Evolutionary Algorithms for Multi-Objective Flexible Job Shop Scheduling Problems

• Main Authors: Wen-Chou Huang, 黃文洲
• Other Authors: 徐旭昇
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/33530092299336882539

碩士 === 元智大學 === 工業工程與管理學系 === 98 === Scheduling problems occur in many production systems, since it is necessary to distribute and sequence the work among many jobs. A job may consist of one or more operations. When scheduling jobs, the management generally does not focus solely on one objective. Roy (1985) pointed out that taking several criteria into account enables us to provide the decision maker with a more realistic solution. This study focuses on solving the multi-objective flexible job shop scheduling problem (MO-FJSSP) with three minimization objectives – total tardiness, total machine workload, and critical machine workload, subject to a production cycle time. The FJSSP is an extension of unrelated parallel machine scheduling and job shop scheduling problems, and has attracted intensive interest from researchers and practitioners. Many meta-heuristics for the FJSSP have been proposed in the literature, but most of them are based on the following approach: determine machine-assignment, and then decide job-sequence (top-down). Our study proposes a different approach: job-sequence first, and machine-assignment second (bottom-up). To solve the MO-FJSSP, a novel local search termed PR-FA (path-relinking front-advancing) is incorporated into two multi-objective evolutionary algorithms, NSGAII and SPEA2. In PR-FA, path-relinking operates in the decision space using similarity for distance measure. The resulting solution representation lists are then mapped into the objective space, and then an effective direction towards the Pareto front is chosen to make further refinements. Three performance metrics – generalizational distance (GD), spread, and hypervolume (HV) – are used to evaluate various algorithms. The numerical results reveal that the bottom-up approach with PR-FA local search surpasses the top-down/bottom-up approach with other local search methods. In addition, a sensitivity analysis is presented using ?-constraint method on various production cycles.