GT yöntemlerinin sınıflandırması, performans ölçütleri, üretimle ilgili verileri kullanan yeni yöntemlere örnekler ve genetik algoritmalar

Taxonomy of GT methods, performance measures,some new GT methods that is able to incorporate pertinent manufacturing data and genetic algorithms

Tez No: 66838
Yazar: HATİCE DERİCİ
Danışmanlar: DOÇ. DR. M. BÜLENT DURMUŞOĞLU
Tez Türü: Yüksek Lisans
Konular: Endüstri ve Endüstri Mühendisliği, Industrial and Industrial Engineering
Anahtar Kelimeler: Belirtilmemiş.
Yıl: 1997
Dil: Türkçe
Üniversite: İstanbul Teknik Üniversitesi
Enstitü: Fen Bilimleri Enstitüsü
Ana Bilim Dalı: Endüstri Mühendisliği Ana Bilim Dalı
Bilim Dalı: Belirtilmemiş.
Sayfa Sayısı: 300

Özet

plans for the parts and additional copies of machines) improves this diagonalization. This aspect has not been adequately dealt with in literature. The model in the 10th Section, which is developed by Adil&Rajamani& Strong (1996) combines the evaluation procedure by considering the minimization of a weighted sum of the voids and the exceptional elements in the objective. This leads to better identification of groupings in the existing data. Also, in the model, changing weights for void and exceptional elements gives the designer the flexibility of forming large loose cells (more voids but less exceptional elements) or small tight cells (less voids ands more exceptional elements). The model has been illustrated with numerical examples. The optimal solutions for these examples are obtained by solving the linearized version of the model. For efficient solution of larger problems a simulated annealing algorithm is used. Genetic algorithms (GA) are search algorithms based on the mechanics of natural selection and natural genetics. The fundamental principle of genetic algorithm is that in each step the string which is a higher objective function value in the old generation has a higher probability of contributing one or more off springs to the new generation. A simple genetic algorithm that yields good results in many practical problems as composed of three main operators: i.e. reproduction, crossover and mutation. Genetic algorithms manipulate decision or control variable presentations at the string level to exploid similarities among high performance strings. Other methods usually deal with functions and their control variables directly. Because genetic algorithms work from a population wide sample points, the probability of reaching a false peak reduced. The transition rules of genetic algorithms are stochastic via sampling; most other methods are deterministic transition rules. Therefore, genetic algorithms are more suitable for multiple-peaks funntions. Although in theory, genetic algorithms cannot guarantee to attain the best solution, in practise, non-inferior solutions can be obtained and sometimes it is possible to get the best solution. Because of this reason, genetic algorithms have been applied in a variety of engineering and optimization applications since 1981, for instance, gas pipeline optimization, oil pump-pipeline system, engineering optimization, blind knapsack problem, job shop scheduling and the travelling salesman problem. The 1 1th Section considers of a new approach of optimizing GT part family formation by Genetic Algorithm {GA) which is based on the principles of natural selection. A genetic algorithm model combining a coding design, a penalty factor and a scaling operation was incorporated (by Hon&Chi, 1994) for this purpose. For comparative purpose, this new genetic algorithm was applied to a set of four different part-families of different sizes, investigated by Ventura,etc. This set of data was used because they allow detailed step by step comparison as each new generation is formed. Secondly, this also provides an opportunity to evaluate genetic algorithm approach against three other heuristic methods, i.e., a subgradient (SUB) method and two other methods based on a simple (KKV1) and a more sophisticated graph theoretic method (KKV1+2). Results generated from four case examples and comparisons with three other heuristic methods demonstrates that the GA approach provides a powerfull and effective numerical tool for the optimization of GT part families, on the other hand, experience gained in this investigation showed that the design of coding of string requires a considerable amount of skill and thought. This is because of the fact that coding could produce a considerable impact on the optimality of results as well as the computational time. XXI

Özet (Çeviri)

In the Section 12, there is a new GA (developed by Billo, 1994) that is able to incorporate other pertinent manufacturing data such as product demand. Results of the comparison of the GA solutions with the SLCA solutions are evaluated. Upon observing this results, it is attempted to identify the reasons for the discrepancies between the two approaches. It is hypothesized that the GA formulation is outperforming the SLCA formulation due to some fallacy in the approach for incorporating a part into a family. This fallacy appears to be the approach used by cluster analysis procedures where a part is incorporated into a larger family based on the best pairwise similarity coefficent between individual parts. This approach incorrectly assumes transitivity to be true for pairwise part similarities: if Part A is similar to Part B, and Part B is similar to Part C, then Part A is similar to Part C. In actuality, it is found this approach for part family incorporation can sometimes lead to less than optimum groupings as measured by lower groupability indices(G). This approach uses a groupability index which is more efficient than pairwise similarity coefficents used by the other methods. In addition the GA techniques collected in this thesis, I investigated a new approcah (Venugopal&Nanendran, 1992) which proposes a bi-criteria mathematical model with a solution procedure based on a GA. This algorithm aims both minimizing the volume of intercell moves and minimizing the total within cell load variation by. managing processing times of parts on machines, available times on machines and production requirements of components in a given period of time. Trials on a sample problem suggest that the proposed algorithm can be a powerful tool that can be gainfully employed in a cellular manufacturing environment. The algorithm is found to be effective in offering collection of satisfactory solutions, which is essential in a multi-objective environment, to be enable the decision maker to choose the best alternative. It is independent of the nature of the objective functions. The algorithm is inherently parallel and is capable of super linear speed-up in multi-processor systems. With the avaüibility of parallel computers, this algorithm will be particularly usefull in solving part-family problem in complex, large scale FMS environments. Depending on once more additional reference, we can say it is possible to optimize neural networks with GAs. (NeuroGENESYS) (Harp&Samad, 1992). NeuroGENESYS has produced exciting results in several small-scale experiments and we can hope to conduct realistic evaluations of this approach in the near future. Finally, in the Section 13, some important issues which affect the success of CFT to be used were underlined once more again. Additionally,, all performance measures and all new CFTs collected in the thesis are evaluated, compared with others and summarized in 2 tables (Table 13.1. and Table 13.2.) by me. The unique objective to gather so independent CFTs was to provide a new and widely insight for examining the real applications robustly and to attempt a whole comparison in this field as just one of some rare studies XXllplans for the parts and additional copies of machines) improves this diagonalization. This aspect has not been adequately dealt with in literature. The model in the 10th Section, which is developed by Adil&Rajamani& Strong (1996) combines the evaluation procedure by considering the minimization of a weighted sum of the voids and the exceptional elements in the objective. This leads to better identification of groupings in the existing data. Also, in the model, changing weights for void and exceptional elements gives the designer the flexibility of forming large loose cells (more voids but less exceptional elements) or small tight cells (less voids ands more exceptional elements). The model has been illustrated with numerical examples. The optimal solutions for these examples are obtained by solving the linearized version of the model. For efficient solution of larger problems a simulated annealing algorithm is used. Genetic algorithms (GA) are search algorithms based on the mechanics of natural selection and natural genetics. The fundamental principle of genetic algorithm is that in each step the string which is a higher objective function value in the old generation has a higher probability of contributing one or more off springs to the new generation. A simple genetic algorithm that yields good results in many practical problems as composed of three main operators: i.e. reproduction, crossover and mutation. Genetic algorithms manipulate decision or control variable presentations at the string level to exploid similarities among high performance strings. Other methods usually deal with functions and their control variables directly. Because genetic algorithms work from a population wide sample points, the probability of reaching a false peak reduced. The transition rules of genetic algorithms are stochastic via sampling; most other methods are deterministic transition rules. Therefore, genetic algorithms are more suitable for multiple-peaks funntions. Although in theory, genetic algorithms cannot guarantee to attain the best solution, in practise, non-inferior solutions can be obtained and sometimes it is possible to get the best solution. Because of this reason, genetic algorithms have been applied in a variety of engineering and optimization applications since 1981, for instance, gas pipeline optimization, oil pump-pipeline system, engineering optimization, blind knapsack problem, job shop scheduling and the travelling salesman problem. The 1 1th Section considers of a new approach of optimizing GT part family formation by Genetic Algorithm {GA) which is based on the principles of natural selection. A genetic algorithm model combining a coding design, a penalty factor and a scaling operation was incorporated (by Hon&Chi, 1994) for this purpose. For comparative purpose, this new genetic algorithm was applied to a set of four different part-families of different sizes, investigated by Ventura,etc. This set of data was used because they allow detailed step by step comparison as each new generation is formed. Secondly, this also provides an opportunity to evaluate genetic algorithm approach against three other heuristic methods, i.e., a subgradient (SUB) method and two other methods based on a simple (KKV1) and a more sophisticated graph theoretic method (KKV1+2). Results generated from four case examples and comparisons with three other heuristic methods demonstrates that the GA approach provides a powerfull and effective numerical tool for the optimization of GT part families, on the other hand, experience gained in this investigation showed that the design of coding of string requires a considerable amount of skill and thought. This is because of the fact that coding could produce a considerable impact on the optimality of results as well as the computational time. XXIIn the Section 12, there is a new GA (developed by Billo, 1994) that is able to incorporate other pertinent manufacturing data such as product demand. Results of the comparison of the GA solutions with the SLCA solutions are evaluated. Upon observing this results, it is attempted to identify the reasons for the discrepancies between the two approaches. It is hypothesized that the GA formulation is outperforming the SLCA formulation due to some fallacy in the approach for incorporating a part into a family. This fallacy appears to be the approach used by cluster analysis procedures where a part is incorporated into a larger family based on the best pairwise similarity coefficent between individual parts. This approach incorrectly assumes transitivity to be true for pairwise part similarities: if Part A is similar to Part B, and Part B is similar to Part C, then Part A is similar to Part C. In actuality, it is found this approach for part family incorporation can sometimes lead to less than optimum groupings as measured by lower groupability indices(G). This approach uses a groupability index which is more efficient than pairwise similarity coefficents used by the other methods. In addition the GA techniques collected in this thesis, I investigated a new approcah (Venugopal&Nanendran, 1992) which proposes a bi-criteria mathematical model with a solution procedure based on a GA. This algorithm aims both minimizing the volume of intercell moves and minimizing the total within cell load variation by. managing processing times of parts on machines, available times on machines and production requirements of components in a given period of time. Trials on a sample problem suggest that the proposed algorithm can be a powerful tool that can be gainfully employed in a cellular manufacturing environment. The algorithm is found to be effective in offering collection of satisfactory solutions, which is essential in a multi-objective environment, to be enable the decision maker to choose the best alternative. It is independent of the nature of the objective functions. The algorithm is inherently parallel and is capable of super linear speed-up in multi-processor systems. With the avaüibility of parallel computers, this algorithm will be particularly usefull in solving part-family problem in complex, large scale FMS environments. Depending on once more additional reference, we can say it is possible to optimize neural networks with GAs. (NeuroGENESYS) (Harp&Samad, 1992). NeuroGENESYS has produced exciting results in several small-scale experiments and we can hope to conduct realistic evaluations of this approach in the near future. Finally, in the Section 13, some important issues which affect the success of CFT to be used were underlined once more again. Additionally,, all performance measures and all new CFTs collected in the thesis are evaluated, compared with others and summarized in 2 tables (Table 13.1. and Table 13.2.) by me. The unique objective to gather so independent CFTs was to provide a new and widely insight for examining the real applications robustly and to attempt a whole comparison in this field as just one of some rare studies XXllplans for the parts and additional copies of machines) improves this diagonalization. This aspect has not been adequately dealt with in literature. The model in the 10th Section, which is developed by Adil&Rajamani& Strong (1996) combines the evaluation procedure by considering the minimization of a weighted sum of the voids and the exceptional elements in the objective. This leads to better identification of groupings in the existing data. Also, in the model, changing weights for void and exceptional elements gives the designer the flexibility of forming large loose cells (more voids but less exceptional elements) or small tight cells (less voids ands more exceptional elements). The model has been illustrated with numerical examples. The optimal solutions for these examples are obtained by solving the linearized version of the model. For efficient solution of larger problems a simulated annealing algorithm is used. Genetic algorithms (GA) are search algorithms based on the mechanics of natural selection and natural genetics. The fundamental principle of genetic algorithm is that in each step the string which is a higher objective function value in the old generation has a higher probability of contributing one or more off springs to the new generation. A simple genetic algorithm that yields good results in many practical problems as composed of three main operators: i.e. reproduction, crossover and mutation. Genetic algorithms manipulate decision or control variable presentations at the string level to exploid similarities among high performance strings. Other methods usually deal with functions and their control variables directly. Because genetic algorithms work from a population wide sample points, the probability of reaching a false peak reduced. The transition rules of genetic algorithms are stochastic via sampling; most other methods are deterministic transition rules. Therefore, genetic algorithms are more suitable for multiple-peaks funntions. Although in theory, genetic algorithms cannot guarantee to attain the best solution, in practise, non-inferior solutions can be obtained and sometimes it is possible to get the best solution. Because of this reason, genetic algorithms have been applied in a variety of engineering and optimization applications since 1981, for instance, gas pipeline optimization, oil pump-pipeline system, engineering optimization, blind knapsack problem, job shop scheduling and the travelling salesman problem. The 1 1th Section considers of a new approach of optimizing GT part family formation by Genetic Algorithm {GA) which is based on the principles of natural selection. A genetic algorithm model combining a coding design, a penalty factor and a scaling operation was incorporated (by Hon&Chi, 1994) for this purpose. For comparative purpose, this new genetic algorithm was applied to a set of four different part-families of different sizes, investigated by Ventura,etc. This set of data was used because they allow detailed step by step comparison as each new generation is formed. Secondly, this also provides an opportunity to evaluate genetic algorithm approach against three other heuristic methods, i.e., a subgradient (SUB) method and two other methods based on a simple (KKV1) and a more sophisticated graph theoretic method (KKV1+2). Results generated from four case examples and comparisons with three other heuristic methods demonstrates that the GA approach provides a powerfull and effective numerical tool for the optimization of GT part families, on the other hand, experience gained in this investigation showed that the design of coding of string requires a considerable amount of skill and thought. This is because of the fact that coding could produce a considerable impact on the optimality of results as well as the computational time. XXIIn the Section 12, there is a new GA (developed by Billo, 1994) that is able to incorporate other pertinent manufacturing data such as product demand. Results of the comparison of the GA solutions with the SLCA solutions are evaluated. Upon observing this results, it is attempted to identify the reasons for the discrepancies between the two approaches. It is hypothesized that the GA formulation is outperforming the SLCA formulation due to some fallacy in the approach for incorporating a part into a family. This fallacy appears to be the approach used by cluster analysis procedures where a part is incorporated into a larger family based on the best pairwise similarity coefficent between individual parts. This approach incorrectly assumes transitivity to be true for pairwise part similarities: if Part A is similar to Part B, and Part B is similar to Part C, then Part A is similar to Part C. In actuality, it is found this approach for part family incorporation can sometimes lead to less than optimum groupings as measured by lower groupability indices(G). This approach uses a groupability index which is more efficient than pairwise similarity coefficents used by the other methods. In addition the GA techniques collected in this thesis, I investigated a new approcah (Venugopal&Nanendran, 1992) which proposes a bi-criteria mathematical model with a solution procedure based on a GA. This algorithm aims both minimizing the volume of intercell moves and minimizing the total within cell load variation by. managing processing times of parts on machines, available times on machines and production requirements of components in a given period of time. Trials on a sample problem suggest that the proposed algorithm can be a powerful tool that can be gainfully employed in a cellular manufacturing environment. The algorithm is found to be effective in offering collection of satisfactory solutions, which is essential in a multi-objective environment, to be enable the decision maker to choose the best alternative. It is independent of the nature of the objective functions. The algorithm is inherently parallel and is capable of super linear speed-up in multi-processor systems. With the avaüibility of parallel computers, this algorithm will be particularly usefull in solving part-family problem in complex, large scale FMS environments. Depending on once more additional reference, we can say it is possible to optimize neural networks with GAs. (NeuroGENESYS) (Harp&Samad, 1992). NeuroGENESYS has produced exciting results in several small-scale experiments and we can hope to conduct realistic evaluations of this approach in the near future. Finally, in the Section 13, some important issues which affect the success of CFT to be used were underlined once more again. Additionally,, all performance measures and all new CFTs collected in the thesis are evaluated, compared with others and summarized in 2 tables (Table 13.1. and Table 13.2.) by me. The unique objective to gather so independent CFTs was to provide a new and widely insight for examining the real applications robustly and to attempt a whole comparison in this field as just one of some rare studies XXll

Benzer Tezler

Tez No
28876
A Part family formation algorithm in GT environment using a multi-objective cluster analysis
GT ortamında çok amaçlı sınıflandırma analizi ile bir parça oluşturma algoritması
HAKAN OKAN BALKÖSE
Yüksek Lisans
İngilizce
1993
Endüstri ve Endüstri Mühendisliği İhsan Doğramacı Bilkent Üniversitesi
Endodonti Ana Bilim Dalı
YRD. DOÇ. DR. SELİM AKTÜRK
Tez No
495307
Splice site prediction using machine learning
Makine öğrenmesi kullanarak uçbirleştirme yeri tahmini
ELHAM PASHAEI
Doktora
İngilizce
2017
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol Yıldız Teknik Üniversitesi
Bilgisayar Mühendisliği Ana Bilim Dalı
PROF. DR. NİZAMETTİN AYDIN
Tez No
56024
İleri imalat teknolojilerinin ekonomik analizi ve esneklik faktörünün sayısallaştırılmasına bulanık kümeler yaklaşımı
Başlık çevirisi yok
CENGİZ KAHRAMAN
Doktora
Türkçe
1996
Endüstri ve Endüstri Mühendisliği İstanbul Teknik Üniversitesi
PROF.DR. ETHEM TOLGA
Tez No
676745
Makine öğrenmesi ile gemi makinelerinin hata analizi
Fault analysis of ship machines with machine learning
FUNDA KAYA İNCEİŞÇİ
Yüksek Lisans
Türkçe
2021
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol Marmara Üniversitesi
Bilgisayar Mühendisliği Ana Bilim Dalı
DR. ÖĞR. ÜYESİ AYÇA AK
PROF. DR. VEDAT TOPUZ
Tez No
452293
Üst oblik felci olgularında torsiyon ölçümü ve farklı cerrahi yöntemlerin torsiyona etkisi
Measurement of torsion in superior oblique palsy cases and effects of different surgical methods on torsion
OKTAY DİNER
Tıpta Uzmanlık
Türkçe
2015
Göz Hastalıkları GATA
Göz Hastalıkları Ana Bilim Dalı
PROF. DR. FATİH MEHMET MUTLU

Geri Dön