This paper treats multi-objective problem for manufacturing process design. A purpose of the process design is to decide combinations of work elements assigned to different work centers. Multiple work elements are ord...This paper treats multi-objective problem for manufacturing process design. A purpose of the process design is to decide combinations of work elements assigned to different work centers. Multiple work elements are ordinarily assigned to each center. Here, infeasible solutions are easily generated by precedence relationship of work elements in process design. The number of infeasible solutions generated is ordinarily larger than that of feasible solutions generated in the process. Therefore, feasible and infeasible solutions are located in any neighborhood in solution space. It is difficult to seek high quality Pareto solutions in this problem by using conventional multi-objective evolutional algorithms. We consider that the problem includes difficulty to seek high quality solutions by the following characteristics： （1） Since infeasible solutions are resemble to good feasible solutions, many infeasible solutions which have good values of objective functions are easily sought in the search process, （2） Infeasible solutions are useful to select new variable conditions generating good feasible solutions in search process. In this study, a multi-objective genetic algorithm including local search is proposed using these characteristics. Maximum value of average operation times and maximum value of dispersion of operation time in all work centers are used as objective functions to promote productivity. The optimal weighted coefficient is introduced to control the ratio of feasible solutions to all solutions selected in crossover and selection process in the algorithm. This paper shows the effectiveness of the proposed algorithm on simple model.展开更多
In this paper, the structure of infeasible solutions to Job Shop Scheduling Problem (JSSP) is quantitatively analyzed, and a necessary and sufficient condition of the deadlock for JSSP is also given. For a simple JSSP...In this paper, the structure of infeasible solutions to Job Shop Scheduling Problem (JSSP) is quantitatively analyzed, and a necessary and sufficient condition of the deadlock for JSSP is also given. For a simple JSSP with 2 machines and N jobs, a formula for calculating the infeasible solutions is proposed, which shows that the infeasible solution possesses the majority of search space and only those heuristic algorithms which do not produce infeasible solutions are valid.展开更多
文摘This paper treats multi-objective problem for manufacturing process design. A purpose of the process design is to decide combinations of work elements assigned to different work centers. Multiple work elements are ordinarily assigned to each center. Here, infeasible solutions are easily generated by precedence relationship of work elements in process design. The number of infeasible solutions generated is ordinarily larger than that of feasible solutions generated in the process. Therefore, feasible and infeasible solutions are located in any neighborhood in solution space. It is difficult to seek high quality Pareto solutions in this problem by using conventional multi-objective evolutional algorithms. We consider that the problem includes difficulty to seek high quality solutions by the following characteristics： （1） Since infeasible solutions are resemble to good feasible solutions, many infeasible solutions which have good values of objective functions are easily sought in the search process, （2） Infeasible solutions are useful to select new variable conditions generating good feasible solutions in search process. In this study, a multi-objective genetic algorithm including local search is proposed using these characteristics. Maximum value of average operation times and maximum value of dispersion of operation time in all work centers are used as objective functions to promote productivity. The optimal weighted coefficient is introduced to control the ratio of feasible solutions to all solutions selected in crossover and selection process in the algorithm. This paper shows the effectiveness of the proposed algorithm on simple model.
基金Supported by The National Natural Science Foundation of China ( No.794 3 0 0 2 2 )
文摘In this paper, the structure of infeasible solutions to Job Shop Scheduling Problem (JSSP) is quantitatively analyzed, and a necessary and sufficient condition of the deadlock for JSSP is also given. For a simple JSSP with 2 machines and N jobs, a formula for calculating the infeasible solutions is proposed, which shows that the infeasible solution possesses the majority of search space and only those heuristic algorithms which do not produce infeasible solutions are valid.
