Genetic Algorithms Embedded with Concise Chromosome Representations for Distributed and Flexible Job-shop Scheduling

• Main Authors: Lu, Po-Hsiang, 盧柏翔
• Other Authors: Wu, Muh-Cherng
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/82tfy6

博士 === 國立交通大學 === 工業工程與管理系所 === 103 === This paper proposes four genetic algorithms (GA_JSA, GA_JS, GA_J, and GA_JCS) for solving distributed and flexible job-shop scheduling (DFJS) problems. A DFJS problem involves three scheduling decisions: (1) job-to-cell assignment, (2) operation-sequencing, and (3) operation-to-machine assignment. Therefore, solving a DFJS problem is essentially a 3-dimensional solution space search problem; each dimension represents a type of decision. The GA_JS algorithm is developed by proposing a new and concise chromosome representation SJOB, which models a 3-dimensional scheduling solution by a 1-dimensional scheme (i.e., a sequence of all jobs to be scheduled). In GA_JS, we develop a decoding method to convert a chromosome into a solution. In addition, given a solution, we use a refinement method to improve the scheduling performance and subsequently use a encoding method to convert the refined a solution into a chromosome. The decoding method is designed to obtain a “good” solution which tends to be load-balanced. In contrast, the refinement and encoding methods of a solution provides a novel way (rather than by genetic operators) to generate new chromosomes, which are herein called shadow chromosomes. Comparing to IGA (De Giovanni and Pezzella, 2010) which is the up-to-date best-performing genetic algorithm in solving DFJS problems, GA_J is distinct in using SJOB chromosome representation; GA_JS is an extension of GA_J by additionally including the development of shadow chromosomes; GA_JSA is an extension of GA_JS by the inclusion of an all-cell refinement procedure which can yield shadow chromosomes with better quality; and GA_JCS is distinct in using SJOB-CELL chromosome representation. GA_JCS is an extension of GA_JS by additionally including the job-to-cell assignment decision. In terms of solution quality, GA_JSA > GA_JS > GA_JCS > GA_J = IGA. Experiment results indicate that the inclusions of shadow chromosomes have a positive effect in solving DFJS problems.