Tam zamanında imalat sisteminin simülasyon ile analizi ve uygulanabilirliğinin etüdü

Abstract

ÖZET İkinci Dünya Savaşı'ndan sonra Japon Toyota Firması yöneticilerinin, maliyetlerini düşürerek dünya piyasasında söz sahibi olabilmek amacıyla salt kendilerine özgü bir sis tem arayışları sonucunda ortaya çıkan `Tam Zamanında Üretim Sistemi`nin incelendiği bu çalışma beş bölümden oluşmaktadır. Birinci bölümde, Tam Zamanında üretim Sistemi'nin (TZÜS) ortaya çıkışı tarihsel açıdan anlatılmış, itme ve çekme sistemleri tanıtılarak Malzeme İhtiyaç Planlaması, İmalat Kaynak Planlaması ve TZÜS hakkında kısa bilgiler verilmiştir. İkinci bölümde TZÜS'nin elemanları ve Kanban Sistemi tanıtılmış, ayrıca Kanban sistemine alternatif gösterilen diğer TZÜ sistemleri de özetler halinde sunulmuştur. Üçüncü bölüm ise yayın taraması ile ilgilidir. Dördüncü bölümde, Tam Zamanında İmalat Sisteminin (TZÎS) uygulandığı düşünsel bir atölyeye işletim şartlarının etkilerinin araştırılması için Q-GERT tekniği ile yapılan simülasyon deneyleri anlatılmıştır. Ayrıca, işletim şartlarından olan değişken talep şartlarının sisteme etkilerini belirlemek için bir deneysel tasarım ve istatistiksel analiz gerçekleştirilmiştir. Bu çalışmanın sonuçlarından yola çıkılarak beşinci bölümde, otomotif endüstrisine çalı şan bir firma için TZİS'nin uygulanabilirliği araştırılmış tır. Bu çalışmanın özgün yanı, ilk simülasyon deneyleri ve istatistiksel analizler sonucu elde edilen sistem tepkilerinin göz önüne alınarak oldukça karmaşık olan gerçek bir ortamda Grup Teknolojisi ile ulaşılan bir pilot bölgeye hangi şartlar altında TZÎS'nin uygulanabilir olduğunun simülasyon yardımıyla ortaya konmasıdır. Grup Teknolojisi ile TZÎS beraberce kullanılmış, sistem parametre ve değişkenleri üze rinde sağlanan çok yönlü hakimiyet sonucu bunların çeşitlen dirilmesi ile elde edilen politikalar zincirinin hangi halkalarının TZÎS 'ne olumlu tepki gösterdiği saptanmıştır. - x -

SUMMARY AN ANALYSIS OF THE JUST IN TIME MANUFACTURING SYSTEM BY SIMULATION AND A STUDY FOR ITS APPLICABILITY The Just in Time (JIT) production system is the production of necessary items in necessary quantities at a necessary time. That is, in JIT production ör manu facturing systems, the succeeding process orders and/or withdraws parts from the storage of the preceding process, only at the rate and at the time it has consumed the items. When this approach is applied to all of the production processes, JIT will achieve its ideal objective. The objectives of this system are to: - minimize the work-in-process (WIP) inventory, - minimize fluctuations in WIP to simplify inventory controls, - minimize production instability by preventing demand fluctuations from one process to another, - provide better control through decentralized shop floor control, - reduce defects, and - reduce manufacturing costs. JIT system has been applied in repetitive manufac turing in electronic, automotive etc. industries, succes- fully. After the World War II, this system was developed by the managers of Toyota Motor Company in Japan who were seeking methods of their own to reduce manufacturing costs and to become more competitive in the world markets. Following the oil crisis in 1973, JIT known as Toyota Production System, attracted the attention of Japanese, American and European countries. The companies in these countries began implementing this system. But, the American and European companies have not been as successful as the Japanese because of certain operational conditions. The causes of the success of the Japanese are the following: - frozen master production schedules, - reduced set-up times, - highly skilled and cross-trained workers, - xi -- utilization of high degree automation and robotics, - proximity of suppliers. This type of system's characteristics generally reduce the system variabilities. In this way, the Japanese companies are capable of reducing system variability, as the customer demand can be estimated and controlled very accurately. The third and fourth specifications, for instance, reduce the variability of processing times. The effects of the system's variabilities on the system perfor mance have been examined, mostly by simulation methods, during the last years. Here, in this thesis, after the explanation of JIT manufacturing system and the Kanban sys tem which is the sub-system of JIT, an experimental analysis is materialized for examining two operational conditions which reduce the system performance in a multi-line, multi stage manufacturing system. Furthermore, a simulation study regarding the applicability of JIT system in a Turkish company is developed. This thesis consists of the following chapters: In the first chapter, the stages of JIT system development are mentioned in chronological order. Besides this, the Pull and the Push Systems are defined and com pared according to the Material Requirements Planning, Manufacturing Resource Planning and JIT concepts. The elements of a successfull JIT production system in a repetitive manufacturing environment are described in the second chapter. These elements are listed belowj - a focused factory, - reduced set-up times, - group technology, - total protective maintenance, - cross- trained workers, - uniform work loads, - just in time delivery of purchased parts, - Kanban. As explaining the Kanban system, the kinds of Kanbans, the full-work system, the rules of Kanban and a manufacturing system with Kanban are mentioned and the manufacturing system is shown by the movements of Kanbans and carts. In addition, the formula which is applied in Toyota Motor Co. for the calculation of the number of Kanbans is given and the fac tors influencing the number of Kanbans are described. Furt hermore, the alternative systems of Kanban are introduced, briefly. Some of these systems are: Integrated Assembly, Micro-Kanban, PYMAC or Synchro MRP, Action Plate (AP), Mini mum Stock and Minimum Standard Time, and Periodic Pull System. The third chapter deals with a literature survey. This chapter is categorized in four sections. In the first - xii -two sections, mathematical and simulation models are des cribed. In 1981, Kimura and Tereda formulated the pull sys tem and built a model simulation of fluctuation in production and inventory, through the whole process in terms of system parameters like batch size, lead time, etc. The pull sys tem is considered as a multi-stage series process with a single item. Kimura and Tereda obtained the following results: - In the case of the Pull System, the size of the order unit has much importance. In cases where the size is small compared with the production quantity level, pro duction fluctuation will not be amplified in the preceding stage. Amplification will be brought about when the size is rather large although, in this case also, the amplifica tion is not further magnified in further preceding stages. - In the case of the Push System, amplifications of production and inventory fluctuations occur under the inf- luance of errors in forecasting. As far as amplification. is concerned, the choice between the Push and Pull Systems is determined by the degree of errors in forecasts. - The other factor in the system parameters of the Pull System, which affects the amplification ratio is the 'lead-time' from the moment when a Kanban is removed from a container to the moment when the production at that stage is completed. The longer the lead-time, the larger is the amplification ratio. In a Periodic Pull System (PPS) as an alternative of a Kanban system, the manual information processing time of a Kanban system is replaced with an instant on-line com puterized processing, and the material lead time is much shorter than that of a Kanban system. This results in better system performance such as less lead-time inventory and faster system response. In 1985, a PPS was formulated mathematically by Kim. Kim considered the Kimura and Tereda' s pull system (1981) and formulated a PPS for this system, then provided an approach for solution for the target stock levels, as well as the analysis of the fluctuations of in-process material flow, on-hand stock levels, target availability, etc. In 1987, Bitran and Chang set up a mathematical programming model for the Kanban system in a deterministic multi-stage assembly production setting. The model assists managers in determining the number of circulating Kanbans and hence the inventory level, at each stage. Contrasting with Kimura and Tereda (1981) the model has no assumptions on the container size and allows limited production capacity, Bitran and Chang investigated a general solution procedure to the model and discussed three special cases of practical interest. - xiii -Miyazaki and Ohta, in 1988, developed a practical algorithm to obtain the optimal order interval and the total number of Kanbans required for the optimal interval. They derived four formulas. Two of these are to calculate the avarage inventory yielded by fixed interval withdrawal Kanban and supplier Kanban, respectively. By using these two formulas, the other two formulas to determine the mini mum number of Kanbans required for the fixed interval with drawal Kanban and the supplier Kanban are also shown. In their algorithm, they proposed to minimize the total opera tion cost composed of the inventory cost and the withdrawal cost for the Kanban system. Therefore, the authors empha sized that the total cost curve in the Kanban model did not reveal the convex function like in the conventional Economic Order Quantity (EOQ) model. In 1983, Huang, Rees and Taylor III, made a simulation study of just-in-time with Kanban technique for a multi line, multi-stage production system in order to determine its adaptability to an American production environment that might include such characteristics as variable processing times, variable master scheduling, and imbalances between production stages. They conducted four simulation experi ments by Q-GERT simulation technique. The purpose of these experiments was to observe the effects of variable proces sing times, stage bottlenecks, variable demand rates, and a combination of these factors on the Kanban process. The simulations results showed that the variability in proces sing times and demand rates significantly affected overtime requirements and production output. To solve bottleneck problems, the authors recommended the training of workers or utilization of more automated and improved machinery. Huang, Rees and Taylor III also suggested a preperation and transition period of at least one year. In 1984, Schroer, Black and Zhang developed a simu lation model to compare with the various levels of in-pro cess inventory and production rates. The objective of this simulation model is to minimize in-process inventory at the various stock points in a hypothetical manufacturing system and to balance this system in terms of resource utilization and queues using a SIMAN simulation language. Also in 1984, Ebrahimpour and Fathi developed a simulation model by DYNCMO simulation language. The pur pose of the model was to test the effect of implementation of Kanban system on the work-in -process levels and to compare these effects with traditional system under similar conditions. Therefore, system behavior under the cyclical demand and constant growth o E demand is analyzed with respect to the two systems. In addition, the Kanban system behaviour is analyzed under the condition of gradual reduc-,` tion of cards during stable demand. Rees, Philipoom, Taylor III and Huang explored a workable method for dynamically adjusting the number of - xiv -Kanbans in a JIT shop in unstable production environment in 1987. They presented their methodology for a hypothe tical shop by Q-GERT simulation technique. In addition, they submitted two examples to demonstrate how well the methodology reacted to the dynamics of the shop operation. The same authors made another simulation study to identify the factors influencing the number of Kanbans required at a workcentre in JIT system with Kanbans. The factors are identified as the throughput velocity, the co efficient of variation in processing times, the machine utilization and autocorrelation of processing times. Furt hermore, they described a methodology for determining the number of Kanbans to use at a workcentre given a dynamic production environment. This methodology is also demons trated by simulation in a Q-GERT language. In 1988, Sarker and Harris developed a simulation model with a SLAM simulation technique. The objective of this model was to analyze the different effects of the imbalance of stage operation times in a real life JIT pro duction system for manufacturing toys. The third section of literature survey is interested in the companies which applied JIT approach. Some benefits of these companies from JIT applications are presented in this section. In the last section, JIT applications on the service systems are mentioned as official paperwork opera tions, accounting, marketing, purchasing, packaging sys tems, etc. In the fourth chapter, an experimental analysis of JIT by the Q-GERT simulation technique is conducted. The purpose of this analysis is to examine the factors negatively influencing a multi-line, multi-stage JIT manufacturing system.. These factors are variable processing times and variable demand. The system which is under consideration consists of three production lines and one line has three, the other line has one and the last one has two cells or operation centers. The units which are processed at each line go to the assembly operation. This system is the same as Huang, Rees and Taylor's (1983) system. In addition to their studies, the total queue lenghts vs. the variability of processing times and the effects of batch sizes vs. different system performance measures and the effects of variability of demand on system performance measure overtime by means of experimental design and statistical analysis are examined. In this study, firstly the two sets of simulation experiments are performed under the condition of constant daily demand to analyze the effects of variable processing times and the batch sizes on the system performance measures. Then, an experimental design and statistical analysis (one way and two-way analysises of variance and multiple regres sion analysis) were carried out to see the effects of - xv -variable demands, the interactions between the variability of demand and the other factors; variable processing times and number of Kanbans, also. As a result of the experimen tal analysis, are the following: - The first set of simulation experiments applied under the condition of constant demand shows ;that the vari ability of process times increases the need for overtime. This increase can be controlled and reduced by raising the number of Kanbans or work-in-process inventories. - Determining the necessary number of Kanbans is proven to be very important for all the system performance measures through the study. - The second set of simulation experiments reveals that if set-up times are negligibly small than the undesi rable effects cause by the variability of process times can be, to some extent, eleminated by smaller batch sizes. - The one-way analysis of variance proves that taken one at a time, each of the factors (number of Kanbans, vari ability in process times, and in demand) drastically influ ences the system. - After the two-way analysis of variance, no consi derable relation can be found between the variability of demand and the other two factors; variability of process times and number of Kanbans. - Multiple regression analysis shows that the vari ability of demand and number of Kanbans are more effective than the variability of proccessing time on the system. In the last chapter, based on the results achieved in the fourth chapter, the application of JIT manufacturing system for a company in the automotive industry in Turkey was researched by simulation, and various policies were developed for evaluating the system performance measures. A company produces propeller shafts for the automotive industry, P.T.O. (Power -Take-Off ) shafts for agricultural equipment, pump drive shafts and their auxilary parts and universal joint kits. The reasons that this company was chosen are because this company is involved in discrete and batch type repetitive manufacturing and has relatively simple work flows. The basic idea of this study is to develope a simu lation model for a selected pilot area and according to the simulation results, to make an evaluation if JIT manufac turing system is applicable for, that area and the rest of the manufacturing system in the company. The applicability survey of JIT system was done by Q-GERT simulation technique and developed the various policies to evaluate the system performance measures. The procedure progressed as follows: - xvi -A pilot area was chosen by using group technology methods and the number of Kanbans required in that area was calculated. According to the batch size and furnace charge capacity selected, the simulation was run and it was obtained that overtime requirements, queue lengths and machine utili zations responded negatively to the question of applicabitity of JIT for the selected area. Because of this response, a chain of policies was formulated by increasing the variety of the values which concern batch sizes, furnace charge capacity, set-up times, number of Kanbans and additional tool investments. As a result of simulation experiments under these policies, which link of the chain responded positively to JIT was. also identified. In other words, under which condi tions JIT could be applied in the pilot area was determined. These experiments demonstrate the original characteristics of this thesis. This thesis concludes with the final results of the study conducted and makes proposals for future research in this area. - xvii -