摘要
在分布式柔性作业车间多目标调度问题的求解过程中,存在调度规模大、多个目标难以协调等缺陷。针对上述缺陷,提出了一种改进的非支配排序遗传算法Ⅱ(NSGA2),并对分布式柔性作业车间多目标调度问题进行了求解。首先,建立了以完工时间、机器负荷、能耗为优化目标的分布式柔性作业车间多目标调度模型;然后,基于帕累托(Pareto)等级特点设计了一种动态双种群搜索策略和种群划分机制,以替代传统的选择操作,并对每个种群采用了不同的搜索策略;针对关键工厂,在第二个种群中设计了局部搜索策略,基于Pareto等级的支配关系设计了Q学习的状态、奖励函数,采用Q学习对双种群的数量比例进行了自适应调整;最后,采用扩展的基准算例对该改进算法的有效性进行了验证,并将其与其他算法进行了对比分析。研究结果表明:采用动态双种群搜索策略改进的NSGA2算法能有效保持种群多样性,且不易陷入局部最优,提高了算法的求解质量。该改进算法与传统NSGA2算法相比,多样性评价指标平均提高了15.34%,收敛性评价指标平均提高了76.37%,证明了该算法在解决分布式柔性作业车间多目标调度问题上的优越性。
Aiming at the problems of large scheduling scale,and difficulty in coordinating multiple objectives in distributed flexible job shop multi-objective scheduling problem,an improved non-dominated sorting genetic algorithm II(NSGA2)was proposed.The multi-objective scheduling problem of distributed flexible job shop was solved.Firstly,a distributed flexible job shop multi-objective scheduling model was established,and the optimization goals were minimizing completion time,minimizing machine load,and minimizing machine energy consumption.Then,a dynamic dual-population search strategy based on Pareto hierarchy of dominance relationships was designed.A population partitioning mechanism was devised to replace the traditional selection operation,and each population executed different search strategies.A local search strategy was specifically designed for the key factories in the second population.The state space and the reward functions for Q-learning were based on Pareto hierarchy of dominance relationships.The Q-learning was used to adaptively adjust the proportion of the dual populations.Finally,the effectiveness of the algorithm improvements was verified through extended benchmark examples,and compared with other algorithms to verify the effectiveness of the improvement.The research results indicate that the improved NSGA2 algorithm,which employing a dynamic dual-population search strategy,is able to effectively maintain population diversity and is less prone to becoming trapped in local optima,thereby enhancing the overall solution quality of the algorithm.Comparing with the NSGA2 algorithm,the diversity evaluation hypervolume of the improved algorithm is increased by 15.34%on average,and the convergence evaluation index inversion generation distance is increased by 76.37%on average.It proves the superiority of the proposed algorithm in solving multi-objective scheduling problems in distributed flexible job workshops.
作者
汪豪
谢辉
李艳武
WANG Hao;XIE Hui;LI Yanwu(School of Electronic&Information Engineering,Chongqing Three Gorges University,Chongqing 404100,China)
出处
《机电工程》
CAS
北大核心
2024年第12期2252-2260,共9页
Journal of Mechanical & Electrical Engineering
基金
重庆市教委科学技术研究项目(KJQN202301216)。
关键词
柔性作业车间调度问题
分布式多目标柔性作业车间
车间多目标调度问题求解
帕累托等级
改进非支配排序遗传算法Ⅱ
动态双种群搜索策略
Q学习
flexible job skop scheduling problem(FJSP)
distributed multi-objective flexible job shop
solution of multi-objective scheduling problem in workshop
Pareto rank
improved non-dominated sorting genetic algorithm II(NSGA2)
dynamic dual population search strategy
Q-learning
