Abstract: With the consideration of setup and transportation operations,the multi-objective hybrid flowshop scheduling problem with variable sublots was studied,which was aimed to simultaneously optimize two conflicting objectives:the makespan and the total number of sublots.A multi-objective mixed integer programming model was developed and the trade-off between the two objectives was evaluated.Since the problem belongs to Non-deterministic Polynomial (NP) problems,the Multi-objective Evolutionary Algorithm (MOEA) was suggested to solve it.To eliminate the biases of previous experience for configuring MOEA,an Automated Algorithm Design (AAD) methodology was introduced to conceive a promising MOEA based on MOEA framework,which was enabled designing the MOEA by determining parameters and their combinations automatically with minimal user intervention.Considering the variable sublots,a dynamic decoding strategy was proposed.With regards to the problem-specific characteristics and the employed algorithm framework,for the categorical and numerical parameters,reasonable value ranges were given.For the AAD methodology,the I/F-Race was employed.Compared with CPLEX ,the automated MOEA was demonstrated much more effective.

Key words: hybrid flowshop scheduling, variable sublots, multi-objective evolutionary algorithm, automatic algorithm design

摘要： 在考虑运输和启动作业的基础上,研究了多目标变分批混合流水车间调度问题,旨在同时优化最大完工时间和子批总数两个相互冲突的目标。建立了多目标混合整数规划模型,验证了两目标间的冲突关系。由于问题属于非确定性多项式困难问题,采用多目标进化算法(MOEA)解决该问题。为消除MOEA在构造过程中受到先前经验偏见的影响,基于MOEA框架,采用自动算法设计方法(AAD)构造了高性能的MOEA。AAD能够通过最小的干预自动确定MOEA的各种参数取值以及最优的参数组合。考虑变分批技术,提出了动态解码策略;针对问题特性和所采用的算法框架,对于可配置的类别参数和数值参数,给出了合理的取值区间;对于AAD方法,采用了I/F-Race方法。最后,通过与CPLEX和已提出的MOEAs对比分析,证明了自动生成的MOEA更加有效。

关键词: 混合流水车间调度, 变分批, 多目标进化算法, 自动算法设计