Selected Topics in Manufacturing: AITeM Young Researcher Award 2021 (Lecture Notes in Mechanical Engineering) 3030826260, 9783030826260

Citation preview

Lecture Notes in Mechanical Engineering

Luigi Carrino Tullio Tolio Editors

Selected Topics in Manufacturing AITeM Young Researcher Award 2021

Lecture Notes in Mechanical Engineering Series Editors Francisco Cavas-Martínez, Departamento de Estructuras, Universidad Politécnica de Cartagena, Cartagena, Murcia, Spain Fakher Chaari, National School of Engineers, University of Sfax, Sfax, Tunisia Francesca di Mare, Institute of Energy Technology, Ruhr-Universität Bochum, Bochum, Nordrhein-Westfalen, Germany Francesco Gherardini , Dipartimento di Ingegneria, Università di Modena e Reggio Emilia, Modena, Italy Mohamed Haddar, National School of Engineers of Sfax (ENIS), Sfax, Tunisia Vitalii Ivanov, Department of Manufacturing Engineering, Machines and Tools, Sumy State University, Sumy, Ukraine Young W. Kwon, Department of Manufacturing Engineering and Aerospace Engineering, Graduate School of Engineering and Applied Science, Monterey, CA, USA Justyna Trojanowska, Poznan University of Technology, Poznan, Poland

Lecture Notes in Mechanical Engineering (LNME) publishes the latest developments in Mechanical Engineering—quickly, informally and with high quality. Original research reported in proceedings and post-proceedings represents the core of LNME. Volumes published in LNME embrace all aspects, subfields and new challenges of mechanical engineering. Topics in the series include: • • • • • • • • • • • • • • • • •

Engineering Design Machinery and Machine Elements Mechanical Structures and Stress Analysis Automotive Engineering Engine Technology Aerospace Technology and Astronautics Nanotechnology and Microengineering Control, Robotics, Mechatronics MEMS Theoretical and Applied Mechanics Dynamical Systems, Control Fluid Mechanics Engineering Thermodynamics, Heat and Mass Transfer Manufacturing Precision Engineering, Instrumentation, Measurement Materials Engineering Tribology and Surface Technology

To submit a proposal or request further information, please contact the Springer Editor of your location: China: Ms. Ella Zhang at [email protected] India: Priya Vyas at [email protected] Rest of Asia, Australia, New Zealand: Swati Meherishi at [email protected] All other countries: Dr. Leontina Di Cecco at [email protected] To submit a proposal for a monograph, please check our Springer Tracts in Mechanical Engineering at http://www.springer.com/series/11693 or contact [email protected] Indexed by SCOPUS. All books published in the series are submitted for consideration in Web of Science.

More information about this series at https://link.springer.com/bookseries/11236

Luigi Carrino · Tullio Tolio Editors

Selected Topics in Manufacturing AITeM Young Researcher Award 2021

Editors Luigi Carrino DICMaPI University of Naples Federico II Napoli, Italy

Tullio Tolio Mechanical Engineering Department Politecnico di Milano Milano, Italy

ISSN 2195-4356 ISSN 2195-4364 (electronic) Lecture Notes in Mechanical Engineering ISBN 978-3-030-82626-0 ISBN 978-3-030-82627-7 (eBook) https://doi.org/10.1007/978-3-030-82627-7 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 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

Europe is playing a major role in worldwide manufacturing and has a particularly good positioning in medium-high-tech products. Manufacturing accounts for 83% of European export and is therefore vital for the European economy. Italy is the second major player in European manufacturing after Germany. Recent crises, namely the financial crisis in 2008 and the pandemic crisis in 2019, have highlighted the importance of manufacturing as the backbone of European and Italian economies and have shown that manufacturing is extremely important to guarantee the resilience of a country. Manufacturing has also important social implications since, on the one hand, it provides occupation with average wages higher than the service sector and, on the other hand, can provide all the products for the society of the future both in normal and in critical times. Indeed, for example, manufacturing is the cornerstone to supply products needed to overcome health problems, as it has been recently appreciated in the pandemic crisis, and as another example, all the products required to fight environmental changes which is one of the most important challenges for the future of the planet. The main idea is that, when big problems are at stake, manufacturing can provide advanced and highly specialized products in big volumes to really make the difference. Therefore, manufacturing is the key to the future well-being in our societies. In order to remain competitive, manufacturing requires continuous innovation to maintain technological leadership. Italy, as well as Europe, does not own significant natural resources, and is based on an advanced society which implies the centrality of safety at work, the well-being of the employees, strict regulation on environmental preservation and enhancement. Therefore, worldwide competition is particularly challenging and needs to be addressed with continuous technological improvement. Indeed, manufacturing is the sector with the highest growth rate in efficiency and industry 4.0 is just the last wave in this process. However, efficiency is not enough, and to guarantee leadership it is necessary to continuously innovate products, manufacturing processes and manufacturing systems. Indeed, the European ambition is to become the world’s most competitive and dynamic knowledge-based economy.

v

vi

Preface

In order to go in this direction, it is necessary that basic research and applied research work hand in hand in order to shorten the period to bring new ideas to potential application and to be more productive and imaginative. Also research and innovation need to be more and more connected, in order to guarantee fast exploitation of new solutions. In this scenario, AITeM the Italian Manufacturing Association (Associazione Italiana delle Tecnologie Manifatturiere) has the role of connecting all the manufacturing professors and researchers belonging to the scientific sector “ING-IND/16 Manufacturing” (Tecnologie e sistemi di Lavorazione) and creating a continuous interaction with industry. With its 290 academic members, AITeM connects all the 38 Italian Universities dealing with manufacturing; also, by the same token, it gets access to a network of manufacturing labs that is by far the widest and more complete experimental structure related to manufacturing technologies and systems available in Italy. This network virtually encompasses all the new and conventional manufacturing technologies and manufacturing system applications. Since each researcher and each lab in the network is connected to many companies operating in the field both as technology providers and technology users, AITeM also represents the strongest Italian relation between manufacturing companies and academia. Finally, AITeM academic members teach all the courses on manufacturing in Italy getting access to and teaching all the cohorts of manufacturing students in Italy. These students represent the future of Italian manufacturing and guarantee a transfer “by head” to Italian and foreign companies of the new findings in manufacturing research and innovation. The actions of AITeM are based on community building, to strengthen and continuously nurture the relations within the network described before, and interaction with industry, to guarantee a smooth and fast transition of new ideas to industry and to formally define new manufacturing problems on the basis of emerging industrial needs. Regarding community building, AITeM organizes every other year a conference where all the Italian scientists participate to discuss the most recent findings in the area of manufacturing, and organizes other more focused conferences and workshops on specific topics of manufacturing science and technology. Also, it defines the strategy of evolution of the Italian sector ING-IND/16 with particular emphasis on the new generations of researchers and professors. Regarding the interaction with industry, AITeM has three main streams of action. The first one aims at offering short courses created for companies, where innovative manufacturing topics are presented both from a theoretical and a practical point of view. These courses are taught by AITeM members but also involve invited national and international scientists and company professionals. The second one supports talent scouting and breakthrough idea generation through an initiative called Manuthon®, which is the first hackathon devoted to manufacturing. It involves manufacturing students (bachelor, master and PhD) from all Italian universities to address a set of challenges proposed by companies. The third one is open innovation, where AITeM gives its scientific support to define open innovation challenges proposed by the big companies and provides the scientific support in screening solutions coming from startups and innovative SMEs.

Preface

vii

All the described instruments are used by the AITeM “Sections” to foster the collaboration between companies and academia. Sections are communities of academic and industrial specialists dealing with a specific cutting-edge manufacturing topic, and each Section has an academic and an industrial chairperson. Within a Section, new courses are generated, Manuthon® challenges and open innovation challenges are identified, and new applications and new ideas are discussed in order to diminish the gap between idea generation and industrial applications. Sections have a predefined time frame and, at the end of their 2-year cycle, generate a white paper showing, among other things, the Italian state of play in the specific topic of the Section. This book is one of the AITeM actions towards the development of science and its presentation to a wider public. It collects the best papers produced by young AITeM members in the area of manufacturing selected by the Editorial Committee and carefully reviewed. In particular, this year’s edition contains contributions covering the areas of additive manufacturing, new processes for new materials, sustainability and circular economy, manufacturing systems design and operations. This book clearly shows the nature of AITeM’s approach which is cross-sectoral, deals with the intimate relation between the technologies and the corresponding machines and manufacturing systems, and supports the creation of new manufacturing paradigms. The book provides a snapshot of the vitality of the Italian research community looking towards the future and a novel contribution to research and innovation. Naples, Italy Milan, Italy

Luigi Carrino President of the Editorial Committee Tullio A. M. Tolio President of AITeM

Contents

Assessing the Effect of a Novel Production Control Policy on a Two-Product, Failure-Prone Manufacturing/Distribution Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Roberto Rosario Corsini, Sergio Fichera, and Antonio Costa Systematic Repeatability Analysis of Nanosecond Pulsed Laser Texturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gianmarco Lazzini, Adrian Hugh Alexander Lutey, and Luca Romoli Scheduling Remanufacturing Activities for the Repair of Turbine Blades: An Approximate Branch and Bound Approach to Minimize a Risk Measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lei Liu and Marcello Urgo

1

21

41

Robust Improvement Planning of Automated Multi-stage Manufacturing Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maria Chiara Magnanini and Tullio A. M. Tolio

61

Characterization of the Chemical Finishing Process with a Cold Acetone Bath of ABS Parts Fabricated by FFF . . . . . . . . . . . . . . . . . . . . . . . Leonardo Riva, Antonio Fiorentino, and Elisabetta Ceretti

77

Comparison Between Micro Machining of Additively Manufactured and Conventionally Formed Samples of Ti6Al4V Alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andrea Abeni, Paola Serena Ginestra, and Aldo Attanasio

91

Development of Single Point Exposure Strategy to Suppress Vapour Formation During the Laser Powder Bed Fusion of Zinc and Its Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 Leonardo Caprio, Fabio Guaglione, and Ali Gökhan Demir

ix

x

Contents

An Experimental–Numerical Analysis of Innovative Aluminum Foam-Based Sandwich Constructions Under Compression Loads . . . . . . 131 Antonio Viscusi, Massimo Durante, and Antonio Formisano Wire Arc Additive Manufacturing Monitoring System with Optical Cameras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Giuseppe Venturini, Francesco Baffa, and Gianni Campatelli Reuse of Composite Prepreg Scraps as an Economic and Sustainable Alternative for Producing Car Components . . . . . . . . . . . 171 Alessio Vita, Archimede Forcellese, and Michela Simoncini

Assessing the Effect of a Novel Production Control Policy on a Two-Product, Failure-Prone Manufacturing/Distribution Scenario Roberto Rosario Corsini, Sergio Fichera, and Antonio Costa

Abstract This research deals with a production control policy for an unreliable manufacturing system handling two different types of product. Since the production line cannot produce both product families at the same time, setup operations are required to change the production mode. The decision on product changeover is guided by the production control policy, coping with capacity and inventory shortages due to long setup times and failure events. Usually, the literature adopts the Hedging Corridor Policy, which is optimal in a one-machine two-product reliable manufacturing system for minimizing the total backlog and inventory costs. However, the complexity of the problem increases whenever companies have to satisfy orders with high variability coming from a distribution chain. To address this problem, an analytical model based on discrete time difference equations and an extended experimental analysis have been accomplished to demonstrate the effectiveness of a new production control strategy, named Adaptive Hedging Corridor Policy. Keywords Production control policy · Unreliable manufacturing systems · Changeover

1 Introduction The production control problem is concerned with finding the best strategy for processing a given number of semi-finished parts in a manufacturing system [1]. However, production operations in manufacturing systems can be interrupted by disruptive events that can involve a decay of production rate or inventory shortages. R. R. Corsini (B) Department of Physics, University of Catania, Catania, Italy e-mail: [email protected] S. Fichera · A. Costa DICAR Department, University of Catania, Catania, Italy e-mail: [email protected] A. Costa e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 L. Carrino and T. Tolio (eds.), Selected Topics in Manufacturing, Lecture Notes in Mechanical Engineering, https://doi.org/10.1007/978-3-030-82627-7_1

1

2

R. R. Corsini et al.

These disruptions include failure events, supply shortages, setup operations, work unavailability, etc. [2]. In the last years, various strategies, defined as production control policies (PCP), have been developed to cope with these adverse events. In general, the literature on the production control problem can be divided into two groups. The first group deals with controlling unreliable single-product manufacturing systems. In this regard, PCP is used to decide the production rate so as to avoid the risk of inventory shortages due to failure events. The second group faces the production control problem of two-product manufacturing systems. In such circumstances, changeover operations, which are required to switch from one type of product to another, represent the main source of disruption. The risk of inventory shortages is relevant since both types of products cannot be produced simultaneously and changeover times are not negligible. Therefore, PCP is used to establish the type of product to process and when a product changeover is needed. Kimemia and Gershwin [3] can be considered as pioneers in production control problem of the single-product scenario since they introduced a hierarchical algorithm for controlling the production of manufacturing systems with unreliable machines. They studied a manufacturing system that produces a family of parts of products requiring similar operations and, thus, the changeover time from one part to the next was assumed negligible. The production control policy proposed in their work makes sure that the inventory reaches a specific target level or inventory threshold as soon as possible in order to be protected against future inventory shortages due to machine failures [4, 5]. The same approach has been later used by Akella and Kumar [6] and Sharifnia [4]. The former addressed a failure-prone single-machine single-product manufacturing system with a constant demand rate. The objective of their work was to minimize the discounted inventory cost, which was described by cost rates for both positive and negative inventories. The optimal solution for this production control problem is characterized by an optimal inventory level or inventory threshold. In fact, the manufacturing systems produces at the maximum production rate until the optimal inventory threshold is achieved by the current inventory level. After that, the production rate of the manufacturing becomes equal to the demand rate. The latter was the first to define this approach as Hedging Point Policy (HPP). HPP has been used by the authors to control the production of a single-product manufacturing system with arbitrary number of failure modes. In particular, the objective was to find the optimum inventory threshold (named hedging point) to minimize the average inventory cost. In the case of manufacturing systems handling two types of products, Elhafsi and Bai [7] developed a production control policy called Hedging Corridor Policy (HCP), which can be considered as the counterpart of HPP for the multi-product scenario. In their study, HCP has been used in a single-machine two-product manufacturing system with constant demand rate. HCP consists of creating a positive inventory level to prevent inventory shortages caused by non-negligible changeover times to switch from one type of product to another. They proved that HCP is optimal for minimizing the total sum of backlog and inventory costs incurred over a finite time horizon.

Assessing the Effect of a Novel Production Control Policy …

3

Bai and Elhafsi [8] used HCP for a single-machine two-product manufacturing system subject to failure events and proved that HCP is optimal to minimize a costbased objective function, which also indirectly involves the changeover costs. Interestingly, they also tested HCP on other performance measures, including the production percentage, which measures the fraction of the satisfied demand expressed as a percentage. A production percentage equal to 100% means that HCP enables satisfying the overall demand throughout the time horizon. Assid et al. [9] used the same performance indicator, defined as Customers Satisfaction Level (CSL), to compare HCP with a modified version in unreliable single-machine two-product manufacturing systems with constant demand rate. Their work showed that HCP requires a lower inventory threshold than the modified version to achieve the same performance in terms of CSL. The mentioned works address the production control problems in manufacturing systems characterized by constant demand rates. However, it could represent a strong simplification since the market demand may be significantly variable or uncertain along the time horizon [10]. Recently, Costa et al. [11] adopted the HCP to investigate the performance of a two-echelon supply chain where the factory’s node consists of a failure-prone two-product manufacturing system characterized by a production line and production flow time. Notably, in their work, the factory has to fulfill a variable demand resulting from several orders placed by a retailer into a finite time horizon. The objective of the research was to optimize the CSL indicator of the entire supply chain. To this end, HCP was used to manage the manufacturing system under consideration, which is simultaneously characterized by two disruptive events, i.e., the non-negligible changeover times to switch from one type of product to another and the failure events. Interestingly, Polotski et al. [12] pointed out that the production control problem with uncertain or variable demand can be faced through two different approaches. The first one, defined as ‘guaranteed approach’, consists of implementing production control policies, such as HPP and HCP, which can be considered ‘good on average’ solutions. This means that these policies guarantee the best possible results when applied several times. The second approach is defined as ‘adaptive approach’ and concerns with a strategy that uses some information about a source of uncertainty to create a rule that varies according to such information. In this regard, they propose an adaptive approach based on a Kalman filter-based technique as an estimator for a failure-prone single-machine single-product manufacturing system where the demand is uncertain. To the best of our knowledge, the adaptive approach has not been used to solve the production control problem of a failure-prone two-product manufacturing systems with demand variability. To fill this gap, a new adaptive strategy called Adaptive Hedging Corridor Policy (AHCP) is proposed in this paper. It can be considered as a variant of HCP, since it consists in building positive inventory levels through a variable inventory threshold that protects against inventory shortages due to production changeovers and failures. The adaptive aspect consists of continuously changing the inventory threshold by estimating the demand. For this purpose, the variable demand

4

R. R. Corsini et al.

of the distribution chain is estimated using the exponential smoothing forecasting technique. Inspired by the work of Costa et al. [11], in this paper AHCP is compared with HCP to maximize the CSL when a highly variable demand arises from the downstream players. For this purpose, an analytical simulation model based on discrete-time difference equations has been used to obtain the outcomes of a full-factorial Design Of Experiment (DOE). AHCP and HCP were compared through several scenarios generated by varying the nominal production capacity, the changeover times and an inventory threshold factor used by each PCP. A proper ANOVA analysis and a series of interval plots at 95% confidence intervals were arranged to evaluate the impact and effectiveness of the two PCPs on the CSL indicator as well as the interactions among the adopted experimental factors. The reminder of the paper can be summarized as follows. Section 2 describes the manufacturing system and reports the mathematical formalization of the analytical model. Section 3 describes the two production control policies. Section 4 deals with the experimental design and the statistical analysis of the results. Finally, Sect. 5 reports the managerial implications and the conclusion of the present paper.

2 Description of the Manufacturing System The manufacturing system under study consists of an unreliable two-product production line. Therefore, it is subject to failures and cannot manufacture both types of products simultaneously. In this regard, changeover operations are required to switch from one type of product to another and the changeover time is not negligible. Before starting the changeover operations, the semi-finished products have to be completed. Both changeover and failure events cause disruptive phenomena such as production rate decay and inventory shortages. A PCP is adopted to cope with these adverse events. The manufacturing system always has raw materials available; thus, does not issue any order to a supplier. Therefore, the production line produces at the maximum production rate. When the production process is concluded, the finished products are stored in inventories used to satisfy the demand. The inventories are limited by two bounds, i.e., an upper and a lower bound named inventory threshold and inventory threshold of lost sales, respectively. The inventory threshold is a target level used to make decisions about the product changeovers. Specifically, the decision on product changeover is made when the forecasted inventory level of finished products overcomes its inventory threshold. On the other hand, the inventory threshold of lost sales is adopted by the manufacturer to limit the backlog quantity. In fact, it is assumed that the cost of backlog is larger than the cost of lost sales when the backlog level exceeds the inventory threshold of lost sales [11, 13, 14]. Finally, the manufacturing system is connected to a distribution chain consisting of a distributor and final users. The demand of the manufacturing plant is variable since it consists of the orders issued by the distributor using the smoothing Order-Up-To replenishment policy [11, 15]. Figure 1 depicts the diagram of the system under investigation and shows that the

Assessing the Effect of a Novel Production Control Policy …

5

Fig. 1 Diagram of manufacturing system and its connection with the distribution chain

manufacturer and the distribution chain are connected by an information flow, i.e., the orders emitted by the distributor, and a material flow, i.e., the quantity of finished products delivered by the manufacturer to the distributor. It is assumed that there is no information sharing among them.

2.1 Notation The analytical simulation system under investigation has been modelled through discrete-time difference equations. Specifically, the time horizon is partitioned into discrete time slots, so-called time-buckets, that have the same size. The following notations are used in this paper: Parameters. i p t F χp α z

Index related to the manufacturing plant and the distribution chain, i.e., i=1 is the manufacturing plant, i=2 is the distributor Index of the type of product p ∈ P = {A, B} Time-bucket (t = 1, . . . , T ) Production flow time Nominal production capacity of product p Forecasting smoothing factor Inventory threshold factor

6

k δ λ LT ε β

R. R. Corsini et al.

Threshold of lost sales factor Changeover time Failure rate Delivery lead-time Safety stock factor Proportional controller

Variables I N P p,t Wi, p,t OU T p,t Ii, p,t Iˆp,t Z p,t C O p, p ,t tr t Ci, p,t T I p,t T W p,t O p,t dˆi, p,t d p,t

Input quantity of product p at time-bucket t Work-in-progress quantity at stage i of product p at time-bucket t Output quantity of product p at time-bucket t Inventory level at stage i of product p at time-bucket t Forecasted inventory level of product p at time-bucket t Inventory threshold of product p at time-bucket t Residual changeover time for switching from product p to product p  at time-bucket t Repair time at time-bucket Delivered units at stage i of finished products p at time-bucket t Target inventory of product p at time-bucket t Target work-in-progress of product p at time-bucket t Orders of product p placed by the distributor at time-bucket t Forecasted orders at stage i of product p at time-bucket t Final users’ demand of product p at time-bucket t

2.2 Mathematical Formalization of the Manufacturing System The time each product needs to pass through the production line is the production flow time, hereinafter denoted by F. In other words, it is the interval between the time feeding the production line with raw materials and removing finished products to be stored in inventories. The number of raw materials entering the production line at time-bucket t is called as input quantity I N P p,t . It is equal to the nominal production capacity χ p unless there is a changeover or a failure event. When the production system is manufacturing the product p, the input quantity I N P p,t is calculated as:   INPp,t = χ p · (1 − trt ) · 1 − C O p , p,t

(1)

Notably, the time to repair tr t is derived from a uniform distribution in the range [0,1] when a failure event occurs with probability λ. The residual changeover time C O p , p,t is equal to δ if the changeover event happens at time-bucket t, otherwise, it is equal to max{C O p , p,t−1 − 1; 0}. The production line is manufacturing the product p in two alternative cases:

Assessing the Effect of a Novel Production Control Policy …

7

• The product p was processed in the previous period t − 1 and no decision on the product changeover has been taken (INP p,t−1 > 0 and C O p, p ,t = 0); • The changeover procedure to switch from the type of product p  to the type of product p is nearly (C O p , p,t ∈]0, 1[) or definitively completed (C O p , p,t−1 ≥ 1 and C O p , p,t = 0); It can be noticed that the value of the residual changeover time C O p , p,t can be in the range ]0, 1[ since it is assumed that both changeover and production operations could occur into the same time-bucket t. The output quantity OU T p,t is the number of finished products removed by the production line after a period equal to the production flow time F: OU T p,t = INPp,t−F

(2)

Therefore, the production work-in-progress (WIP) level W1, p,t can be calculated as follows: W1, p,t = W1, p,t−1 + INPp,t − OU T p,t

(3)

To further explain the work-in-progress, Fig. 2 describes the two-product manufacturing system and Fig. 3 shows the variation of the WIP level. In both cases, it is

Fig. 2 Production system with non-negligible changeover times

8

R. R. Corsini et al. WA Serie1

00

W Serie2 WIP−CAPA

00

Work In Progress

00

χA

00

WIP−CAPB

00 00 00 0

00

χB

χA

χB

202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219Time 220

Fig. 3 Variation of the work-in-progress

considered a scenario where both the production flow time (F) and the changeover time (δ) are equal to 2 periods. Under steady-state conditions (i.e., without failures), the WIP level reaches a maximum constant level, which is called as WIP-CAP. There is a different WIP-CAP for each type of product, since it is computed as F · χ p . On the other hand, if a decision on the product changeover is taken, the WIP level turns to zero in F periods. The WIP level of the other product starts to increase when the changeover operation is completed, i.e., after δ periods. The finished products removed from the production line at time-bucket t increase the inventory level I1, p,t . At the same time-bucket t, the inventory level I1, p,t decreases to fulfill the demand, i.e., the distributor’s orders O p,t−1 . If the inventory level is not enough to completely fulfill the demand, a stock-out is generated and partial replenishment is used. In addition, the backlog is allowed without exceeding a threshold of lost sales [13, 14], defined as −k · μd p . In fact, it is assumed that, in this case, the cost of backlog is larger than the cost of lost sales and, thus, the manufacturer does not accept to further increase the backlog level (i.e., negative inventory level). In general, the inventory level can be computed as follows:   I1, p,t = max I1, p,t−1 + OU T p,t − O p,t−1 ; −k · μd p

(4)

The units delivered to the distributor are represented by C1, p,t . The delivery process is characterized by an unlimited capacity. The delivered units are calculated as:   C1, p,t = max min(I1, p,t−1 + OU T p,t ; O p,t−1 ); 0

(5)

Assessing the Effect of a Novel Production Control Policy …

9

2.3 Mathematical Formalization of the Distribution Chain The distributor sends orders to the manufacturing plant throughout the time horizon. The smoothing Order-Up-To replenishment policy is adopted by the distributor to define the order quantity O p,t of product p at time-bucket t. It is defined as follows: O p,t = max{dˆ2, p,t + β · (TI p,t − I2, p,t ) + β · (T W p,t − W2, p,t ); 0}

(6)

The logical operator “max” is included since returns of finished products are not allowed. It can be noticed that that the order quantity O p,t is the sum of three components. The first component of O p,t is the forecast of the final users’ demand dˆ2, p,t . It is calculated by using the exponential smoothing technique as follows: dˆ2, p,t = α · d p,t + (1 − α) · dˆ2, p,t−1

(7)

where α is the smoothing factor, which can assume values in the range of [0,1] and d p,t is the final users’ demand of product p at time-bucket t. d p,t is derived from a normal distribution with mean μd p and variance σd2p . The second component of O p,t is the difference between the target inventory T I p,t and the current inventory level I2, p,t . The target inventory is the product of a safety stock factor ε and the forecast of the final users’ demand dˆ2, p,t : TI p,t = ε · dˆ2, p,t

(8)

The inventory level I2, p,t of product p at time-bucket t depends on d p,t , which is the final users’ demand to satisfy at time-bucket t, and C1, p,t−L T , which is the amount of finished products delivered by the manufacturing plant after the delivery lead-time LT. If the inventory level is not enough to completely satisfy the final users’ demand, backlogs are allowed and limited by the threshold of lost sales. The inventory level is defined as: I2, p,t = max{I 2, p,t−1 +C 1, p,t−L T −d p,t ; −k · μd p }

(9)

It can be noticed that the difference between the target inventory and the current inventory level is adjusted by a proportional controller β that can assume values in the range [0,1]. The last component is the difference between the target delivery work-in-progress T W p,t and the current delivery work-in-progress level W2, p,t . This difference is also adjusted by the proportional controller β. The target work-in-progress T W p,t is the product of a delivery lead-time LT and the forecast of the final users’ demand dˆ2, p,t : T W p,t = L T · dˆ2, p,t

(10)

10

R. R. Corsini et al.

The delivery work-in-progress W1, p,t is defined as: W2, p,t = W2, p,t−1 + C1, p,t − C1, p,t−L T

(11)

Finally, it is noteworthy that if the inventory level is not enough to fulfill the final users’ demand, the distributor adopts partial replenishment to deliver the units of products C2, p,t , defined as follows: C2, p,t = max{min{I2, p,t−1 + C1, p,t−L T ; d p,t }; 0}

(12)

3 Production Control Policies A production control policy (PCP) is required by the manufacturing system to cope with inventory shortages due to production changeovers and failure events. In each period t, the PCP determines the type of product to be manufactured and whether a changeover operation is required. In this paper, two production control policies are compared: Hedging Corridor Policy (HCP) and Adaptive Hedging Corridor Policy (AHCP). Since the manufacturing system is characterized by a production flow time F that involves semi-finished products in the production line, the decision to switch from one type of product to another should consider the future inventory level, hereinafter defined as forecasted inventory level Iˆp,t and calculated as follows:    Iˆp,t = max I1, p,t + F · χ p − dˆ1, p,t ; −k.μd p

(13)

where dˆ1, p.t is the forecasted demand of product p at time-bucket t. It is calculated using the exponential smoothing technique, as follows: dˆ 1, p,t = α · O p,t−1 + (1 − α) · dˆ 1, p,t−1

(14)

The production control policies studied in this work are described in the following sections.

3.1 Hedging Corridor Policy HCP consists of creating a positive inventory level that protects the manufacturing plant against inventory shortages due to changeover or failure events. According to HCP, the decision on product changeover is made by comparing the forecasted inventory level with an upper bound, called the inventory threshold (Z p ). In HCP,

Assessing the Effect of a Novel Production Control Policy …

11

Fig. 4 Variation of current inventory level and forecasted inventory level with HCP

Z p is fixed and depends on a control parameter, named the inventory threshold factor (z), and the mean value of the final users’ demand (μd p ), which is known in advance. It is defined as follows: Z p = z · μd p

(15)

Figure 4 shows the variation of the forecasted inventory level Iˆp,t when HCP is chosen as the production control policy of the manufacturing system. A product changeover event occurs when the forecasted inventory of the product type in process is equal to or larger than its inventory threshold (i.e., Iˆp,t ≥ Z p ). Consequently, the forecasted inventory level of product p decreases as the output quantity turns to zero, while, the forecasted inventory level of the alternative product p’ increases since it is being manufactured.

3.2 Adaptive Hedging Corridor Policy AHCP is a new version of HCP proposed in this paper to deal with the uncertainty of the variable demand coming from the distribution chain. Different from HCP, Z p,t is calculated at each time-bucket t and depends on the inventory threshold factor (z), which can be adjusted by managers, and the forecasted demand dˆ1, p,t , which is calculated in each period by using the exponential smoothing method, as in Eq. (14). Therefore, the inventory threshold Z p,t is defined as: Z p,t = z · dˆ1, p,t

(16)

12

R. R. Corsini et al.

Fig. 5 Variation of current inventory level and forecasted inventory level with AHCP

Figure 5 shows the variation of the forecasted inventory level Iˆp,t when AHCP is applied. The changeover event occurs if the forecasted inventory level of the product type in process exceeds the inventory threshold (i.e., Iˆp,t ≥ Z p,t ). It is noteworthy that the threshold is not linear but assumes a variable trend. As in HCP, when a decision on the product changeover is taken, the forecasted inventory level of product p decreases and, on the other hand, the forecasted inventory level of the other product type p’ increases.

4 Experimental Campaign and Results In order to assess the effect of the novel AHCP compared to the well-known HCP as production control policy of a failure-prone two-product manufacturing system, a full-factorial Design Of Experiment (DOE) was conducted, as described in Sect. 4.1. The results of the experimental campaign are reported and analyzed in Sect. 4.2.

4.1 Design of Experiments The proposed DOE entails four independent variables: production control policy (PCP), the ratio between nominal production capacity and the mean of the final users’ demand (χ p /μd p ), changeover time (δ) and inventory threshold factor (z). The production control policy is varied at two levels (HCP and AHCP) while the other independent variables at three levels. A number of 2 · 33 = 54 scenarios were considered. To avoid randomness biasing the results, 30 replications were considered for each scenario. Therefore, 30 · 54 = 1620 simulation runs were performed. The

Assessing the Effect of a Novel Production Control Policy …

13

Table 1 Model parameters Model parameters

Symbol

Values

Machin repair time

tr

U ∈ [0,1]

Smoothing forecasting factor

α

0.30

Standard deviation of demand/Mean demand

σd p /μd p

0.10

Delivery lead time

LT

2.00

Production flow time

F

2.00

Failure rate

λ

0.10

Threshold of lost sales factor

k

2.00

Proportional controller

β

0.20

Safety stock factor

ε

1.00

Mean demand of product A

μd A

100

Mean demand of product B

μd B

50

Table 2 Design of experiments Independent variables

Symbol

Levels I

II

III

Production control policy

PCP

HCP

AHCP

Nominal production capacity/Mean demand

χ p /μd p

3.00

3.50

– 4.00

Changeover time

δ

1.00

2.00

3.00

Inventory threshold factor

z

7.00

9.00

11.00

values of the model parameters, shown in Table 1, and the independent variables, reported in Table 2, were set in accordance with common values widely used in the relevant literature [11, 15, 16]. The production control policies were tested by investigating the CSL as response variable of the experimental campaign, which is defined as follows: C S L p = (μC2, p /μd p ) · 100

(17)

where μC2, p is the mean of the units of products delivered by the distributor to the final users into the time horizon (T ) of the simulation, while μd p is the mean of the final users’ demands. Notably, the simulation time T is equal to 2,000 periods, also including a warm-up time of 200 periods. The simulation model based on discretetime difference equations, coded in Matlab r2020®, was used to obtain the results of the experimental campaign. The simulation runs were carried out on a workstation equipped with a INTEL i9-9900 3.6 GHz 10 core CPU, 32 Gb DDR4 2,666 MHz RAM and Win 10 PRO OS.

14

R. R. Corsini et al.

4.2 Analysis of Results To evaluate the significance of the model and the independent variables on the CSL, an ANOVA analysis with 95% confidence interval was performed by using the commercial package Minitab 17® as a statistical tool. The outcomes resulting from this analysis are shown in Fig. 6. For the sake of simplicity, statistical analysis just refers to the results of the first product (hereinafter denoted as product A) since no significant difference was found between the two types of products. Firstly, it can be observed that the p-value of the model and the blocks (i.e., replications) are equal to 0 and 1, respectively. These demonstrate that the proposed simulation model is statistically significant and the randomness does not affect the results. R-sq is equal to 95.15% and therefore the quadratic model fit is effective. Moreover, the predicted R-sq is in reasonable agreement with the adjusted one since their difference is less than 20% (R-sq(adj) - R-sq(pred) = 0.18%). Finally, all independent variables and 2-way interactions have p-values lower than 0.05, thus resulting statistically significant at a 95% confidence level. Figure 7 depicts the main effects plots of the ANOVA analysis and Table 3 reports the mean values of the CSL for each level of the independent variables studied. The most interesting findings come out from the main effects plot of the production control policy PCP. In general, the analysis shows that the new AHCP strategy is Analysis of Variance Source Model Blocks Linear PCP χ_p⁄μ_(d_p ) δ z 2-Way Interactions PCP*χ_p⁄μ_(d_p ) PCP*δ PCP*z χ_p⁄μ_(d_p )*δ χ_p⁄μ_(d_p )*z δ*z Error Total

DF 54 29 7 1 2 2 2 18 2 2 2 4 4 4 1565 1619

Adj SS 39.5008 0.0062 32.3601 2.4039 2.1296 19.2207 8.6059 7.1346 0.0097 1.2982 0.6453 0.9930 0.0874 4.1009 2.0135 41.5143

Adj MS 0.73150 0.00021 4.62286 2.40392 1.06479 9.61034 4.30294 0.39637 0.00486 0.64912 0.32267 0.24824 0.02185 1.02523 0.00129

Model Summary S 0.0358690

R-sq 95.15%

Fig. 6 ANOVA table

R-sq(adj) 94.98%

R-sq(pred) 94.80%

F-Value 568.56 0.17 3593.12 1868.45 827.61 7469.64 3344.46 308.08 3.77 504.53 250.80 192.95 16.98 796.86

P-Value 0.000 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.023 0.000 0.000 0.000 0.000 0.000

Assessing the Effect of a Novel Production Control Policy … χ_p⁄μ_(d_p )

PCP

100%

15

90%

CSL

80% 70%

HCP

AHCP

3.0

3.5

δ

100%

4.0

z

90% 80% 70%

1

2

7

3

9

11

Fig. 7 Main effects plots

Table 3 Mean values of CSL for each level of DOE Independent variables

Symbol

Mean CSL Lev. I

Lev. II

Lev. III

Production control policy

PC P

82.69%

90.39%

Nominal production capacity / Mean demand

χ p /μd p

81.60%

87.81%

90.20%

–

Changeover time

δ

99.61%

87.06%

72.95%

Inventory threshold factor

z

77.06%

87.78%

94.78%

more effective than the HCP rule. Indeed, AHCP allows gaining a mean value of CSL of 90.39%, while HCP allows achieving a mean value of CSL of 82.69%. It is worth highlighting the effect of the nominal production capacity (χ p ) on CSL. A higher nominal production capacity (χ p ) protects the manufacturing plant against inventory shortages. In fact, an adequate production capacity allows the inventory levels of each type of product to quickly reach the inventory threshold (Z p,t ). Therefore, the manufacturing plant easily meet the distributor’s orders (O p,t ) throughout the time horizon, thus increasing the CSL indicator of the distribution chain. In confirmation of this, the mean value of CSL is equal to 90.20% when the nominal production capacity of the manufacturing system is set to the highest level (i.e., χ p /μd p = 4.00). Moreover, it can be noticed that the CSL indicator is more sensitive to lower values of χ p /μd p . In fact, CSL deeply decreases moving from χ p /μd p = 3.50 (where the mean value of CSL is equal 87.81%) to χ p /μd p = 3.00 (where the mean value of CSL is equal to 81.60%). Looking at the F-values in the ANOVA table (see Fig. 6), the changeover time (δ) is the most influential independent variable. A higher changeover time (δ) has a

16

R. R. Corsini et al.

negative impact on the CSL indicator. Confirming this, there is a relevant difference in terms of CSL (i.e., C S L ∼ = 27% ) between the lowest value of changeover time (δ = 1.00) and highest one (δ = 3.00). It is noteworthy that when the changeover time (δ) is equal to one, disruptive phenomena are not significant since the CSL indicator is close to 100%. Interestingly, the inventory threshold factor (z) positively affects the performance indicator at hand. Although it might be better to maintain a lower inventory level (Ii, p,t ), a higher inventory threshold (Z p,t ) assures a better performance in terms of CSL. In fact, the CSL indicator moves from a value of 77.06%, when the inventory threshold factor (z) is set to 7 to a value almost equal to 95% when z is set to 11. Finally, in order to make the comparison between production control policies more robust, a series of interval plots illustrating the interactions between HCP and AHCP with the adopted experimental factors are presented in the following lines. More specifically, Fig. 8 reports the interaction between PCP with the ratio between the nominal production capacity and the mean demand (χ p /μd p ), Fig. 9 shows the interval plot involving the changeover time (δ) while Fig. 10 depicts the interaction with the inventory threshold factor (z). Interesting insights are obtained from the interval plot illustrated in Fig. 8. In general, the trends emerged by the main effects plot (depicted in Fig. 7) of the production control policies (PCP) and the nominal production capacity (χ p ) are confirmed in the interval plot. Indeed, the interval plot points out that the adaptive approach, i.e., AHCP, outperforms the traditional rule, i.e., HCP, in terms of CSL for each value of nominal production capacity (χ p ). Furthermore, it is worth noting that the AHCP rule allows the manufacturing plant and the distribution chain to achieve a CSL larger than 85% for each value of 100%

AHCP HCP

95%

CSL

90%

85%

80%

75%

3.0

3.5

4.0

χp μd

p

Fig. 8 Interval plot of production control policies for different levels of χ p /μd p to analyze the effect on the CSL

Assessing the Effect of a Novel Production Control Policy …

17

100%

AHCP HCP

CSL

90%

80%

70%

60%

1

2

3

δ Fig. 9 Interval plot of production control policies for different levels of δ to analyze the effect on the CSL 100%

AHCP HCP

95%

CSL

90% 85% 80% 75% 70% 65%

7

9

11

z Fig. 10 Interval plot of production control policies for different levels of z to analyze the effect on the CSL

the ratio between the nominal production capacity and the mean demand (χ p /μd p ). Interestingly, the performance achieved by AHCP with χ p /μd p = 3 is mostly equal to the CSL achieved when the manufacturing system with χ p /μd p = 4 adopts the traditional HCP. This means that the adaptive rule allows the manufacturing system to achieve the same performance of a manufacturing system with a larger nominal production capacity (χ p ) that adopts the HCP as production control policy.

18

R. R. Corsini et al.

The effectiveness of the production control policy also strictly depends on the time taken to perform the changeover operations (δ). When the changeover time is low (δ= 1), there is no difference in the CSL, regardless of the production control policy (PCP) adopted. In fact, in this case, the CSL is almost equal to 100% when both HCP and AHCP are used. On the other hand, the highest changeover time (δ) leads to a large decrease in the CSL indicator. Such detrimental effect depends on the production control policy chosen. In fact, when the changeover time (δ) is larger than one, Fig. 9 shows that AHCP is always more effective in terms of CSL. Finally, regarding the inventory threshold factor (z), lower values of z emphasize the difference in terms of CSL between AHCP and HCP. As shown in Fig. 10, the difference in CSL is more than 10% when z = 7, while, when z = 11, the performances of the two production control policies are similar.

5 Managerial Implications and Conclusion This paper addresses the evaluation of production control policies in a failure-prone two-product manufacturing system interacting with a distribution chain. This paper proposes a new PCP, called Adaptive Hedging Corridor Policy (AHCP). AHCP was compared with the well-known Hedging Corridor Policy (HCP), which is often used in manufacturing systems where the changeover time to switch from one type of product to another is not negligible. The interactions between PCP and experimental variables that characterize the production capacity and inventory of the manufacturing plant were considered. The objective of the paper is to identify the influence of these two PCPs in relation to the CSL indicator. Here, the main findings that emerge from the comparison of the two PCPs are summarized. Some of these findings can be considered relevant as they extend the literature of production control of two-product manufacturing systems by evaluating the effect of an adaptive strategy on the CSL performance measure. Such findings are: (1)

The adaptive AHCP strategy outperforms traditional HCP to improve the CSL performance measure.

This research should motivate practitioners to implement the proposed adaptive production control strategy to manage multi-product manufacturing systems with non-negligible changeover times. The results of the statistical analysis show that significant benefits may arise from employing an adaptive PCP focused on the CSL. Such benefits were emphasized by comparing two alternative PCPs, namely AHCP and HCP, in various scenarios where nominal production capacity, changeover time and inventory threshold have been varied.

Assessing the Effect of a Novel Production Control Policy …

(2)

19

The adaptive PCP approach is able to reduce shortages due to a lower production capacity.

The results of the analysis show that the AHCP can achieve higher values of the CSL indicator than the HCP for each value of nominal production capacity. It is worth noting that, the AHCP with χ p /μd p = 3 assures almost the same CSL of HCP when the ratio χ p /μd p is set to 4. Therefore, practitioners managing production systems with lower production capacity might be encouraged to adopt an adaptive control approach rather than investing in expanding the capacity of the manufacturing system. (3)

The choice of production control strategy requires a strong consideration of the changeover time required by the production system.

Depending on the changeover time (δ), the production control strategies can involve different effects on the CSL indicator. The results of this study show that there is no significant difference between the use of AHCP or HCP when the manufacturing system is characterized by a low changeover time (e.g. δ = 1). On the other hand, the production control strategy assumes a crucial role when the manufacturing system is described by long changeover times (e.g. δ = 3) and, thus, the adaptive approach allows achieving higher values of CSL then the HCP. (4)

Setting the inventory threshold is a key-factor to improve the CSL indicator.

The choice of an appropriate inventory threshold factor (z) is relevant. In fact, higher inventory threshold allows enhancing the CSL performance measure. As concerns the production control policies, the analysis of results highlights that AHCP emerges to be an effective strategy for each value of inventory threshold factor (z) considered in the investigated DOE. (5)

Investigating explicit multi-product models of unreliable manufacturing systems is needed to faithfully evaluate the impact of the production control policies in industrial contexts.

From a theoretical viewpoint, our study reveals that modelling multi-product manufacturing systems may provide new findings regarding the production control policies that cannot be discovered in aggregated single-product industrial scenarios. Then, the present paper would suggest that further studies may contribute to capture the impact of production control policies in unreliable multi-product manufacturing systems. Motivated by the above-mentioned findings, potential future studies may extend the knowledge of production control strategies in failure-prone two-product manufacturing systems. The study of AHCP strategy, which protects the manufacturing plant with variable demand from decay of production rate and inventory shortages, may be thorough by evaluating other performance indicators. Finally, it is worth highlighting that, in this paper, the smoothing forecasting technique has been used to estimate the future demand coming from the distribution chain. However, it may be

20

R. R. Corsini et al.

worthwhile to compare different forecasting techniques to evaluate if the effectiveness of the adaptive approach in terms of CSL performance measure can be further improved.

References 1. Hajji A, Gharbi A, Kenné JP, Hidehiko Y (2009) Production and changeover control policies of a class of failure prone buffered flow-shops. Prod Plan Control 20(8):785–800 2. Yang J, Qi X, Yu G (2005) Disruption management in production planning. Naval Res Logist (NRL) 52(5):420–442 3. Kimemia J, Gershwin SB (1983) An algorithm for the computer control of a flexible manufacturing system. AIIE Trans 15(4):353–362 4. Sharifnia A (1988) Production control of a manufacturing system with multiple machine states. IEEE Trans Autom Control 33(7):620–625 5. Gharbi A, Kenné JP, Hajji A (2006) Operational level-based policies in production rate control of unreliable manufacturing systems with set-ups. Int J Prod Res 44(3):545–567 6. Akella R, Kumar P (1986) Optimal control of production rate in a failure prone manufacturing system. IEEE Trans Autom Control 31(2):116–126 7. Elhafsi M, Bai SX (1996) Optimal production and setup control of a dynamic two-product manufacturing system: Analytical solution. Math Comput Model 24(3):57–78 8. Bai SX, Elhafsi M (1997) Scheduling of an unreliable manufacturing system with nonresumable setups. Comput Ind Eng 32(4):909–925 9. Assid M, Gharbi A, Hajji A (2014) Joint production and setup control policies: an extensive study addressing implementation issues via quantitative and qualitative criteria. Int J Adv Manuf Technol 72(5):809–826 10. Polotski V, Kenne JP (2017) Adaptive control of manufacturing systems with incomplete information about demand and inventory. IFAC-PapersOnLine 50(1):15598–15603 11. Costa A, Cannella S, Corsini RR, Framinan JM, Fichera S (2020) Exploring a two-product unreliable manufacturing system as a capacity constraint for a two-echelon supply chain dynamic problem. Int J Prod Res:1–29 12. Polotski V, Kenne JP, Gharbi A (2020) Kalman filter based production control of a failure-prone single-machine single-product manufacturing system with imprecise demand and inventory information. J Manuf Syst 56:558–572 13. Sajadi SM, Esfahani MMS, Sörensen K (2011) Production control in a failure-prone manufacturing chain using discrete event simulation and automated response surface methodology. Int J Adv Manuf Technol 53(1–4):35–46 14. Disney SM, Ponte B, Wang X (2020) Exploring the nonlinear dynamics of the lost-sales order-up-to policy. Int J Prod Res:1–22 15. Cannella S, Ciancimino E (2010) On the bullwhip avoidance phase: Supply chain collaboration and order smoothing. Int J Prod Res 48(22):6739–6776 16. Cannella S, Dominguez R, Ponte B, Framinan JM (2018) Capacity restrictions and supply chain performance: Modelling and analysing load-dependent lead times. Int J Prod Econ 204:264–277

Systematic Repeatability Analysis of Nanosecond Pulsed Laser Texturing Gianmarco Lazzini, Adrian Hugh Alexander Lutey, and Luca Romoli

Abstract The fabrication of specific surface features requires an adequate level of repeatability to be applied in industrial production environments. Although reference is frequently made to repeatability as a strength of many micro-machining techniques, few works in the literature have been dedicated to systematic and quantitative study of this parameter. The purpose of this work was therefore to quantitatively analyze the repeatability of textured surfaces with an original approach. Aluminum alloy substrates were textured with nanosecond laser pulses to produce micro-dimple arrays. Such features are widely employed to enhance wettability and tribological behavior, for which analysis was performed in terms of interfacial area and void volume. To give the work a more general character, two other parameters based on the definition of the Pearson’s Correlation Coefficient were also tested and compared to find the most suitable parameter for assessing repeatability for a given application. Finally, a study of the point-to-point repeatability of single surface features was conducted to detect variations in process repeatability in different portions of the same processed area. For the specific laser texturing process considered, it was found that an increase in total energy dose led to improved process repeatability. Keywords Repeatability · Micro-dimples · Wettability · Tribology

List of Symbols and Abbreviations Symbol T Tn ¯ h¯ h n , h,

Definition Set of three-dimensional coordinates representing topographic information Set of three-dimensional coordinates representing topographic information associated with dimple n Height of dimple n, arithmetic average height, standard deviation of height

G. Lazzini (B) · A. H. A. Lutey · L. Romoli Department of Engineering and Architecture, University of Parma, 43124 Parma, Italy e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 L. Carrino and T. Tolio (eds.), Selected Topics in Manufacturing, Lecture Notes in Mechanical Engineering, https://doi.org/10.1007/978-3-030-82627-7_2

21

22

¯  A¯ Anp , A, An V n , V¯ , V¯ PCC RPT rTnk rCnk r T  r C  E HAZ PPAM PPSTD

G. Lazzini et al.

Projected area of dimple n, arithmetic average of projected area, standard deviation of projected area Surface area of dimple n Volume of dimple n, arithmetic average volume, standard deviation of volume Pearson’s Correlation Coefficient Ring Projection Transform PCC between sets Tn and Tk PCC between the RPTs associated with sets Tn and Tk Arithmetic mean of rnk C over all 2-element combinations of dimples without repetition Arithmetic mean of rCnk over all 2-element combinations of dimples without repetition Energy dose per dimple Heat Affected Zone Point-to-Point Arithmetic Average Point-to-Point Standard Deviation

1 Introduction Repeatability is defined as the degree of agreement between the results of successive measurements of the same quantity carried out under the same conditions. This concept is an essential feature to guarantee the required production uniformity in many technological fields related to surface micro-machining, including micro-fluidics [1], biomedicine [2], tribology [3] and optics [4]. In particular, surface texturing has emerged as a novel and powerful approach to induce relevant surface properties on a substrate. In fact, it is known that a specific surface property is the result of effects induced on a skin layer when manufacturing surface topography. Textured surfaces with relevant functional properties require elevated repeatability. Such a characteristic is firstly required to grant sufficient uniformity of surface properties over the entire processed area. In addition, for some applications such as optics, repeatability is essential due to physical phenomena relating to interactions between the surface and light. Sophisticated surface texturing and micro-machining approaches have been developed to achieve highly regular surface morphology, including laser machining [5], photolithography [6], photopolymerization [7] and nano-imprinting [8]. Although reference is frequently made to repeatability as a strength of these techniques [9, 10], few works have been dedicated to systematic and quantitative study of this parameter [11]. Several factors can potentially affect the repeatability of a surface structuring process. The material properties of the treated surface can play a relevant role in determining the final surface morphology. This aspect is particularly crucial for metal alloys, which are often characterized by local variations in chemical composition

Systematic Repeatability Analysis of Nanosecond …

23

and crystalline structure that are randomly distributed throughout the volume [12]. Other factors affecting process repeatability can instead be traced back to intrinsic characteristics of the fabrication technique. For example, local fluctuations in the intensity profile of a laser beam or deviations in the angle of incidence can lead to measurable variability in the final surface morphology. Another example can be found in electrochemical machining (ECM), where process repeatability can be affected by electrolyte stagnation in the machining zone [13]. Deviations in the regularity of surface features have the potential to affect several functional surface properties relevant in many industrial applications. Given a patterned surface comprising features that are repeated many times over the workpiece surface, a characteristic length L can be associated with each repeating unit in relation to the functional property of interest. For example, if the repeating units have a dimple-like shape, an appropriate definition of L is the dimple depth. Therefore, the dispersion of L due to deviations in process repeatability can be estimated by the ratio L/L, where L is the mean value of L associated with multiple repeating units and L is the corresponding standard deviation. If the functional surface property of interest can be correlated to a change in the interfacial area provided by the machining process, the theory of errors allows the  dispersion associated with the 2 ( L 2 ) 1 ∂ (L )  L = 2L/L. An example interfacial area to be estimated as 2 = 2 ∂ L  L L L=L of this case is the wettability of rough surfaces, which is influenced by the contact area between the liquid and substrate according to the Wenzel and Cassie-Baxter theories. Analogously, if the property of interest depends on quantities related to the void   3 volume, the corresponding dispersion can be estimated by  L 3 /L = 3L/L. An example of this case includes the tribological properties of surfaces patterned with micro-dimples [14]. In this case, the void volume provided by such structures acts as a reservoir for lubricant and wear debris, increasing efficiency and durability, and reducing surface damage. These examples demonstrate that small losses in topographic repeatability can in principle lead to large errors depending on the macroscopic property of interest. In this work, a quantitative analysis approach is developed for assessing the repeatability of micro-dimple arrays produced by nanosecond pulsed laser ablation of aluminum alloy. Nanosecond pulsed laser ablation represents a fast single-step texturing process that can be performed on flat or curved surfaces in air or other gaseous environments without the need for clean room facilities. Though the use of nanosecond laser pulses leads to strong absorption, precise material removal and limited heat accumulation in the workpiece, significant amounts of molten material are ejected from the irradiated zone by strong hydrodynamic forces, giving rise to irregular rings or burrs around the edges of ablation craters, which can lead to loss of repeatability. Therefore, nanosecond laser texturing is characterized by lower repeatability than other manufacturing approaches such as photolithography [6], nano-imprinting [8] and ultrashort (femtosecond) pulsed laser ablation. However, its high throughput, ease of implementation and low setup cost make nanosecond laser processing the most widespread marking and surface structuring technique in industry and the most suitable method for creating patterns on large areas.

24

G. Lazzini et al.

An additional reason for choosing nanosecond pulsed laser irradiation to test the developed approach is that this technique belongs to a family of machining approaches characterized by complex physical phenomena due to rapid and nonuniform changes in the target properties. This complexity makes the development of detailed theoretical models for the prediction of surface morphology difficult. The study of process repeatability and regularity of surface features therefore cannot be performed by comparing experimental topographic data with a reference shape that is known a priori. The procedure developed in this work was therefore conceived to meet this requirement, allowing analysis to be performed on the similarity of surface features without the need for a reference shape. The work was carried out in line with the following methodology: • Aluminum alloy specimens were irradiated with a nanosecond pulsed laser to generate micro-dimple arrays. Different experimental conditions were adopted to obtain different dimple geometries, in particular crater depth; • The topography of all processed surfaces was acquired with an optical profiler; • A protocol for the segmentation and identification of topographic information of interest from the initial set of data was developed; • A set of morphological parameters was chosen to quantitatively study process repeatability. Parameters related to changes in surface and void volume were employed to study repeatability with reference to wettability and tribological properties. • A preliminary test was conducted on theoretical surfaces to assess the properties and behavior of the defined parameters. This step was essential to facilitate the interpretation of values calculated based on experimental topographic data; • A study on the point-to-point repeatability of laser-textured surfaces was conducted to understand variations in process repeatability within different portions of each surface.

2 Experimental Procedure AA6082 aluminum alloy specimens, 100 mm × 40 mm × 2 mm in size, were used for experiments. An Ytterbium-doped fiber laser with emission wavelength of 1064 nm and pulse duration of 104 ns was employed to perform laser texturing experiments [12]. The beam was scanned over the surface of each specimen with a galvanometric scanning head with 160 mm focal length f-theta lens. Texturing experiments were performed in ambient air with a focused laser spot diameter of 60 µm, scanning speed of 2400 mm/s and hatch distance of 120 µm, leading to completely separate ablation craters, or dimples. The average laser power was 16 W and the repetition rate 20 kHz, resulting in a pulse energy of 800 µJ. During the scanning procedure, treated portions of each specimen were subjected to 1–6 pulses in the same position by scanning the laser beam over the target surface multiple times. The energy dose, E, was defined as the average energy delivered to generate a single dimple per unit area. This quantity was used as a reference parameter to study process repeatability

Systematic Repeatability Analysis of Nanosecond … Table 1 Summary of the relevant processing parameters employed to generate the laser-textured surfaces

25

Parameter

Value

Repetition rate (kHz)

20

Power (W)

16

Total energy per pulse (µJ)

800

Focused spot diameter (µm)

60

Number of pulses per dimple

1, 2, 3, 4, 5, 6

Energy dose per dimple (E) (J/cm2 )

28.3, 56.6, 84.9, 113.2, 141.5, 169.8

under different experimental conditions. Relevant laser parameters are summarized in Table 1. The topography of each textured surface was acquired with a Taylor-Hobson CCI-MP Coherence Scanning Interferometer (CSI) equipped with a 50 × objective (NA = 0.55, WD = 3.4 mm), allowing a lateral resolving power of