A Continuos Simulation Approach

Simulation Modelling Practice and Theory 15 (2007) 185–198

www.elsevier.com/locate/simpat

A continuous simulation approach for supply chains inthe automotive industry

Henri Pierreval *, Romain Bruniaux, Christophe Caux

Laboratoire d’Informatique de Modelisation et d’Optimisation des Systemes UMR CNRS 6158, Research Group in

Manufacturing Systems of IFMA IFMA, Campus des Cezeaux, BP 265, F-63175 Aubiere, France

Available online 6 December 2006

Abstract

In the automotive industry supply chains, several factories collaborate to manufacture a product (car, engines, etc.). Inorder to fulfill customers’ needs, they have to be designed and organized in the proper way. The dynamic analysis of theirbehavior through simulation provides important information to improve their performances. Most existing research worksaddressing the modeling and simulation of supply chains are generally based on a discrete event worldview. We are con-cerned here with medium or long term decision problems, which necessitate ‘‘macroscopic’’ models of the supply chain. Atthese levels, the representation of the individual flows of the numerous parts that circulates in the supply chain being quitedifficult, given the objectives considered, we chose a continuous worldview. The models are based on Forrester’s systemdynamic paradigms. The proposed approach is actually applied to a large French company, in the automotive industry.The supply chain presented in this article is composed of five existing plants, located in two different production areas. Theresults show the concrete benefits that can be achieved. Several research directions are suggested.� 2006 Elsevier B.V. All rights reserved.

Keywords: Simulation; Intra firm network; Supply chain; Distributed enterprise; Modeling; System dynamic; Extended enterprise;Manufacturing systems; Automotive industry

1. Introduction

Nowadays, many companies have to collaborate through a network of production units, so as to providethe customer with the desired products. Supply chains (SCs), which are more precisely addressed in this article,generally refer to a set of networked organizations working together to source, produce and distribute prod-ucts and services to the customer [36,40]. In this article, emphasis is put on the sector of automotive industry,which includes various types of production units (e.g., forge, foundries, mechanics and assembly) and of com-plex products, which are composed of numerous sub-components. The production rates reach several thou-sands products per day and the production is managed according to a just in time strategy. Because of thehigh competition that exists in this sector, it is very important that the costs and the production lead times

1569-190X/$ - see front matter � 2006 Elsevier B.V. All rights reserved.

doi:10.1016/j.simpat.2006.09.019

* Corresponding author. Tel.: +33 473 28 81 06; fax: +33 473 28 81 00.E-mail addresses: [email protected] (H. Pierreval), [email protected] (R. Bruniaux), [email protected]

(C. Caux).

mailto:[email protected]



186 H. Pierreval et al. / Simulation Modelling Practice and Theory 15 (2007) 185–198

can be reduced, and that the due dates can be met. In this respect, these last decades, tremendous efforts havebeen put on better designing and managing each production facilities individually. It clearly appears, bothfrom the increasing number of publications in the area, and from the directions taken by most enterprises,that working globally on the improvements of the supply chain has become a major issue.

This article is concerned with the dynamic analysis of the supply chain through simulation modeling. Theresearch reported here, has been carried out during three years, from 1997 to 2000, to study the behavior of alarge network of plants producing cars, so as to point out directions of improvements and assist decision mak-ers in the global management of the supply chain. This research constituted the basis of Romain Bruniaux’sunpublished dissertation [8].

Simulation of supply chains has shown of growing interest these last years [47,44]. The area of automotiveindustry, presents several specific features. When the objective is to understand the global behavior of the net-work of plants, then we have to model the flows of products inside and between the production units. How-ever, the parts, or more precisely the components that circulate from one plant to another, are complex and ofnumerous types, so that capturing their dynamic behavior turns out to be quite difficult. In this respect, thesystem modeling approach considered in this article is macroscopic: we are not interested in following ortracking the behavior of particular entities (moreover, that would be quite difficult due to the too large numberof components that flow in the network). Rather, we aimed at addressing the following objectives:

(1) To highlight global dynamic trends of the global system behavior, considered as a collection of manu-facturing units that collaborate.

(2) To understand the major reactions of the whole network to particular situations, such as suddenincreases or decreases in the demand of certain classes of products, and the impact of major technicalproblems in given production units,

(3) to address questions about how different strategies to increase the production capacities (e.g., employeeovertime) of certain production units could contribute to avoid overloads or bullwhip effects.

To address such macroscopic objectives, a continuous simulation worldview was used. Continuous simu-lation approaches to tackle supply chains seem to be little reported in the related literature [38], where discreteevent models, often developed at a more microscopic level are mainly found.

The article is organized as follows. After introducing current works in the area, the suggested approach tomodel supply chains and networks of production units is presented. It is based on Forester’s system dynamic(SD) principles, which are briefly recalled. Then the application of this approach to a real industrial supplychain is described and discussed. Several research perspectives will finally be given.

2. Modeling and simulating the supply chain

2.1. Discrete world view

Concerning the research on supply chain behavior and characteristics, Holweg and Bicheno [21], observedthat the fundaments were laid by Forester [18], with his works on industrial dynamics. He was probably thefirst to highlight the major concepts and issues related to the modeling of extended enterprises, consideringflows of different natures. Unfortunately, this area of research seems to have been then less investigated inthe next decade, and articles addressing more specifically the dynamic modeling of the supply chain weremainly published since the late 1990s. The necessity to evaluate dynamically the performance of the differentproduction units involved from the suppliers of the raw material to the final customer was pointed out in agrowing number of articles (e.g., [25,23,24,5,41,36,7,44]). The simulation studies aim at analyzing such prob-lems as the possible overload of production units, the behavior of the inventory and the possible supply short-ages, or the well know bullwhip effect [28]. The models can be used to better understand how the supply chaindynamically behaves [21] and as decision support, to determine the impact of possible allocation strategies forhuman and technological resources, such as employees’ overtime, and new investments [32]. Thus simulationhas appeared as a powerful approach to study and design global strategies for the enterprise. Bruniaux [8], alsopointed out that simulation is a very relevant approach to study both the flexibility and the reactivity of the

H. Pierreval et al. / Simulation Modelling Practice and Theory 15 (2007) 185–198 187

supply chains to unexpected event, at a macroscopic level (e.g., logistic perturbations, raw material shortages,etc.). Other authors have put the emphasis on studying the efficiency of the logistic system, which insures thetransportation of parts, raw material or components between each location of the production activities [42].

These potential benefits have led to intensive publications [44], which address modeling methods for supplychains and for networks of plants. Object oriented modeling has been suggested by Changchien and Shen [11],and Biswas and Narahari [6] to model SC, for evaluating and analyzing reengineering proposals, and for deci-sion support purposes. Because of the uncertainty of the production environment, fuzzy sets theory has beenincorporated in the modeling approach [35]. Of course, more classical discrete event simulation packages havebeen used to model the SC and to provide animation capabilities. For example, ARENA was used at differentlevels of abstractions [22] for evaluating the business processes and inventory control parameters of a logisticand distribution SC. Benjamin et al. [4] have used WITNESS and PROSIM within the toolkit they suggest toprovide intelligent support to SC. The use of other well known simulation tools, such as Taylor II and Auto-mod is reported in [44]. The growing interest of both researchers and industrial decision makers in the mod-eling and simulation of SC has led to the development of dedicated environment based on such simulationtools as ARENA or EXTEND [37].

To cope more specifically with the integrated management issues, with the exchanges of information, or tointegrate software modules that already exist, a Java-based approach is proposed by Chatfield et al. [12] andZeigler et al. [48], propose to use the DEVS methodology and a CORBA framework. The high level architec-ture (HLA) has also been used to develop discrete event simulation based on SCOR models of the supplychain [3].

A few case studies, addressing real problems have been published. For example, an Ericsson’s SC of themobile communication industry was simulated [34] to compare several designs, with respect to quality,lead-times and costs, as well as to learn about the interrelations among parameters in the SC. Banerjeeet al. [2] study a three echelons SC network to examine the effect of transshipment approaches. Rota [41] stud-ies the behavior of a complex SC in the aerospace industry to provide assistance with the planning problems.Telle et al. [43] have simulated, using an agent-based discrete approach the customer–supplier relationships inan integrated supply chain. Higuchi and Troutt [20] report some lessons derived from their experience of sim-ulating the Tamagotchi supply chain

2.2. Continuous world view

As noted by Lee et al. [26], most of these published works are based on a discrete event worldview. Typ-ically, in classical manufacturing system simulations, one models the flow of individual products through a setof production resources (e.g., machines, operators, and automated guided vehicles), waiting in queues if nec-essary. At the supply chain level one studies the flow of batches of products (instead of products) that flowbetween production unit or work center (instead of machines), which wait in inventories and flow from oneunit to another using logistic resources (e.g., trucks). To help in the modeling, communication models areused, which can be based on business flowcharts, IDEF3, or SCOR models, as it is often done when modelingmanufacturing systems.

As a consequence, very few works have addressed the simulation modeling of supply chains using a con-tinuous simulation worldview. In a very specific area [33], have addressed a very specific problem using a‘‘master equation’’ to model attitudes between individuals in a network of small and medium enterprises.Lee et al. [26] suggest a combined discrete event-continuous approach. In fact, they argue that ‘‘SCs are nei-ther completely discrete nor continuous’’. If the ‘‘continuous nature’’ is not obvious (we will explain furtherwhy a continuous approach is relevant), their article demonstrates that part of the dynamic behavior of the SCcan be described in a relevant way using equations, especially when one is concerned with strategic levels.

The interest and potential of system dynamics in several areas of production engineering has already beenidentified. Baills and Vessilier [1] indicate that this approach was useful to solve several types of problems inthe automotive industry, in particular in marketing research and forecasting issues (see also [29]). Its applica-tion in the simulation of manufacturing facilities has already been discussed [17,45]. Movahedkhah [31] sug-gests SD to study the interrelation of performance indicators in food industries. Its more specific applicationto the supply chain [7] has been addressed in few research publications. Minegishi and Thiel [30] present the


simulation of a ‘‘supply chain’’ in the food industry. However, in their article, very little is said about the SC.Except [18], who clearly identified the possibility of his method to address various types of links between sup-pliers, factories, distributors and retailers, the simulation of supply chains using SD has been addressed inHaffez et al. [19,24]. In the latter, the potential of SD is underlined and a Vensim simplified model of certainaspects of the supply chain is given. Haffez et al. [19] and Degres et al. [13,15], present real applications of SDin steel industry, which clearly demonstrate the potential of SD in modelling real world SCs. Rabelo et al. [38]shown that SD can be combined with neural networks and eigenvalues to analyze supply chains. Theirapproach is illustrated by a case in the electronics manufacturing industry.

This work addresses a very different type of system and has different objectives. We are concerned with theautomotive area, which is characterized by numerous products with assembly operations and convergent billsof material [10]), while the steel industry is closer to process industries, with divergent bill of material. In thenext section, our research orientations are more precisely explained and justified.

3. System dynamic modelling of the supply chain

3.1. Global principles

In this work, emphasis is put on decisions to be made that concern the next 3–6 months and that are relatedto the behavior of plants that interact to manufacture or assemble a product. Hence, the manufacturing pro-cesses are represented at a ‘‘macroscopic level’’ [45]. We are not interested here in evaluating the average uti-lization of a machine, nor in determining when a given task will be completed. Instead, we would like to knowwhat will be the global dynamic behavior of entire sectors of plants.

System dynamics [18,39,7] privileges the study of the important flows of products or components throughthe main production areas of the network, rather than a detailed study of the flow of each product through theset of resources available in the network. Indeed, because of the large sizes of the plant considered in the auto-motive industry and of the numerous parts and components, a detailed model may not be realistic [45,17,24].

3.2. Main features of system dynamics

Forrester’s System Dynamics [18] has been used in many fields, such as management, finance, agri-foodindustries [46], production systems [45], tactical strategies for automotive industry [29], supply chain [7], auto-motive industrial plant networks [8] and steel industry [15]. In this approach, much attention is paid to theflows that control the organization, and therefore not only to the entities that compose it. To develop a model,Forrester suggests four main concepts, which are as follows:

• Levels describe accumulations within the system and are drawn as tanks. Levels represent the present valuesof the variable they contain that have resulted from the accumulated difference between inflows andoutflows.

• Flows, which transport the content of one level to another.• Decision functions, which control the rates of flows between levels (drawn as valves).• Information channels, which connect the levels to the decision functions.

These concepts are associated with graphical representations, which allow diagrams to describe the studiedsystem (Fig. 1).

Fig. 1 depicts a very simple structure of a reservoir or level, with an inflow and an outflow. To specify thedynamic behavior, a system of equations is defined. It consists of two types of equations, which correspond tolevels and decision functions (rates). Equations control the changing interactions of a set of variables, as timeadvances. The continuous advance of time is broken into small intervals of equal length dt. For example theequations describing the state of the level in Fig. 1 is

LevelðtÞ ¼ Levelðt � dtÞ þ ðDecision function 1� Decision Function 2Þ � dt

INITLevel ¼ 0ð1Þ

Decision function 1 Decision function 2Level

Information

Input flow Output flow

Fig. 1. System Dynamics basic symbols.


Level in Fig. 1 at time t depends on its own value at time t � dt and the value going in from decision function 1minus the value going out to decision function 2. It is necessary to give the initial value of it to solve this equa-tion. There are as many equations as variables. To determine the variables’ behavior, the differential equationssystem is integrated. Several software tools exist that support this approach (for example Vensim, I think orPowersim).

3.3. Modeling approach

To apply these system dynamic paradigms to the supply chain, we decompose each production plant intologistic units, which are defined as follows [8]: ‘‘a logistic unit concerns a physical sector of the plant (which canbelong to the same enterprise or be an external supplier), which has an upstream storage area, generally usedfor raw material or part components, and its own human and technical resources and facilities, to performtasks on the products’’. These logistic units are key concepts that are used to define the adequate level ofdetails (granularity) of the simulation models. This granularity may vary according to the types of problemsstudied and to the simulation objectives.

We are interested here in studying decisions at strategic or tactical levels (some months) rather than opera-tional levels. As a consequence this macroscopic level of modelling is also required to model the numerous partsor components that flow through the logistic units of the network of plants. Therefore, it is not desirable (itwould be anyway extremely difficult), to integrate the details of part routings, bills of materials and of the pro-duction orders involved in the studied supply chain. Parts are modelled through families of products that sharecommon characteristics and that have the same dynamic behavior in the global system. These families are builtthrough an aggregation procedure, whose purpose is to identify macroscopic flows, which is presented in Bruni-aux et al. [9]. The basic principles of this aggregation approach rely on a number of rules and constraints thatneeds to be taken into account, and on similarity criteria between the flow (rate, routings, etc.) of the consideredproducts. For example, 2 products that do not use the same resources of a given logistic unit cannot be groupedin the same family. Therefore, in a first step, the product flows that are relevant to consider regarding the sim-ulation objectives are identified, and the other will not be considered. Then, basic rules (such as the one givenabove) are used in an iterative procedure that converges towards a admissible decomposition of the set of prod-uct flows. If the resulting decomposition does not lead to a suited level of details, then statistical clusteringmethods [9] or optimization methods [14] can be used to search for an adequate number of classes.

The equations used in the simulation models are deduced from several types of analysis: from a systemicanalysis of the supply chain [18], from the identification and the analysis of the different logistic units involved,from the partition of the set of product flows and from the production rates and the production calendars.

This modelling approach has been used to study the industrial case in the automotive industry presented inthe next section.

4. Case study

4.1. The system

The global study was carried out on a number of networked industrial production facilities belonging to awell known cars producer in France, with the objective of identifying weaknesses, studying the flexibility of the


network and its global reactivity to disturbances and changes of the customers’ demand. Both for the sake ofconcision and confidentiality, we focus here on a particular supply chain, which is a subset of the global net-work. Detailed data and results cannot be provided. From a methodological point of view, in a preliminarystep, a systemic approach was conducted to identify the key variables and their relationships. Several staticmodels were built using the Modular Analysis of Systems methodology [27], and data flow diagrams [16],so as to identify the flow of products and information and define the logistic units involved. Causal diagramswere also used to highlight the interrelation between the variables. The static models have led us to decomposethe studied system in 5 logistic units, located in two different production sites, which are connected to the sup-pliers (upstream) and to the customers (downstream), as depicted in Fig. 2. The global system transforms rawmaterial, provided by the suppliers, into engines, delivered to customers. The activity, location and output ofeach logistic unit is given in Table 1. A calendar, which details what are the days on and off, the daily produc-tion rate (i.e., number of parts per day) is also associated with each unit.

The considered supply chain being managed according to a just in time strategy, customers send Kanbancards to unit 1, which sends cards to unit 2 and so on. These cards allow the production or shipment of eachunit to be controlled.

Several objectives are assigned to the simulation experiments. We are interested in studying the dynamicbehavior of inventories on a time horizon of four months. One important objective is to analyze possible pen-uries of products, which constitutes inputs and outputs of the different units. We are also concerned with thenumber of orders waiting in each unit and with comparing the possible (i.e., theoretical) throughput rates ofeach unit with the actual throughput rates. With regard to these objectives, several assumptions have beenmade. The suppliers are supposed to be able to provide unit 5 with enough raw material, and the effect of qual-ity defects on the flow of parts is neglected (in reality, the corresponding rate of defect is lower than 0.01 forthe machining of cases and for the assembly of engines). The daily capacity of shipping is assumed to be infi-nite (units 1 and 4). Therefore, these units work according to their production calendar (generally from Mon-day to Saturday).

The initial conditions turn out to have a great influence on the dynamic behavior of the supply chain, inparticular, the initial level of stocks. The volume of orders on the time period of the study and the way they

Flow of products Flow of orders

Unit 1 Unit 3Unit 2 Unit 4 Unit 5Rawcases

Raw cases

Machinedcases

Engines

Suppliers

Rawmaterial

Customers

Engines

Plant 2 Plant 1

Fig. 2. Supply chain modeled.

Table 1Characteristics of the logistic units

Logistic units Activity Location Output products

Unit 1 Shipping Site 1 EnginesUnit 2 Assembly Site 1 EnginesUnit 3 Machining Site 1 Machined casesUnit 4 Shipping Site 2 Raw casesUnit 5 Production Site 2 Raw cases


fluctuate constitutes essential factors. In the results given in the following, the orders generated in the modelhave been defined according to commercial forecasts on four months, as well as the production calendar ofeach unit. The initial state has been stated in a pessimistic ways, with low hypothesis about the stock levels,in order to better point out the system reactions to difficult situations.

4.2. Supply chain modelling

The model was built using the simulation system I think�. Fig. 3 presents the flow diagram of plant 2(defined as the logistic unit 4), which is in charge of shipping row cases. Given the very high number of existingproduct references, the data have had to be aggregated, as introduced in the previous section. After aggrega-tion, 7 families of products were defined and approved by the managers of the system.

Below are examples of important model equations

Engagement shippping plant 2½Cases� ¼ MINðStock raw cases plant 2½Cases�=dt;

Orders raw cases plant 2 waiting½Cases�=dt;

Throuput maxi shipments plant 2½Cases� � Calendar shipping plant 2Þ

Eq. (2) states that the shipments actually performed depends upon the available quantity of raw cases instock, of the raw cases waiting (in progress), upon the maximum production rate and upon the shipmentscalendar

Fig. 3. Example of flow diagram : unit 4.


End shipping plant 2½Cases� ¼ðWIP shipments plant 2½Cases�=Transportation time from plant 2½Cases�Þ� Calendar shipping plant 2 ð3Þ

Eq. (3) characterizes the transportation time between plant 2 (unit 4) and site 1 (unit 3).

Conso cases plant 2½Cases� ¼ Engagement shippping plant 2½Cases� ð4Þ

Eq. (4) means that the stock level of raw cases decreases according with the shipments performed.

5. Results

Given the very large model size and the confidentiality inherent to this study, it is not possible to give manydetails about the obtained results. Hence we have selected the most representative family of product (type 3engines), in a particular initial context (initial upstream stock levels, availability of resources), which is notfavorable to a quick completion of orders.

First, let us notice that each graph given in the following presents oscillations that correspond with openingand closing periods of the production calendars. If we focus on the analysis of the dynamic behavior of theplants modelled, Fig. 4 shows a shortage of raw transmission cases in unit 4. This shortage means that area4 would be capable of processing more cases if the upstream area (unit 5), could deliver more parts. Then, thisshortage is propagated to the downstream units. Fig. 4 shows that a shortage occurs soon after in unit 3. Thena shortage of transmission cases occurs in unit 2 and to unit 1, which finally has no more engine to produce.Therefore the consequences are particularly severe because the final customer will have to face a lack ofavailable engines.

Figure Caption Missing


We can now try to focus on the causes that have induced these shortages. Thanks to the simulation model,it is possible to analyze the behavior of each stock of the logistic units. Fig. 5 shows that the number of rawtransmission cases becomes lowest at the beginning of the considered period. The stock in unit 3 compensatesthis shortage for a limited period of time only. The same type of behavior occurs for the stocks in units 1 and2, which compensate the lack of delivery of products of upstream units and then become too low. Therefore,these intermediary stocks do not allow these shortages in the supply chain to occur. From these simulationresults, it can be deduced that the origin of these problems is in unit 4: the stock level of cases turns outnot to be sufficient.

The simulation model also allows the evolution of the set of orders waiting in each unit of the supply chainto be studied. Fig. 6 shows the orders waiting in the unit, not sent in production (their number is limitedbecause of the pull strategies used). Fig. 6 also demonstrates that orders are accumulated in units 1 and 2at the end of the studied period since these two units cannot process the orders they receive.

Finally let us consider the production rates. The charts given in Fig. 7 allow the possible production rate(i.e., what is theoretically possible to produce if every required component was available) and the effectiveproduction rate (i.e., what can be effectively achieved given the available components) to be compared.From these results, we can note that units 2 and 3 do not use the total of their production capacities:the resources available to produce are sufficient: there is no bottleneck. Unfortunately, in unit 5, we cansee that the 2 curves (possible and effective rates) are quite close, which means that the production facilitiesare saturated.

Fig. 5. Stock levels available along the supply chain.

Fig. 6. Orders waiting at the different stages.


Fig. 7. Comparison between possible production rates and actual production rates.



Therefore, as a conclusion, we can suggest that, in order to eliminate or reduce the shortages pointed outfrom these simulation results, it is necessary for the firm to check in priority whether the capacity of unit 5 canbe increased. This would contribute to help to absorb the orders waiting in this unit and to increase the stocklevels of the downstream units.

6. Conclusion

In this article, a continuous worldview has been suggested and used to model supply chains. In thisapproach, the production units are considered at a macroscopic level. The dynamic behavior families of prod-ucts flowing through logistic units, which are managed in just in time, is studied so as to better understand theglobal performance of the supply chain and how it reacts in particular circumstances. This approach has beensuccessfully applied to a network of factories located in different places in France. We have reported the sim-ulation results of a subset of this network, a supply chain that produces engines. The simulation results haveturned out to be very useful. In particular, they have pointed out that the global system can have an undesiredbehavior, depending on the initial level of inventories and the volumes of the orders, which results in deliveryshortages or in increased lead-times, due to insufficient available capacities.

This approach also allows the possibility to study the effects of variations in the demand, which are modeledthrough equations representing more or less sudden and more or less cyclic variations of customer orders[8,15].

System dynamic turns out to be relevant in contexts where the system is observed at a low level of details. Infact, as far as large and complex networks of production facilities are concerned, detailed modelingapproaches can be difficult to implement. The large amount of data necessary to describe the numerous prod-ucts and the processes can be extremely difficult to collect, and the effort required to develop detailed modelsof each production unit and of their interrelations can appear unrealistic in many cases. As a counterpart, onehas to note that, at this macroscopic level, the analyst expectations about the results are concerned with iden-tifying global trends of the system behavior, rather than computing estimated of particular performance mea-sures, as it is commonly done in discrete event simulation of manufacturing systems.

There are several research issues that can be addressed as future works. More sophisticated methodologiescan be used to study the sensitivity of the SC to disturbances and to variation of the demand. Chaotic behaviormight be identified in certain cases. Based on Forester’s theory and methodology the causal relationshipsbetween the variables can be studied more in depth, to identify and diagnose, from a more qualitative pointof view, typical cause-effect phenomena (analyzing loops, etc.). Let us finally note that the use of alternativemodeling approaches, such as bound graphs or continuous (or hybrid) Petri nets can be contemplated.

References

[1] G. Baills, J-C. Vessilier, Les applications industrielles de la dynamique des systemes chez Renault, Revue internationale de systemique5 (1991) 119–125.

[2] A. Banerjee, J. Burton, S. Banerjee, A simulation study of lateral shipments in single supplier, multiple buyers supply chain networks,International Journal of Production Economics 81–82 (2003) 103–114.

[3] M.W. Barnett, C.J. Miller, Analysis of the virtual enterprise using distributed supply chain modelling and simulation: an applicationof e-SCOR, in: J.A. Joines, R.R. Barton, K. Kang, P.A. Fishwick (Eds.), Proceedings of the 2000 Winter Simulation Conference,Wyndham Palace Resort and Spa, Orlando, USA, 2000, December.

[4] P. Benjamin, M. Graul, R. Mayer, M. Painter, C. Marshall, Supply chain design toolkit (SCPDT). in: B.A. Peters, J.S. Smith, D.J.Medeiros, M.W. Rohrer (Eds.), Proceedings of the 2001 Winter Simulation Conference, 2001, pp. 1146–1154.

[5] S. Bhaskaran, Simulation analysis of a manufacturing supply chain, Decision Sciences 29 (3) (1998) 633–657.[6] S. Biswas, Y. Narahari, Object oriented modelling and decision support for supply chains, European Journal of Operational Research

153 (2004) 704–726.[7] R. Bruniaux, H. Pierreval, C. Caux, Simulating supply chains with system dynamics, in: Proceedings of the ESS’99 European

Simulation Symposium, Nuremberg, Germany, October 26–28, 1999, pp. 209–213.[8] R. Bruniaux, Simulation continue de reseaux de sites industriels. Application aux chaınes logistiques dans le secteur Automobile.

Ph.D. Dissertation, in French, Universite Blaise Pascal, Clermont-Ferrand, France, 2000.[9] R. Bruniaux, C. Caux, H. Pierreval, An aggregation procedure for the continuous simulation of production units organised in

network. in: Proceedings of the IFAC Conference Management and Control of Production and Logistics, July 5–8, Grenoble, France,2000.


[10] R. Bruniaux, H. Pierreval, C. Caux, Simulation par la dynamique des systemes de chaines logistiques : etude d’un cas dans l’industrieautomobile, in: Proceedings of the 4th International Conference in Industrial Engineering, Aix en Provence, France, June 12–15,2001.

[11] W. Changchien, H.Y. Shen, Supply chain reengineering using a core process analysis matrix and object-oriented simulation,European Journal of Operational Research 153 (2002) 704–726.

[12] D.C. Chatfield, T.P. Harrison, J.C. Hayya, SISCO: a supply chain simulation tool utilizing SILK and XML, in: B.A. Peters, J.S.Smith, D.J. Medeiros, M.W. Rohrer (Eds.), Proceedings of the 2001 Winter Simulation Conference, 2001, pp. 614–622.

[13] L. Degres, H. Pierreval, C. Caux, Simulation of Steel Industry using System Dynamics. in: Proceedings of International Federation ofAutomatic Control 11th IFAC Symposium on Information Control Problems in Manufacturing (INCOM’04), April 5–7, Salvador deBahia (Brasil), 6 pages , CD-ROM, 2004a.

[14] L. Degres, C. Caux, H. Pierreval, ‘‘Agregation de produits metallurgiques pour la simulation en dynamique de Forrester’’. in: Proc.5th Conference Francophone de Modelisation et Simulation (MOSIM’04), September 1–3, Nantes (France), 2004b, pp. 89–96.

[15] L. Degres, H. Pierreval, C. Caux, Analysis of the effects of the variation in the demand of products in a Steel Industry: a SystemDynamics approach, International Conference on Industrial Engineering and Systems Management IESM 2005, May 16–19,Marrakech, Maroc, 2005, pp. 894–903.

[16] T. De Marco, Structured Analysis and System Specification, Prentice-Hall, New Jersey, 1978.[17] J. Edghill, D. Towill, The use of system dynamics in manufacturing system engineering, Transactions of the Institute of Measurement

and Control 11 (4) (1989) 208–216.[18] J.W. Forrester, Industrial Dynamics, Productivity Press, 1961.[19] K. Hafeez, M. Griffiths, J. Griffiths, M.M. Naim, Systems design of a two-echelon steel industry supply chain, International Journal

of Production Economics 45 (1-3) (1996) 121–130.[20] T. Higuchi, M.D. Troutt, Dynamic simulation of the supply chain for a short life cycle product—Lessons from the Tamagotchi case,

Computers and Operations Research 31 (7) (2004) 1097–1114.[21] M. Holweg, J. Bicheno, Supply chain simulation-a tool for education, enhancement and endeavour, International Journal of

Production Economics 78 (2002) 163–175.[22] Jain, Collins, Workman, Erwin, Lathrop, Development of a high level supply chain simulation model, in: B.A. Peters, J.S. Smith, D.J.

Medeiros, M.W. Rohrer (Eds.), Proceedings of the 2001 Winter Simulation Conference, 2001, pp. 1129–1137.[23] J-P. Kieffer, J.A. Arekion, K. Ayadi, D. Baudoin Simulation de processus de cooperation entre les systemes d’ordonnancement

dynamique de deux ateliers interdependants. Actes de la journee d’etude sur l’utilisation de la simulation en gestion de production,Clermont-Ferrand, France, June, 1995, pp. 22–26.

[24] U.H. Konig, Simulating Multidimensional Supply Chains – A Vensim based Model. in: Proceedings of 15th International SystemDynamics Conference, Istanbul, Turkey, August, 1997, pp. 447–450.

[25] P. Le Page, Analyse des relations clients-fournisseurs par une approche multi-agent. Ph.D. dissertation, ENSAE, Toulouse, France,December, 1993.

[26] Y.H. Lee, M.K. Cho, S.J. Kim, Y.B. Kim, Supply chain simulation with discrete continuous combined modeling, Computers andIndustrial Engineering 43 (1–2) (2002) 375–392.

[27] J. Melese, L’Analyse Modulaire des Systemes: AMS. Edition Hommes et Techniques, Paris, France, third ed., 1982.[28] Y. Merkuryev, J. Petuhova, M. Buikis, Simulation-based analysis of the Bullwhip effect insupply chains, in: Abstract in the

Proceedings of the Joint International Meeting ‘‘New opportunities for operations research‘‘, Istanbul, Turkey, July, 2003.[29] S. Minegishi, performance et fiabilite dans les delais de livraison de produits industriels assembles: application a l’industrie

automobile, Ph.D. dissertation, in French, Ecole Centrale de Paris, 2003.[30] S. Minegishi, D. Thiel, System dynamics modeling and simulation of a particular food supply chain, Simulation Practice and Theory 8

(5) (2000) 321–339.[31] M. Movahedkhah, Proposition d’un modele generique de dynamique des systemes pour la simulation de l’evolution des indicateurs de

performance des industries alimentaires, Ph.D. dissertation, in French, Ecole Centrale de Paris, 2002.[32] J.L. Paris, L. Tautou-Guillaume, H. Pierreval, Dealing with design options in the optimization of manufacturing systems: an

evolutionary approach, International Journal of Production Research 14 (5) (2001) 1081–1094.[33] D.W. Pearson, P. Albert, B. Besombes, M.-R. Boudarel, E. Marcon, G. Mnemoi, Modelling enterprise networks: a master equation

approach, European Journal of Operational Research 138 (3) (2002) 663–670.[34] F. Persson, J. Olhager, Performance simulation of supply chain designs, International Journal of Production Economics 77 (2002)

231–245.[35] D. Petrovic, Simulation study of supply chains behaviour and performance in an uncertain environment, International Journal of

Production Economics 71 (2001) 429–438.[36] D. Petrovic, R. Roy, R. Petrovic, Modelling and simulation of a supply chain in an uncertain environment, European Journal of

Operational Research 109 (1998) 299–309.[37] R.A. Phelps, D.J. Parsons, A.J. Siprelle, The SDI industry product suite: simulation from the production line to the supply chain, in:

J.A. Joines, R.R. Barton, K. Kang, P.A. Fishwick (Eds.), Proceedings of the 2000 Winter Simulation Conference, 2000, pp. 208–214.[38] L. Rabelo, H. Magdy, C. Lertpattarapong, Analysis of supply chains using system dynamics, neural networks and eigenvalues, in:

J.A. Joines, R.R. Barton, K. Kang, P.A. Fishwick (Eds.), Proceedings of the 2004 Winter Simulation Conference, December,Washington, DC, USA, 2004.

[39] J-L. Rascle, CAO : une aide graphique a la modelisation des systemes dynamiques3rd Cycle doctoral dissertation, in French,University Claude Bernard, Lyon, France, 1980, October.


[40] K. Rota, C. Thierry, G. Bel, Supply chain management: a supplier perspective, Production Planning and Control 13 (4) (2002) 370–380.

[41] K. Rota, Coordination temporelle de centres gerant de facon autonome des ressources. Application aux chaınes logistiques integreesen Aeronautique. PhD. dissertation, in French, ENSAE, Toulouse, France, 1998.

[42] D. Shunk, B. Plott, Using simulation to analyze supply chains, in: Proceedings of the 2000 winter simulation conference, 2000, pp.1095–1100.

[43] O. Telle, T. Pistre, C. Thierry, G. Bel, Simulation d’une relation Client / Fournisseur au sein d’une chaıne logistique integree : Mise enœuvre industrielle ’’, in: Proceedings of 4th French speaking conference on modeling and simulation ‘‘Conception, Analyse et Gestiondes Systemes Industriels, MOSIM’03, Toulouse, France, April, 2003.

[44] S. Terzi, S. Cavalieri, Simulation in the supply chain context: a survey, Computers in Industry 53 (2004) 3–16.[45] D. Thiel, Analysis of the behaviour of production systems using continuous simulation, International Journal of Production Research

34 (11) (1996) 3227–3251.[46] D. Thiel, La dynamique des systemes, complexite et chaos, Hermes Science publications, Paris, 1998.[47] C. Thierry, Gestion de chaınes logistiques : Modeles et mise en oeuvre pour l’aide a la decision a moyen terme, Dissertation for the

‘‘Habilitation a diriger des recherches ’’, in French, Toulouse, France, 25 juin 2003.[48] B.P. Zeigler, D. Kim, S.J. Buckley, Distributed supply chain simulation in a DEVS/CORBA execution environment, in: Proceedings

of the 1999 Winter Simulation Conference, Phoenix, Arizona, USA, 1999, pp. 333–1340.