Abstract: To study multi-objective flexible job-shop problem with heterogeneous Dejong learning effect,variable neighborhood invasive weed optimization was proposed.Based on random key method,one two-section weed coding method was used to solve scheduling,routing and assignment problems.The fitness of invasive weed was calculated according to grey entropy correlation grade.New parent population was produced based on preference and crowded distance.Three kinds of neighborhood structure were constructed to search elite individuals with variable neighborhood search strategy at the end of iteration.The algorithm was used to solve Kacem benchmark problems and flexible job-shop problem with Dejong learning effect.Results showed that the proposed algorithm could solve above problems,and the total completion time was more sensitive to M value.Decision makers should pay attention to the heterogeneous learning effect in the flexible job-shop problem.Suitable workers should be assigned to the right station to improve the operation level and reduce the labor cost.

Key words: invasive weed optimization, variable neighborhood search, behavior management, multi-objective optimizing, flexible job-shop

摘要： 针对存在异质性员工的多目标柔性作业车间调度问题,构建了具有Dejong学习效应的调度模型,并提出变邻域杂草算法求解该问题。为解决工序排序、机器选择和员工指派3个子问题,基于随机键编码方式对杂草个体进行编码。采用灰熵关联方法给出杂草的适应度值,根据杂草之间的偏序关系对杂草群体的拥挤距离进行排序,从而产生新的父代群体。构造了3种邻域结构,在迭代后期对精英个体进行变邻域搜索。最后,将算法用于Kacem柔性作业车间基准问题和具有异质性员工的柔性作业车间调度问题求解。案例分析表明,算法可有效求解基准问题和多个“不可压缩因子”F值的柔性车间调度问题,其总完工时间对F值的敏感度更高。

关键词: 杂草算法, 变邻域搜索, 行为运作, 多目标优化, 柔性作业车间