两级排序遗传算法在柔性工作车间调度中的应用  

Genetic Algorithms Based on Two Ranks for Flexible Job Shops Scheduling

作　　者：陈华平[1] 谷峰[1] 古春生[1] 卢冰原[1] 

机构地区：[1]中国科学技术大学信息管理与决策科学系,合肥230026

出　　处：《系统仿真学报》2006年第6期1717-1720,共4页Journal of System Simulation

基　　金：安徽省自然科学基金(050460404)

摘　　要：由于在遗传算法的搜索寻优过程中种群有收敛于单一个体的趋势,为了减轻这种趋势,在Pareto多目标遗传算法的基础上做了一些改进,即用Pareto最优概念对种群进行第一级排序,然后计算种群中每个个体与同Pareto级别所有个体之间的全局拥挤距离作为该个体的次要属性进行第二级排序,根据这两级排序的结果进行联赛制选择操作和交叉变异操作。为了验证算法的性能,以多目标柔性工作车间调度问题作为实例并针对柔性工作车间调度问题的特点设计了相应的交叉变异方法。仿真结果表明该算法可以产生更多的分布在非劣解前沿上的解。Because the populations tend to converge to a single solution over the course of the genetic search process, Pareto multi-objective genetic algorithm was improved in order to alleviate the trend, At first, the populations were ranked based on Pareto optimal concept; secondly, the crowding distance of individuals was computed which have the same Pareto level as secondary attribution for the second rank. The result of the two ranks is the foundation of the league matches selection, crossover and mutation. In order to valuate the algorithm＇s performance, an example of flexible job shop scheduling and design the method of crossover and mutation were used according to the characteristics of flexible job shop scheduling. The experiment results show the algorithm yields more solutions in the non-dominated front.

关 键 词：柔性工作车间调度 两级排序 多目标 遗传算法 

分 类 号：O221.6[理学—运筹学与控制论]