关键词: 柔性作业车间调度, 遗传算法, 多目标调度, 选择策略, 信息熵

Abstract: Multi-objective Flexible Job-shop Scheduling Problem（Mo-FJSP）is a kind of important scheduling problems in complex manufacturing system. Aiming at the problem that the majority of algorithms used for Mo-FJSP tend to have the characteristics of poor stability， insufficient search depth and inability to deeply search a single object in a multi-objective problem，this paper improved the general Genetic Algorithm（GA）by designing a new crossover strategy and deleting the selection operator. On this basis，a method called two-phase GA was proposed. This paper introduced the conception of information entropy to analyze the population optimized by this algorithm and from the respects of minimizing maximal completion time of machines，minimizing workload of the most loaded machine and minimizing the total workload of all machines，several classic examples were tested by this algorithm. The two-phase GA achieved 31 non-dominated answers which can dominate other answers got by other the same kind algorithms and 96 new non-dominated answers. Experimental results show that the proposed algorithm has a good ability of deep searching and can jump out of local optimization while retaining the diversity of population， which shows the reliability of this algorithm.

Key words: Flexible Job-Shop Scheduling Problem (FJSP), Genetic Algorithm (GA), multi-objective scheduling; selection strategy, information entropy