Geri Dön

Hücresel imalatın başlangıç aşamaları için uzman sistem yaklaşımı

An Expert systems approach to the early stages of cellular manufacturing systems design

  1. Tez No: 39872
  2. Yazar: UFUK CEBECİ
  3. Danışmanlar: PROF.DR. ATAÇ SOYSAL
  4. Tez Türü: Doktora
  5. Konular: Endüstri ve Endüstri Mühendisliği, Industrial and Industrial Engineering
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 1994
  8. Dil: Türkçe
  9. Üniversite: İstanbul Teknik Üniversitesi
  10. Enstitü: Fen Bilimleri Enstitüsü
  11. Ana Bilim Dalı: Belirtilmemiş.
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: 213

Özet

ÖZET Hücresel İmalat (Hİ) Sistemlerinin Modern İmalat Sistemleri arasında önemli bir yeri vardır. Bir üretim sisteminin yapısı uygun olduğu taktirde, Hİ verimliliği önemli ölçüde arttıran bir sistem olmaktadır. Klasik üretim sistemlerinden başlayıp esnek üretim sistemlerine kadar uzanan geniş bir yelpazede uygulanma imkanı bulmaktadır. -v Kümelendirme yöntemleri ve üretim sistemleriyle ilgili özelliklerin çok sayıda olması, Hİ'ın uygulanacağı üretim sistemi şartlarına göre uygun bir yöntemin seçilmesini ve uygulanmasını zorlaştırmaktadır. EXCLUSTER adı yerilen, geliştirilen Uzman Sistem Yaklaşımı (USY), bir işletme için Hİ'ın başlangıç aşamalarıyla ilgili faaliyetlerde yardımcı olur. Hİ'ın uygulanması için hangi yöntem veya yöntem sonucunun seçilmesi gerektiğini belirler. Ayrıca Hİ, kümelendirme yöntemleri ve bu yöntemlerin nasıl kullanılmaları gerektiği hakkında, çeşitli bilgiler sunar ve önerilerde bulunur. Hİ Sistemi tasarımı 5 aşamadan oluşur: 1) Parça nüfuslarının seçimi ve parçaların ailelere gruplandırılması, 2) Makina ve proses nüfuslarının seçimi ve hücrelere ayrılması, 3) Takımların, paletlerin ye tertibatların seçimi, 4) Malzeme taşıma ekipmanının seçimi, 5) Ekipman yerleştirilmesi. Geliştirilen USY bu tasarım aşamalarından ilk ikisi ile yoğun olarak ilgilidir. Ayrıca sonraki aşamalarla ilgili tavsiyelerde bulunur. Kullanıcıdan işletmeyle ilgili veriler alınır ve işletmenin özel şartlarına göre ne tür bilgilerin sisteme girileceği belirlenir. USY, sorgulama işlemi sırasında bilgi tabanındaki kurallar yardımıyla açıklamalarda bulunarak kullanıcıyı aydınlatabilir. Ayrıca, uzman sistemin çalışmasında gerekli olacak işletmeyle ilgili parametreler (hücrede istenen maksimum veya minimum makina sayısı, performans ölçütleri ve ölçütlerin ağırlıklarının bir bölümünün seçimi v.b.) belirlenir. Kullanıcıdan elde edilen bu verilerin analizi sonucunda“işletmenin gerçekleri”elde edilmiş olur. Bundan sonraki aşamada, çıkarım mekanizması bütün kümelendirme algoritmalarını inceler ve işletme için uygun olmayan algoritmaları eler. Geriye kalan algoritmalar için, gerekli veri kümesi kullanıcı tarafından sisteme girilir. Kullanıcının oluşturması gereken değerler için önerilerde bulunur. Kullanıcıyla karşılıklı etkileşim içinde çalışarak ölçütlerin ağırlıklarını belirler. ELECTRE Yöntemini kullanan bir Çok Ölçütlü Karar Verme Modülü, işletme için en uygun kümelendirme yöntem(ler)ini ve şartlarını seçer ve Hİ sistemi tasarımının mevcut ve daha sonraki aşamalarıyla ilgili olarak kullanıcıya genel ve seçilen kümelendirme yöntemlerine özgü tavsiyelerde bulunur. viü

Özet (Çeviri)

SUMMARY i AN EXPERT SYSTEM APPROACH TO THE EARLY STAGES OF CELLULAR MANUFACTURING SYSTEM DESIGN The purpose of this research is to support and make easier the solution of the problem of Cellular Manufacturing System Design for early stages (part families and machine grouping). Cellular Manufacturing System (CMS) Design is a very difficult study, because too many manufacturing data should be evaluated and manufacturing system conditions are rather variable from one company to another. Also, most of these data are ill-structured and requires expert knowlegde and experince. The Clustering Algorithms are data dependent and does not exist the best one for every type company. The most important stages are early stages of CMS design. (Sundaram ve Lian, 1990). For these reasons, an expert system approach is developed for early stages of CMS design. Cellular Manufacturing (CM), has an important role in manufacturing industry. CM gained a great popularity after Flexible Manufacturing Systems (FMS) were introduced. However, it can be applied in conventional manufacturing. The concept of Group Technology (GT) leads to formation of part families that are similar in design or manufacture. The machines required to produce a part family are grouped together in a manufacturing cell. Cell formation is the first, and the most important phase of GT application. This initial decision influences all other decisions involved in the design of Cellular Manufacturing (CM) Systems. The following methods for cell formation in GT have been studied mainly:. Production Flow Analysis (Burbidge 1971).. Cluster analysis (Mc Auley 1972).. Graph theoretic approach (Rajagopalan and Batra 1975).. Rank Order Clustering method, ROC, (King and Nakornchai 1982).. Machine-component cell formation in group technology: MACE ( Waghodekar, and S ahu ).. ZODIAC: an algorithm for concurrent formation of part families and machine cells (Chandrasekharan, and Rajagopalan 1987).. The generalized group technology concept (Kusiak 1987. a).. An assignment model for the part-families problem in group technology by Srinivasan, Narendran, and Mahadevan (1990).. An integer programming approach proposed by Gunasingh and Lashkari (1989).. Polyhedral dynamics as a tool for machine-part group formation, i.e. q analysis (Robinson and Duckstein 1986).. The similarity coefficient method subsequently employed by Seifoddini (1989).. An approach for designing cellular manufacturing systems (Sundaram and Lian 1990).. An efficient heuristic proposed by Harhalakis, Nagi, and Proth (1990). Compared with functional layout in a job shop, the results of successful application of GT in manufacturing ensure as follows: IX1. Reduction in machine set-up times 2. Smooth work flow due to the possibility of small batch manufacturing 3. Reduced work-in-process inventories 4. Reduction in throughput times 5. Reduction in handling costs 6. Simplification in production planning and control 7. Reduction in tooling investment 8. unification of responsibilities 9. Increased quality 10. Improved human relations 11. Reduced paper work 12. Automation. GT is the first evolutionary step in automation (Burbidge, 1992). All of these advantages are mainly realized by simplifying the work flow in machine shop applied to 6T principles. However, some disadvantages occur in GT. The major disadvantages are: 1. Extensive analysis, along with the accompanying difficulties in the selection of the appropriate methodology, is required for the formation of part families and machine groups. 2. The purchase of additional machines may be required in order to realize the gains inherent in the cell structure. 3. Chu (1989) concluded that the multitude of clustering algorithms, clustering criteria, and measures of performance make it difficult to evaluate and select a proper or better clustering method. 4. Different clustering criteria may produce different grouping results even if the same algorithm and data are used. 5. Although minimizing the number of exceptional elements has been widely used as a measure, their appropriateness is not proven. 6. The determination of an optimal number of manufacturing cells is a controversial issue. An expert system approach is developed to reduce these disadvantages and to select the most appropriate clustering algorithm(s) and get advice including data collection and other preliminary works for the user. This expert system approach is called EXCLUSTER. System overview The information module of EXCLUSTER enables its user to get help about Group Technology, the current clustering algorithms and the expert system, any time. Further, it provides necessary suggestions for an effective use of the expert system. After this preliminary work, the user and the system are ready to work together. The data input module acquires the facts of the company, and determines what kind of data set has to be entered to the system according to the specific features of the company. At this stage, the inference engine examines all clustering algorithms, and eliminates the inefficient algorithms for the company. For the left clustering algorithms, the necessary data set is entered to the system by the user. Since some clustering algorithms are only represented by their facts, the clustering results cannot be obtained. However, the algorithms that have computer programs, are executed individually and grouping efficiencies are calculated. For thealgorithms with satisfactory efficiencies, EXCLUSTER evaluates their performance by asking some options to the user interactively and modifying the performance criterion. A multiple criteria module using the BLECTRE method (Benayoun, 1966) selects the most proper solution or solutions for the company and may give suggestions for the current and late stages of designing cellular manufacturing systems. After computing grouping efficiencies and the other related efficiency measures, the inference engine might result not to apply cellular manufacturing systems. It means that the company is not convenient for Group Technology. The forward chaining production system was selected, because, as noted by Brownston (1985), forward chaining systems are most appropriate when there are several equally acceptable goal states and a single initial state. Cell formation algorithms are in this type systems and nothing is known about the data set. EXCLUSTER presently contains approximately 120 rules and 20 clustering algorithms. Because of its dynamic structure, new rules can be added to its knowledge base. EXCLUSTER consists of five components:. Data base. Knowledge base. Inference engine. Clustering algorithms. ELECTRE multiple criteria decision making module. Database The database contains the information about the company that is listed below.. the number of machines for each type,. the number of machine types,. annual depreciation costs of machines,. the machines that should be placed closely,. the machines that should be placed apart from each other,. the machine-part incidence matrix of the manufacturing system,. the number of parts,. material handling cost: depending upon the weight, size, shape, or other attributes of a part.. the annual demands of the parts,. processing times of the parts for each machine,. the parts that can be subcontracted,. maximum and minimum numbers of the machines in a cell,. maximum and minimum numbers of the cells,.. the variability of the annual demand from one year to another for the parts,. the variability of the demand from one part to another,. the variability of the processing times from one part to another,. the variability of the annual depreciation costs of the machines,. alternative routes of parts, In order to achieve a reasonable solution, it is not necessary to have all these facts of the company for EXCLUSTER. However, the more data is entered, the better solution and the more suggestions will be accepted. The contents of the database of EXCLUSTER are either provided by the user as input data or generated by the expert system. xiOne of the further issues is to collect some of the data such as route information, and processing times of parts by means of an MRP II package. Because, one of the reasons most companies hesitate to explore the possibility of Group Technology application is the time and effort involved in collecting part and work-centre details. Knowledge base EXCLUSTER is a rule-based expert system. Kusiak and Chen (1988.b) noted that rule-based expert systems are being more frequently applied in manufacturing planning. This is due to the IF-THEN rules are easily acceptable because they are similar to the common sense logic. Each production rule has the following format: (Rule number ) (IF conditions THEN actions ) (the reference of the rule) ) In addition, a detailed description of the rule is given to the user by expert system, any time. Though the knowledge base of this version of EXCLUSTER includes about 120 rules, a few rules are not used by the system. These are stored for the later probable use. Fuzzy Logic and Variability Concepts Fuzzy Logic was used to determine the characteristics of some data and the facts. For example, to determine the variability of processing times, both the Coefficient Of Variation (COV) and Fuzzy Logic are used. The Fuzzy Set Theory and the Variabilty Concept are very practical to convert the numeric values of production data to the linguistic terms in a manufacturing system. In this section, how these data are gathered and converted are explained. The application of fuzzy set theory and variability from statistics (Sanders 1990, p. 150-151) in the valve manufacturing company are discussed in the“Application”Section. Variability The variability plays an essential role in the behavior and performance of a manufacturing system. Variable processing times are only one source of variation in early stages of Cellular Manufacturing Systems for this application. Other sources are:. demand of the parts,. demand in time,. the similarity of process sequence of the parts,. processing times of the parts on the same machine,. volume or weight of the parts,. setup times from one machine to another for the same part. xuFuzzy Set Theory“The motivation for the use of words or sentences rather than numbers ^is that linguistic characterizations are, in general, less spesif ic than numerical ones”(Zadeh 1973). A special kind of fuzzy sets are linguistic variables. These are variables which can take words or terms of words as values. By the use of linguistic variables we are able to model and process qualitative and verbally expressed statements in mathematical models. The membership functions used here are discrete. The production rules in EXCLUSTER are acquired from one expert in group technology, the literature and the author of this research. The knowledge base related to the clustering algorithms consists of two classes of production rules: The eliminating rules, The priority rules. The eliminating rules rejects improper solutions of clustering algorithms in the knowledge base. For example, the rule: RULE: IF VARIABILITY OF PROCESSING TIMES FROM ONE PART TO ANOTHER IS HIGH THEN CHOOSE THE ALGORITHMS TAKING INTO ACCOUNT PROCESSING TIMES. EXPERT. eliminates improper algorithms in the solution set. Other examples of the eliminating rules are given below: RULE: IF THE NUMBER OF MACHINES IS GREATER THAN 30 AND THE NUMBER OF PARTS IS GREATER THAN 60 THEN THE MATRIX IS LARGE. (KUSIAK, A. 1988 a). RULE: IF THE MATRIX IS LARGE THEN ELIMINATE MATHEMATICAL PROGRAMMING CLUSTERING APPROACHES. (KUSIAK, A. 1988 a). In this case, heuristic algorithms are only selected by inference engine of EXCLUSTER and mathematical programming techniques are omitted. RULE: IF THE WORKING RANGE OF CTsjES LESS THAN 0.2 OR THE WORKING RANGE OF OgjTS GREATER THAN 0.35 THEN REJECT THE ALGORITHM FOR THIS COMPANY. (CHANDRASEKHARAN, M.P. RAJAGOPALAN, R. 1989). xiii-Chandrasekharan and Rajagopalan stated that data are ill-structured, too sparse or too dense if they fall outside this range and concluded that such matrices can safely be rejected as unsuitable for GT applications. An example for the priority rule of clustering algorithms is : RULE: IP THE COMPUTATIONAL EFFICIENCIES OF THE ALGORITHMS ARE IMPORTANT OR VERY IMPORTANT AND LONG EXECUTION TIMES ARE EXPECTED THEN GIVE FIRST PRIORITY THE ALGORITHMS OF WHICH EXECUTION TIMES ARE RELATIVELY SHORT. EXPERT. This rule gives a priority to the algorithms of which execution times are short. However, it does not eliminate any algorithm. Another example for these kind of rules is the frequency of using of the algorithm by the researchers and practitioners. These kind of rules are also used to get advice and to select the proper algorithm more easily and quickly. Some rules are not directly related to the cell formation techniques. The examples are below : RULE: IF THE MATERIAL HANDLING COST IS TOO HIGH THEN ASSIGN A VALUE TO THE INTER-CELL FLOW EFFICIENCY GREATER THAN 0.50. EXPERT Inference Engine After selecting candidate solutions of clustering algorithms, the computer programs of algorithms are executed and cells are formed by every algorithm. If the algorithm is improper for group formation, then it is eliminated and for the rest, ELECTRE Method is applied. The efficiency measures of the solutions of the clustering algorithms are two parts of the grouping efficiency defined in Chandrasekharan and Rajagopalan (1989), minimum extra machines, the satisfaction of important customers, material handling cost, the number of exceptional elements, etc. ELECTRE multiple criteria decision making module ELECTRE method is selected for multiple criteria decision making module, because it is very similar to fuzzy logic and discrete. Application The method was applied to a valve manufacturing company. Problem size is 42 machines and 50 parts. The results are satisfactory and reasonable. The system finds one the best solution and a good solution and also suggests useful information for CMS Design. The company mainly produces engine valves and has the largest production capacity in the country. Despite approximately 400 type valves being produced, 95 % of the annual production is only for 50 parts. For this reason, the number of parts is 50. xivThe expert system does not use fuzzy logic only to derive the facts of production system but also in the ELECTRE multiple criteria decision making modul. Fuzzy Logic and ELECTRE method can be used together succesfully. Conclusion In this thesis a rule-based expert system approach for solving the problem of grouping machines and forming part families was explained. EXCLUSTER does not only find a solution for the company but also gets some valuable information by using a number of clustering algorithms. The purpose of this research was to support and make easier the solution of the problem of Cellular Manufacturing System Design for early stages. The expert system approach eliminates unnecessary manufacturing data and saves time. Since manufacturing system conditions are rather variable from one company to another, it evaluates these different ones by means of standardized efficiency measures. The Clustering Algorithms using 0-1 data and the others that use other data such as production volume, production time, are not satisfactory alone. They need Expert Knowledge and System Analysis Approach. The developed expert system approach is general and can be used for solving the problems of other scientific areas by just changing the rules and efficiency measures of the expert system. The approach is flexible. In time, advanced or new clustering algorithms and new rules can be added or updated to the database. The approach is very valuable for the late stages of CMS Design. For instance The ELECTRE Multiple Criteria Module can be used for choosing the best layout. This is a strong argument that the reduction of variability is an essential objective on the way to increased manufacturing efficieny. Both the variability and fuzzy logic concepts are useful for expert system design. Without fuzzy logic, it is very difficult to obtain the facts of production system when designing a Cellular Manufacturing System. These concepts convert qualitative data to quantitative and the opposite. The data is also normalized by using variability. When using expert system, the decisions can be made easily and quickly. Cellular Manufacturing Systems should not be designed without taking into account the related systems. The applications and the experience show that Total Quality Management and Just In Time Systems should be integrated with CMS. Further research can be made in these fields: A Simulation Expert System (Sabuncuoğlu ve Hommertzheim 1988) is very useful for minimization of work-in-progress in the cells, Just In Time applications. A knowledge based system can be used for scheduling and control of a manufacturing cell (Sepulveda ve Sullivan 1988). xv

Benzer Tezler

  1. Tez No

    896525

    Spiral presleme makinesi tasarım ve imalatı

    Design and manufacturing of spiral pressing machine

    GÖKHAN KÜÇÜK

    Yüksek Lisans

    Türkçe

    Türkçe

    2024

    Makine MühendisliğiKocaeli Üniversitesi

    Makine Mühendisliği Ana Bilim Dalı

    DR. ÖĞR. ÜYESİ ALP EREN ŞAHİN

  2. Tez No

    144681

    Hücresel imalatın GERT tekniği ile planlaması ve bir uygulama denemesi

    Cellular manufacturing with GERT technique and an aplication trial

    PELİN MİNSİN

    Yüksek Lisans

    Türkçe

    Türkçe

    2004

    İşletmeBalıkesir Üniversitesi

    İşletme Ana Bilim Dalı

    PROF. DR. ADEM ÇABUK

  3. Tez No

    416984

    U tipi montaj hattı dengeleme problemlerinde üretim oranının artırılmasına yönelik yeni çözüm yaklaşımları

    New solution approaches for U shaped assembly line balancing problems with maximizing production rate

    MURAT ŞAHİN

    Yüksek Lisans

    Türkçe

    Türkçe

    2016

    Endüstri ve Endüstri MühendisliğiGazi Üniversitesi

    Endüstri Mühendisliği Ana Bilim Dalı

    YRD. DOÇ. DR. TALİP KELLEGÖZ

  4. Tez No

    357219

    Üretim akış analizi ve hücresel imalat uygulaması

    Production flow analysis and cellular manufacturing application

    BEKİR VOLKAN AĞCA

    Yüksek Lisans

    Türkçe

    Türkçe

    2014

    Makine MühendisliğiBaşkent Üniversitesi

    Makine Mühendisliği Ana Bilim Dalı

    YRD. DOÇ. DR. YUSUF TANSEL İÇ

  5. Tez No

    95886

    A Simulation based comparison of functional and cellular layouts in a manufacturing plant

    Bir imalat işletmesinde fonksiyonel ve hücresel yerleşimlerin benzetim yardımıyla karşılaştırılması

    MERT ALTINKILINÇ

    Yüksek Lisans

    İngilizce

    İngilizce

    2000

    Endüstri ve Endüstri MühendisliğiDokuz Eylül Üniversitesi

    Endüstri Mühendisliği Ana Bilim Dalı

    DR. BAYRAM ALİ SU

Geri Dön