Grup teknolojisi imalat sistemleri tasarımı için bir metodoloji ve bu metodolojinin endüstri'de uygulanması

Abstract

Avrupa topluluğu ile gümrük birliğine girmeyi planlayan Türkiye'nin dış pazarlardaki rekabet gücünü arttırılabilmesi, ülkemizdeki endüstriyel kuruluşların istenilen performansı yakalayabilmesi için Grup Teknolojisine dayalı "Hücresel Üretim" felsefesine uygun şekilde sistemlerini yeniden düzenleyerek dünya pazarlarındaki yerlerini almak zorundadırlar. Ancak hücresel üretimin başarısı, mevcut sistemi basitleştirebilme düzeyine, yani tasarımın etkinliğine ve değişime inanmaya bağlıdır. Böyle bir projenin hazırlanması bir ekip çalışmasını gerektirir ve zaman alır. Hücresel imalat sistemleri geleceğin endüstri kuruluşları için yapılmış önemli bir stratejidir. Bu çalışma geleneksel üretim sistemleriyle çalışan sistemlerin Hücresel imalata dönüşümü sırasında yapılması gereken hazırlıkların bilimsel tabana oturtulması amacıyla gerçekleştirilmiştir. İmalat sistemi içinde basit ve etkin bir malzeme akışı kurulmaya çalışılmıştır. Bu hazırlık, tezde dört aşamalı bir metodoloji halinde sergilenmektedir: Bilindiği üzere, günümüzde imalat sektöründeki gelişmelere paralel olarak yaygın bir şekilde, CNC tezgahlar kullanılmaya başlanmıştır. Bunun getirdiği rotalama esnekliğinden, imalat sistemlerinin tasarımında yararlanmak gerekmektedir. Bunun için önerilen metodolojinin birinci aşamasında, minimum maliyet esasına göre, bir matematiksel model yardımıyla ürün rotası seçimi sonucu, rotalama esnekliği kullanılmıştır, ikinci aşamada minimum maliyetli rotalar kullanılarak yapay sinir ağı yardımı ile imalat hücreleri oluşturulmuştur. Üçüncü aşamada, hücre konfigürasyonuna son sekli vermek üzere, yeni yatırım, üretme/satın alma ve hücreler arası akışa izin verme gibi karar verme sürecine yardımcı olacak bir matematiksel model kurulmuştur. Son aşama ise uygulamaya alınacak olan hücrelerin iç düzenlemesinin yapılması ile ilgilidir. Tezde, birinci, ikinci, dördüncü aşamada literatürde mevcut olan yöntemler seçilerek algoritmaları bilgisayarda kodlanmış ve endüstriyel bir firmaya uygulanmıştır. Bilinen ve farklı çalışmalarda ayrı ayrı kullanılmış olan söz konusu yöntemlerin bir metodoloji halinde toplanarak uygulanmış olması, tez çalışmanın özgün yönü olarak gösterilebilir. Üçüncü aşamada hücre konfigürasyonuna son şekli vermek üzere yeni yatırını, üretme/satın alma ve hücreler arası akışa izin verme gibi karar verme sürecine yardımcı olacak bir matematiksel model geliştirilmiştir. Bu da tez çalışmasının özgün bir yönü olarak gösterilebilir. Çalışmanın teorisinin uygulama açısından getireceği yararların sağlıklı bir şekilde tespit edilebilmesi için, araştırma bir uygulama ile desteklenmiştir. Uygulama Ülkemizde cam sanayine kalıp üretimi yapan Şişe Cam'a bağlı Camiş Makina ve Kalıp San. A.Ş.' de gerçekleştirilmiştir. Yöneticilere kararlarında destek olması açısından önemli ipuçları içerdiği söylenebilir.

Group technology (GT) is a technique that is proved to be quite effective in solving many of the problems in discrete part batch manufacture. In the past, industrial engineers have tended to view each part produced in a company as being unique. Group technology brings similar parts together in families, each member is unique but each has also some characteristics similar to the others. The part-machine grouping problem forms one of the first and most important steps in the design of a cellular manufacturing system. It was first identified by BURBIDGE (1963) as 'group analysis' within the context of production flow analysis (PFA). A description of the overall cell-formation problem is provided. Cellular manufacturing (CM) involves processing a collection of similar parts on a dedicated cluster of dissimilar machines. CM has been found beneficial to some companies because it can reduce lead time, material handling, setup time, work in process, etc (BLACK 1983). In batch manufacture, the concept of GT leads to the formation of families of parts that are similar in design or manufacture. The machines are required to produce a family of parts are grouped together to form a manufacturing cell. Such a grouping of machines, which is known as cellular manufacturing reduces material handling and enables a smooth flow of parts in the plant. Cellular manufacturing, whether automated or not, provides many advantages over the traditional approach of functional layout. Shorter lead times, reduced work-in-process inventories, and improved product quality are some of the benefits which have been reported (HYEROF 1987). WEMMERLOV and HYEROF (1987) divide the design phase of the cellular manufacturing system into five stages. These five stages are: (1) selection of part populations and grouping of parts into families; (2) selection of machine and process populations and grouping of these into cells; (3) selection of tools fixture and pallets; (4) selection of material handling equipment; (5) choice of equipment layout. In the first chapter, production systems are defined by the system approach and the role of manufacturing systems in determined. Later on, manufacturing systems are divided in two classes; Traditional Manufacturing Systems Cellular or GT Manufacturing Systems. Traditional manufacturing systems are often classified into three categories: Job shop Flow shop Project shop The job-shop uses flexible, general-purpose machines, and it is a process in which units for different orders follow different paths (individual routes or sequences) through processes or machines. The flow-shop is characterised by product flow layout that uses special- purpose, single-function machines. Flow shop is a process in which successive units undergo the same sequence of operations with more specialised, dedicated equipment. The third category of traditional manufacturing is a project shop. This system is directed toward creating a product which is very large or one-of-a-kind, which typically must be produced in some specified sequence. The men, material, and the machines all come to the project site for assembly or processing. Cellular manufacturing (group production) and family programming are basically manufacturing process which produces families of parts within a single line or cell of machines operated by machinists or robots that operate only within the line or cell. Some of the benefits of Cellular Manufacturing Systems are: (1) Elimination of or decrease in setup time and setup cost. (2) Greater manufacturing flexibility. (3) Reduced work-in-process and lower inventory (just-in-time). (4) Less floor space around the machines. (5) Lower raw materials. (6) Reduction in the cost of goods produced. (7) Capability of using high-investment machinery in the production process. (8) Reduction in direct labour cost. (9) Higher productivity. The part-machine grouping problem involves a reorganisation of the rows and columns of a part-machine incidence matrix to obtain a block diagonal form. The intractability of this subproblem has led to the development of numerous heuristic procedures. CHU (1989) classified the literature under the categories: (1) array- based methods; (2) hierarchical clustering techniques; (3) non-hierarchical clustering methods; (4) mathematical programming formulations; (5) graph-theoretic approaches; and (6) other heuristic methods. In the second chapter, the larger problem involves the consideration for several real-world complexities, including, the presence of multiple copies of machines of each type, and load balancing among various cells. In addition, recent analytical models have highlighted the need to address the loss of pooling synergy when partitioning job shop work centres to form cells (SURESH 1991, 1992, SURESH and MEREDITH, forthcoming). Artificial neural network (ANN) models are characterised by their properties, viz., the structures of the network (topology), how and what the network computes (computational property) and how and what the network learns to compute (learning or training property). The learning factor, more than any other, is responsible for exhibiting the characteristics of the brain. Learning is the process in which a set of input vectors is presented sequentially to the input of the networks and the network weights are adjusted such that similar vectors activate the same output neuron. Learning strategies are categorised as supervised and unsupervised. Supervised training requires the pairing of each input vector with a target vector representing the desired output. An input vector is applied and the output of the network is calculated. This type of learning indicates whether the computed output is right or wrong. It has been criticised as being biologically implausible. In unsupervised learning, the training set consists of input vectors only. The output is determined by the network during the course of training. The training procedures construct internal models that capture regularities in their input vectors without receiving any additional information. In this thesis, we pay attention to unsupervised learning. In the third chapter, the four stages have been desenbed for the integrated manufacturing design. In the first stage, a mathematical model has been described for the minimal on route selection. In the second stage, this stage presents a neural network clustering method for the part-machine grouping problem in group technology. Among the several neural networks, a CARPENTER-GROSSBERG network is selected due to the fact that this clustering method utilizes binary-valued inputs and it can be trained without supervision. It is shown that this adaptive leader algorithm offers the capability of handling large, industry-size data sets due to the computational efficiency. In the third stage, an integer mathematical model has been developed to aid the decision process to minimize exceptional elements, new machine tool purchasing, subcontracting, intercell movements between the cells. In the forth stage, two model have been presented to layout the machines and equipment in the cell and then alternative layouts have been compared. In the fourth chapter, the methodology developed has been applied to test the algorithm in a real industrial company. We can say that it contains important insigths for giving managements an overall information. At the first stage, alternate part routing are considered to determine the best part routing that minimises the operating cost. Integer programming model was formulated and solved to select the optimal part routing for every part. This integer programming model was solved using LINDO (SCHRAGE 1987) on the personnel computer. At the second stage, the results from the first stage from 0-1 part-machine matrix. The binary is used as the input to an ARTİ neural network based cell formation module to group machines into a specified number of cells. The algorithm can be effectively utilized for part-machine grouping especially for large, industry size problems. The algorithm was coded in Visual C++ and run on the personnel computer. At the third stage, Cell formation solutions often contain exceptional elements (EE). EE create interactions between two manufacturing cells. They can be considered the result of bottleneck machines, i.e. machines required by two or more part families. Conversely, EE can be viewed as parts that require processing on machines in two or more cells. The interaction between cells caused by the existence of EE runs counter to the express CM philosophy of creating independent cells. As will be discussed later, this interaction creates disruptions in a CM environment, for which tangible and intangible costs are incurred. If the part involved in EE cannot be re-designed or altered, two major options have been suggested for eliminating EE. Duplicates of the bottleneck machine can be acquired. This action would eliminate the need for transferring parts between cells in order for them to receive processing on bottleneck machines. In some instances, the part can be subcontracted, removing it from the CM production environment. This thesis presents a mathematical programming approach for dealing with exceptional elements in CM. The approach suggests whether it is cost-effective to intercellular transfers caused by the EE should remain in the cell formation. It provides an optimal solution for resolving the interaction created by EE in the initial cell formation solution. In addition, the model recognises potentially advantageous mixed strategies by previous approaches. Integer programming model can be used to minimize the costs associated with intercellular transfers, part subcontracting and bottleneck machine duplication. Integer programming model was formulated and solved using LINDO (SCHRAGE 1987) on the personnel computer. At the fourth stage, In optimising the layout design of a multi-product assembly environment, the analysis of the material flow is a vital ingredient. In a multi-product production environment, products are usually grouped together into families. First Category, the ANEKE and CARRIE (1986) presented a travel-chart method. This method, similar to Hollier's link-analysis, constructs the flowline from both ends simultaneously. This construction method is usually good in situations where a family of products have similar initial and final work station requirements, and processes in which the difference occur mainly in the intermediate process. The algorithm was coded in Visual C++ and run on the personnel computer. In the second category, Heuristic pattern matching model is shown. This method was the traditional line structure as a basic for flowline construction process of the heuristic pattern matching method emphasises sequence similarity among products and uses sequence similarity information to construct the flow line. There are three main stages in the heuristic pattern matching method. The first stage is the selection stage where a candidate product is selected. The sequence of the candidate product is accommodated into the constructed flowline, which is a flow path that consists of different operations (machines/workstations). The selection of the candidate product is mainly based on the sequence similarity between the product sequence and the sequence of the constructed flowline. The product that has the highest sequence similarity with the constructed flowline is selected. A sequence similarity measurement called 'compliant index' is introduced in this method. The second stage is the construction stage where a new flowline is constructed by modifying its sequence so that the sequence of the candidate product can be included into the flowline. The third stage is the elimination stage where the elimination of the infeasible or uneconomic workstations/machines from the constructed flowline is performed. The algorithm was coded in Visual C++ and run on the personnel computer. The performance measure used here is total flow distance. In order to calculate flow distance, it is assumed that the length of the links connecting two workstations is one unit. From the test experiments, it is observed that the method of ANEKE and CARRIE (1986) is sensitive to the initial and the final operations of the products. This is because their method selects the operations from either end of the products, and adds the selected operations to the corresponding end of the flowline. The disadvantage of this is that the products may travel through many unnecessary intermediate workstations, resulting in larger flow distance. On the other hand, the heuristic pattern matching algorithm emphasizes the overall sequence similarity in its building processes, and less sensitive to the initial and final operational of the products, which explain its better performances.