Simulation-based optimization approach with scenario-based product sequence in a reconfigurable manufacturing system (RMS): A case study

9th IFAC Conference on Manufacturing Modelling, Management and 9th Control 9th IFAC IFAC Conference Conference on on Manufacturing Manufacturing Modelling, Modelling, Management Management and and 9th IFAC Conference on Manufacturing Modelling, Management and Control online at www.sciencedirect.com 9th IFAC Conference on Manufacturing Modelling, Management and Berlin, Germany, August 28-30, 2019 Available Control Control 9th IFAC Conference on Manufacturing Modelling, Management and Berlin, Germany, August 28-30, 2019 Control Berlin, Germany, August 28-30, 2019 Berlin, Germany, August 28-30, 2019 Control Berlin, Germany, August 28-30, 2019 Berlin, Germany, August 28-30, 2019

ScienceDirect

IFAC PapersOnLine 52-13 (2019) 2638–2643

Simulation-based optimization approach with scenario-based product sequence in a Simulation-based optimization approach with scenario-based product sequence in a Simulation-based optimization approach with scenario-based product in reconfigurable manufacturing system (RMS): A case studysequence Simulation-based optimization approach with scenario-based product sequence in in aaa reconfigurable manufacturing system (RMS): A case studysequence Simulation-based optimization approach with scenario-based product reconfigurable manufacturing system (RMS): A case study manufacturing (RMS): A study S. reconfigurable Ehsan Hashemi Petroodi*, Amélie Beauvillesystem Dit Eynaud** ***, Nathalie reconfigurable manufacturing system (RMS): A case caseKlement**, study

S. Amélie Beauville S. Ehsan Ehsan Hashemi Hashemi Petroodi*, Petroodi*, Amélie Beauville Dit Dit Eynaud** Eynaud** ***, ***, Nathalie Nathalie Klement**, Klement**, Reza Tavakkoli-Moghaddam**** S. Amélie Beauville Dit Dit Eynaud** Eynaud** ***, ***, Nathalie Nathalie Klement**, Klement**, S. Ehsan Ehsan Hashemi Hashemi Petroodi*, Petroodi*, Amélie Beauville Reza Tavakkoli-Moghaddam**** Reza Tavakkoli-Moghaddam****  S. Ehsan Hashemi Petroodi*, Amélie Beauville Dit Eynaud** ***, Nathalie Klement**, Reza Tavakkoli-Moghaddam****  Reza Tavakkoli-Moghaddam****  4, rue Alfred Kastler - B.P. 20722, F-44307 Nantes Cedex 3, *IMT Atlantique, LS2N - UMR CNRS 6004, La Chantrerie, Reza Tavakkoli-Moghaddam****  4, rue Alfred Kastler - B.P. 20722, F-44307 Nantes Cedex 3, *IMT CNRS 6004, La  4, rue Alfred Kastler - B.P. 20722, F-44307 Nantes Cedex 3, *IMT Atlantique, Atlantique, LS2N LS2N -- UMR UMR CNRS 6004,[email protected]) La Chantrerie, Chantrerie, France (e-mail:  4, rue Alfred Kastler - B.P. 20722, F-44307 Nantes Cedex 3, *IMT Atlantique, LS2N UMR CNRS 6004, La Chantrerie, *IMT Atlantique, LS2N UMR CNRS 6004, La Chantrerie, rue Alfred Kastler8,-boulevard B.P. 20722, F-44307 NantesLille, Cedex 3, France (e-mail: [email protected]) France (e-mail: [email protected]) **LISPEN, Arts et Métiers, Ecole Nationale Supérieure4, et Métiers. Louis XIV, 59046 France *IMT Atlantique, LS2N -HeSam. UMR CNRS 6004, La Chantrerie, 4,d'Arts rue Alfred Kastler - B.P. 20722, F-44307 Nantes Cedex 3, France (e-mail: [email protected]) FranceEcole (e-mail: [email protected]) **LISPEN, Nationale Supérieure d'Arts et Métiers. 8, boulevard Louis XIV, 59046 Lille, France **LISPEN, Arts Arts et et Métiers, Métiers, HeSam. HeSam. Ecole Nationale Supérieure d'Arts et Métiers. 8, boulevard Louis XIV, 59046 Lille, France (e-mail: [email protected]) France (e-mail: [email protected]) **LISPEN, HeSam. Ecole Nationale Supérieure d'Arts et Louis **LISPEN, Arts Arts et et Métiers, Métiers, HeSam. Ecole Nationale et Métiers. Métiers. 8, 8, boulevard boulevard Louis XIV, XIV, 59046 59046 Lille, Lille, France France (e-mail: [email protected]) (e-mail: [email protected]) *** Groupe PSA, Route de Supérieure Gisy, 78943d'Arts Vélizy-Villacoublay Cedex, France **LISPEN, Arts et Métiers, HeSam. Ecole Nationale Supérieure d'Arts et Métiers. 8, boulevard Louis XIV, 59046 Lille, France (e-mail: [email protected]) (e-mail: [email protected]) Groupe Route de Gisy, 78943 Vélizy-Villacoublay Cedex, France *** Groupe PSA, PSA, Route de Gisy, 78943 Vélizy-Villacoublay Cedex, France **** School*** of Industrial Engineering, College of Engineering, University of Tehran, Tehran, Iran (e-mail: [email protected]) *** Groupe Route 78943 Vélizy-Villacoublay Cedex, France Groupe PSA, PSA, Route de de Gisy, Gisy, 78943 Vélizy-Villacoublay Cedex, FranceTehran, **** of Engineering, College of University of Iran **** School School*** of Industrial Industrial Engineering, College of Engineering, Engineering, University of Tehran, Tehran, *** Groupe PSA, Route de Gisy, 78943 Vélizy-Villacoublay Cedex, FranceTehran, Iran **** School of Industrial Engineering, College of Engineering, University of Tehran, **** School of Industrial Engineering, College of Engineering, University of Tehran, Tehran, Tehran, Iran Iran **** School Industrial College of Engineering, Tehran, Tehran, Iran Abstract: In thisof study, we Engineering, consider a production planning andUniversity resourceofallocation problem of a Abstract: study, aa(RMS). production planning and allocation problem of Abstract: In In this thisManufacturing study, we we consider consider production planningscenarios and resource resource allocationfor problem of aa Reconfigurable System Four general are considered the product Abstract: In study, we aa(RMS). production planning and allocation problem of Abstract: In this thisManufacturing study, we consider consider production planning and resource resource allocation problem of ofaa Reconfigurable System Four scenarios are considered for product Reconfigurable System aims Four general general scenarios are considered for the product arrival sequence. The objective function to minimize total completion time of jobs. For a the given set Abstract: In thisManufacturing study, we consider a(RMS). production planning and resource allocation problem of a Reconfigurable Manufacturing System (RMS). Four general scenarios are considered for the product Reconfigurable System we (RMS). Four scenarios are considered foraa the product arrival sequence. The aims to time jobs. For given set of arrival sequence.Manufacturing The objective objective function to minimize minimize total completion time ofthe jobs. For given set of input parameters defined by the function market, want to findgeneral thetotal bestcompletion configuration forof production line with Reconfigurable Manufacturing System aims (RMS). Four general scenarios are considered fora the product arrival sequence. The objective function aims to minimize total completion time of jobs. For given set of arrival sequence. The objective function aims to minimize total completion time of jobs. For a given set ofa input parameters defined by the market, we want to find the best configuration for the production line with input parameters defined by the function market, we want to find on thetotal bestcompletion configuration forofthe production lineset with respect to the number of resources and their allocation workstations. In order to solve the problem, arrival sequence. The objective aims to minimize time jobs. For a given of input parameters defined by the market, we want to find the best configuration for the production line with input parameters defined by the market, we want to find the best configuration for the production line with respect to the number of resources and their allocation on workstations. In order to solve the problem, respect to the number of resources and their allocation on workstations. In order to solve the problem, aa hybridization approach based on simulation and optimization (Sim-Opt) is proposed. In simulation input parameters defined by the market, we want to find the best configuration for the production line with respect to the number of resources and their allocation on workstations. In order to solve the problem, a respect to the number of resources and their allocation on workstations. In order to solve the problem, hybridization approach based on simulation and optimization (Sim-Opt) is proposed. In simulation hybridization approach based on simulation and optimization (Sim-Opt) is proposed. In the simulation phase, atoDiscrete Eventofbased Simulation (DES) model is developed. On the other hand, simulated annealingaa respect the approach number resources and their allocation on workstations. Inisorder toa solve the problem, hybridization on simulation and optimization (Sim-Opt) proposed. In simulation hybridization approach based on simulation and is optimization (Sim-Opt) proposed. In the simulation phase, Discrete Event Simulation (DES)tomodel model is developed. On the the other hand, simulated annealing phase, aa Discrete Simulation (DES) developed. On hand, aa simulated annealing (SA) algorithm isEvent developed in Python optimize the solution. In other thisis the results of the hybridization approach based on simulation and is optimization (Sim-Opt) is approach, proposed. In the simulation phase, aa Discrete Simulation (DES) On hand, aa simulated annealing phase, Discrete Event Simulation (DES) model is developed. developed. On the the other hand, simulated annealing (SA) algorithm isEvent developed in model. Python tomodel optimize the performance solution. In other thisthese approach, theareresults results offrom the (SA) algorithm is developed in Python to optimize the solution. In this approach, the of the optimization feed the simulation On the other side, of solutions copied phase, a DiscreteisEvent Simulation (DES)tomodel is developed. On the other hand, a simulated annealing (SA) algorithm developed in Python optimize the solution. In this approach, the results of the (SA) algorithm isto developed in model. Python tothe optimize the performance solution. In best thisthese approach, thecan results offrom the optimization feed the simulation On other side, of solutions are copied optimization feed the simulation model. On the other side, performance of these solutions are copied from simulation model the optimization model. The best solution with the performance be achieved (SA) algorithm isthe developed in model. PythonOn tothe optimize the performance solution. In of thisthese approach, theareresults offrom the optimization feed simulation other side, solutions copied optimization feed the simulation model. On the other side, performance of these solutions are copied from simulation model to the optimization model. The best solution with the best performance can be achieved simulation model the optimization model. The best solution the best performance be from achieved by this manually cyclic approach. The proposed approach iswith applied a real case can study the optimization feed to the simulation model. On the other side, performance ofon these solutions are copied from simulation model to the optimization model. The best solution with the best performance can be achieved simulation model tocyclic the optimization model. The best solution is with the best be from achieved by this approach. The proposed approach applied on aa real study the by this manually manually approach. The proposed approach applied on performance real case case can study the automotive industry. Copyright © 2019 IFAC simulation model tocyclic the optimization model. The best solution is with the best performance can be from achieved by this manually cyclic approach. The proposed approach is applied on aa real case study from the by this manually cyclic approach. The proposed approach is applied on real case study from the automotive industry. Copyright © 2019 IFAC automotive industry. Copyright © 2019 IFAC by this manually cyclic approach. The proposed approach is applied on a real case study from the automotive industry. Copyright © 2019 IFAC Keywords: Simulation-Based Optimization, Reconfigurable Manufacturing System (RMS), © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. automotive industry. Copyright © 2019 IFAC Keywords: Simulation-Based Optimization, Reconfigurable Manufacturing System (RMS), automotive industry. Copyright © 2019 IFAC Keywords: Simulation-Based Optimization, Reconfigurable Manufacturing System (RMS), Reconfigurability, Simulated Annealing (SA). Keywords: Simulation-Based Optimization, Reconfigurable Keywords: Simulation-Based Optimization, Reconfigurable Manufacturing Manufacturing System System (RMS), (RMS), Reconfigurability, Simulated (SA). Reconfigurability, Simulated Annealing Annealing (SA). Keywords: Simulation-Based Optimization, Reconfigurable Manufacturing System (RMS), Reconfigurability, Simulated Annealing (SA).  Reconfigurability, Simulated Annealing (SA). Reconfigurability, Simulated Annealing (SA).  This study focuses on proposing a manufacturing system  1. INTRODUCTION This study focusesofon on proposing manufacturing system  This study focuses manufacturing enabling assembly twoproposing different aaatypes of products:system diesel 1. INTRODUCTION INTRODUCTION  This study focuses on proposing manufacturing system 1. This study focuses on proposing a manufacturing system enabling assembly of two different types of products: diesel INTRODUCTION enabling assembly of two different types of products: diesel engines (abbreviated DVR in the following paragraphs) and Nowadays, rapid and1. global data transformation, open market This study focusesofon proposing atypes manufacturing system 1. INTRODUCTION enabling assembly two different of products: diesel enabling assembly of two different types of products: diesel 1. INTRODUCTION engines (abbreviated DVR in the following paragraphs) anda Nowadays, rapid and global data transformation, open market (abbreviated DVR in the following paragraphs) and Nowadays, rapid and global compel data transformation, opensystems market engines gasoline engines (EB). A production planning problem of and changeable demand manufacturing enabling assembly of two different types of products: diesel engines (abbreviated DVR in the following paragraphs) and Nowadays, rapid and global data transformation, open market engines (abbreviated DVR in the following paragraphs) and gasoline engines (EB). A production planning problem of Nowadays, rapid and global data transformation, open market and changeable demand compel manufacturing systems gasoline engines (EB). A production planning problem of and changeable demand RMS is investigated. To calculate completion time ofof theaaa structure to rapid change. According themanufacturing new technologies, new engines (abbreviated DVR in the total following paragraphs) and Nowadays, and global compel datatotransformation, opensystems market gasoline engines (EB). A production planning problem and changeable demand compel manufacturing systems gasoline engines (EB). A production planning problem of RMS is investigated. To calculate total completion time of the and changeable demand compel manufacturing systems structure to change. According to the new technologies, new RMS is investigated. To calculate total completion time of the structure to change. According to the new technologies, new system, four general scenarios fortotal theplanning product problem sequence areaa product variety and high demand fluctuation for different gasoline engines (EB). A production ofthe and changeable demand compel manufacturing systems RMS is investigated. To calculate completion time of structure to change. According to the new technologies, new RMS is investigated. To calculate total completion time of the system, four general scenarios for the product sequence are structure to change. According to the new technologies, new product variety and high demand fluctuation for different system, four general scenarios for the product sequence are product variety and high demand fluctuation for different considered. Task assignment tofortotal thetheworkstations and oftheir products, production systems should be more flexible. They is investigated. To calculate completion time the structure to change. According to thefluctuation new technologies, new RMS system, four general scenarios product sequence are product variety and high demand for different system, four general scenarios for the product sequence are considered. Task assignment to the workstations and their product variety and high demand fluctuation for different products, production systems should be more flexible. They considered. Task assignment to the workstations and their products, production systems should be more flexible. They sequence in each station are already known. Lower and upper must be variety able to and reactsystems and response quickly inflexible. face of They these system, fourTask general scenariostoforthetheworkstations product sequence are product high demand fluctuation for different considered. assignment and their products, production should be more considered. Task assignment to the workstations and their sequence in each station are already known. Lower and upper products, production systems should be more flexible. They must be able to react and response quickly in face of these station already Lowerareand upper must be able to reactsystems and response face of these bound forin theeach number of are machines inknown. each station defined. changes. The traditional manufacturing were notThey able sequence considered. Task assignment to the workstations and their products, production shouldquickly besystems morein flexible. sequence in each station are already known. Lower and upper must be able to react and response quickly in face of these sequence station alreadyin Lowerare upper bound forin theeach number of are machines inknown. each station station areand defined. must be able to react and response quickly inManufacturing face of changes. The traditional manufacturing systems were notthese able bound for the number of machines each defined. changes. The traditional manufacturing systems not able to handle these challenges. The Dedicated sequence in each station are alreadyin known. Lowerareand upper must be able to react and response quickly in were face of these bound for the number of machines each station defined. changes. The traditional manufacturing systems were not able 2. LITERATURE REVIEW bound for the number of machines in each station are defined. changes. The traditional manufacturing systems were not able to handle these challenges. The Dedicated Manufacturing to handle these challenges. The Dedicated System (DMS) has a high throughput but very Manufacturing low of machines in each station are defined. changes. The traditional manufacturing systems wereflexibility not able bound for the number 2. LITERATURE LITERATURE REVIEW to handle these challenges. The Dedicated Manufacturing 2. REVIEW to handle these challenges. The System (DMS) has aa high throughput but low LITERATURE REVIEW System (DMS) has high throughput but very very Manufacturing low flexibility to produce different product partDedicated families. Theflexibility Flexible According to the2. evolution of manufacturing 2. LITERATURE REVIEW systems, each of handle these challenges. The Dedicated Manufacturing System (DMS) has aa high throughput but very low flexibility 2. LITERATURE REVIEW System (DMS) has high throughput but very low flexibility to produce different product part families. The Flexible According to the evolution of manufacturing systems, each each In of to produce different product part families. Theautomated Flexible According to theown evolution ofadvantages manufacturing systems, of Manufacturing System (FMS) consists of full them have their special and disadvantages. System (DMS) has a high throughput but very low flexibility to produce different product part families. The Flexible According to the evolution of manufacturing systems, each of to produce different product part families. The Flexible According to the evolution of manufacturing systems, each of Manufacturing System (FMS) consists of full automated them have their own special advantages and disadvantages. In Manufacturing System (FMS) consists of full automated them have their own special advantages and disadvantages. In components high flexibility to response afull stable demand According the dedicated manufacturing system (DMS), automation to produce with different product part families. Theautomated Flexible to the evolution ofadvantages manufacturing systems, eachand of Manufacturing System (FMS) consists of them have their own special and disadvantages. In Manufacturing System (FMS) consists of full automated them have their own special advantages and disadvantages. In components with high flexibility to response a stable demand the dedicated manufacturing system (DMS), automation and components with high flexibility to response a stable demand the dedicated manufacturing system (DMS), automation and of products. The Reconfigurable Manufacturing System structure are fixed. A possibility to produce in a high Manufacturing System (FMS) consists of full automated them have their own special advantages and disadvantages. In components with high flexibility to aa stable the dedicated manufacturing system automation and components with flexibility to response response stable demand demand the dedicated systemof(DMS), (DMS), automation and of The Reconfigurable System structure are fixed. A possibility to in of products. products. Thehigh Reconfigurable Manufacturing System are fixed. A advantage possibility totheproduce produce in aa high high (RMS) can overcome the limitations ofManufacturing these systems. RMS is structure throughput is manufacturing the main DMS (Koren components with high flexibility to response a stable demand the dedicated manufacturing system (DMS), automation and of products. The Reconfigurable Manufacturing System structure are fixed. A possibility to produce in a high of products. The Reconfigurable Manufacturing System structure are fixed. A possibility to produce in a high (RMS) can overcome the limitations of these systems. RMS is throughput is the the Moreover, main advantage advantage of the DMS (Koren and (RMS) can overcome the limitations of these systems. RMS is throughput is main of the DMS (Koren and part of Industry 4.0, aiming to cover high flexibility of FMS Shpitalni, 2010). the fixed structure of DMS is of products. The Reconfigurable Manufacturing System structure are fixed. A advantage possibilityoftotheproduce in a high (RMS) can overcome the limitations of these systems. RMS is throughput is the main DMS (Koren and (RMS) can overcome theDMS. limitations ofhigh these systems. RMS is Shpitalni, throughput is the main advantage of various the DMS (Koren and part of Industry 4.0, aiming to cover flexibility of FMS 2010). Moreover, the fixed fixed structure of DMS DMS is part of Industry 4.0, aiming to cover high flexibility of FMS Shpitalni, 2010). Moreover, the structure of is and high throughput of It is also able to adjust rapidly unchangeable either to manufacture products or to (RMS) can overcome the limitations ofhigh these systems. of RMS is throughput is the Moreover, main advantage of the DMS (Koren and part of Industry 4.0, aiming to cover flexibility FMS Shpitalni, 2010). the fixed structure of DMS is part of Industry 4.0, aiming to cover high flexibility of FMS Shpitalni, 2010). Moreover, the fixed structure of DMS is and high throughput of DMS. It is also able to adjust rapidly unchangeable either to manufacture various products or to andcontext high throughput of DMS. Itcover is also able tobyadjust rapidly either to the manufacture various products orThe to in of functionality and productivity rearranging increase throughput of system (Koren et al., 2017). part of Industry 4.0, aiming toIt high flexibility ofrapidly FMS unchangeable Shpitalni, 2010). Moreover, the fixed structure of DMS is and high throughput of DMS. is also able to adjust unchangeable either to manufacture various products or to and high throughput of DMS. It is also able to adjust rapidly unchangeable either to manufacture various products or to in context of functionality and productivity by rearranging increase throughput of the system (Koren et al., 2017). The in context of functionality and productivity byadjust rearranging throughput system (Koren al., 2017). The existing needed in the new increase Flexible manufacturing system (FMS) hadet been introduced and high components. throughput ofBecause DMS. ItRMS is alsoisable toby rapidly unchangeable either of to the manufacture various products orThe to in context of functionality and productivity rearranging increase throughput of the system (Koren et al., 2017). in context of andsystems, productivity rearranging increase throughput of the system (Koren etbeen al., 2017). The existing components. Because RMS is in the manufacturing system (FMS) had introduced existing components. Because RMS is needed needed in the new new Flexible Flexible manufacturing system (FMS) had been introduced generation of functionality manufacturing it isby important after DMS (Katz, 2007). These systems can adapt to many in context of functionality and RMS productivity byanin rearranging increase throughput of the system (Koren etbeen al., 2017). The existing components. Because is needed the new Flexible manufacturing system (FMS) had introduced existing components. Because RMS is needed in the new Flexible manufacturing system (FMS) had been introduced generation of manufacturing systems, it is an important after DMS (Katz, 2007). These systems can adapt to many generation of manufacturing systems, it is an important after DMS (Katz, 2007). These systems can adapt to many subject. The RMS is capable to cover a variable demand. manufacturing requirement easily and quickly (Abele et al., existing components. Because RMS is needed in the new Flexible manufacturing system (FMS) had been introduced generation ofRMS manufacturing systems, is after DMS (Katz, 2007). These systems can to many generation manufacturing systems, it variable is an an important important after DMS (Katz, 2007). These systems can adapt adapt to et many subject. The is capable to aait demand. manufacturing requirement easily and quickly (Abele al., subject. this Theof RMS is capable to cover cover variable demand. manufacturing requirement easily and possibility quickly (Abele et al., Hence, system is a dynamic system. According to the 2006). FMS has been extended to make structure to generation of manufacturing systems, it is an important after DMS (Katz, 2007). These systems can adapt to many subject. The RMS is capable to cover aa variable demand. manufacturing requirement easily and quickly (Abele et al., subject. The RMS is capable to cover variable demand. manufacturing requirement easily and quickly (Abele et al., Hence, this system is a dynamic system. According to the 2006). FMS has been extended to make possibility structure to According to the Hence, this system is a dynamic system. 2006). FMS has been extended to make possibility structure to variablethis demand in is real world systems, of manufacturing adjust and scale the capacity quickly subject. The RMS capable to cover areconfigurability variable demand. requirement easily and within quicklyproducing (Abele et part al., Hence, system is a dynamic system. According to the 2006). FMS has been extended to make possibility structure to Hence, this system is a world dynamic system. According to the 2006). FMS has been extended toRMS makehas possibility structure to variable demand in real systems, reconfigurability of adjust and scale the capacity quickly within producing part in real world systems, reconfigurability of variable demand adjust and scale the capacity quickly within producing part machines and system structure is an effective point of RMS. families (ElMaraghy, 2007). The been introduced by Hence, this system is a world dynamic system.reconfigurability According to the 2006). and FMSscale has been extended quickly to make within possibility structure to variable demand in of the capacity producing part variable demand in real real systems, of adjust adjust the The capacity quickly within machines and system structure issystems, an effective effective point of RMS. RMS. families 2007). The RMS introduced by machines and structure is an point of familiesand (ElMaraghy, 2007). The RMS has been introduced by Moreover, withsystem respect to world the extension of reconfigurability market competition et(ElMaraghy, al.scale (1999). efficiency is has highbeen in producing the contextpart of variable demand in real world systems, reconfigurability of Koren adjust and scale the capacity quickly within producing part machines and system structure is an effective point of RMS. families (ElMaraghy, 2007). The RMS has been introduced by machines and is using an effective point of RMS. families 2007). The RMS been introduced Moreover, withsystem respectstructure to the the extension of market competition al. The efficiency is high in the the context of Moreover, with respect to extension of market competition Koren et et(ElMaraghy, al. (1999). (1999). The efficiency high in context of and decreasing production costs, RMS leads system to Koren responsiveness to sudden changes of is thehas market (Battaïa et by al, machines and system structure is an effective point of RMS. families (ElMaraghy, 2007). The RMS has been introduced by Moreover, with respect to the extension of market competition Koren et al. (1999). The efficiency is high in the context of Moreover, with respect to the extension of market competition Koren et al. (1999). The efficiency is high in the context of and decreasing production costs, using RMS leads system to responsiveness to sudden changes of the market (Battaïa et al, and decreasing production costs, using of RMS leads system to responsiveness to sudden changes of thehigh market (Battaïa et al, progress inwith this situation. 2017). The is an system regarding capacity Moreover, respect to the extension market competition Koren et al.RMS (1999). Theadjustable efficiency is in the context of and decreasing production costs, using RMS leads system to responsiveness to sudden changes of the market (Battaïa et al, and decreasing production costs, using RMS leads system to responsiveness to sudden changes of the market (Battaïa et al, progress in this situation. 2017). The RMS is an adjustable system regarding capacity progress in this situation. The RMS is an adjustable system regarding capacity and decreasing production costs, using RMS leads system to 2017). responsiveness to sudden changes of the market (Battaïa et al, progress in this situation. 2017). The RMS is an adjustable system regarding capacity progress in this situation. 2017). The RMS is an adjustable system regarding capacity progress in this situation. RMS Ltd. is anAlladjustable system regarding capacity Copyright IFAC 26932017). 2405-8963 © 2019 2019, IFAC (International Federation of Automatic Control) Hosting The by Elsevier rights reserved. Copyright 2019 2693 Peer review© responsibility of International Federation of Automatic Copyright ©under 2019 IFAC IFAC 2693Control. Copyright © 2019 IFAC 2693 10.1016/j.ifacol.2019.11.605 Copyright © 2019 IFAC 2693 Copyright © 2019 IFAC 2693

2019 IFAC MIM Berlin, Germany, August 28-30, 2019 S. Ehsan Hashemi Petroodi et al. / IFAC PapersOnLine 52-13 (2019) 2638–2643

and functionality. In fact, the RMS covers advantages of both previous manufacturing systems (Bi et al., 2008) and also overcome some disadvantages of the FMS such as high cost, obsolescence, unfavourable tools, and unreliability (Mehrabi et al., 2000). One of the main differences between FMS and RMS is the customized flexibility of RMS and the general flexibility of FMS. Customized flexibility means that the system can be changed whenever it is needed (Wiendahl et al., 2007). A whole comparison of these three manufacturing systems have been proposed by (Zhang et al., 2006), (Koren and Shpitalni, 2010) and (Koren et al., 2017). Researchers worked on process planning in RMS; Chaube et al. (2012) proposed an NSGA-2 algorithm to solve an RMS process planning problem. Firstly, they assigned tasks to a set of reconfigurable machines, and then optimized the completion time and cost by scheduling these tasks. In this study, machines are not the same, and they have their own set of reconfigurations. In an other study, Prasoon et al. (2011) worked on optimization of reconfigurable set-up plans in a dynamic production system by studying an algorithm portfolio approach. Basically, simulation modelling is an efficient approach to handle the complex systems under uncertainty (Borshchev and Filippov, 2004). Simulation should be used as an approach to evaluate complex systems (Juan et al., 2015). Discrete Event Simulation is the procedure of modelling by considering different changes over time (Chica et al., 2017). On the other hand, optimization gives the possibility to find the best parameter combination in order to run the system efficiently. The right optimization method should be selected depending on the faced problem. Exact methods provide optimal solution for small size problems and metaheuristics can solve NP-hard problems (nondeterministic polynomial time problems), providing near optimal solution. For instance, Dou et al., (2009) proposed a genetic algorithm (GA) to find some of the best configurations among all the optimal configurations that had been obtained by some feasible generated sequences. A systematic approach to generate different feasible configurations for the single-product RMS was developed. Combination of simulation models and these optimization methods enhances the solution. Hybridization of simulation and metaheuristics provides interesting results because of its ability to provide high quality solutions for NP-hard real problems in reasonable calculation time (Juan et al., 2015). Gansterer et al. (2014) proposed a simulation-based optimization method to assess some parameters in production planning. Fu (2002) proposed a classification of hybridization approaches based on simulation and optimization, namely, simulation-based optimization and optimization-based simulation. Optimization and simulation phases connect with each other by different interfaces. These interfaces may be user-defined (Dehghanimohammadabadi et al., 2017) (Attar et al., 2017) or general tools like Excel software (Dehghanimohammadabadi et al., 2017). Simul8 is a good software to implement a Discrete Event Simulation (DES) model (Carteni and De Luca, 2012). Imran et al. (2017) linked DES and different metaheuristics.

2639

3. PROPOSED APPROACH 3.1 Problem Description Among the different paradigms to build a reconfigurable production system, the integration of mobile robots on the production system has been selected in order to enable easy reconfiguration of the production line. This choice is made based on the assumption that there exists a safety system enabling the integration of a collaborative robot on a movable platform. The reconfiguration of the production system consists in the reallocation of the movable robots on workstations. This occurs when the system is subject to fluctuations of the economic context, leading to changes of the production demand in volume or product variety. This paper focuses on product variety. In this study, we consider two objective functions aimed to minimize total completion time. The system contains m workstations with specific assigned tasks to produce p product types. The problem is about allocation of n identical mobile robots to the workstations. s = 1,2,…,m , r = 1,2,…,n and k = 1,2,…,p respectively are sets of stations, robots and product types. As parameters of the model, P k and crs are respectively the price of the product k and cost of per percent utilization of robot r on station s. These parameters are fixed. On the other hand, two other parameters are obtained by simulation: the number of final products k (Nk) and the utilization percentage of assigned robots to station s (us). In the proposed model, two binary decision variables xrs and yrs are 1 if robot r is assigned to station s respectively for producing DVR and EB, and one continuous variable CT representing total completion time. Task sequence of each product and processing time are already known. Lower and upper bound are considered for the number of robots in the stations. To calculate completion time, different scenarios are defined for product arrival. The objective function maximizing profit is calculated by (1) and the objective function minimizing completion time is presented by (2). 𝑝𝑝 𝑛𝑛 𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧1 = ∑𝑘𝑘=1 𝑁𝑁𝑘𝑘 𝑃𝑃𝑘𝑘 − ∑𝑚𝑚 𝑠𝑠=1 𝑢𝑢𝑠𝑠 ∑𝑟𝑟=1(𝑥𝑥𝑟𝑟𝑟𝑟 + 𝑦𝑦𝑟𝑟𝑟𝑟 ) 𝑐𝑐𝑟𝑟𝑟𝑟 𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧2 = 𝐶𝐶𝑇𝑇

(1) (2)

3.2 Methodology To model the proposed problem, an efficient DES model was developed using Simul8 tool. The simulation model is running with respect to the demand of specific period, in our study during one week. Tasks sequence, tasks duration, product mix ratio, and generated robot allocation are input parameters of the simulation model. Utilization of robots (in percentage of the total simulated time) and the number of final products are output parameters of this model which are considered as input parameters of the optimization phase (Fig. 1). The optimization module generates a new resource allocation, which is used as input for the next simulation run, and so on. Input parameters of the simulation model, as Excel file, are imported in Simul8. The simulation model is developed to provide insights into the workflow process, calculate the utilization percentage of resources, and obtain the number of final products.

2694

2019 IFAC MIM 2640 Berlin, Germany, August 28-30, 2019S. Ehsan Hashemi Petroodi et al. / IFAC PapersOnLine 52-13 (2019) 2638–2643

is repeated. After a predefined number of iterations, the algorithm stops. SA is applied to solve the two objective functions, minimizing completion time and maximizing the profit value. Torabi and Hassini (2008) introduced the so-called TH method, based on a fuzzy approach, used to solve the bi-objective model. 4. APPLICATION TO A CASE STUDY 4.1 Case Study: an Automotive Assembly System Fig. 1. Schema of the proposed hybridization approach. A Simulated Annealing (SA) algorithm was developed in Python for the optimization phase, presented Fig. 2.

In this study, we consider the engine assembly system of an automotive company as a case study. The factory aims to improve its current system and make a reconfigurable manufacturing system. In this company, two types of engines, diesel and gasoline, are assembled. These engines are respectively called DVR and EB. Four assembly station are investigated in this study. The assigned tasks to workstations and their sequence is shown by Fig. 4.

Fig. 2. Schema of the optimization phase

Fig. 4. Precedence of the assigned tasks to the workstations. Reconfigurability is reached by having mobile collaborative robots on AGVs. These robots are the resources we target to assign to workstations, depending on the current product mix and product sequence. The considered problem is production planning and resource allocation problem for the reconfigurable system. Actually, we want to evaluate the effect of reconfigurability on the current system. Thereby, two models with two different objective functions are considered in optimization phase. The objective functions are maximizing the profit of company and minimizing the total completion time. The assumptions considered in this study are:

Fig. 3. Flow chart of SA In the proposed SA, detailed in Fig. 3, a random generated solution is used as initial solution with an initial temperature, and neighbourhood solutions are generated by changing robot allocation on different stations according to a swap structure. If the newly generated solution improves the objective function, this last one is saved. However, keeping a worse solution is authorized according to a predefined probability, which decreases at each iteration of the algorithm. The temperature of the SA is updated, which reduces the acceptation of a worse solution for the next step, and the cycle

 Each robot should be assigned to only one workstation at the same time.  Conveyor starts moving when the last product is released in the stations.  The total completion time is calculated with respect to the four proposed scenarios and the number of assigned robots to the workstations.  There is an upper-bound for the number of assigned robots to the workstations. The maximum number of robots which can be assigned to each workstation is equal to the maximum number of parallel tasks in the workstation.  At maximum one robot can be assigned to workstation 10.  At maximum two robots can be assigned to workstation 11.  The maximum number of robots which can be used in workstations 12 and 13 are 5.

2695

2019 IFAC MIM Berlin, Germany, August 28-30, 2019 S. Ehsan Hashemi Petroodi et al. / IFAC PapersOnLine 52-13 (2019) 2638–2643

 At least one robot should be assigned to each workstation.  The maximum number of existing product in each workstation at each time is one product.  The assigned tasks to the workstations are already known.  DVRs and EBs have some common tasks and some specific tasks.  Tasks sequence for each type of product are already known. Different scenarios are considered for the product sequence in assembly line. For example, in a scenario DVRs are assembled and then EBs can be assembled (Fig. 5), and this scenario can be considered vice versa as second scenario (Fig. 6). In other scenarios, products might be assembled alternately, in which they can be started with a DVR (Fig. 7) or an EB (Fig. 8). In these scenarios the ratio of the number of DVR engines and EB engines can be different. In these figures, 1 robot in workstation 10, 2 robots in workstation 11, 5 robots in workstation 12 and 13 are assumed. Considering the relative ratio between the products is the main point in these scenarios. For example, the number of DVRs may be greater than EBs or vice versa.

2641

4.2 Results and Discussion A SA algorithm is developed to solve separately each objective function. A feasible random initial solution is generated. After 2000 iterations of the SA algorithm, the best profit value is obtained. The solutions for robot allocation to the stations are shown for DVRs (Fig. 9) and EBs (Fig. 10).

Fig. 9. Best robot allocation to the workstations maximizing profit value for assembling DVRs (X*).

Fig. 10. Best robot allocation to the workstations maximizing profit value for assembling EBs (Y*).

Fig. 5. Assembly line for assembling all DVRs before EBs.

After 2000 iterations, the minimal completion time is 2269 min with a 50-50 ratio of the two types of products (Fig. 11). Obtained solutions for robot allocation to the stations are shown for DVRs (Fig. 12) and EBs (Fig. 13). Scenario 2 is the best scenario for product sequence with respect to given number of products.

Fig. 6. Assembly line for assembling all EBs before DVRs.

Fig. 7. Assembly line for assembling products alternately, starting with DVR.

Fig. 11. Completion time values (10-2 min). Fig. 8. Assembly line for assembling products alternately, starting with EB. According to the number of assigned robots to the workstations, the defined scenarios and the processing time of tasks, total completion time of each station can be calculated. These values, presented in Table 1, are considered in 10-2 min. Table 1. Completion time of each station with respect to the number of assigned robots (Unit: 10-2 min) Nb of robots DVR EB Nb of robots DVR EB

WS 10 >=1 CT 42 CT 42

CT CT

1 110 0

WS 11 1 >=2 56 31 0 0 WS 13 2 3 70 60 0 0

1 85 79

2 51 62

4 50 0

>=5 40 0

WS 12 3 4 34 34 45 45

Fig. 12. Best robot allocation to the workstations minimizing completion time for assembling DVRs (X*).

>=5 17 45 Fig. 13. Best robot allocation to the workstations minimizing completion time for assembling EBs (Y*). To implement TH method for solving the bi-objective model, Z-positive ideal solution (ZPIS) and Z-negative ideal solution (ZNIS) for two objective functions are needed (Table 2). 2696

2019 IFAC MIM 2642 S. Ehsan Hashemi Petroodi et al. / IFAC PapersOnLine 52-13 (2019) 2638–2643 Berlin, Germany, August 28-30, 2019

Table 4, it is preferable to use scenario 1 (assembling all DVRs before EBs) when the EB ratio is higher, and use scenario 3 (assemble products alternately) when the DVR ratio is higher.

Table 2. ZPIS and ZNIS for minimizing completion time Second model (minimize completion time function) (min)

ZPIS 2269

ZNIS 3501

5. CONCLUSION AND FUTURE STUDY

Table 3. Values of the objective function with respect to different parameters of TH 𝒁𝒁𝟐𝟐 (min) 0.2 (0.2, 0.8) 0.579 0.999 2269 0.2 (0.5, 0.5) 0.610 0.999 2269 0.2 (0.8, 0.2) 0.686 0.670 2668 0.5 (0.2, 0.8) 0.642 0.999 2269 0.5 (0.5, 0.5) 0.648 0.999 2269 0.5 (0.8, 0.2) 0.650 0.670 2668 0.8 (0.2, 0.8) 0.562 0.999 2269 0.8 (0.5, 0.5) 0.668 0.999 2269 0.8 (0.8, 0.2) 0.690 0.670 2668 According to Table 3, it can be concluded that if the importance of an objective function (𝝋𝝋𝒊𝒊 ) increases, this objective function will be improved and vice versa. It also should be said that the satisfaction degree of each objective function will increase by changing 𝝑𝝑. Moreover, the Pareto frontier of the problem was provided. 𝝑𝝑

(𝝋𝝋𝟏𝟏 , 𝝋𝝋𝟐𝟐 )

𝝁𝝁𝟏𝟏

𝝁𝝁𝟐𝟐

The objective function values for different product ratios are listed in Table 4. These different values with respect to the different numbers of DVRs and EBs are shown Fig. 14.

In this paper, a general concept of hybridization approach based on simulation and optimization is implemented. It is developed to solve production planning and resource allocation problem in a reconfigurable manufacturing system (RMS). Solutions obtained in the optimization phase are then evaluated using the developed simulation model. In this approach, some parameters are obtained from simulation and imported into the optimization to give an efficient resource allocation structure. A metaheuristic (SA) has been applied on the problem to determine the best configuration to achieve a minimal completion time with maximization of the profit function. Consequently, it can be concluded that the proposed approach provides a good performance for design and production planning problems of RMS. Moreover, regarding defined scenarios for product sequence, the best scenario can be chosen with respect to the demand variety of product families. As a future study, other optimization algorithms could be tested to solve this problem. Furthermore, the methodology could be applied to an extended use case with a higher number of workstations and resources, meaning a higher degree of complexity. The hybridization between simulation and optimization has also to be continued, as the communication between the different modules is not fully automatized yet.

Table 4. Completion time values for different number of products Number of DVRs 0 1000 2000 2349 3000 4000 5000

Number of EBs 5000 4000 3000 2438 2000 1000 0

Completion time value (min) 2251 2301 2351 2269 2061 1031 182

ACKNOWLEDGEMENTS This work took place in the framework of the OpenLab ‘Materials and Processing’ combining ENSAM network, GeorgiaTech Lorraine, LIST Luxemburg and Groupe PSA company.

Best scenario 1 1 1 2 3 3 3

REFERENCES

Fig. 14. Completion time values for different number of products. The best scenario of product sequence with respect to the demand variety can be selected. Regarding results depicted in

Abele, E., Wörn, A., Martin, P., & Klöpper, R. (2006, July). Performance evaluation methods for mechanical interfaces in reconfigurable machine tools. In International Symposium on Flexible Automation, Osaka, Japan. Attar, A., Raissi, S., & Khalili-Damghani, K. (2017). A simulation-based optimization approach for free distributed repairable multi-state availability-redundancy allocation problems. Reliability Engineering & System Safety, 157, 177-191. Battaïa, O., Dolgui, A., & Guschinsky, N. (2017). Decision support for design of reconfigurable rotary machining systems for family part production. International Journal of Production Research, 55(5), 1368-1385. Bi, Z. M., Lang, S. Y., Shen, W., & Wang, L. (2008). Reconfigurable manufacturing systems: the state of the art. International Journal of Production Research, 46(4), 967-992. Borshchev, A., & Filippov, A. (2004, July). From system dynamics and discrete event to practical agent based modeling: reasons, techniques, tools. In Proceedings of the 22nd international conference of the system dynamics society (Vol. 22). Oxford: System Dynamics Society.

2697

2019 IFAC MIM Berlin, Germany, August 28-30, 2019 S. Ehsan Hashemi Petroodi et al. / IFAC PapersOnLine 52-13 (2019) 2638–2643

Cartenì, A., & De Luca, S. (2012). Tactical and strategic planning for a container terminal: Modelling issues within a discrete event simulation approach. Simulation Modelling Practice and Theory, 21(1), 123-145. Chaube, A., Benyoucef, L., & Tiwari, M. K. (2012). An adapted NSGA-2 algorithm based dynamic process plan generation for a reconfigurable manufacturing system. Journal of Intelligent Manufacturing, 23(4), 1141-1155. Chica, M., Juan Pérez, A. A., Cordon, O., & Kelton, D. (2017). Why simheuristics? Benefits, limitations, and best practices when combining metaheuristics with simulation. Dehghanimohammadabadi, M., Rezaeiahari, M., & Keyser, T. K. (2017, December). Simheuristic of patient scheduling using a table-experiment approach: simio and matlab integration application. In Proceedings of the 2017 Winter Simulation Conference (p. 238). IEEE Press. Dou, J. P., Dai, X., & Meng, Z. (2009). Precedence graphoriented approach to optimise single-product flow-line configurations of reconfigurable manufacturing system. International Journal of Computer Integrated Manufacturing, 22(10), 923-940. ElMaraghy, H. A. (2007). Reconfigurable process plans for responsive manufacturing systems. In Digital enterprise technology (pp. 35-44). Springer, Boston, MA. Fu, M. C. (2002). Optimization for simulation: Theory vs. practice. INFORMS Journal on Computing, 14(3), 192215. Gansterer, M., Almeder, C., & Hartl, R. F. (2014). Simulationbased optimization methods for setting production planning parameters. International Journal of Production Economics, 151, 206-213. Imran, M., Kang, C., Lee, Y. H., Jahanzaib, M., & Aziz, H. (2017). Cell formation in a cellular manufacturing system using simulation integrated hybrid genetic algorithm. Computers & Industrial Engineering, 105, 123-135. Juan, A. A., Faulin, J., Grasman, S. E., Rabe, M., & Figueira, G. (2015). A review of simheuristics: Extending metaheuristics to deal with stochastic combinatorial optimization problems. Operations Research Perspectives, 2, 62-72. Katz, R. (2007). Design principles of reconfigurable machines. The International Journal of Advanced Manufacturing Technology, 34(5-6), 430-439. Koren, Y., Heisel, U., Jovane, F., Moriwaki, T., Pritschow, G., Ulsoy, G., & Van Brussel, H. (1999). Reconfigurable manufacturing systems. CIRP annals, 48(2), 527-540. Koren, Y., & Shpitalni, M. (2010). Design of reconfigurable manufacturing systems. Journal of manufacturing systems, 29(4), 130-141. Koren, Y., Gu, X., & Guo, W. (2018). Reconfigurable manufacturing systems: Principles, design, and future trends. Frontiers of Mechanical Engineering, 13(2), 121136. Mehrabi, M. G., Ulsoy, A. G., & Koren, Y. (2000). Reconfigurable manufacturing systems: Key to future manufacturing. Journal of Intelligent manufacturing, 11(4), 403-419. Prasoon, R., Das, D., Tiwari, M. K., & Wang, L. (2011). An algorithm portfolio approach to reconfigurable set-up

2643

planning. International Journal of Computer Integrated Manufacturing, 24(8), 756-768. Torabi, S. A., & Hassini, E. (2008). An interactive possibilistic programming approach for multiple objective supply chain master planning. Fuzzy sets and systems, 159(2), 193-214. Wiendahl, H. P., ElMaraghy, H. A., Nyhuis, P., Zäh, M. F., Wiendahl, H. H., Duffie, N., & Brieke, M. (2007). Changeable manufacturing-classification, design and operation. CIRP annals, 56(2), 783-809. Zhang, G., Liu, R., Gong, L., & Huang, Q. (2006). An analytical comparison on cost and performance among DMS, AMS, FMS and RMS. In Reconfigurable manufacturing systems and transformable factories (pp. 659-673). Springer, Berlin, Heidelberg.

2698