Production System Analysis: a Simulation Based Approach

FACULDADE DE ENGENHARIA DA UNIVERSIDADE DO PORTO Production System Analysis: a Simulation Based Approach Edgar Filipe da Cruz Bento DISSERTATION FINAL VERSION MESTRADO INTEGRADO EM ENGENHARIA ELECTROTÉCNICA E COMPUTADORES Orientadores: Jorge Pinho de Sousa e Samuel Moniz July 2016 © Edgar Filipe da Cruz Bento, 2016 i Abstract With the rise of company’s competiveness and a growing global market, the necessity of tools and strategies that can deliver higher product or service value are on demand [1]–[3]. In a time were there had been a converging levelling of production philosophies and tools, like Lean, the necessity for the next step is raising [4]–[6]. Many like Schroer et al. [7] believe that’s exactly where simulation can help. Because not all tools can be applied in the same standard way or neither can they all be of simple comprehension regarding the pursue of improvements, because of the need for decision and strategies support in opposition to conjectures or guesses, it is why manufacturing systems simulation and modelling could and should be the next evolutionary step. In that matter it must be acknowledge that some companies already use these tools with great sophistication and knowledge like General Electric, Intel, AirBus Group [8], Price Water Coopers (PWC), Infineon [9], Port of Hamburg [10], Telefonica and Alcatel Lucent [11] but that is still short. Because of the low cost and de great advantages, improvements and cost reduction [12], in short term the majority of companies should converge to this reality. Some of the reasons for this delay have been pointed out by McLean et al. [13], Fowler et al. [14] e Abdulmalek et al. [15], being the major one the lack of commitment and confidence of managers with these tools, usually replaced with more complex and less effective ones. The main goal of this dissertation was to analyse and evaluate simulation software through implementation of production system tools aiming to improve the facility layout design. Followed by the creation of a tool that could determine optimal facility layout results for complex routings, specific input and output points, fixed and heterogeneous shapes with different department proportions. The objectives are the investigation of the benefits and difficulties of using these simulation tools as well as the improvements that could be implemented in a scenario like this. For this purpose, it will be used a case study of a production system modeled with discrete events through AnyLogic simulation software. iii Acknowledgments I would like to thank my supervisors, Professor Jorge Pinho de Sousa and Samuel Moniz for the support and guidance during this dissertation. Secondly, to my girlfriend Filipa Ferreira for the unconditional help and love specially on the difficult times. To my father, for the continuous support all these years even when it was hard, my mother for the positive attitude and belief in me and to my sister for the friendship, this achievement is also yours. To the Real República dos LYS.O.S., my second home and to all the brothers I have made in the last years of the course. I’m grateful for the all the professors and colleagues that shared knowledge and friendship aiding to the development of what I’m today. And finally to Sociedade de Debates da Universidade do Porto for helping me develop my communication skills. Edgar Bento iv “One of the greatest discoveries a person makes, one of their great surprises, is to find they can do what they were afraid they couldn't do.” - Henry Ford v Contents Abstract .................................................................................................................. i Acknowledgments ................................................................................................ iii Contents ................................................................................................................. v List of figures ....................................................................................................... ix List of tables ......................................................................................................... xi Abbreviations ..................................................................................................... xiii Chapter 1 ................................................................................................................ 1 Introduction ....................................................................................................... 1 1.1 Keywords .............................................................................................. 1 1.2 Contextualization and Problem ........................................................... 1 1.3 Motivation and Goals .......................................................................... 2 1.4 Methodology........................................................................................ 3 1.5 Planning .............................................................................................. 4 1.6 Document Structure ............................................................................ 5 Chapter 2 ...............................................................................................................7 Literature review ................................................................................................7 2.1 Inventory ..............................................................................................7 2.2 Buffer ................................................................................................... 8 2.3 Push and Pull Production Systems ..................................................... 9 2.3.1 Push ............................................................................................................................. 9 2.3.2 Pull ............................................................................................................................... 9 2.3.3 Comparison between Pull and Push Production Systems ....................................... 10 2.3.4 Conclusions ............................................................................................................... 10 2.4 Facility Layout .................................................................................... 11 2.4.1 Basic Production Layout Formats ............................................................................. 11 2.4.2 Facility Layout Problem ............................................................................................ 13 2.4.3 Solution methodology ............................................................................................... 15 2.4.4 Computerized Relative Allocation of Facilities Technique - CRAFT ...................... 18 2.5 Simulation .......................................................................................... 19 2.5.1 Introduction .............................................................................................................. 19 2.5.2 Methodology for a Simulation Process.................................................................... 20 2.5.3 Steps in a Simulation Study ...................................................................................... 21 vi Introduction 2.5.4 Simulation Models Characteristics........................................................................... 23 2.5.5 Simulation Benefits and Disadvantages ................................................................... 23 2.5.6 When Simulation is not appropriate ........................................................................ 25 2.5.7 Areas of Application .................................................................................................. 26 2.5.8 Systems and System Environment Definition ......................................................... 27 2.5.9 Examples of Components of a System .....................................................................28 2.5.10 Simulation Modeling Approaches ..........................................................................28 2.6 Production System Tools and Methods Review using Discrete Event Simulation 31 Chapter 3..............................................................................................................37 Methodology ....................................................................................................37 3.1 Case Study ..........................................................................................37 3.1.1 Characterization of the system .................................................................................38 3.2 Facility Layout Design ....................................................................... 44 3.2.1 Facility Layout Approach .......................................................................................... 44 3.2.2 Layout Construction ................................................................................................. 44 3.2.3 Layout Evaluation ..................................................................................................... 54 3.2.4 Layout Improvement ................................................................................................ 59 3.3 AnyLogic Simulation Software .......................................................... 63 3.3.1 User Interface ............................................................................................................ 64 3.3.2 Process Modelling Library Blocks ............................................................................ 69 3.4 Simulation Modeling ......................................................................... 70 3.4.1 System requirements ................................................................................................ 70 3.4.2 Model Creation .......................................................................................................... 70 3.4.3 Final Buffer Simulation Model ................................................................................. 77 Chapter 4 ............................................................................................................ 79 Results analysis ............................................................................................... 79 4.1 Facility Layout Results Analysis........................................................ 79 4.1.1 Iteration Results ........................................................................................................ 79 4.1.2 Final Analysis ............................................................................................................86 4.2 Buffer Sizing Simulation Analysis ..................................................... 89 4.2.1 Buffer Physical Space Determination ...................................................................... 90 4.3 Facility Layout and Buffer merged Analysis ..................................... 94 Chapter 5 ............................................................................................................. 97 Conclusions and Further Research ................................................................. 97 5.1 Conclusions ....................................................................................... 97 5.2 Further Research ............................................................................... 99 Appendixes .................................................................................................... 100 References ......................................................................................................103 1.1 - Keywords vii Additional Bibliography ................................................................................. 107 xiv Introduction 1 Chapter 1 Introduction This dissertation was developed under the Integrated Master in Electrical and Computer Engineering, Faculty of Engineering of Porto (FEUP). The present chapter introduces the project, its contextualization regarding the subject, problem and research question, the objectives, the adopted methodology and planning and finally the document structure. 1.1 Keywords Manufacturing Systems Design, Facility Layout Problem, CRAFT, Buffer Sizing, Manufacturing System Simulation, Discrete Event Simulation. 1.2 Contextualization and Problem At the present time and due to a competition growth associated with a global market it is important to guarantee an efficient, effective and fast production system management given the rapid changes in the world’s context and scenarios. Earlier, the evolution jump was attached to the implementation of a new management philosophy. “Lean Manufacturing” started in Japan with the Toyota Production System, looking to reduce waste and therefore increase productivity and gains. Nowadays, with the methodology convergence the margin for error has decreased. Thus, it can be observed a growth in the need to test new hypothesis, diminish errors, anticipate scenarios, and to have the data and conclusions which support and justify the decision making and strategies [4]–[6]. In response to these felt needs, solutions have appeared, each time more sophisticated and complete than before [16]–[18]. In this context the simulation starts to win even more a bigger relevance. Today hold like one of the most powerful tools to be used in production system analysis. It can assess the impact on systems parameters variations and increase the chance for success of informed decisions based on multiscenarios [19]. Even though manufacturing simulation is believed to have great advantages its utilization is not growing accordingly. Some of the reasons for this delay have been pointed out by McLen et al. [13], Fowler et al. [14] e Abdulmalek et al. [15], being the major one the lack of commitment and confidence of managers with these tools, usually replaced with more complex and less effective ones. Regarding the facility layout problem, the optimal design of the physical layout is one of the most important issues to be considered in the early stages of the design of a manufacturing system. Tompkins et al. [20] estimated that 15± 70% of the total 2 Introduction operating expenses within manufacturing systems are attributed to material handling, and that these costs can be reduced by at least 10± 30% through a good layout planning. Furthermore, the system efficiency and work-in process inventory are also significantly affected by layout design. Therefor this dissertation tries to evaluate one of these tools regarding the implementation of several production system methodologies and addresses the development of a tool for layout optimization given the specific case study. Research questions: 1. What is the potential of simulation software in the manufacturing systems? 2. What kind of constrains and difficulties can be found to the implementations of these models? 3. How can simulation aid the improvement of the facility layout design? 1.3 Motivation and Goals The drive of this project lies in the possibility to develop a simulation of discrete events model that can help in a company’s production system analysis. Thus, this practical case study, through distinct scenarios creation can allow real improvements which could be an interesting feature for the verification of theoretical concepts interaction with the actual practical application. Making, this way, possible to evaluate and analyze changes of several production variables, look for improvements, understand certain events, predict future situations and ultimately help the decision making so that is sustained, informed and riskless. Furthermore, regarding the facility layout, Drira et al. [21] estimated that 20– 50% of the manufacturing costs are due to the handling of parts and then a good arrangement of handling devices might reduce them for 10–30%. Thus developing a tool that can aid the optimization process and incorporate the simulation results can deliver great outcomes that could point the right path to an increase in productivity and profit. 1.4 - Methodology 3 1.4 Methodology Figure 1 - Project Methodology The methodology for this project can be divided in four main parts. Firstly, the process started with the search, reading and gathering of literature that relates to this project and its goals. The major topics addressed were: Lean Manufacturing Systems, Discrete Event Simulation, Manufacturing System Simulation, and Facility Layout Design. This initial step is one of the more importance given it gives un understanding of all concepts, a background in history and evolution of them and defines a clear path of choices that can be taken. The second was the analysis production system tools and their capacity to be simulated with the project constrains. Given that some could be easily calculated with non-simulated methods, and for the purposes of the global improving it was concluded that the study of the facility layout design could be a great value added in this step. Thirdly, the simulation part, it was conduced a deep study and learning of AnyLogic software and JAVA language, the later was needed given that AnyLogic software uses JAVA as the lower level language to be embedded in the properties and functionalities, resulting in a better and closer to reality model. Following by the model implementation, test and record of results, being this process developed with PDCA methodology because of the continuous need for change and adaptation. Lastly the fourth part groups the results analysis and the conclusions of the project. Given that the part 2 and 3 were related, it was needed a new part regarding the analysis of the results as one to evaluate their relations and in that matter what changes could be triggered by the findings in the previous results. This process terminates with the conclusions of the all project. 1 - Literature Review 2 - Analysis of Simulation Capability of Production Sytem Tools 3 - Simulation Learning and Modeling 4 - Results Analysis and Conclusions 4 Introduction 1.5 Planning Next follows the description of the activities and provided planning Figure 2, given the expected duration of the dissertation.  Research and gathering of state of art towards discrete events simulation in production systems concepts and methodologies.  JAVA and AnyLogic software learning followed by the modelation of the production system  Results study and analysis.  Writing the Dissertation.  Preparation of the final presentation  Final Review of the Dissertation Figure 2 - Gantt Chart W3 W 4W5 W1 W2 W3 W 4W1 W2 W3 W 4W1 W2 W3 W 4W1 W2 W3 W 4 State of the art revision Modeling the production system methodologies Results analysis Initial discussion of the choices made Dissertation writing Preparation of the final presentation Final Review of the Dissertation July Tasks - Months March April May June 1.6 - Document Structure 5 1.6 Document Structure Besides this introduction, this dissertation is presented with 5 more chapters. The second chapter describes the state of art of the subjects being study and gives some historical view over the evolution of the lean methodologies in the manufacturing systems. Chapter 3, methodology, presents and discusses the methods undertaken to answer the research questions. The fourth chapter analyses the results given by the previous chapter and explains how can the practical approach and the simulation one can relate and complement each other to a greater solution in the end. The last chapter presents the overall conclusions of the work followed by the appendixes, with auxiliary information regarding code, the model, the excel evaluation tool and the case study, and with the references and the additional bibliography. Figure 3 - Document Structure Chapter 1 •Introduction Chapter 2 •Literature Review Chapter 3 •Methodology Chapter 4 •Results Analysis Chapter 5 •Conclusions 7 Chapter 2 Literature review This chapter presents the major theoretical concepts that support the following work. Firstly, a description of inventory and buffers is given, demonstrating the importance and necessity of each subject in production systems. Secondly it is shown different production strategies and the respective characteristics, when should be applied, benefits and disadvantages. Thirdly the facility layout literature review gives a deeper insight to the topic and resolution methods. Fourthly the Simulation literature is addressed presenting the major advantages and drawbacks of its use. Lastly it is presented the benefits of discrete event simulation regarding its use to model production system tools and methods. 2.1 Inventory Nicholas Chase et al. [22] presents the definitions, “Inventory is the stock of any item or resource used in an organization”, and “An inventory system is the set of policies and controls that monitor levels of inventory and determine what levels should be maintained, when stock should be replenished, and how large orders should be.” An inventory can be helpful to a company in various ways [22]:  To maintain independence of operations – where a line of production of a part doesn’t depend directly of the line of production of other parts and keep its own production flow.  To meet variation in product demand  To allow flexibility in production scheduling  To provide a safeguard for variation in raw material delivery time  To take advantage of economic purchase order size Any alterations made to the inventory size should at all times consider some costs [22]:  Holding costs  Setup costs  Ordering costs  Shortage costs 8 Literature review 2.2 Buffer The buffer storage emerged through the necessity to reduce machine unpredictability, variability of break downs, unbalance processing times and fluctuated production requirements. Therefor it serves to decouple machines and mitigate these variables. On the other side, the implementation of buffer storage has an impact on the performance characteristics such as productivity, flexibility, and space utilization so the buffer size calculation is an important aspect of a manufacturing system [23]. Figure 4 - Machine flow line with buffers [24] Figure 4 represents production line with buffer storage, where the squares, letters M, represent the machines and the circles, letters B, represent the buffers. 2.3 - Push and Pull Production Systems 9 2.3 Push and Pull Production Systems At this point it will be presented, discussed and compared two production strategies, Push and Pull. 2.3.1 Push Push production had its origin with mass production era. Zheng and Xiaochum [25] agree that this kind of production starts with the forethought, followed by fabrication process development and lastly the management system implementation and production control. Push production or make-to-stock (MTS), search for a given master production schedule (MPS) based on demand study and forecast, to push the merchandize to the market making it available for consumers. For that to happen, the same push process is made to push the supplies since the start, through several processes till the system end, then being storage in the warehouse and available to distributors. Zheng et al., Krishnamurthy et al. and Zhou et al. [25]–[27] acknowledge also that one of the major advantages of this system is the ability to increase output and equipment usage. The main disadvantage lies in the considerable increase of product inventory and risk of mistaken forecasts says Zheng. Figure 5 illustrates the push production where you can easily see that for a certain demand forecast the production is started, ending usually in the merchandize storage in warehouse which then go to distribution. Figure 5 - Push Type Production [25] 2.3.2 Pull In order to eliminate one of the wastes, Toyota created a system called Just-InTime, JIT, allowing inventory reduction considered a form of waste by his own definition, the muda. As Zhen and Xiaochun [25] explain, this waste reduction, caused by excessive production for instance, is made through JIT, using Kanban’s which together demand a pull production. A pull production or make-to-order (MTO) is based on consumer’s current demand. For this reason the production process begins with the income of consumer orders which make a pull, of the supply necessary to the processes within the production system. Hopp et al., Savsar et al., Spearman et al. e Deleersmyden et al. and Zheng et Xiaochun [25], [28]–[31] highlight several advantages, being the inventory reduction the major one, but on the other hand there is a possibility to increase the delivery’s delay. Figure 6 represents the functioning of pull production system. 16 Literature review Accordingly to our bibliographic review, we can say that despite all the continuous evolution in computers and computing ability, in which Moore’s Law [43] stipulates the doubling in circuit complexity every 18 months, the issue still lies. Since exact approaches are often found not to be suited for large size problems, numerous researchers have developed heuristics and meta-heuristics [21]. 2.4.3.2 Heuristics Heuristic algorithms can be classified as construction type algorithms [36]. Construction approaches build progressively the sequence of the facilities until the complete layout is obtained whereas improvement methods start from one initial solution and they try to improve the solution with producing new solution [21]. Construction based methods are considered to be the simplest and oldest heuristic approaches to solve the QAP from a conceptual and implementation point of view, but the quality of solutions produced by the construction method is generally not satisfactory. Improvements based methods start with a feasible solution and try to improve it by interchanges of single assignments. Improvement methods can easily be combined with construction methods.[36] CRAFT is a popular improvement algorithm that uses pairwise interchange [44] later on, this specific method will be addressed further. These heuristics are classified as adjacency and distance based algorithms.[36] The difference between these two algorithms lies in the objective function. The objective function for adjacency based algorithms is given as equation ( 1): Max ∑∑(𝑟𝑖𝑗)𝑥𝑖𝑗 𝑗𝑖 ( 1) Where xij is 1 if department ‘i’ is adjacent to department ‘j’ and else 0. The basic principle behind this objective function is that the material handling cost is significantly reduced if the two departments have adjacent boundaries. The objective function of distance based algorithms is given as equation( 2): Min(TC)=1 2 ∑∑ ∑ ∑𝐶𝑖𝑘 ∗𝐷𝑗𝑙 ∗𝑋𝑖𝑗 ∗𝑋𝑘𝑙 𝑛 𝑙=1 𝑛 𝑘=1 𝑛 𝑗=1 𝑗≠𝑙 𝑛 𝑖=1 𝑖≠𝑘 ( 2) The underlying philosophy behind this objective function is that the distance increases the total cost of traveling. Cik can be replaced by Fik depending on the objective. Equation ( 3) is used as an objective function when the facility layout is designed for multi-floor. min ∑∑∑ ∑(𝐶𝑖𝑘𝐻 ∗𝐷𝑗𝑙𝐻 ∗𝐶𝑖𝑘𝑉 ∗𝐷𝑗𝑙𝑉)∗ 𝑛 𝑙=1 𝑛 𝑘=1 𝑋𝑖𝑗 ∗𝑋𝑘𝑙 𝑛 𝑗=1 𝑗≠𝑙 𝑛 𝑖=1 𝑖≠𝑘 ( 3) 2.4 - Facility Layout 17 Where, CikH and DjlH stand for horizontal material handling cost and horizontal distance, respectively. The same meanings are applicable for CikV and DjlV but in vertical directions. 18 Literature review 2.4.3.3 Meta-heuristics Various meta-heuristics such as simulated annealing (SA), genetic algorithm (GA), and ant colony are currently used to approximate the solution of very large FLP. The SA technique originates from the theory of statistical mechanics and is based upon the analogy between the annealing of solids and solving optimization problems [36]. GA gained more attention during the last decade than any other evolutionary computation algorithms; it utilizes a binary coding of individuals as fixed-length strings over the alphabet {0,1}. GA iteratively search the global optimum, without exhausting the solution space, in a parallel process starting from a small set of feasible solutions (population) and generating the new solutions in some random fashion. Performance of GA is problem dependent because the parameter setting and representation scheme depends on the nature of the problem. Tabu search (TS) is an iterative procedure designed to solve optimization problems. The method is still actively researched, and is continuing to evolve and improve. Recently, a few papers have appeared where an ant colony algorithm has been attempted to solve large FLP [36]. Other approaches which are also currently applied to FLP are neural network, fuzzy logic and expert system. 2.4.4 Computerized Relative Allocation of Facilities Technique - CRAFT Computerized Relative Allocation of Facilities Technique CRAFT is the archetypal improvement-type approach and was developed by Armour and Buffa [44] in 1963. CRAFT begins by determining the centroid of each department in the initial layout. It then performs two-way or three-way exchanges of the centroids of non-fixed departments that are also equal in area or adjacent in the current layout. For each exchange, CRAFT will calculate an estimated reduction in cost and it chooses the exchange with the largest estimated reduction (steepest descent). It then exchanges the departments exactly and continues until there is not any estimated reduction due to twoway or three-way exchanges. Constraining the feasible department exchanges to those departments that are adjacent or equal in area is likely to affect the quality of the solution, but it is necessary due to its exchange procedure. [45] The objective of the algorithm is to minimize total cost (TC). The function is represented by the following equation ( 4) TC= ∑∑𝐷𝑖𝑗 ×𝑊𝑖𝑗 ×𝐶𝑖𝑗 𝑛 𝑗=1 𝑛 𝑖=1 ( 4) Dij is the distance from departments i to department j. Wij is the interdepartmental traffic from departments i to department j Cij is the handling cost between departments i and department j [46] 2.5 - Simulation 19 2.5 Simulation At this point is made a presentation of the main simulation aspects applied to production systems. Negahban et al. and Smith [18], [47] identify in their studies about design simulation literature and production systems operation that simulation has been having a fundamental part in analysis and optimization of industrial management area, such in design as in operation of production systems. They conclude as well that this is a growing reality due to the need to evaluate lean philosophy’s implementations and to preview future alterations, changes or new strategies, diminishing the risk, increasing scenario development and strengthening decisions. 2.5.1 Introduction The word simulation can be defined in several ways:  Ingals [48] introduces simulation as “a powerful tool if understood and used properly”.  Banks [49] says that “ simulation is the imitation of the operation of a realworld process or system over time. Simulation involves the generation of an artificial history of the system, and the observation of that artificial history to draw inferences concerning the operating characteristics of the real system that is represented.”  Shannon [50] defines simulation as “the process of designing a model of a real system and conducting experiments with this model for the purpose of understanding the behavior of the system and /or evaluating various strategies for the operation of the system”. As simulation is according to these authors the imitation or the system drawing process or industrial processes, in this case, using close to reality models, it matters also to define model and system.  Maria [51] defines model as “a representation of the construction and working of some system of interest. A model is similar to but simpler than the system it represents.  Shannon [50] claims that “by a model we mean a representation of a group of objects or ideas in some form other than that of the entity itself.” Regarding a system, the same author declares:  Shannon [50] that “by a system we mean a group or collection of interrelated elements that cooperate to accomplish some stated objective. 20 Literature review 2.5.2 Methodology for a Simulation Process The creative process of a simulation model varies with the need or application, but generally there are several similar steps. Fowler [14] suggests that a production system analysis using simulation involves the following process:  Model Design o Identify the issues to be addressed o Plan the project o Develop the conceptual model  Model Development o Choose a modeling approach o Build and test the model o Verify and validate the model  Model Deployment o Experiment with the model o Analyze the results o Implement the results for decision making Maria [52] identically identifies the phases of this kind of project and explains in what way the simulation can be used continuously promoting an also continuous improvement as shown in Figure 11. Figure 11 - Simulation Study Design [52] 2.5 - Simulation 21 2.5.3 Steps in a Simulation Study Figure 12 - Steps in a Discrete Simulation Study [53] 22 Literature review Table 3 - Steps in a Discrete Simulation – Definitions [53] Steps Definition Problem formulation  Requires the definition of the problem ensuring its clear understanding.  Sometimes the problem needs to be reformulated due to the course of the study. Objectives and Overall Project Plan  Indicates questions to be answered by simulation.  Revision of the methodology to apply.  Includes: different stages of the study, time required for each stage, cost of study, number of people needed, and expected results. Model Conceptualization  Starting point with a basic model and built upon it  Get the essence of the real system Data Collection  Important to collect from the beginning because it is time consuming  “constant interplay between construction of the model and data collection Model Translation  Transforming the real problem into computational form  Choice of language to program the model Verified?  Achieved naturally through common sense  Verify if everything is running properly Validated?  Accepted certainty level in which the model represents the real system  Same expected outcome than in real system  Minimizing the discrepancies between the model and the real system Experimental Design  Try and fail/succeed to consider all the alternatives  Reach the final model with the best choices Production Runs and Analysis  Estimate measures of performance for the simulated scenarios. More Runs?  Deciding if it is necessary to run more simulations Documentation and reporting  Program documentation – if a program has multiple users; easier to understand how it works; keeping track of modifications  Progress documentation – Chronology of the project; checking if whether the work is up to date or not  Reports – Insight of others on the progress; early catch of any issues or doubts and easy solutions Implementation  Depends on the quality of the previous steps;  If a previous step was neglected , some issues will surface during this step 2.5 - Simulation 23 2.5.4 Simulation Models Characteristics The simulation models can be characterized as statics or dynamic, deterministic or stochastic and continuous or discrete, according to several authors as explain Reeb e Leavengood [54]. A static model represents the system in a given moment whereas a dynamic one represents how the system evolves through time. The static model examples are, for instance casino games simulation such as roulette, cards, dice, etc. Here, the time factor is irrelevant cause doesn’t conditions in any way the simulation. Regarding to dynamic models, typically are all those who represent a process behavior through time, a boiler warm-up, the making of a given part, etc. are dynamic. In a deterministic model, doesn’t exist variation in the model parameters or in its variables, if it is fed the same values on its way in, it will always calculate the same exit. This way it can simulate the trajectory of a baseball ball including the laws of physics involved in the model. On the other hand, a stochastic model contains at least one random variable to describe the process within the system of study. This difference results that the exit results are mere estimates of the true model characteristic. These situations happen, for instance, in the randomness in which a customer arrives to a bank balcony among other similar. Regarding a continuous system its main characteristic is the status variables to vary continuously. The examples are simulation of vehicle movement, liquid flows, chemical reactions, electronic circuits and econometric models, etc. Lastly Reeb et al [54] explains that in a discrete system the variables change only in a given number of points in time. Examples include traffic control, distribution system and stock control, production lines simulation, production systems as a whole, etc. Analyzing the problem involved it can be envisaged that it will be a dynamic simulation model by discrete events, eventually stochastic due to error randomness, damages and other simulating factors. 2.5.5 Simulation Benefits and Disadvantages According to Shannon [50] these are the major simulation advantages:  We can test new designs, layouts, etc. without committing resources to their implementation.  It can be used to explore new staffing policies, operating procedures, decision rules, organizational structures, information flows, etc. without disrupting the ongoing operations.  Simulation allows us to identify bottlenecks in information, material and product flows and test options for increasing the flow rates.  It allows us to test hypothesis about how or why certain phenomena occur in the system  Simulation allows us to control time. Thus we can operate the system for several months or years of experience in a matter of seconds allowing us to quickly look at long time horizons or we can slow down phenomena for study. 24 Literature review  It allows us to gain insights into how a modeled system actually works and understanding of which variables are most important to performance.  Simulation's great strength is its ability to let us experiment with new and unfamiliar situations and to answer "what if" questions. The same author refers that, though the simulation has many advantages it also has some disadvantages. Mainly being:  Simulation modeling is an art that requires specialized training and therefore skill levels of practitioners vary widely.  The utility of the study depends upon the quality of the model and the skill of the modeler.  Gathering highly reliable input data can be time consuming and the resulting data is sometimes highly questionable. Simulation cannot compensate for inadequate data or poor management decisions.  Simulation models are input-output models, i.e. they yield the probable output of a system for a given input. They are therefore "run" rather than solved. They do not yield an optimal solution, rather they serve as a tool for analysis of the behavior of a system under conditions specified by the experimenter. In this point Maria [52] also presents the benefits and traps of these model simulations. Of note, the following traps especially:  Unclear objective  Using simulation when an analytic solution is appropriate  Invalid model  Simulation model too complex or too simple  Erroneous assumptions  Undocumented assumptions. This is extremely important and it is strongly suggested that assumptions made at each stage of the simulation modeling and analysis exercise be documented thoroughly.  Using the wrong input probability distribution  Replacing a distribution (stochastic) by its mean (deterministic).  Using the wrong performance measure  Bugs in the simulation program  Using standard statistical formulas that assume independence in simulation output analysis.  Initial bias in output data  Making one simulation run for a configuration  Poor schedule and budget planning  Poor communication among the personnel involved in the simulation study. 2.5 - Simulation 25 2.5.6 When Simulation is not appropriate Table 8 summarizes what Banks et al. [53] and Banks and Gibson [55] thought about when simulation would be a problem. For that they created a 10 rule approach to help the simulation model developers determine whether to use or not simulation in a specific problem. Table 4 - 10 Rules When Simulation is not Appropriate # Rules Description 1 When the problem can be solved using common sense 2 When the problem can be solved analytically 3 When it is easier to perform direct experiments 4 When the costs exceed the savings 5 When the resources are not available 6 When the time is not available 7 When there is not any data available 8 When it is not possible to verify or validate the simulation model 9 When the power of simulation is overestimated 10 When the system behavior is too complex or cannot be defined Adapted from [53], [55]. 32 Literature review KPI (Key Performance Indicator) Metric designed to register and encourage the progress of critical goals to the organization. Strongly promoted KPI’s can be behavior motors and so it is extremely important to choose the desired KPI’s. The effects can’t be simulated but results and information can be extracted to help this metric’s construction, whether in the identification of the best and most suitable indicators or in the study of those indicators. Muda It refers to all within the production process that doesn’t add value from the customer’s perspective. The main focus in a lean production system is cutting out the waste. Relatively to the Toyota Production System (TPS) there are 7 muda’s.  Transportation – Each time a product is moved it stands the risk of being damaged, lost, delayed, etc. as well as being a cost for no added value. Transportation does not make any transformation to the product that the consumer is willing to pay for.  Inventory – Inventory be it in the form of raw materials, work-in-progress (WIP), or finished goods, represents a capital outlay that has not yet produced an income either by the producer or for the consumer. Any of these three items not being actively processed to add value is waste.  Motion – Refers to the damage that the production process inflicts on the entity that creates the product either overtime (wear and tear for equipment and repetitive strain injuries for workers) or during discrete events (accidents that damage equipment and/or injure workers).  Waiting – Whenever goods are not in transport or being processed they are waiting. In traditional processes, a large part of an individual product’s life is spent waiting to be worked on.  Over-processing – Occurs any time more work is done on a piece other than what is required by the customer. This also includes using components that are more precise, complex, higher quality or expensive than absolutely required. (Traditional notion of waste, as exemplified by scrap that often results from poor product or process design).  Over-production – Occurs when more products are produced than is required at that time by your customers. One common practice that leads to this muda is the production of large batches, as often consumer needs change over the long times large batches require. Overproduction leads to excess inventory, which then requires the expenditure of resources on storage space and preservation, activities that do not benefit the customer.  Defects – Whenever defects occur, extra-costa are incurred reworking the part, rescheduling production, etc. This results in labor costs, more time in the WIP. Defects in practice can sometimes double the cost of one single product. This should not be passed on to the consumer and should be taken as a loss. Through simulation all the waste described can be verified and simulated. Even though it is important to consider that for the simulation and identification of all of these 2.6 - Production System Tools and Methods Review using Discrete Event Simulation 33 to happen, it would take a rather complex and realistic system, which held information about a great number of variables. Overall Equipment Effectiveness (OEE) It is a type of metric used to measure the productivity loss to a certain productive process. Three kinds of losses are evaluated, availability (down time), performance (slow cycles) and quality (rejected). Offers a base line/reference and the means to register the waste elimination progress, 100% means a perfect production; producing good parts, as fast as possible without any down time due to damages. It can be simulated through model. It can be a good evaluation measure for the various productive scenarios as shown by Gibbons [63]. Root Cause Analysis It is a problem resolution methodology that focuses on the problem source instead the quick fixing of symptoms. Helps insure that the problem is truly eliminated through corrections on the problem source. The simulation due to its ability to control time and isolate certain processes can be a help in the identification of the problems and its sources. Six Big Losses This refers to six categories of productivity loss which are felt almost universal in production systems:  Malfunction  Setup time  Small stops  Speed reduction  Rejection of the first pieces  Rejects due to production This permits identifying and attacking the most common causes of waste in production systems. These issues can be simulated in order to identify the consequences of these kinds of productivity losses looking for solutions and avoid their existence. Takt Time Guides the way in which the feedstock goes through processes. Takt means compass, rhythm. This tool application gives a simple, consistent and intuitive way of giving rhythm to the production. This technique can be simulated in order to get the application consequences or takt time changes [7]. 34 Literature review Value Stream Mapping A tool used to visually map the production flow and also the future procedure status in order to reach improvement opportunities. Exposes the waste present in current procedures and presents a guiding way to its improvement [64]. It can be reached through simulation as explained by Jarkko et al. and Abdulmalek et al. [5], [15]. Buffer Sizing Several studies have explored the buffer problem [23], [24], [65]–[67] regarding different methods and approaches including the analytic method, gradient search, experimental design, and heuristics. Furthermore the apparent computational difficulty has led to the majority of buffer sizing approaches to be heuristic-oriented [23], but more recent articles point the benefits of computational studies which allows the study of any production line configuration, which is hard (or sometimes even impossible) to analyse using theoretical approaches and allows the study of cases in which the random variables that govern the behaviour of the system are characterized by any general distribution. [24]. Lee et al. also concludes that the one computational approach can is efficient and flexible for determining buffer storage in both serial production lines and more complex manufacturing systems. [23]. Simulation of Facility Layout Problems Aleisa and Lin [68] raise the important question in their study “For effectiveness facilities planning: Layout optimization then simulation or vice-versa?” The resume can be found in the table Table 7 - Layout than Simulate or vice versa Table 7 - Layout than Simulate or vice versa [68] Paradigm Layout then simulate Simulate then layout Belief Simulation analysis is local, where layout optimization analysis is global Simulation prior layout study produces layouts that are efficient and realistic Benefits Time efficient Provides accurate estimate of flow for layout optimization simulation Application (Best for) • Improving existing layout • Resolving congestion and bottlenecks in layout • Only minor system’s process’ parameters need to be adjusted • Technology embraced requires special layout type and simulation for verification • Insignificant stochastic behaviour • Focus is on minimizing travelled distance • Creating a new layout for a system that exhibit significant: − stochastic behaviour/demand • and/or − complex interactions • Major operational policies/technologies are not predetermined or need to be justified prior layout optimization • Simulation is used to generate random flow to be fed for a layout routine • Solving flow congestions and bottlenecks have higher priority than reducing distances 2.6 - Production System Tools and Methods Review using Discrete Event Simulation 35 To conclude they believe that the choice of the approach depends on the objectives and the characteristics of the system. 37 Chapter 3 Methodology Previously it was defined the research questions for this project: 1. What is the potential of simulation software in the manufacturing systems? 2. What kind of constrains and difficulties can be found to the implementations of these models? 3. How can simulation aid the improvement of the facility layout design? Regarding these questions, the methodology chapter discusses the how, the why and the what. How was the research made to answer these questions? Why was it made with these approaches and not others? What was specifically developed to achieve those goals? Firstly, a brief description of the case study used in this project is presented. Secondly, and according to the previous analysis, given the constrains encountered and the rules that will be described it is developed a tool for facility layout design that can further be improved using additional information from simulation modelling focusing on question 4. Lastly an overview of the software is made followed by the simulation model construction that trough them deliver a pathway with results to answer the first and second research questions. 3.1 Case Study This case study has the purpose of serving as data input to the facility layout improving tool developed further and also to the simulation and modelling of the buffers and test of the limitations of a manufacturing system simulation, with a large array of products, materials with a high annual demand. This data was presented in a need to know bases because of the necessity to respect the confidential information of the original case study. In that matter some information regarding the type of industry, what type of products produced, initial layout and other information that could present a comparison bases and confirm some of the guesses and premises where not available. 38 Methodology 3.1.1 Characterization of the system The object of this case study was a company of a given sector in which:  The annual search of the manufactured products was of 357.010 units;  Range of available products was of 68 types The products are composed by:  An array of material, sometimes more than one kind of material  The products composition is available through the Table 29 Materials:  Each type of material follows a defined route within the workstations  The annual demand of materials was of 38.462.971  There are 256 types of materials in this company The simulation model needed to be able to handle more than 250 different types of materials and about 70 different types of products. Each one consists of a set of materials sometimes more than one amount of the same type of material. Product Demand Table 8 - Input Orders by ReleaseDate idDemand idComponent releaseDate Total Production 1 274 05-01-2015 00:00 1789 2 279 05-01-2015 00:00 1777 3 284 05-01-2015 00:00 1334 4 257 05-01-2015 00:00 1314 5 286 05-01-2015 00:00 1073 6 273 05-01-2015 00:00 992 7 259 06-01-2015 00:00 835 8 282 06-01-2015 00:00 820 9 298 06-01-2015 00:00 722 10 281 06-01-2015 00:00 658 --- --- --- --- 358 295 20-02-2015 00:00 94  “idDemand” – unique order identifying number  “idComponent” – refers to product type, this allows to build a population of products, then used to create materials (in push production)  “releaseDate” – day of the arrival of the order used as a trigger for each new arrival  “Total Production” – is the amount of products needed to fulfil the order 3.1 - Case Study 39 Table 9 - Total Products Ordered and Variety in a Year There are 68 types of product with an annual demand of 357.010. Table 10 - Materials Annual Demand idComponent length Rework Pcs/pal Scrap Yearly demand 1 1,8 0,01 476 0,028 132.893 2 1,8 0,01 476 0,028 132.893 3 2,1 0,01 1666 0,028 265.786 4 2,1 0,01 451 0,028 132.893 5 2,1 0,01 3608 0,028 199.339 6 1,8 0,01 276 0,028 132.893 7 1,5 0,01 238 0,028 43.174 8 1,5 0,01 238 0,028 23.650 9 1,8 0,01 238 0,028 52.622 10 2,1 0,01 102 0,028 66.446 --- --- --- --- --- --- --- 256 1,8 0,01 68 0,028 1.087 Table 11 - Total Materials Yearly Demand Total Material Variety Total Material Yearly Demand 256 38.462.971  “idComponent” – material type  Total Material variety of 256 different material types  Total material annual demand of 38.462.971 Total Product Variety Total Products Ordered in a Year 68 357.010 40 Methodology Bill of Materials Bill of materials(BOM) is a list of raw materials or unassembled parts and quantities that constitute each product. Table 12 - Bill of Materials idProduct idMaterial materialDescription BOM multiplier 232 69 Material_69 2 232 70 Material_70 2 233 13 Material_13 2 233 14 Material_14 2 --- --- --- --- 298 65 Material_65 2 298 236 Material_236 2 298 78 Material_78 1 298 43 Material_43 1 298 16 Material_16 1 298 75 Material_75 1 298 74 Material_74 1 298 54 Material_54 1 298 52 Material_52 1 299 77 Material_77 2 299 67 Material_67 2 299 50 Material_50 1 299 17 Material_17 1  “idProduct” – Product type  “idMaterial” – Material type  “BOM multiplier” – quantity of each material to compose a product Routings Table 13 - Routings idM ater ial Alt idInputM achine idMachin eString idMachine StringAlt Processin gTime Processing TimeAlt 1 0 M03R1 0,0023 1 0 M11R1 0,0176 1 0 M12R1 0 1 1 M14R2 M14R2 M13R1 0,0526 0,1754 1 0 M19R2 0 1 0 M21R1 0 3.1 - Case Study 41 idM ater ial Alt idInputM achine idMachin eString idMachine StringAlt Processin gTime Processing TimeAlt 1 0 Sink1 0 --- --- --- --- --- --- --- 255 1 M17R2 M17R2 M18R2 0,3922 0,3509 255 0 M21R1 0,0283 255 0 Sink1 0 256 1 M17R2 M17R2 M18R2 0,3922 0,3509 256 0 M21R1 0,0283 256 0 Sink1 0  “idMaterial” – represents the material type  “Alt” – Is a variable that states that there is a workstation alternative  “idInputMachine” – represents a string sequence of workstations  “idMachineStringAlt” – represents the workstation alternative sequence  “ProcessingTime” – The time that each workstation takes to process a specific material  “ProcessingTimeAlt” - The time that the alternative workstation takes to process a specific material Setup Time Table 14 - Setup Time idMachine Number MachineI d Resource Workstati on Name Machine setup times WorkCentr e capacity 1 A M01R1 Cutting 0,00 0,00 2 B M02R1 Coating 0,00 0,00 3 C M03R1 Sawing 1,00 1,00 4 D M04R1 Wrapping 15,00 1,00 5 E M05R1 Cross cutting 10,00 1,00 6 F M06R1 4 side 10,00 2,00 7 G M07R2 Drilling line 15,00 2,00 8 H M08R1 2 sides 30,00 1,00 9 I M09R1 Corner cutting 5,00 1,00 10 J M10R1 Profile wrapping 35,00 1,00 11 K M11R1 Wrapping line 30,00 1,00 12 L M12R1 Cutting machine 5,00 1,00 13 M M13R1 Edge banding 10,00 1,00 14 N M14R2 Edge banding 10,00 2,00 15 O M15R2 Hot dowling 15,00 2,00 16 P M16R2 Friulmac 15,00 2,00 17 Q M17R2 Frame assembly 10,00 2,00 48 Methodology Figure 14 - CIRCOS - Circular Flow Chart 2 In this figure, each letter represents a workstation as pointed above, and then the flow is represented from pair connections between them, being the width representative of the flow by qualitative and quantitative means. Data input used in this flow chart creator can be viewed on Appendixes Table 41 Some characteristics can be easily identified:  The greatest flow happens between work stations “CJ”, “CK”, “JL”, “SU”;  Almost 70% of the initial flow, from work station “C”, goes to “J” and “K”, respectively 40% and 30%;  All the flow that goes to “J” and “K” goes to “L” and represents almost 100% of the input materials of “L”;  Again work station “L” has the greatest amount of connections; 2 www.circos.ca 3.2 - Facility Layout Design 49 Figure 15 - Sankey Flow Chart 3 Data input used in this flow chart creator can be viewed on Appendixes Table 42. 3 Sankey Flow Chart - http://sankeymatic.com/ 51 Figure 15 - Sankey Flow Chart was another tool implemented to give a greater knowledge and visualization of the flow of materials trough the manufacturing floor. This chart has the advantage of easily show the flow between the different work stations and the weight of it. Some of the previous considerations could also be taken trough the analysis of this chart. Without the evaluation part of the project this chart can easily be the first layout to be feed to the next steps, the evaluation and improving part, giving a reasonable starting point for the iterations that follow. One of the drawbacks of this chart is that for purposes of visualization it forces flow crisscrosses so that it can create a more pleasant and curvy chart. That can be pointed out in connections “D” and “K” and “O” and “F” for example, were it could be switched to prevent the crisscrosses. 3.2.2.4 Creation of a virtual image representation of workstations with proportionality to each other in excel Given the relative measures and shapes of the workstations provided in Figure 16 it was need the creation of a virtual image representation of each one in excel. Figure 16 - Workstations relative measures To create an automatic evaluation tool it were considered several requirements that could affect how the excel representation was created. These requirements are: 1. Keep proportion and shapes as is. 2. Each workstation should be a multiple of equal and square individual excel cells 3. Should be represented entering and exiting points 4. Should have a maximum sum of with and length of 50 and 200 cells respectively because of future CRAFT developments. 52 Methodology To obtain the results based on the requirements it could just be made by multiplying a factor and rounding it to the closest integer. In this specific case that would work because of the relatively small initial measures. Even so, the calculation presented below works for all the cases, especially when the measures are greater than the expected excel representation. Thus the results were calculated through a multiplying factor by the relative width and length. The formulas behind this approach are as follows: 𝐴 = 𝑊 𝑥∗ 𝐿𝑦 ( 5)  A – Area  W – Width  L – Length 𝑊𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 = 𝑤𝑖𝑛𝑖𝑡𝑖𝑎𝑙 ∑𝐴𝑛 𝑖=1 𝑖=𝑛 ( 6)  Wrelative – Relative width  Winitial – Initial width  An – Area of the nth workstation Equation ( 6) is the same for relative length calculation with the respective changes. 𝑊𝑓𝑖𝑛𝑎𝑙 =𝑊𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 ∗𝛼 ( 7)  Wfinal – Final width round to the closest integer  Wrelative – Relative width  α – Multiplicative factor To achieve the fourth goal, and given there were no direct formulas discovered, a process of trial and error was made and the results are presented in Figure 17. 3.2 - Facility Layout Design 53 Figure 17 - Workstation resize and transformation Several trials were made, and several solutions discarded because of not fulfilling some or several requirements. Even so the last result highlighted in green was the one that grants the fulfilment of all of the goals previously set. Figure 18 – Excel workstations representation Length With Relative Proportion Y X Y X Y X Y X Y X Y X 1 M01R1 0,37 0,97 0,3589 0,004722 0,01238 0 1 1 2 1 2 1 3 A 2 M02R1 0,37 0,73 0,2701 0,004722 0,009317 0 1 1 1 1 2 1 2 B 3 M03R1 0,76 0,96 0,7296 0,0097 0,012252 1 1 1 2 2 2 2 3 C 4 M04R1 1,55 7,3 11,315 0,019782 0,093168 2 9 3 14 418 420 D 5 M05R1 1,55 7,3 11,315 0,019782 0,093168 2 9 3 14 418 420 E 6 M06R1 0,41 5,88 2,4108 0,005233 0,075045 1 8 1 11 114 117 F 7 M07R2 0,41 5,88 2,4108 0,005233 0,075045 1 8 1 11 114 117 G 8 M08R1 0,41 4,97 2,0377 0,005233 0,063431 1 6 1 10 112 114 H 9 M09R1 0,38 0,72 0,2736 0,00485 0,009189 0 1 1 1 1 2 1 2 I 10 M10R1 0,84 4,17 3,5028 0,010721 0,05322 1 5 2 8 2 10 212 J 11 M11R1 0,85 4,43 3,7655 0,010848 0,056539 1 6 2 8 2 11 212 K 12 M12R1 0,36 0,75 0,27 0,004595 0,009572 0 1 1 1 1 2 1 2 L 13 M13R1 0,36 0,75 0,27 0,004595 0,009572 0 1 1 1 1 2 1 2 M 14 M14R2 0,36 0,75 0,27 0,004595 0,009572 0 1 1 1 1 2 1 2 N 15 M15R2 0,87 1,19 1,0353 0,011104 0,015188 1 2 2 2 2 3 2 3 O 16 M16R2 0,36 0,75 0,27 0,004595 0,009572 0 1 1 1 1 2 1 2 P 17 M17R2 0,36 1,32 0,4752 0,004595 0,016847 0 2 1 3 1 3 1 4 Q 18 M18R2 0,36 1,32 0,4752 0,004595 0,016847 0 2 1 3 1 3 1 4 R 19 M19R2 0,36 0,75 0,27 0,004595 0,009572 0 1 1 1 1 2 1 2 S 20 M20R1 0,58 0,99 0,5742 0,007402 0,012635 1 1 1 2 1 2 2 3 T 21 M21R1 2,42 5,43 13,1406 0,030886 0,069301 3 7 5 10 613 715 U 22 M22R3 2,79 7,56 21,0924 0,035608 0,096486 4 10 514 718 821 V 23 M23R1 1,19 1,53 1,8207 0,015188 0,019527 2 2 2 3 3 4 3 4 W Total 18,27 66,4 78,3534 0,233174 0,847442 21 86 39 124 46 161 49 186 # Area Workstation Workstation Factor x220 Factor x190 Factor x150 Factor x100 A1 AA2 B1 B2 K1 K K K K K K K K K K K2 K K C1 CC2 CCC L1 L2 D D M1 M2 D1 D D D D D D D D D D D D D D D D D D D2 D D N1 N2 D D OOO2 OO1 E1 E E E E E E E E E E E E E E E E2 O1 O O2 F1 F F F F F F F F F F F F F F F F2 P1 P2 P1 P2 G1 G G G G G G G G G G G G G G G G2 Q1 Q Q Q2 H1 H H H H H H H H H H H H H2 R1 R R R2 I1 I2 S1 S2 J1 J J J J J J J J J J J2 T1 TT2 J J T T T U1 U U U U U U U U U U U U U U2 V1 V V V V V V V V V V V V V V V V V V V V2 W1 W W W2 54 Methodology Figure 18 shows the converted departments created with multiples of one single square cell, with proportion and shapes relative to each other. For better visualization purposes each department was made with a different colour and the respective letter addressed earlier. The method used for calculation of entering and exit points was by representing those points by numbers, “1” for entering and “2” for exit. It can also be pointed out how easily de departments can be rotated or inverted as shown by departments “O” and “P” double representation on the figure. 3.2.3 Layout Evaluation Concerning the layout design several analytical approaches were studied, as previous mentioned before in the literature review, like the graph-based method and the pairwise exchange. But these methods have their own throwbacks especially when the problem is constituted by a great amount and very heterogeneous, regarding shapes and proportions, workstations and also, when specific entering and exiting points are defined. Thus, it was needed an approach that could tackle these downsides. Important questions to create an evaluation tool for the layout design problem:  What to evaluate;  How to evaluate. 3.2.3.1 What to evaluate The heuristic approach usually takes one of these two methods, the distancebased scoring and the adjacency-based scoring. They were described before, but for argument purposes they are summed up here again. Adjacency-based scoring objective is to maximize the sum of all weights, previously given by the relationships between the pair departments. Distance based objective is to minimize the total cost of transporting materials among all departments in a facility, normally based on rectilinear distance from centroid to centroid. Because of the data that was available, that could easily deliver the material quantity flow between departments, the problem in hands relates do the later approach, the distance-based one. Distance-based scoring like pairwise exchange and CRAFT methods use the following equation for evaluation of the layouts. minTC= ∑∑𝐷𝑖𝑗 ×𝑊𝑖𝑗 ×𝐶𝑖𝑗 𝑛 𝑗=1 𝑛 𝑖=1 ( 8) TC is the Total cost; Dij is the distance from departments i to department j; Wij is the interdepartmental traffic from departments i to department j; Cij is the handling cost between departments i and department j. 3.2 - Facility Layout Design 55 3.2.3.2 How to evaluate The first step of the evaluation process is to calculate how far the departments distance from each other. On that subject there are means of achieving this calculation depending on the type of layout expected and requirements. The rectilinear and Euclidian are two types of distance calculation that also relate to the point where the distance is measure being centroids the most used method. 3.2.3.2.1 Rectilinear Distance between i and j: 𝐷= |𝑥𝑖+𝑥𝑗|+|𝑦𝑖+𝑦𝑗| ( 9) Figure 19 - Rectilinear distance This method presents a calculation that tries to approximate to the real route of the materials from one department to another. Distance between two facilities is measured along path that is orthogonal to each other 3.2.3.2.2 Euclidian Distance between i and j: 𝐷= √(𝑥𝑖−𝑥𝑗)2+(𝑦𝑖−𝑦𝑗)2 ( 10) Figure 20 - Euclidian distance Distance is measured along straight-line path between the two facilities. 56 Methodology 3.2.3.2.3 Centroid The centroid of a plane figure is the arithmetic mean position of all the points in the shape. Figure 21 - Centroid example 𝑥 = ∑𝑥𝑖∗𝐴𝑖 𝑛 𝑖=1 ∑𝐴𝑖 𝑛 𝑖=1 ( 11) Where 𝑥 is the abscissa of the centroid and using the same equation with the respective changes would have 𝑦 as the ordinate of the centroid, 𝐴𝑖 is the area and 𝑥𝑖 is the abscissa of the geometric decomposition of the figure. 3.2.3.2.4 Point of exit to entering point One of the requirements brought by the analysis of the facility shapes and measures was the necessity of the calculation of the distances regarding the point of entering and exit of each department so that the results would approximate the real case. Thus an evaluation method would have to accomplish that. This is one of the great differences from the methods available. Figure 22 - Point of exit to entering point example 3.2.3.3 Layout evaluation tool Given the department shapes created in the Figure 18 – Excel workstations representation, the equation and objective method of distance-based methods, the rectilinear distance calculation between the point of exit and entering of workstations, a excel based evaluation tool was created to aid the improvement process and account for all of the formulas and considerations taken previously. The following topics sum up the final tool. 3.2 - Facility Layout Design 57 Figure 23 - Part of 2D map for layout calculation Figure 23 presents part of the two dimensions excel map created for deployment of the layout figures calculated before. So each department is placed on the 2D map and then the distance is calculated using rectilinear equation and accounting the entering and exit points in each figure, represented in this case by the G1 and G2 letters respectively. Table 20 – Part of Interdepartmental Flow Matrix Table 13 represents part of the global interdepartmental flow matrix which holds the total material movements between departments. Some departments as shown above with zero movements do not have material transit trough them respectively. Table 21 - Part of Distance Calculation Distance CD 6,0 CJ 2,0 CK 5,0 CL 15,0 CP 25,0 CS 23,0 DE 2,0 DF 20,0 DH 17,0 EF 2,0 EH 5,0 FI 4,0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 1 2 3 4 5 6 7 8 9 10 G1 G G G G G G G G G G G G G G G G2 11 12 A B C D E F G H I J K L A 0 0 0 0 0 0 0 0 0 0 0 0 B 0 0 0 0 0 0 0 0 0 0 0 0 C 0 0 0 8641488 0000015817669 9736345 306144 D 0 0 0 0 8488416 153072 0 0 0 0 0 0 E 0 0 0 0 0 6553643 01934773 0 0 0 0 F 0 0 0 0 0 0 0 0 445034 0 0 0 G 0 0 0 0 0 0 0 0 0 0 0 0 H 0 0 0 0 0 0 0 0 365420 0 0 0 I 0 0 0 0 0 0 0 0 0 0 0 0 J 0 0 0 0 0 0 0 0 0 0 0 15817669 K 0 0 0 0 0 0 0 0 0 0 0 9736345 L 0 0 0 0 0 1662777 01473529 0 0 0 0 64 Methodology In this software model is graphically specified as a process flowchart whereas blocks represent operations. The flowchart typically starts with a “source” type of block that generates the agents and inserts them in the process and ends with a “sink” type of block that removes them. The service time and the agents arrival is typically stochastic, and because they are generated from a probability distribution, the discrete event models are themselves also stochastic. In practical terms, results from the need of the model to run certain time or complete a set of replicates in order to produce significant results. The author ends referring that typically results of a simulation model by discrete events can have as exits the resource utilization rate, time spent within the system or part by an agent, waiting times, queue size, system throughput and also bottlenecks. 3.3.1 User Interface Figure 26 - AnyLogic User Interface At the very top of the window the menu is located, under the menu - the toolbar providing the easy access to the most frequently used commands. At the bottom you can see the status bar. By default the following components are shown in the workspace:  Graphical editor - The place to edit graphical diagrams of agents and experiments.  Projects view - Provides access to AnyLogic models currently opened in the workspace. The workspace tree provides easy navigation throughout the models.  Palette view - Provides the list of model elements grouped by categories in a number of stencils (palettes).  Properties view - Allows viewing and modifying the properties of currently selected model item(s).  Problems view - Displays errors found during model development and compilation. 3.3 - AnyLogic Simulation Software 65 3.3.1.1 Pallete The Palette view provides the list of graphical model elements grouped by categories in a number of stencils (palettes) and it is the place to find any AnyLogic graphical element to be add onto a graphical diagram of some agent class or experiment. Figure 27 - AnyLogic Pallete The Palette view consists of a number of stencils:  Agent - The stencil contains elements for defining dynamics of the model, its structure and data.  Presentation - The stencil contains shapes (line, oval, rectangle, polyline, curve, etc.), that you can use to draw presentations and 3D animations of your models and also a set of elements (3D window, camera, light) required to construct 3D animation scene.  System Dynamics - The stencil contains elements frequently used by System Dynamics modellers.  State-chart - The stencil contains elements of state-charts.  Action-chart - The stencil contains blocks of action-charts - structured block charts allowing defining algorithms graphically.  Analysis - The stencil contains elements, used for collecting, viewing and analysing output data.  Controls - The stencil contains controls (button, slider, checkbox, etc.) providing ability for creating interactive active object presentations.  Connectivity - The stencil contains tools for database connectivity.  Pictures - The stencil contains a set of pictures of frequently modelled objects.  3D Objects - The stencil contains a set of 3D images of frequently modelled objects. 66 Methodology 3.3.1.2 Properties The Properties view is used to view and modify the properties of a currently selected model item(s). When something is selected the Properties view displays the properties of the selection. Figure 28 - Item Properties The Properties view contains several sections. Every section contains controls such as edit boxes, check boxes, buttons, etc., used to view and modify properties. The number of pages and their appearance depend on the type of a selected object. 3.3.1.3 Problems AnyLogic supports on-the-fly checking of types, parameters, and diagram syntax. AnyLogic may automatically detect some problems or errors as the model is being developed. The errors found during code generation and/or compilation are displayed in AnyLogic Problems view. For each error, the Problems view displays description and location. 3.3 - AnyLogic Simulation Software 67 Depending on the error, opening it may result in displaying different views. If, for example, it is a graphical error, the corresponding diagram is opened in the graphical editor with invalid shapes highlighted. The Problems view displays information about problems of two types: errors and warnings.  Error - a critical problem that makes the model non-working and should be necessarily fixed.  Warning - information about some non-critical issue that may potentially lead to some problems or just an advice how to optimize the implementation (e.g. information about use of deprecated function). Warnings do not to prevent you from running the model. 3.3.1.4 Agents According to Grigoryev in the book “AnyLogic 7 in Three Days” [69] modulation by discrete events requires a modulator who thinks in the modulating system as a process, an operation sequence made by agents. This modulation can include operations that include delays, services by several features, selection of process branches, divisions and many others. As long as the agents compete for limited resources and can suffer delays, the rows will make part of almost all discrete events models. The agents defined in this book were originally called transitions in General Purpose Simulation System (GPSS) or entities in other simulation software’s. They can represent clients, parts, products, computation transactions, vehicles, tasks, projects, ideas among others. While resources on the other hand represent staff, operators, workers, servers, CPU’s, computer memories, equipment and transport. This software model is graphically specified as a process flowchart whereas blocks represent operations. The flowchart typically starts with a “source” type of block that generates the agents and inserts them in the process and ends with a “sink” type of block that removes them. The service time and the agents’ arrival are typically stochastic, and because they are generated from a probability distribution, the discrete event models are themselves also stochastic. In practical terms, results from the need of the model to run certain time or complete a set of replicates in order to produce significant results. The author ends referring that typically results of a simulation model by discrete events can have as exits the resource utilization rate, time spent within the system or part by an agent, waiting times, queue size, system throughput and also bottlenecks. Within an agent it can define variables, events, state-charts, system dynamics stock and flow diagrams, you can also embed other agents, add process flowcharts and as many types in the model as there are different types of agents. Design of an agent typically starts with identifying its attributes, behavior and interface with the external world. In case of large number of agents with dynamic connections (such as social networks) agents can communicate by calling functions. The agent internal state and behavior can be implemented in a number of ways. The state of the agent can be represented by a number of variables, by the state-chart state, etc. The behavior can be so to say passive (e.g. there are agents that only react to message arrivals or to function calls and do not have their own timing), or active, when 68 Methodology internal dynamics (timeouts or system dynamics processes) of the agent causes it to act. In the latter case agents most probably would have event and/or state-chart objects inside. 3.3.1.5 Events Event is the simplest way to schedule some action in the model. Thus, events are commonly used to model delays and timeouts. There are three types of events: 1. Timeout triggered event. It is used when an action is schedule at some particular moment of time (or some particular date).The event occurs exactly in timeout time after it is started. Timeout triggered event has even more features: you can specify that it expires either once or cyclically, or is fully controlled by the user. 2. Condition triggered event is used to monitor a certain condition and execute an action when this condition becomes true. 3. Rate triggered event is used to model a stream of independent events (Poisson stream). It is frequently needed to model arrivals: e.g. customer arrivals in queuing systems, transaction arrivals in server-based network models, etc. 3.3.1.6 Variables Agent can contain variables. Variables are generally used to store the results of model simulation or to model some data units or object characteristics, changing over time. AnyLogic supports two types of variables – variables and collections. Collections are used for defining data objects that group multiple elements into a single unit. Variable is a simple variable of an arbitrary scalar type or Java class. It always has some value assigned. Java variables can be declared in the Additional class code field in the Advanced Java properties section of the agent type. Variables declared in the code can also be accessed within this object, but defining them visually using variables is much more efficient. Alike other simulation tools AnyLogic supports variables of primitive types: double, integer, Boolean, but only AnyLogic gives infinite possibilities in defining data units by supporting variables of any Java classes. 3.3 - AnyLogic Simulation Software 69 3.3.2 Process Modelling Library Blocks Agents contained in the Process Modelling Library are the building blocks that can be used to construct flowcharts. As usual, objects generate agents, control agent flow, process agents, work with resources, and transport agents. In this reference guide, they are described in the following categories: Table 27 - Process Modelling Library Blocks Library Blocks Description Source – Generates agents. Sink – Disposes incoming agents. . Delay – Delays agents by the specified delay time . Queue – Stores agents in the specified order SelectOutput – Forwards the agent to one of the output ports depending on the condition. SelectOutput5 – Routes the incoming agents to one of the five output ports depending on (probabilistic or deterministic) conditions. Hold – Blocks/unblocks the agent flow. . Assembler – Assembles a certain number of agents from several sources (5 or less) into a single agent Conveyor – Moves agents at a certain speed, preserving order and space between them. ResourcePool – Provides resource units that are seized and released by agents. Seize – Seizes the number of units of the specified resource required by the agent. Release – Releases resource units previously seized by the agent. . Service – Seizes resource units for the agent, delays it, and releases the seized units Enter – Inserts agents created elsewhere into the flowchart. . Exit – Accepts incoming agents TimeMeasureStart – TimeMeasureStart as well as TimeMeasureEnd compose a pair of objects measuring the time the agents spend between them, such as "time in system", "length of stay", etc. This object remembers the time when an agent goes through. TimeMeasureEnd – TimeMeasureEnd as well as TimeMeasureStart compose a pair of objects measuring the time the agents spend between them. For each incoming agent this object measures the time it spent since it has been through one of the corresponding TimeMeasureStart objects. 70 Methodology 3.4 Simulation Modeling This topic will explain the simulation model construction, the steps taken and the major reflections taken. Previously it were studied several production tools and methods regarding their capability to be simulated using discrete events, afterwards, and given these conclusions it was developed a facility layout method to aid the determination of good solutions. Because of the raised importance in this project of layout determination and the analysis obtained of the capability of buffer simulation and their benefits it was considered a major improvement if the simulation could aid to achieve a better result and more realistic layout. Recalling: “The determination of buffer size also has a bearing on the performance characteristics such as productivity, flexibility, and space utilization for a manufacturing system.” Thus it was believed that the simulation of the case study could point out the need for additional factory floor space to account the need for buffers, and that this assessment would affect the final layout solution. Therefore, the following topics will address the construction of a tool to calculate buffer size given a typical demand of products input. 3.4.1 System requirements The objective is to develop a simulation model capable of running the data input from a case study to obtain the expected size of the workstation buffers. Therefor providing further information relative to the space needed so that the layout initial solution can be improved in a greater realistic way. 3.4.1.1.1 Functional Requirements  Determine buffer size for each department 3.4.1.1.2 Non-Functional Requirements  Use AnyLogic Simulation Software  Run simulation in less than 5 minutes  Use case study data as input  Extract results 3.4.2 Model Creation Given the system requisites and objectives, this topic will address the major steps taken to create a simulation model that can respond to the needs. 3.4 - Simulation Modeling 71 3.4.2.1 Buffer Simulation Model Flowchart For better interpretation of the simulation process it was developed a flowchart. Figure 29 - Buffer Simulation Model Flowchart Figure 29 represents the simulation model logic behind the implementation. The simulation starts with the load of the information from the excel sheet, and stored in several variables and collections like the demand, the bill of materials, the routings, the processing times and other information. Then, given that the demand happens at a specific frequency, an event is triggered every day at that specific time, searching for the release date of the orders. Whenever an order has that specific release date, the materials are injected with the respective quantity and type given by the bill of materials from the relation with the expected final product that needs them. Every material follows the specific route of machine sequences until the end, and for every different type, the respective processing times are applied. The simulation ends when the period/time given for the simulation, in this case 1 year, finishes. 3.4.2.2 Data Input Table 21 shows the partial component demand. Where “idDemand” is the unique order identifying number; “idComponent” refers to product type; “releaseDate” refers to 72 Methodology the date arrival of the order; “Total Production” is the amount of products needed to fulfil the order. Other tables exist but were not used in this simulation. There are 68 types of product with an annual demand of 357.010. Table 28 - Partial Component Demand idDemand idComponent amount lotsize Priority releaseDate Total Production 1 274 1 1789 0 05-01-2015 00:00 1789 2 279 1 1777 0 05-01-2015 00:00 1777 3 284 1 1334 0 05-01-2015 00:00 1334 --- --- --- --- --- --- --- 358 295 1 94 0 20-02-2015 00:00 94 This model for the buffer sizing calculation implements a push production that pushes orders in the system that pushes products, which in the end push materials. Even though later it tried a pull system, this was the best way to create a discrete event that released a waterfall of functions and creations of agents/materials. Bill of materials is a list of raw materials or unassembled parts and quantities that constitute each product. Table 29 - Bill of Materials idProduct idMaterial materialDescription BOM multiplier 232 69 Material_69 2 232 70 Material_70 2 233 13 Material_13 2 233 14 Material_14 2 --- --- --- --- 299 77 Material_77 2 299 67 Material_67 2 299 50 Material_50 1 299 17 Material_17 1 In Table 22, “idProduct” stands for the product type, “idMaterial” for the material type and “BOM multiplier” for the quantity of each material to compose a product. When a product enters the model a function creates a population of materials based on the “productType”, bill of materials and quantity needed, parameters loaded from product variables, the BOM from BOM sheet and the routing from the materials sheet. This way we create a need for materials in the production line. 3.4 - Simulation Modeling 73 Figure 30 - AnyLogic Model Variables Routings represent several important factors that aid to give reality to the model, like the sequence of workstations, the setup time, the processing time and the alternatives. Table 30 – Partial Routing Sequence idMaterial Alt idInputMa chine idMachine String idMachine StringAlt Processing Time Processing TimeAlt 1 0 M03R1 0,0023 1 0 M11R1 0,0176 1 0 M12R1 0 1 1 M14R2 M14R2 M13R1 0,0526 0,1754 1 0 M19R2 0 1 0 M21R1 0 1 0 Sink1 0 --- --- --- --- --- --- --- 256 1 M17R2 M17R2 M18R2 0,3922 0,3509 256 0 M21R1 0,0283 256 0 Sink1 0 In Table 23, “idMaterial” represents the material type, “Alt”is a variable that states that there is a workstation alternative, “idInputMachine” represents a string sequence of workstations, “idMachineStringAlt” represents the workstation alternative sequence, “ProcessingTime” the time that each workstation takes to process a specific material and “ProcessingTimeAlt” the time that the alternative workstation takes to process a specific material. Given the need for easier and fast access, issue addressed afterwards, this table was used to construct two specific excel sheets, one for the routing Table 26, and another for the processing time matrix Table 24. 80 Figure 37 - First Iteration Figure 37 denotes the first iteration. For the layout construction it was added one cell space around each department representing for example material flow paths or the real scenario in a production floor. This layout was constructed has previous mentioned in Chapter 3 using the interdepartmental flow information. Table 34 - First Iteration Cost Matrix Table 27 shows the cost matrix first iteration. For analysis purposes it was applied a filter, the red values represent a greater cost, where a light yellow represents a minor cost. T1 TT2 K1 KKKKKKKKKKK2 O O O2 G1 GGGGGGGGGGGGGGGG2 TTT A1 AA2 K K O1 M1 M2 Q1 Q Q Q2 I1 I2 C1 CC2 J1 JJJJJJJJJJJ2 L1 L2 U1 UUUUUUUUUUUUUU2 V1 VVVVVVVVVVVVVVVVVVVV2 W1 WWW2 C C C J J N1 N2 H1 HHHHHHHHHHHHH2 P1 P2 D D R1 R R R2 S1 S2 B1 B2 D1 DDDDDDDDDDDDDDDDDDD2 D D E1 EEEEEEEEEEEEEEEE2 D D F1 FFFFFFFFFFFFFFFF2 Cos t Matrix A B C D E F G H I J K L M N O P Q R S T U A 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 B 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 C 0 0 0 34565952 0 0 0 0 0 15817669 29209035 3979872 0 0 0 6511640 0 0 1546524 0 0 D 0 0 0 0 8488416 3979872 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 E 0 0 0 0 0 281806649 083195239 0 0 0 0 0 0 0 0 0 0 0 0 0 F 0 0 0 0 0 0 0 0 2225170 0 0 0 0 0 0 0 0 0 0 0 15848916 G 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 H 0 0 0 0 0 0 0 0 1461680 0 0 0 0 0 0 0 0 0 0 0 9128646 I 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3241816 J 0 0 0 0 0 0 0 0 0 0 0 15817669 0 0 0 0 0 0 0 0 0 K 0 0 0 0 0 0 0 0 0 0 0 29209035 0 0 0 0 0 0 0 0 0 L 0 0 0 0 0 28267209 025049993 0 0 0 0 2284286 7007820 6467013 17416788 0 0 26020390 0 0 M 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4568572 0 0 N 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 35039100 0 0 O 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 19401039 P 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6456760 Q 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7282287 R 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4507672 S 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 13435437 T 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 U 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4.1 - Facility Layout Results Analysis 81 Table 35 - First Iteration Total Cost Table 28 represents the final result of equation ( 8) which will serve for comparison evaluation of the different iterations. The objective will be the reduction of the total cost. Table 36 - First Iteration Distance Calculation Table 29 denotes the rectilinear distance calculation between the exit point of one station and the entering point of another. Iteration Total Cost 11308188800 Distance CJ 2,0 JL 2,0 SU 6,0 CK 5,0 KL 5,0 CD 6,0 DE 3,0 FU 9,0 LN 3,0 NS 20,0 EF 18,0 LO 3,0 OU 27,0 LS 24,0 PU 4,0 HU 8,0 LP 22,0 QU 31,0 EH 22,0 LF 14,0 LH 6,0 LM 3,0 MS 22,0 RU 9,0 IU 4,0 FI 10,0 HI 5,0 CP 38,0 CL 15,0 DF 5,0 CS 40,0 DH 3,0 EH 22,0 MP 20,0 82 Results analysis The main objective of this method is to minimize the equation ( 8) which relates the distance, the interdepartmental traffic and the the handling cost of those movements. Given that there was no information regarding different costs in different movements it was considered that the cost was the same for all the movements. Thus only two factors of the equation remain the interdepartmental traffic and the distance. Because Table 19 is ordered from greater amount of traffic between two departments, the less amount of distance from the top results in a smaller total cost. So one of the best objectives is to get the top departments as close as possible they can be. Therefore, it can be observed that with the first iteration the top 3 connections, “CJ” and “JL” are as close they can be. Because of “CJ” and “JL” connections “CK”, “KL” and “CD” apparently cannot be any closer. But “DE”, “SU”, “NS” and “EF” can be closer by arranging the layout. As affirmed because another layout improving process with the results from buffer simulation is needed, this analysis is brief. Thus the table that follows illustrates 5 iterations. Table 37 - Layout Iterations Iteration 1 Distance Iteration 2 Distance Iteration 3 Distance Iteration 4 Distance Iteration 5 Distance CJ 2,0 CJ 2,0 CJ 2,0 CJ 2,0 CJ 2,0 JL 2,0 JL 2,0 JL 2,0 JL 2,0 JL 2,0 SU 6,0 SU 2,0 SU 2,0 SU 2,0 SU 2,0 CK 5,0 CK 5,0 CK 5,0 CK 5,0 CK 5,0 KL 5,0 KL 5,0 KL 5,0 KL 5,0 KL 5,0 CD 6,0 CD 12,0 CD 6,0 CD 6,0 CD 6,0 DE 3,0 DE 3,0 DE 2,0 DE 2,0 DE 2,0 FU 9,0 FU 7,0 FU 5,0 FU 5,0 FU 5,0 LN 3,0 LN 3,0 LN 2,0 LN 3,0 LN 4,0 NS 20,0 NS 3,0 NS 4,0 NS 3,0 NS 2,0 EF 18,0 EF 5,0 EF 2,0 EF 2,0 EF 2,0 LO 3,0 LO 2,0 LO 2,0 LO 2,0 LO 2,0 OU 27,0 OU 7,0 OU 5,0 OU 5,0 OU 5,0 LS 24,0 LS 3,0 LS 7,0 LS 7,0 LS 7,0 PU 4,0 PU 4,0 PU 4,0 PU 4,0 PU 4,0 HU 8,0 HU 8,0 HU 9,0 HU 9,0 HU 9,0 LP 22,0 LP 5,0 LP 9,0 LP 9,0 LP 9,0 QU 31,0 QU 12,0 QU 10,0 QU 10,0 QU 10,0 EH 22,0 EH 7,0 EH 5,0 EH 5,0 EH 5,0 LF 14,0 LF 19,0 LF 25,0 LF 25,0 LF 25,0 LH 6,0 LH 17,0 LH 26,0 LH 26,0 LH 26,0 LM 3,0 LM 8,0 LM 6,0 LM 6,0 LM 6,0 MS 22,0 MS 10,0 MS 4,0 MS 4,0 MS 4,0 RU 9,0 RU 5,0 RU 5,0 RU 5,0 RU 5,0 IU 4,0 IU 8,0 IU 10,0 IU 6,0 IU 6,0 FI 10,0 FI 6,0 FI 16,0 FI 4,0 FI 4,0 HI 5,0 HI 11,0 HI 20,0 HI 8,0 HI 8,0 CP 38,0 CP 21,0 CP 25,0 CP 25,0 CP 25,0 CL 15,0 CL 15,0 CL 15,0 CL 15,0 CL 15,0 DF 5,0 DF 24,0 DF 20,0 DF 20,0 DF 20,0 CS 40,0 CS 19,0 CS 23,0 CS 23,0 CS 23,0 DH 3,0 DH 26,0 DH 17,0 DH 17,0 DH 17,0 EH 22,0 EH 7,0 EH 5,0 EH 5,0 EH 5,0 MP 20,0 MP 12,0 MP 2,0 MP 2,0 MP 2,0 4.1 - Facility Layout Results Analysis 83 After identifying the pairs that highly affect the total cost portrayed in Table 31 several exchanges were made. From first iteration to the second “SU”, “FU”, “NS”, “EF”, “LO”, “OU, “LS”, “LP” and “QU” got a major improvement with the expense of “CD” positioning that got worst, even so resulting in a major reduction of the total cost. The pair “CD” is one with great flow weight, so from the second to the third iteration the focus it. Therefore improving “CD”, but also “DE”, “FU”, “LN”, “EF” and others. From the top “NS”, “OU” and “LS” got worst results, which, in the global run, resulted in reduction. The fourth iteration analysis focused in the middle to bottom pair connections, but for that to happen a swap was made to the pair departments “LN” and “NS”. Because these stations have the same flow, maintaining the distance proportion of both has no affect in the cost. This move allowed for deeper reduction of the stations “IU”, “FI” and “HI” drastically, resulting in cost reduction. The final iteration resulted in no reduction of the cost, so for the analysis purpose and given that there will be a future and deeper analysis, the iteration process improving was stopped. Table 38 - Iterations Total Cost Iteration Total Cost 5657346365 4657346365 3670313629 2727871931 11308188800 84 Results analysis Figure 38 - Layout Iteration 5 Figure 38 shows the final layout iteration in this part of the project. Figure 39 - U Shape Experiment R2 R R R1 Q1 Q Q Q2 G1 GGGGGGGGGGGGGGGG2 M1 M2 P1 P2 B1 B2 K1 K K K K K K K K K K K2 N1 N2 S1 S2 U1 UUUUUUUUUUUUUU2 V1 VVVVVVVVVVVVVVVVVVVV2 K K O O O2 C1 CC2 J1 J J J J J J J J J J J2 L1 L2 O1 C C C J J I1 I2 F2 FFFFFFFFFFFFFFFF1 D D T1 TT2 A1 AA2 D1 DDDDDDDDDDDDDDDDDDD2 E1 EEEEEEEEEEEEEEEE2 W1 W W W2 T T T D D D D H2 HHHHHHHHHHHHH1 I1 I2 F2 FFFFFFFFFFFFFFFF1 V1 V V V V V V V V V V V V V V V V V V V V1 U2 U U U U U U U U U U U U U U1 S2 S1 Q2 Q Q Q1 T1 TE2 T T W1 W W W2 E P2 P1 R2 R R R1 T2 T E E H2 H H H H H H H H H H H H H1 E E O2 O O E O1 E N2 M2 E K1 K K K K K K K K K K K2 N1 M1 E A1 AA2 K K E L2 E C1 CC2 J1 J J J J J J J J J J J2 L1 E C C C J J E E B1 B2 D D E D1 D D D D D D D D D D D D D D D D D D D2 E1 D D D D G1 GGGGGGGGGGGGGGGG2 4.1 - Facility Layout Results Analysis 85 Figure 40 - L Shape Experiment Typically heuristic methods start from an initial solution, therefor, as previously stated, the first solution as high implication of the future iterations especially on the design/shape. To avoid this drawback, it was created different kinds of initial solutions so that through experience learning some characterization could be taken. Thus Figure 39 and Figure 40, demonstrate a “U” and an “L” shape layout design with interesting results. D D D1 D D D D D D D D D D D D D D D D D D D2 E1 D D E A1 AA2 D D E E C1 CC2 J1 J J J J J J J J J J J2 L1 L2 F1 E C C C J J F E M1 N1 F E B1 B2 K1 K K K K K K K K K K K2 M2 N2 FH1 E K K F H E O1 O F H E G1 G G G G G G G G G G G G G G G G2 O F H E O2 F H E T1 T F H E T T F H E T2 TQ1 F H E R1 Q F H E R Q F H E2 I1 RQ2 F H I2 R2 F H S1 F H P1 P2 S2 F2 H2 U1 U U U U U U U U U U U U U U2 W1 WV1 W V W2 V V V V V V V V V V V V V V V V V V V2 86 Results analysis Figure 41 - Different Shapes Total Cost Figure 41 determines the total cost of the preceding layouts. 4.1.2 Final Analysis This topic will initially address the evolution process that was taken to get to the previous results, and then discuss the findings. Figure 42 - Layout Determination Process 4.1.2.1 Development of an Layout Evaluation Tool Firstly it all started with the analysis of the data from the case study, followed by the need to get an optimal facility layout design as a plus for the dissertation project. Therefor with the data provided it were sorted and constructed new tables and graphs to help this objective, like Table 12, Table 13, Figure 14 and Figure 15. The result was the Table 32 with a 2 dimension map for layout evaluation. This tool used rectilinear distance calculation. The objective is the same, the minimization of the sum of the movement cost plus the amount plus the distance given by equation ( 8). The approach for the layout construction was a trial and error using the distance reduction of the weightiest interdepartmental flow. Iteration Total Cost U Shape 901819777 L Shape 910912250 4.1 - Facility Layout Results Analysis 87 Table 39 - Initial Layout Evaluation Tool Several trails were made and the results were promising and the analysis can be viewed in Table 33. 4.1.2.2 CRAFT method for Layout Improvement The search for automatic tools that could confirm, calculate or improve the layout lead to the CRAFT method and the Facility Layout add-in developed by Paul A. Jensen 4 . Figure 43 - CRAFT Facility Layout Excel Add-in The excel add-in tries to find the layout of the departments within the facility that minimizes the total cost of material handling. It accepts as data: the list of departments, the physical sizes of departments, part flows between departments, material handling costs between departments and the size of a proposed facility. It is a powerful tool, but given the specifications of the case study layout it was difficult to implement and get the expected results. The major drawbacks were: difficult to handle a big number of departments with great differences in shapes and proportions, to lock the shapes, to get the flow from specific points of entering and exit and the size limitation of the layout. This lead to interesting results but not approximated to reality. Yet it was a great starting point to raise the awareness of the variables and calculations within a heuristic model like this one. 4 https://www.me.utexas.edu/~jensen/ORMM/omie/computation/unit/lay_add/lay_create.htm l Layout 1 2 3 4 5 6 7 8 9 10 1 A D C J P B 2 H E K L N Q T 3 I F M O S R G 4 5 6 7 8 9 10 88 Results analysis Table 40 - First Evaluation Tool and CRAFT Total Cost Analysis Table 33 presents the results from topic 4.1.2.1 and 4.1.2.2 were it can be examined different layout solutions from the evaluation tool and the CRAFT method. The optimal solution encountered was actually discovered by the both methods, the solution layout of 3 units of length by 7 of width, with a total cost of 118725145. The CRAFT tool delivered a better solution, the last experiment with total cost of 117931012, but unless the facility layout was in a 3 dimensions design, which in this case it was not considered as a possibility, the placement of the entering department in the middle provoked the disregard of this solution. 4.1.2.3 Development of a New Facility Layout Method The development of a new facility layout method that could overpass the drawbacks of the methods analysed was taken in consideration given the necessity to approximate the results to the real case scenario. Therefor several methods were studied as stated in topic 3.2.1, and the closest to the desired objectives was the pairwise method. This method and many similar to it as the objective function to minimize the sum of distances plus material flow plus the movement cost, equation ( 2). This method uses distance calculation from centroid and typically uses a gluttony method where only 1 pair is analysed. It is very simple if departments are of same size and shapes. The solution was to develop a method similar that could tackle these limitations. The objective function is similar, equation ( 8), which uses rectilinear distance calculation from point of exit to entering, allowing also the rotation and inversion of the departments. The final constrain was related to the shapes and sizes, so an operation was developed to relativize the proportions and with the help of an excel 2 dimensions map tackle the shapes, so the method uses a set of rules that with the visual awareness of the departments given to the modeler results in a realistic layout through a set of iterations. Experiment Layout Total Cos t 14,00 4X5 117931012 13,00 3X6 118725145 12,00 4X5 121681212 11,00 4X5 119186389 10,00 2X10 130752608 9,00 10X2 142740636 8,00 2X10 164523594 7,00 3X7 118725145 6,00 4X5 119049330 5,00 3X7 118725145 2,00 3X7 197084721 4,00 2X10 145508776 3,00 2X10 160813913 1,00 2X10 222563064 CRAFT Trial and Error 4.2 - Buffer Sizing Simulation Analysis 89 4.2 Buffer Sizing Simulation Analysis Buffer sizing simulation analysis presents the results gathered from the modelation of the case study. As previous affirmed, to achieve the goal, the buffer calculation, it was used a set of data inputs and AnyLogic function blocks complemented with java functions to get the expected behaviour and a case study yearly product demand for the experimental run purposes. After several implementations and corrections the final result was validated through a series of tests and trials. The tests made were, for example, printing to the console information relative to the agent’s parameters and AnyLogic function blocks expected outputs. The final set of results was validated with a series of repetitive runs/simulations where the outcome was exactly the same. Table 41 - Maximum Buffer Simulation Results Buffer Number Maximum Buffer buffer1 0 buffer2 0 buffer3 13594 buffer4 1 buffer5 1 buffer6 1 buffer7 0 buffer8 1 buffer9 217 buffer10 1 buffer11 1 buffer12 1 buffer13 1 buffer14 1 buffer15 1 buffer16 1 buffer17 1555 buffer18 777 buffer19 1 buffer20 0 buffer21 2 buffer22 0 buffer23 0 Because of the workstation/laptop lack of resources the simulation was implemented with a factor. This issue will be addressed further in the dissertation, but essentially one major constrain in the simulation was the quantity of agents that flowed in the model. That led to a great increase in simulation time, and often the breakdown of the run/simulation. Therefor a factor was applied to reduce the number of agents, but to maintain an approximation to the real scenario that factor was also applied to the 96 Results analysis Figure 46 - Final Layout with Buffer Input Figure 46 demonstrates the final optimal solution for the case study analysed with the addiction of the buffer analysis from the simulation approach. It is important to affirm, as stated by the literature review, that this kind of approaches lead to optimal solutions not the best solution. Nevertheless given the results it can be strongly affirmed that this layout is realistic and reduces the total cost of the material handling which was one of the objectives of the project. Q1 R1 Q R Q R Q R Q R Q R QR2 Q2 G1 GGGGGGGGGGGGGGGG2 I1 I I I I I I2 B1 B2 K1 KKKKKKKKKKK2 O O O2 S1 S2 U1 UUUUUUUUUUUUUU2 V1 VVVVVVVVVVVVVVVVVVVV2 K K O1 N1 N2 P2 C1 CCCCCCCCCCCCC2 J1 JJJJJJJJJJJ2 L1 L2 P1 C C C J J M1 M2 F2 FFFFFFFFFFFFFFFF1 D D T1 TT2 D1 DDDDDDDDDDDDDDDDDDD2 E1 EEEEEEEEEEEEEEEE2 W1 WWW2 T T T A1 AA2 D D D D H2 HHHHHHHHHHHHH1 97 Chapter 5 Conclusions and Further Research This last chapter presents an overview of all the work done for this dissertation, as well as recommendations and further research. 5.1 Conclusions Modelling and simulation of production systems have been growing in the global context, following the evolution of computers and the need for the industry to evolve and improve, striving to achieve productive leaps that differentiate them from competition. The main goal of this dissertation is to analyse and evaluate simulation software through implementation of production system tools aiming to improve the facility layout design. Below it will be address primarily the answers to the research questions. One of the objectives of this dissertation was the study of the simulation software AnyLogic and its application in the manufacturing systems, thus evaluating the potential and capabilities. It was identified that by supporting the most common simulation methods nowadays, namely System Dynamic, Process-Centric/Discrete Events and Agent based modulation, which can be used simultaneously and combined is a major advantage and delivers great flexibility. The new Process Modelling Library allows the simple and rapid implementation of manufacturing models because of the building blocks that can easily modulate the behavior of the system and thus easily simulate a great majority of problems. Whenever the building blocks fall short to approximate the model to reality, the usage of java as an additional programing language solves it. This capability permits the adjustment of the properties of the system and the respective blocks, being one of the most popular programing languages which simplify learning. Nevertheless, it was observed that even dough the learning curve of AnyLogic and Java is smooth and relative rapid when the modeler wants the implementation of higher complexity models with particular behaviors the programming itself raises complexity also. Overall the findings and the previous research lead to the previous conclusions but it is also important to state that given the great possibilities and variety of production system tools and methods this objective cannot be fully determined with just one case study implementation and few previous studies. 98 Conclusions and Further Research Even dough there are great advantages for using simulation, there are still constrains and difficulties that hurt the rapid spread of these tools. The simulation model is an approximation of reality and so the assumptions made during implementation should be taken very carefully in order not to produce armful errors that could compromise results. They are as accurate as the input data and very often the gathering of reliable data is difficult. The utility of the outcomes depends on the skill of the modeler and lastly the time and cost of using these tools are important. The implementation of a complex model requires a skilled modeler, probably a team with multifaceted capabilities to aid the process, the gather of accurate data, acquisition of specific software and hardware and usually training lessons for the maintenance of the model. Therefore highly costly and time consuming, but on the other hand the cost of not using simulation should also be taken in consideration, especially the great advantage to have it as a support for decision making diminishing the risk. Several studies have pointed out the tradeoff, and suggested a set of rules to evaluate when to use it, but that could not be enough because of the rapid change of the markets and respective competition. Therefore an analysis of short and long term benefits and disadvantages should also be made to give further information to the decision. Several simulation approaches have appeared that address the facility layout problem but still with situations like this case study complexity, more than 50 products composed by more than 250 materials with unique routings through 25 workstations, the models would have been highly complex, time-consuming and should require a high skill on the matter. Nevertheless the heuristics approaches like pair-wise exchange and CRAFT continue to deliver optimal solutions, but regarding the specifics of the problem, the exotic shapes, highly distinct proportions and the particular entry and exit point of the department’s layout, even dough they could tackle these needs with additional programing, they are not normally that sophisticated. Thus the necessity to solve these issues leaded to the development of a similar tool that by using several equations and conceptions of CRAFT solves the problem through an additional set of rules and experiments. The results were promising and highly realistic, granting great confidence in the results and also in the process developed. The steps made are independent of the case study, so this approach can easily be used in other complex layout where the distance is the key problem in the material handling. But if the facility layout was not calculated trough simulation how can it aid the design. The answer to this question leaded to the study of different methodologies and the buffer sizing was the right candidate because the determination also has a bearing on the performance characteristics such as productivity, flexibility, and specially space utilization for a manufacturing system. In this case would take a major influence in the facility layout design regarding the additional space needed. Therefor it was built, using the case study, a model that simulates the behaviour of the manufacturing system and determines the maximum buffer size of a typical annual demand using a push type production. There were some drawbacks, as previously discussed, that halted the improving of the simulation, being the lack of computational capacity the major one, which lead to simplifications that prevented further developments that would need more time-consuming functions to emulate other behaviour or production types. Nevertheless the method undertaken has proven to be rigorous and delivers important data, even dough further advances could identify greater 5.2 - Further Research 99 improvements and approximate even further the model to the real scenario. The fact that the initial layout design was unknown did not allow a comparison analysis. Overall the facility layout determination aided by buffer sizing simulation proved to be a major advantage to the final result giving a deeper analysis to the study, a more realistic solution closer to the reality of the production system. Therefore offering arguments that could support with data and conclusions the decision making process for the continuous improvement. 5.2 Further Research Although the main objectives were fulfilled, several aspects can be exhaustively studied in the future. The buffer determination could be simulated with the additional setup time and without any reduction factor multiplier permitting a greater approximation to the real scenario. The simulation of different production strategies pull, with CONWIP and Kanban, and hybrid pull-push could arise more information for production system continuous improvement and buffer sizing determination. Implementation of the packaging line with a decoupling point between it and production with supermarket sizing calculation could deliver important information to a wider case study analysis. Incorporation of sifts with brake’s and lunch time could give more information and statistics of the working force and workstation utilization. Modelation of work station break times could also allow better risk management, test stress scenarios and take conclusions to develop mechanisms and methods to eliminate, reduce and predict problems. Lastly, AnyLogic’s functionalities were not completely studied specially the combination of different simulations methods like agent based with discrete event that could allow a deeper analysis. Appendixes 100 100 Appendixes Table 47 - Work Station Sifts Resource Description Shift name Break time (hours) M01R1 Cutting A 3,25 M02R1 Coating A 3,25 M03R1 Sawing A 3,25 M04R1 Wrapping D 0,25 M05R1 Cross cutting A 3,25 M06R1 4 side B 2,75 M07R2 Drilling line B 2,75 M08R1 2 sides D 0,25 M09R1 Corner cutting C 0,75 M10R1 Profile wrapping D 0,25 M11R1 Wrapping line D 0,25 M12R1 Cutting machine A 3,25 M13R1 Edge banding C 0,75 M14R2 Edge banding C 0,75 M15R2 Hot dowling C 0,75 M16R2 Friulmac C 0,75 M17R2 Frame assembly A 3,25 M18R2 Auto frame assembly A 3,25 M19R2 Drilling C 0,75 M20R1 CNC A 3,25 M21R1 Buffer D 0,25 M22R3 Packing A 3,25 M23R1 Rework A 3,25 M24R2 Product stacking A 3,25 Table 48 - Circus Flow Map Data Input labels A B C D E F G H I J K L M N O P Q R S T U V W X Y A 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 B 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 C 0 0 0 8641488 0 0 0 0 0 15817669 9736345 306144 0 0 0 325582 0 0 81396 0 0 0 0 0 0 D 0 0 0 0 8488416 153072 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 E 0 0 0 0 0 6553643 0 1934773 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 F 0 0 0 0 0 0 0 0 445034 0 0 0 0 0 0 0 0 0 0 0 7924458 0 0 0 0 G 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 H 0 0 0 0 0 0 0 0 365420 0 0 0 0 0 0 0 0 0 0 0 3042882 0 0 0 0 5.2 - Further Research 101 I 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 810454 0 0 0 0 J 0 0 0 0 0 0 0 0 0 0 0 15817669 0 0 0 0 0 0 0 0 0 0 0 0 0 K 0 0 0 0 0 0 0 0 0 0 0 9736345 0 0 0 0 0 0 0 0 0 0 0 0 0 L 0 0 0 0 0 1662777 0 1473529 0 0 0 0 1142143 7007820 6467013 2902798 0 0 5204078 0 0 0 0 0 0 M 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1142143 0 0 0 0 0 0 N 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7007820 0 0 0 0 0 0 O 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6467013 0 0 0 0 P 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3228380 0 0 0 0 Q 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2427429 0 0 0 0 R 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1126918 0 0 0 0 S 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 13435437 0 0 0 0 T 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 U 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 38462971 V 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 W 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Y 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Table 49 - Sankey Flow Chart Data Input C [8641488] D C [15817669] J C [9736345] K C [306144] L C [325582] P C [81396] S D [8488416] E D [153072] F E [6553643] F E [1934773] H F [445034] I F [7924458] U H [365420] I H [3042882] U I [810454] U J [15817669] L K [9736345] L L [1662777] F L [1473529] H L [1142143] M L [7007820] N L [6467013] O L [2902798] P L [5204078] S M [1142143] S 102 Appendixes N [7007820] S O [6467013] U P [3228380] U Q [2427429] U R [1126918] U S [13435437] U References 103 103 References [1] S. Terzi and S. Cavalieri, “Simulation in the supply chain context: A survey,” Comput. Ind., vol. 53, no. 1, pp. 3–16, Jan. 2004. [2] N. V. K. Jasti and R. Kodali, “Lean production: literature review and trends,” Int. J. Prod. Res., no. August, pp. 1–19, 2014. [3] H. C. W. Lau, B. Jiang, F. T. S. Chan, and R. W. L. Ip, “An innovative scheme for product and process design,” J. Mater. Process. Technol., vol. 123, no. 1, pp. 85– 92, 2002. [4] A. Mahfouz, J. Crowe, and A. Arisha, “Integrating Current State and Future State Value Stream Mapping with Discrete Event Simulation : a Lean Distribution Case Study,” no. c, pp. 161–168, 2011. [5] E. Jarkko, W. Lu, S. Lars, and O. Thomas, “Discrete Event Simulation Enhanced Value Stream Mapping : An Industrialized Construction Case Study,” Lean Constr. J., vol. 10, pp. 47–65, 2013. [6] K. Venkat and W. Wakeland, “Using simulation to understand and optimize a lean service process,” Simul. Ser., pp. 242–248, 2006. [7] B. J. Schroer, “Simulation as a Tool in Understanding the Concepts of Lean Manufacturing,” Simul. Trans. Soc. Model. Simul., vol. 80, no. 3, pp. 171–175, 2004. [8] “AnyLogic - Manufacturing Case Studies.” [Online]. Available: http://www.anylogic.com/casestudies/?TAGS=manufacturing&NUMBER_TAGS=35. [Accessed: 15-Sep2015]. [9] “AnyLogic - Logistics and Supply Chain Case Studies.” [Online]. Available: http://www.anylogic.com/case-studies/?TAGS=logistics-and-supplychains&NUMBER_TAGS=37. [Accessed: 16-Sep-2015]. [10] “Anylogic - Ports and Containers Terminals Case Studies.” [Online]. Available: http://www.anylogic.com/case-studies/?TAGS=ports-and-containerterminals&NUMBER_TAGS=1508. [Accessed: 16-Sep-2015]. [11] “AnyLogic - Markets and Competition.” [Online]. Available: http://www.anylogic.com/case-studies/?TAGS=markets-andcompetition&NUMBER_TAGS=275. [Accessed: 16-Sep-2015]. [12] A. Arisha and M. El Baradie, “On the Selection of Simulation Software for Manufacturing Application,” 2002. [13] C. Mclean and S. Leong, “The Role of Simulation in Strategic Manufacturing,” Simulation, 1998. [14] J. W. Fowler and O. Rose, “Grand Challenges in Modeling and Simulation of Complex Manufacturing Systems,” Simulation, vol. 80, no. 9, pp. 469–476, 2004. [15] F. a. Abdulmalek and J. Rajgopal, “Analyzing the benefits of lean manufacturing and value stream mapping via simulation: A process sector case study,” Int. J. Prod. Econ., vol. 107, no. 1, pp. 223–236, 2007. [16] A. Gupta, K. Singh, and R. Verma, “A critical study and comparison of 104 References manufacturing simulation softwares using analytic hierarchy process,” J. Eng. Sci. Technol., vol. 5, no. 1, pp. 108–129, 2010. [17] M. Jahangirian, T. Eldabi, A. Naseer, L. K. Stergioulas, and T. Young, “Simulation in manufacturing and business: A review,” Eur. J. Oper. Res., vol. 203, no. 1, pp. 1–13, May 2010. [18] A. Negahban and J. S. Smith, “Simulation for manufacturing system design and operation: Literature review and analysis,” J. Manuf. Syst., vol. 33, no. 2, pp. 241– 261, 2014. [19] “AnyLogic.” [Online]. Available: http://www.anylogic.com/. [Accessed: 31-Aug2015]. [20] J. A. Tompkins, J. A. White, Y. A. Bozer, and J. M. . A. Tanchoco, Facilities Planning. 2010. [21] A. Drira, H. Pierreval, and S. Hajri-Gabouj, “Facility layout problems: A survey,” Annu. Rev. Control, vol. 31, no. 2, pp. 255–267, Jan. 2007. [22] N. Chase, J. Aquilano, and R. B, Operations Management for Competitive Advantage (The Mcgraw-Hill/Irwin Series Operations and Decision Sciences). . [23] S.-D. Lee and S.-H. Ho, “Buffer sizing in manufacturing production systems with complex routings,” Int. J. Comput. Integr. Manuf., vol. 15, no. 5, pp. 440–452, 2002. [24] T. Tezcan, “Optimal buffer allocation in production lines Using an Automata Search,” IIE Trans., vol. 35, no. 1, pp. 1–10, 2003. [25] N. Zheng and X. Lu, “Comparative study on push and pull production system based on anylogic,” Proc. - 2009 Int. Conf. Electron. Commer. Bus. Intell. ECBI 2009, pp. 455–458, 2009. [26] A. Krishnamurthy, R. Suri, and M. Vernon, “Push can perform better than pull for flexible manufacturing systems with multiple products,” Ind. Eng. Res. …, pp. 1– 7, 2000. [27] K. Z. Zhou, C. K. Yim, and D. K. Tse, “The Effects of Strategic Orientations on Technologyand Market-Based Breakthrough Innovations,” J. Mark., vol. 69, no. 2, pp. 42–60, 2005. [28] W. J. Hopp and M. L. Spearman, “To Pull or Not to Pull: What Is the Question?,” Manuf. Serv. Oper. Manag., vol. 6, no. 2, pp. 133–148, 2004. [29] M. Savsar, “Simulation analysis of a pull-push system for an electronic assembly line,” Int. J. Prod. Econ., vol. 51, no. 3, pp. 205–214, Sep. 1997. [30] M. L. Spearman and M. a. Zazanis, “Push and Pull Production Systems: Issues and Comparisons,” Oper. Res., vol. 40, no. 3, pp. 521–532, 1992. [31] J.-L. Deleersnyder, T. J. Hodgson, H. Muller, and P. J. O’Grady, “kanban Controlled Pull Systems: An Analytic Approch,” Manage. Sci., vol. 35, 1989. [32] J.-P. Timsit, A. Castiaux, Y. Truong, G. A. Athaide, and R. R. Klink, “The effect of market-pull vs. resource-push orientation on performance when entering new markets,” J. Bus. Res., vol. 68, no. 9, pp. 2005–2014, Feb. 2015. [33] A. De Toni, M. Caputo, and A. Vinelli, “Production Management Techniques: Push-Pull Classification and Application Conditions,” Int. J. Oper. Prod. Manag., vol. 8, no. 2, pp. 35–51, 1988. References 105 [34] J. Geraghty and C. Heavey, “A comparison of Hybrid Push/Pull and CONWIP/Pull production inventory control policies,” Int. J. Prod. Econ., vol. 91, no. 1, pp. 75– 90, Sep. 2004. [35] Francis, Mcginnis, and White, Facility Layout And Location: An Analytical Approach. . [36] S. P. Singh and R. R. K. Sharma, “A review of different approaches to the facility layout problems,” Int. J. Adv. Manuf. Technol., vol. 30, no. 5–6, pp. 425–433, 2006. [37] M. R. Garey and D. S. Johnson, Computers and Intractability: A Guide to the Theory of NP-Completeness. 1979. [38] Y. Zhu and F. Wang, “Study on the general plane of log yards based on systematic layout planning,” 2009 Int. Conf. Inf. Manag. Innov. Manag. Ind. Eng. ICIII 2009, vol. 3, pp. 92–95, 2009. [39] F. R. Jacobs and R. B. Chase, Operations and supply chain management, vol. Global ed. New York: McGraw-Hill. [40] P. Shewale, “Improvement In Plant Layout Using Systematic Layout Planning (Slp) For Increased Productivity,” … J. Adv. …, vol. 4, no. 12, pp. 1382–1386, 2012. [41] P. Kouvelis and M. W. Kim, “Unidirectional Loop Network Layout Problem in Automated Manufacturing Systems,” Oper. Res., vol. 40, no. 3, pp. 533–550, 1992. [42] V. Maniezo, A. Colorni, and M. Dorigo, “The Ant System Applied to The Quadratic Assignment Problem,” vol. 11, no. 5, pp. 769–778, 1994. [43] R. R. Schaller, “Moore’s law: past, present and future,” Spectrum, IEEE, vol. 34, no. 6, pp. 52–59, 1997. [44] G. C. Armour and E. S. Buffa, “A Heuristic Algorithm and Simulation Approach to Relative Location of Facilities,” Manage. Sci., vol. 9, no. 2, pp. 294–309, 1963. [45] R. D. Meller and K.-Y. Gau, “The facility layout problem: Recent and emerging trends and perspectives,” J. Manuf. Syst., vol. 15, no. 5, pp. 351–366, Jan. 1996. [46] N. H. Prasad, G. Rajyalakshmi, and a. S. Reddy, “A Typical Manufacturing Plant Layout Design Using CRAFT Algorithm,” Procedia Eng., vol. 97, pp. 1808–1814, 2014. [47] J. S. Smith, “Survey on the use of simulation for manufacturing system design and operation,” J. Manuf. Syst., vol. 22, no. 2, pp. 157–171, Jan. 2003. [48] R. G. Ingalls, “Introduction to Simulation,” Proc. 2008 Winter Simul. Conf., pp. 17–26, 2008. [49] J. Banks, “Proceedings of the 2000 Winter Simulation Conference J. A. Joines, R. R. Barton, K. Kang, and P. A. Fishwick, eds.,” pp. 9–16, 2000. [50] R. E. Shannon, “Introduction to the art and science of simulation,” 1998 Winter Simul. Conf. Proc. (Cat. No.98CH36274), vol. 1, pp. 7–14, 1998. [51] A.-C. Fagerlind Ståhl, M. Gustavsson, N. Karlsson, G. Johansson, and K. Ekberg, “Lean production tools and decision latitude enable conditions for innovative learning in organizations: a multilevel analysis.,” Appl. Ergon., vol. 47, pp. 285– 91, Mar. 2015. [52] A. Maria, “Introduction to modeling and simulation,” Proc. - Winter Simul. Conf.,