembly. Procedia CIRP 81, 926–931 (2019). https://doi.org/10.1016/j.procir.2019.03.229 16. Lotter, B., Wiendahl, H.-P.(Hrsg): Montage in der industriellen Produktion, 2 Auflage. Springer, Berlin, Heidelberg (2012) 17. Korn, O.: Context-aware assistive systems for augmented work. A framework using gamification and projection. Universität Stuttgart (2014)
Requirements for an Incentive-Based Assistance System
553
18. Hinrichsen, S., Riediger, D., Unrau, A.: Assistance systems in manual assembly. In: Proceedings 6th International Conference Production Engineering and Management, 29 September–30 September 2016, in Lemgo, Germany, pp. 3–14 (2016) 19. DIN EN ISO 9421-110: Ergonomie der Mensch-Maschine-Interaktion. Teil 110: Grundsätze der Dialogestaltung, Germany, Berlin (2006) 20. Hinrichsen, S., Bendzioch, S.: How digital assistance systems improve work productivity in assembly. Adv. Intell. Syst. Comput. 781, 332–342 (2019). https://doi.org/10.1007/978-3319-94334-3_33 21. Reisinger, G., Komenda, T., Hold, P., Sihn, W.: A concept towards automated data-driven reconfiguration of digital assistance systems. Procedia Manuf. 23, 99–104 (2018). https:// doi.org/10.1016/j.promfg.2018.03.168 22. Apt, W., Bovenschulte, M., Priesack, K., Weiß, C., Hartmann, E.A.: Einsatz von digitalen Assistenzsystemen im Betrieb, Berlin (2018) 23. Schulz, V.: Nichtmaterielle Anreize als Instrument der Unternehmensführung. Deutscher Universitätsverlag (2000) 24. Van Knippenberg, D.: Work motivation and performance: a social identity perspective. Appl. Psychol. 49, 357–371 (2000). https://doi.org/10.1111/1464-0597.00020 25. Deterding, S., Dixon, D., Khaled, R., Nacke, L.: From game design elements to gamefulness: defining “gamification,” Zugl.: dis. Wilhelm Fink Verlag, München (2011) 26. Mekler, E.D., Brühlmann, F., Tuch, A.N., Opwis, K.: Towards understanding the effects of individual gamification elements on intrinsic motivation and performance. Comput. Hum. Behav. 71, 525–534 (2017). https://doi.org/10.1016/j.chb.2015.08.048 27. Warmelink, H., Koivisto, J., Mayer, I., Vesa, M., Hamari, J.: Gamification of production and logistics operations: Status quo and future directions. J. Bus. Res. 1–10 (2018). https://doi. org/10.1016/j.jbusres.2018.09.011 28. Pötters, P., Klöckner, I., Leyendecker, B.: Gamification in der Montage - Untersuchung von Motivations- und Performancesteigerung bei Mitarbeitern. ZWF Zeitschrift für wirtschaftlichen Fabrikbetr 112, 163–167 (2017). https://doi.org/10.3139/104.111679 29. VDI Verein Deutscher Ingenieure: VDI guideline 2860 - Montage- und Handhabungstechnik: Handhabungsfunktionen, Handhabungseinrichtungen; Begriffe, Definitionen, Symbole (1990) 30. Beinke, T., Freitag, M., Schamann, A., Feldmann, K.: Gamification im E-Learning in Verbindung mit individueller Spieleapplikation für die mitarbeiterorientierte Weiterbildung der Zukunft. Ind. 4.0 Manag. 2, 13–17 (2019)
Evaluation of Human-ComputerInteraction Design in Production and Logistics by Using Experimental Investigations Moritz Quandt1(B) , Hendrik Stern1 , Supara Grudpan3 , Thies Beinke2 , Michael Freitag1,2 , and Rainer Malaka3 1
2
BIBA - Bremer Institut f¨ ur Produktion und Logistik GmbH at the University of Bremen, Bremen, Germany {qua,ste}@biba.uni-bremen.de Faculty of Production Engineering, University of Bremen, Bremen, Germany {ben,fre}@biba.uni-bremen.de 3 Digital Media Lab, TZI, University of Bremen, Bremen, Germany {sgrudpan,malaka}@uni-bremen.de
Abstract. HCI, as the cooperation of humans and machines in sociotechnical systems, is playing an increasingly important role in production technology and logistics applications. This cooperation will lead to a higher occurrence of interaction among these actors in production and logistics work. One of the central challenges in this context is the development of interfaces between humans and machines. In this paper, the evaluation of HCI systems is examined based on three use cases. The participation of users in the evaluation of HCI system development is investigated by means of the design of human work in Cyber-physical production systems, the use of Augmented Reality as a promising new technology in production scenarios, as well as on the example of a serious game-based decision support in urban logistics. The implementation of user-based evaluation studies in the use cases shows that such studies can increase user acceptance of HCI systems in the context of production and logistics. Especially in connection with new technologies, this can be achieved by the early development of prototypes and the participation of the user over the entire development process. Keywords: User-centered evaluation Experimental investigations
1
· Human Computer Interaction ·
Introduction
Cyber-physical systems (CPS) are technological systems in which computers are integrated into physical elements (embedded systems). These CPS are linked with each other and share information. This results in an increased product and c The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 554–566, 2020. https://doi.org/10.1007/978-3-030-44783-0_51
Human-Computer-Interaction Design in Production and Logistics
555
system intelligence [1,2]. There are various applications for CPS in production. For example, autonomous cooperating production processes, usage of collaborative robots, or preventive maintenance can be used and implemented [3–5]. Subsequently, human work in future manufacturing and logistics systems will be transformed by CPS due to the technical and organizational changes as described above. Experts expect the development of hybrid production systems, which rely on collaborative work between people and machines [6]. Here, Human-Computer Interaction (HCI) will be ubiquitous. HCI, as the cooperation of humans and machines in socio-technical systems, is, therefore, playing an increasingly important role in production technology and logistics applications. The collaboration between humans and machines will lead to increased importance and occurrence of interaction among these actors in production and logistics work. One of the promising technologies in this context is Augmented Reality (AR). AR is attributed a high potential for the provision of work process integrated support as well as context-related information to support people directly in the work process [7,8]. In contrast to other fields of application such as medicine, AR has so far only been used in a few industrial application scenarios. This is due to the often high complexity, mobility demands, and special requirements of the working environment [9,10]. The successful use of AR solutions in industrial practice is directly related to the acceptance of AR solutions by end-users [10,11]. Therefore, increased user participation in the development of AR solutions for industrial practice is desirable. In logistics, due to the complexity of planning and optimization problems, there is an increased demand for collaboration among all stakeholders [12,13]. For example, in the field of urban logistics, stakeholders (such as city authorities or shipping companies) have different requirements and goals [14], which lead to different technologies and tools used [15,16]. Among others, digital board games using gamification techniques are attributed promising capabilities in order to foster collaboration and a common understanding among the different stakeholder groups by a better understanding of the underlying mechanics [17]. The quality of HCI at these new production and logistics work and training systems will be of crucial importance. Here, among others, the design of user interfaces can be named as a central challenge [18]. This raises the research question, how HCI can be evaluated in the described scenarios? We claim that a user-centered approach is required to design the devices in a way, which is leading to improved performance and human-oriented work. In practice, this has so far often been contrasted by a technology-centered development that does not adequately meet the needs of users [19]. In this paper, we give an overview of applying HCI in production and logistics and highlight some developments that address research challenges in these fields, based on the scenarios of human work in future manufacturing systems, AR for work process support, and gamification in urban logistics. The mentioned use cases have a common method. This method involves the user in the design process and allows to evaluate the behavior of the users with the aim of improving the system. We first describe evaluation methods for HCI systems.
556
M. Quandt et al.
These scenarios share the idea of a user-centered evaluation approach. Then, we provide three use cases that apply HCI to production and logistics application scenarios. They show possible application scenarios and benefits of this approach to production and logistics. Finally, we summarize how user-centered evaluation addresses the challenges in production and logistics and how it can promote the development of human-oriented systems instead of technology-centered systems only.
2
User-Centered Evaluation of HCI Systems
HCI is a part of every socio-technical work system in which people and technology collaborate. This refers both to systems in which the machines are merely human work equipment and to systems in which humans and machines act as collaborating, independent actors. It enables two-way dependent communication between humans and machines [20]. HCI design principles focus on the accessibility of human-machine interfaces and intuitive operation [21]. The evaluation of interactive systems serves to identify weak points during development and to compare different system designs. Evaluation studies have to be carried out throughout the development phase of an interactive system. This serves the fulfillment as well as the adaptation and extension of the system requirements. At the beginning of system development, guidelines and standards can be used to define the requirements and procedures for the development of user interfaces. However, an increasing degree of standardization is accompanied by a low level of detail, so that these standards and guidelines can serve as a good orientation to avoid errors in the development of user interfaces, but do not provide a detailed basis for the concrete design of the user interface. An example of such a standard in the development of human-system-interaction is DIN EN ISO 9241-110, which describes the principles of dialog design. At the end of the development phase, evaluations are carried out to identify and correct any remaining errors in the applications. User-centered methods are used, in which evaluation studies are carried out with representative users as well as expert-based methods. Expert-based methods include that usability experts check the application according to development guidelines and for compliance with the requirements and then make suggestions for improvement. Usually, the characteristics of the respective application and the budget determine the evaluation methods used [22]. An overview of the common evaluation methods for interactive systems is shown in the following figure (see Fig. 1). In the following, we focus on user-centered evaluation and will explain the underlying process of conducting experiments briefly. Here, we will also give a short introduction to query techniques such as interviews and questionnaires as a part of this process. In order to keep the entry-level of data acquisition low, we will exclude techniques, which monitor physiological responses. This way, no additional measurement equipment is needed, and the number of participants can be raised more easily. A user-based evaluation reduces the risk of developing systems that do not meet users’ needs. In particular, requirements that could not be identified in the
Human-Computer-Interaction Design in Production and Logistics
557
Fig. 1. Evaluation methods for interactive systems [23]
preceding requirements analysis can be identified. In this way, prototypes can both be tested at an early stage in the later application environment, and the fulfillment of the requirements by the later users can be confirmed at the final delivery of the system. In addition, feedback during the use phase can support problem identification and enable immediate improvement or consideration in a later system version [24]. A standard method for evaluating interactive systems are experimental evaluation studies [25]. They can be used to show whether there are causal relationships between the design of a human-machine system and effects on the humans involved, the system, or the system performance. Therefore, in an experimental study, one or more variables are modified in order to induce observable effects [25,26]. The gain of knowledge is mainly achieved through social scientific data acquisition techniques such as interviews or questionnaires [27]. The goal of a subsequent statistical evaluation of the experiments is to answer the question whether the selected independent variables (factors) have an influence on the dependent variables. Interviews and questionnaires are techniques that can be used for gathering information about users [28]. They can be used to discover user’s opinions and to assess underlying mental processes by using a set of pre-defined questions. Interview techniques allow us to ask questions to users in a direct manner. Interviews are useful when the topic of inquiry relates to issues that require complex questioning and considerable probing while the questionnaire is a technique for retrieving primary results and leads to first insights in order to conduct
558
M. Quandt et al.
further studies. It also allows us to raise the number of participants easily. In order to gather reliable results, we need to prepare the set of questions to be clear and concise. In production and logistics scenarios, we normally focus on the usability of the system in order to improve the working process of the interaction between people and technology. Therefore, questionnaires for assessing the perceived usability of an interactive system such as the Questionnaire for User Interface Satisfaction [29], the Computer System Usability Questionnaire [30], and the System Usability Scale [31] can be used. Additionally, observations are typically used to assess the behavior of users. Here, observation refers to a data acquisition technique (in contrast to observation techniques, as shown in Fig. 1, which follow a different approach). Observations in the laboratory or in the field can be used to investigate processes and to record unconscious behavior, e.g., during the use of a system or during group processes [27]. They enable the collection of information about work performance and errors on the one hand, and about usability on the other [32]. Both descriptive and closing statistics methods are used for analyzing experimental investigations [26]. Here, descriptive statistics often include the representation of mean values and standard deviations. Typically, the values of the dependent variables are determined, and statements are made about their variance [26]. On this basis, an analysis of variance (ANOVA) is often performed (if different conditions are met, see [33]). It provides a statistical test that takes into account the observed mean values of the different groups and allows a decision to be made whether these differences are significant enough to conclude that there are similar differences in the underlying populations [34].
3
Exemplary Use Cases of Human-Machine-Interaction Design and First Results of Evaluation Studies
In this section, we show three examples of applying HCI to production and logistics research fields. We explain the methods for conducting the experiments by using HCI evaluation techniques in the different approaches. The examples include: First, design of user interfaces in Cyber-physical production systems (CPPS). Second, design of a training device for collaboration improvements and third, design of AR systems for production environments. All use cases pursue aspects of human-oriented design such as user perception and overall system performance in a general way. Besides, in a more specific way, each use case aims at particular aspects of human-oriented design (e.g. usability, workload, solution quality or cooperation behaviour). 3.1
Design of the User Interface in Cyber-Physical Production Systems
In cyber-physical production systems, cognitive work becomes predominant while the share of physical work decreases. As a result, crucial new work design
Human-Computer-Interaction Design in Production and Logistics
559
elements need to be developed to meet these changing requirements. Among others, the design elements usability, assistance system, and feedback could play a crucial role in work system design in CPPS [35]. Here, usability describes a facilitated usage of the interface for the worker. This is characterized by increased effectiveness, efficiency, and satisfaction [36]. By using assistance systems, workers are supported in the processing of their work tasks. Such tools can represent the human-machine-interface themselves but can also be designed as a function on an existing interface (e.g., on a tablet computer) [37,38]. Via feedback workers can be offered information both about the activities of the cyber-physical work system and about the quality of their work results [20,21]. A human factors study was performed to illustrate the effects triggered by specific design elements. Here, the design elements selected (feedback, assistance system, usability) serve as independent variables (factors), which are either set to active or inactive. This results in eight experimental conditions of a factorial 23 design [25,26,33]. The dependent variables represent several aspects of work performance and perception of work. The experiments were carried out as laboratory experiments to control external influences. A total of 68 persons participated in the study (n = 68) [35]. We used an experiment platform for carrying out the investigation. The platform is mainly based on a Raspberry Pi 2 B single-board computer combined with a 7-in. touch display [39]. The participants are able to communicate with the system via the touch display. Prior to the experiment, the investigator sets up the work task or the work setup to be investigated. During the experiment, the system automatically collects data about the test persons performance (observation) and records questionnaire answers. Two different task types and two levels of task complexity were used in the study. In the following, first results are presented, which correspond to tasks, where participants had to solve machine scheduling problems actively at a low complexity level (see [35] for further information on the study details and outcomes). We performed a subsequent three-factor variance analysis to examine the statistical significance of the independent variables. Significant effects of all three design elements investigated were found. We interpret the results in a way that the participants received valuable information via feedback, which resulted in a quicker and more efficient way of solving the given tasks. Here, the processing time was shorter, and the number of display touches was lower. Hence, we assume that feedback should play a key role in CPPS work design. We also conclude that usability supported the perception of information and thus helped to solve the tasks in a meaningful way. Here again, several significant effects were observed: an increase in the evaluation of the satisfaction with the user’s own, final solution, and a similar effect for the evaluation of the quality of the solution. Finally, assistance led to a negative observation since it extended the processing time. Here, we assume that the intended idea of the worker assistance did not meet the user’s requirements and led to confusion [35].
560
M. Quandt et al.
These findings need to be considered in the context of the given tasks, the chosen measures, and the limited number and variety of study participants. Thus, we do not claim general applicability to all CPPS work design use cases. Nevertheless, the results indicate effect sizes and directions of work design elements in CPPS, which can serve as starting points for future investigations [35]. 3.2
Design of AR Systems for Production Environments
For the heating and air-conditioning of large infrastructures, complex systems are installed due to growing environmental awareness and the simultaneous increase in energy costs. The installation and operation of these systems require the service technicians to have extensive specialist knowledge and complete documentation of the system in order to be able to carry out the work efficiently. In particular, inefficiencies occur in the maintenance of the heating, ventilation, and air conditioning systems due to insufficient documentation. The revision plans documented during the planning phase are based on the ground plan of the building and serve as the basis for the assembly activities. Due to the participation of numerous disciplines in the execution of large construction projects, work is not always carried out in accordance with the planning. Therefore, there are discrepancies between the actual installation of individual components and the planning. Due to the high time pressure during assembly, some changes made are not recorded in the plans. To carry out a maintenance task, the latest version of these plans serves as the basis for order processing. This regularly leads to delays in the workflow, as certain components, such as fire dampers, are installed behind other structures, which is not noted in the plans. If the installation location deviates from the position documented in the revision plans, high effort is also required to find the required component. We are therefore developing an AR-based assistance system to support service technicians in carrying out maintenance work on these systems. For this purpose, the real environment is superimposed with the current version of the revision plan. A gesture-based control enables interaction with the virtual revision plan. In this way, assemblies can be correctly arranged, deleted, or newly created. In this way, the service technicians are supported in the documentation, and search efforts during the execution of maintenance tasks are reduced. The AR-based assistance system is implemented on a Microsoft HoloLens. This data goggle preserves the freedom of movement of the service technicians and enables additional virtual information to be displayed directly in the technicians’ field of vision. By using the 3D spatial model generated by the data glasses, the revision plans can be adapted to the respective room geometry. The interaction with the AR device takes place via 3D gesture control. This form of interaction, as well as the representation of virtual information in space, places new demands on the users in the operation of the assistance system. Therefore a human-centered development and evaluation approach was pursued. The consideration of the specific requirements of the industrial application environment is a prerequisite for the successful development of an AR-based assistance system [40]. Acceptance by the users is closely related to usability and
Human-Computer-Interaction Design in Production and Logistics
561
user experience [41]. Therefore, we targeted high user participation in the development of AR solutions for the introduced industrial application. The high number of possible forms of interaction, hardware configurations, and the possibility of addressing different senses (visual, auditory, tactile, etc.) make it challenging to select generally valid, comprehensive evaluation methods for AR applications [42]. We initially decided to carry out a preliminary study in the development phase, which provides a moderated system test of a prototype in a laboratory environment. Subsequently, the user feedback can be integrated into further system development. Finally, user-based studies in the application environment are planned with the adapted and fully implemented AR assistance system. The preliminary study is accompanied by an observing moderator to avoid operating errors and to test all software functionalities existing at this time. Before operating the AR-based assistance system, the test persons execute a training program to learn the hand gesture-based interaction required for operating the AR solution. The actions of the users are logged and subsequently evaluated. Following the software test, the experienced usability and the workload associated with the use of the software are documented by the users using questionnaires. The system usability scale is used for the usability test and the NASA TLX questionnaire for the workload classification. First results from the preliminary study show that the novel interaction represents a challenge for the user, but can be learned with a little practice. In some cases, an imprecise execution of the hand gestures does not immediately activate the desired functionality. The AR solution is currently being further developed for practical testing. When carrying out the practical test, the same methods are used to enable comparability between laboratory and practical tests. For the practical test, in particular, the later system users are selected as test users, i.e., service technicians for heating, ventilation, and air-conditioning systems. 3.3
Design of a Training Device for Collaboration Improvements
One of the many complex issues of urban logistics is the involvement of various stakeholders with independent requirements. Urban logistics needs the involvement of all stakeholders (city authorities, logistics providers, and citizens) to identify suitable solutions for urban mobility. The different needs of the stakeholders in urban mobility often lead to complications in finding a solution to the problem [13,43]. A key challenge refers to understanding the cooperation of the stakeholders [44]. Games are one of the tools that can foster human-human or human-games interaction [45]. Due to the games, players are enabled to make decisions in an experimental, risk-free environment in logistics [46]. Even if some logistics games are designed as role-playing games so that the players can experience different perspectives [47,48], it is difficult to understand and analyze the influence of the game mechanics versus the human-human interaction and the influence of these components on collaboration [49,50]. This leads to the importance of measuring the experience of the players and their interaction. Therefore, to understand the interaction of player-game and player-player as well as their experience during the game, HCI is of interest.
562
M. Quandt et al.
From our previous study [17], we analyzed the cooperation coming from the game-mechanics and from the human interactions with each other in a cooperative game to investigate the possibility of using entertainment games for learning purpose. For this purpose, we conducted an experiment based design. We first modified the digital board game, which has similar decision-making processes with an urban logistics scenario. Then, HCI techniques were used for evaluating the concept of using digital board games as tools for supporting understanding of stakeholders in urban logistics. Our study focuses on evaluating the possibility of using an entertainment board game for training users to have more awareness of cooperation in urban logistics. The results can be used as guidelines to develop a game for supporting cooperative activities in urban logistics in the future. We conducted the experiment to investigate the mechanics of the game, which is to motivate players to cooperate as well as to make the players understand the role of their partner. Then, we evaluated the game by using a structured interview with 30 participants. The participants were randomly paired as a team to play in one session. Each participant went through tutorials on the games rules before playing two different modes of the game (regular and special ability mode). The difference between the two modes is that in the first mode, both the players in each session have the same abilities and role while in the latter mode, the players have individual roles and different abilities. Finally, each player was interviewed after the playing session. From the results, we can prove that the mechanics of cooperative board games can be used for motivating players to realize the importance of having another person as a partner for planning and making decisions. The evidence showed that the players realize that they require a partner, especially when the players were in tough situations. This finding can be used to emphasize the need for understanding stakeholders regarding their involvement in the decision-making process related to urban logistics. Additionally, the conditions or situations that make stakeholders feel that the contribution from another stakeholder is required. This can be applied to urban logistics. We found that role-play can motivate players to understand the role of their partner. The results have been shown to be significant as the participants can compare between normal mode and the special ability mode. Participants found it important to have another partner with different tasks to win the game. In the future, we plan to include the experts in an urban area to test the game in order to have feedback related to urban logistics content.
4
Conclusion
This paper focused on the evaluation of human-machine interfaces in production engineering and logistic applications. By using three use cases, the authors were able to show different user-centered evaluation approaches. The first use case deals with the design of human work in future work environments and examines the effects of usability, assistance, and feedback on users in this context. The second use case deals with the development and introduction of an AR-based
Human-Computer-Interaction Design in Production and Logistics
563
assistance system for the maintenance of heating, air-conditioning, and refrigeration systems in large-scale infrastructures. This has been evaluated in a preliminary study and then in the real field of application with the later system users. In the third use case, a game based training environment has been developed and evaluated the game mechanics, which used for supporting the joint decisionmaking of the stakeholders in urban logistics. A questionnaire-based evaluation will be used to investigate cooperation behavior and the understanding of users’ roles. The user-based evaluation studies carried out in the presented use cases led to initial results that support the involvement of users in the development and evaluation process of production-technical and logistical HCI applications. From the evaluation studies, valuable conclusions could be drawn in the use cases for future work design in CPPS, the use of AR in maintenance scenarios, and decision support in urban logistics. Nevertheless, these results cannot claim to be generally applicable, as suitable evaluation methods have to be selected for each application, and all influencing factors have to be taken into account. The specific characteristics of the use cases argue in favor of a user-based evaluation as opposed to expert evaluation. In the case of the HCI in CPPS, the design of the working environment focuses on the human being as the object of observation. With the introduction of new technologies such as AR, the technology-dependent development of new solutions is particularly opposed to user-oriented development. Therefore, the first prototypes of a solution must be made available at an early stage in order to make the possibilities of the technology tangible and then to be able to include the requirements of the users in further developments. Furthermore, there are no guidelines available for an expert-based evaluation of technologies that include new interaction patterns. In supporting collaborative decision-making processes, the interaction between people and technology, as well as the interaction between stakeholders, play an essential role. These examples underline the need to integrate users into development at an early stage in order to develop needs-based solutions. This can be illustrated on the example of the human-centered design process according to DIN EN ISO 9241-210 for an Augmented Reality application (see Fig. 2): since in this case already the hardware selection determines the usable interaction forms, these must be aligned with the needs of the users, before the actual application development begins. Only by involving the users at an early stage, they can become familiar with the possibilities of the new technology. This can result in further requirements for the development process. By developing and evaluating prototypes at an early stage in cooperation with the users, the potential of the technology for the application can be exploited. To sum up, the implementation of user-based evaluation studies showed promising results in relation to the investigated use cases. Therefore, these usercentered methods should be used in the future for further application scenarios in order to develop better HCI solutions for production and logistics. This leads to increased user acceptance of the solutions, even if this is opposed by a high effort for the implementation of the evaluation studies. As has been shown, prototypical solutions have to be developed for the object of new technologies for
564
M. Quandt et al.
HCI, respectively, for the inclusion of humans as the object of consideration for the individual phases of the development process. These phases require the development of a portfolio of application-specific evaluation methods.
Fig. 2. Human-centered design process in accordance with DIN EN ISO 9241-210 adapted for the development of an AR solution [51] Acknowledgment. The authors would like to thank the German Federal Ministry of Economic Affairs and Energy (BMWi) for their support within the project “Bauen 4.0 - KlimAR” (grant number 16KN062830).
References 1. Huber, W.: Industrie 4.0 in der Automobilproduktion. Springer Fachmedien Wiesbaden, Wiesbaden (2016) 2. Kagermann, H., Wahlster, W., Helbig, J.: Umsetzungsempfehlungen f¨ ur das Zukunftsprojekt Industrie 4.0 (2013) 3. K¨ ohler-Schute, C.: Industrie 4.0: Ein praxisorientierter Ansatz. KS-Energy Verlag, Berlin (2015) 4. Andelfinger, V., H¨ anisch, T.: Industrie 4.0 Wie cyber-physische Systeme die Arbeitswelt ver¨ andern. Bundesministerium f¨ ur Bildung und Forschung (2017) 5. Mertens, P., Barbian, D., Baier, S.: Digitalisierung und Industrie 4.0 - eine Relativierung. Springer Fachmedien Wiesbaden, Wiesbaden (2017) 6. Spath, D., Ganschar, O., Gerlach, S., H¨ ammerle, M., Krause, T., Schlund, S.: Produktionsarbeit der Zukunft-Industrie 4.0. Fraunhofer Verlag Stuttgart (2013) 7. Wang, X., Ong, S.K., Nee, A.Y.C.: A comprehensive survey of augmented reality assembly research. Adv. Manuf. 4(1), 1–22 (2016)
Human-Computer-Interaction Design in Production and Logistics
565
8. Kipper, G., Rampolla, J.: Augmented Reality: An Emerging Technologies Guide to AR, 1st edn. Syngress, Waltham (2013) 9. Palmarini, R., Erkoyuncu, J.A., Roy, R., Torabmostaedi, H.: A systematic review of augmented reality applications in maintenance. Robot. Comput.-Integr. Manuf. 49, 215–228 (2018) 10. Syberfeldt, A., Danielsson, O., Holm, M., Wang, L.: Visual assembling guidance using augmented reality. Proc. Manuf. 1, 98–109 (2015) 11. de Paiva Guimaraes, M., Martins, V.F.: A checklist to evaluate augmented reality applications. In: 2014 XVI Symposium on Virtual and Augmented Reality, pp. 45–52. IEEE (2014) 12. Taniguchi, E.: City logistics for sustainable and liveable cities. In: Green Logistics and Transportation, pp. 49–60. Springer (2015) 13. Lindholm, M.: Enabling sustainable development of urban freight from a local authority perspective. Chalmers University of Technology (2012) 14. Mu˜ nuzuri, J., Larra˜ neta, J., Onieva, L., Cort´es, P.: Solutions applicable by local administrations for urban logistics improvement. Cities 22(1), 15–28 (2005) 15. Cosimato, S., Troisi, O.: Green supply chain management: practices and tools for logistics competitiveness and sustainability. The DHL case study. TQM J. 27(2), 256–276 (2015) 16. Grudpan, S., Baalsrud Hauge, J., Thoben, K.D.: Challenges with tools and technologies supporting collaboration among stakeholders in urban logistics (2018) 17. Grudpan, S., Baalsrud Hauge, J., Malaka, R.: Playful training for understanding activities, roles, and stakeholder in urban logistics (2019) 18. Hirsch-Kreinsen, H.: Welche Auswirkungen hat “Industrie 4.0” auf die Arbeitswelt? WISO direkt, pp. 1–4 (2014) 19. Khosrowpour, M.: Human-Computer Interaction: Concepts, Methodologies, Tools, and Applications (2015) 20. DIN9241: Ergonomics of human-system interaction - Part 392: Ergonomic recommendations for the reduction of visual fatigue from stereoscopic images (ISO 9241-392:2015); German version EN ISO 9241-392:2017 (2017) 21. Butz, A., Kr¨ uger, A.: Mensch-Maschine-Interaktion. De Gruyter Studium. De Gruyter, Berlin (2017) 22. Preim, B., Dachselt, R.: Interaktive Systeme: Band 2: User Interface Engineering, 3D-Interaktion, Natural User Interfaces, 2nd edn. EXamen Press, Springer, Berlin (2015) 23. Dix, A., Finlay, J., Abowd, G.D., Beale, R.: Human-Computer Interaction, 3rd edn. Pearson Education, Harlow (2004) 24. Heinecke, A.M.: Mensch-Computer-Interaktion: Basiswissen f¨ ur Entwickler und Gestalter. 2., u ¨ berarbeitete und erw. aufl. edn. X.media.press, Springer, Heidelberg (2012) 25. Jacko, J.A., Yi, J.S., Sainfort, F., McClellan, M.: Human factors and ergonomic methods. Handb. Hum. Factors Ergon. 4, 289–329 (2012) 26. Wickens, C.D., Gordon, S.E., Liu, Y.: An Introduction to Human Factors Engineering. Longman, New York (2014) 27. Badke-Schaub, P., Hofinger, G., Lauche, K.: Human Factors. Springer, Heidelberg (2012) 28. Lazar, J., Feng, J.H., Hochheiser, H.: Research Methods in Human-Computer Interaction. Morgan Kaufmann, Burlington (2017) 29. Chin, J.P., Diehl, V.A., Norman, K.L.: Development of an instrument measuring user satisfaction of the human-computer interface. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 213–218. ACM (1988)
566
M. Quandt et al.
30. Lewis, J.R.: Ibm computer usability satisfaction questionnaires: psychometric evaluation and instructions for use. Int. J. Hum.-Comput. Interact. 7(1), 57–78 (1995) 31. Brooke, J., et al.: Sus-a quick and dirty usability scale. Usabil. Eval. Ind. 189(194), 4–7 (1996) 32. Stanton, N.A., Hedge, A., Brookhuis, K., Salas, E., Hendrick, H.W.: Handbook of Human Factors and Ergonomics Methods. CRC Press, Boca Raton (2004) 33. Montgomery, D.C.: Design and Analysis of Experiments, 7th edn. Wiley, Hoboken (2009) 34. Fahrmeir, L., Heumann, C., K¨ unstler, R., Pigeot, I., Tutz, G.: Statistik: Der Weg zur Datenanalyse. Springer, Heidelberg (2016) 35. Stern, H., Becker, T.: Concept and evaluation of a method for the integration of human factors into human-oriented work design in cyber-physical production systems. Sustainability 11(16), 4508 (2019) 36. Jordan, P.W.: An Introduction to Usability. CRC Press, Boca Raton (1998) 37. Delisle, S., Moulin, B.: User interfaces and help systems: from helplessness to intelligent assistance. Artif. Intell. Rev. 18(2), 117–157 (2002) 38. Nee, A., Ong, S., Chryssolouris, G., Mourtzis, D.: Augmented reality applications in design and manufacturing. CIRP Ann. 61(2), 657–679 (2012). https://www.sciencedirect.com/science/article/pii/S0007850612002090 39. Stern, H., Becker, T.: A variable low-cost platform for conducting work design experiments. In: CEUR Workshop Proceedings, vol. 2009, pp. 126–130 (2017) 40. Quandt, M., Knoke, B., Gorldt, C., Freitag, M., Thoben, K.D.: General requirements for industrial augmented reality applications. In: Procedia CIRP 51st CIRP Conference on Manufacturing Systems, pp. 1130–1135 (2018) 41. Dey, A., Billinghurst, M., Lindeman, R.W., Swan, J.E.: A systematic review of 10 years of augmented reality usability studies: 2005 to 2014. Front. Robot. AI 5, 1–28 (2018) 42. D¨ unser, A., Billinghurst, M.: Evaluating augmented reality systems. In: Furht, B. (ed.) Handbook of Augmented Reality, pp. 289–307. Springer, New York (2011) 43. Lagorio, A., Pinto, R., Golini, R.: Research in urban logistics: a systematic literature review. Int. J. Phys. Distrib. Logist. Manage. 46(10), 908–931 (2016) ¨ 44. Graham, G., Osterle, I., Aditjandra, P.T., Vaghi, C., Grea, G., Zunder, T.H.: The role of a structured stakeholder consultation process within the establishment of a sustainable urban supply chain. Supply Chain Manage.: Int. J. (2015) 45. Malaka, R.: How computer games can improve your health and fitness. In: International Conference on Serious Games, pp. 1–7. Springer (2014) 46. Hauge, J.B., Riedel, J.C.: Evaluation of simulation games for teaching engineering and manufacturing. Proc. Comput. Sci. 15, 210–220 (2012) 47. Citylogisticsgame.com. https://citylogisticsgame.com/contact.html. Accessed 30 July 2019 48. Raghothama, J., Baalsrud Hauge, J., Meijer, S.: Evaluating city operations design using interactive simulations. J. Simul. Gaming (2017) 49. Siu, K., Zook, A., Riedl, M.O.: Collaboration versus competition: design and evaluation of mechanics for games with a purpose. In: FDG (2014) 50. Wendel, V., Gutjahr, M., G¨ obel, S., Steinmetz, R.: Designing collaborative multiplayer serious games. Educ. Inform. Technol. 18(2), 287–308 (2013) 51. DIN 9241: Ergonomics of human-system interaction – Part 210: Human-centred design for interactive systems (ISO 9241-210:2019); German version EN ISO 9241210:2019 (2019)
Author Index
A Agostino, Ícaro Romolo Sousa, 387 Ali Mohamadlou, Moslem, 166 Alkhudary, Rami, 159 Alvim, Silvio Luiz, 274 Antons, Oliver, 227, 245 Arias, Jaime Andres Cardona, 254 Arlinghaus, Julia C., 137, 227, 245 Aßmann, Uwe, 420 Aziz, Mohd Azizi Abd, 186 B Balachandra, Kavith, 516 Bäumler, Ilja, 286 Beinke, Thies, 527, 541, 554 Berrios, Paulina, 363 Bidisse, Adama, 148 Blank, Felix, 321 Bode, Dennis, 32 Borstell, Hagen, 462 Broda, Eike, 409 Brusset, Xavier, 159 Buer, Tobias, 62 Buscher, Udo, 420 C Cao, Liu, 462 Castillo, Raúl, 363 Cen, Marco, 441 Chankov, Stanislav M., 374
D Darom, Noraida Azura Md, 186 Daschkovska, Kateryna, 331 Dashkovskiy, Sergey, 331 Dastyar, Haniyeh, 166, 341 De Cursi, José Eduardo Souza, 254 De Marco, Alberto, 21 Decker, André, 32 Depner, Thomas, 462 E Elfaham, Haitham, 451 Engbers, Hendrik, 236 Engesser, Sven, 420 Epple, Ulrich, 451 F Fakhry, Hussein, 21 Feldmann, Carsten, 493 Feldt (geb. Wagner), Julia, 398 Fenies, Pierre, 159 Fibrianto, Henokh Yernias, 62 Fischer, Juliane, 451 Franzkeit, Janna, 70 Frazzon, Enzo Morosini, 351, 387, 409 Freitag, Michael, 42, 175, 196, 236, 409, 441, 527, 541, 554 G Galipoglu, Erdem, 204 Gerken, Paul, 79, 286
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 M. Freitag et al. (Eds.): LDIC 2020, LNLO, pp. 567–569, 2020. https://doi.org/10.1007/978-3-030-44783-0
568 Görges, Michael, 42 Gouanlong Kamgang, Nadege Ingrid, 148 Grafelmann, Michaela, 52 Greulich, Christoph, 32 Groneberg, Maik, 101 Grudpan, Supara, 554 H Haasis, Hans-Dietrich, 62, 175 Hacid, Hakim, 21 Hagemann, Vera, 504 Hardi, Elisabeth, 441 Herzog, Otthein, 363 Hishamuddin, Hawa, 186 Hong, Soondo, 62 Huynh, Van-Nam, 309 I Icarte, Gabriel, 363 J Jahn, Carlos, 52, 70, 114 Jung, Eva, 3 K Kang, Bonggwon, 62 Kastner, Marvin, 114 Keiser, Dennis, 541 Khalid, Raheen, 374 Khan, Ayesha, 175 Kim, Bosung, 62 Kim, Kap Hwan, 62 Kinra, Aseem, 299 Klumpp, Matthias, 504 Kontny, Henning, 398 Kotzab, Herbert, 79, 89, 196, 204, 215, 286 Kretzschmar, Johannes, 263 Kreutzfeldt, Jochen, 52 L Lang, Walter, 441 Lange, Ann-Kathrin, 52, 114 Leohold, Simon, 236 Lima Jr., Orlando F., 254 M Malaka, Rainer, 554 Mammar, Zakaria, 21 Marbach, Annika, 62 Mohammadi, Ali, 166 Mrutzek, Bastian, 204
Author Index N Nellen, Nicole, 52 Neubert, Andreas, 431 Nieberding, Bernd, 263 Novaes, Antônio G. N., 254 P Pache, Hannah, 70 Pannek, Jürgen, 341 Perera, H. Niles, 516 Petljak, Kristina, 215 Petzoldt, Christoph, 541 Pfoser, Sarah, 3 Postan, Mykhaylo Ya., 331 Postorino, Marco, 21 Praneetpholkrang, Panchalee, 309 Putz, Lisa-Maria, 3 Q Quandt, Moritz, 554 R Rahman, Mohd Nizam Ab, 186 Redeker, Magnus, 477 Reich, Juri, 299 Ristow, Charles, 387 Rodriguez, Carlos Manuel Taboada, 274 S Samani, Omid, 124 Santos Jr, José Benedito Silva, 254, 274 Sardoux Klasen, Andre, 493 Schaper, Martina, 504 Schindler, Thimo, 32 Schönberger, Jörn, 124, 420 Schuldt, Arne, 32 Schütz, Artur, 101 Shah, Sayed Mehdi, 175 Shahmoradi-Moghadam, Hani, 124 Specht, Patrick, 89 Stern, Hendrik, 554 T Taboada Rodriguez, Carlos Manuel, 387 Takeda-Berger, Satie L., 409 Thibbotuwawa, Amila, 516 Thoben, Klaus-Dieter, 32 Tsapi, Victor, 148 U Unseld, Hans G., 79 Urbas, Leon, 420
Author Index V Veigt, Marius, 441 Vieira, Guilherme Ernani, 351 Vogel-Heuser, Birgit, 451 W Wagenitz, Axel, 477 Wahab, Dzuraidah Abd, 186 Wakolbinger, Tina, 299
569 Wilhelm, Jasper, 527 Woltering, Tim, 493
Z Zahner, Melanie, 137 Zhao, Ziqi, 227 Ziegenbein, Justin, 52 Zimmermann, Manuel, 137