基于协同优化算法的分布式装配车间调度  

Distributed Assembly Workshop Scheduling Based on Collaborative Optimization Algorithm

作　　者：杜松霖 仵大奎 余云涛[2] 刘亚 周文举[1] DU Songlin;WU Dakui;YU Yuntao;LIU Ya;ZHOU Wenju(School of Mechatronic Engineering and Automation,Shanghai University,Shanghai 200444,China;China Electronics Standardization Institute,Beijing 100007,China;Anhui Synkrotron Technologies,Co.,Ltd.,Hefei 230088,Anhui,China)

机构地区：[1]上海大学机电工程与自动化学院,上海200444 [2]中国电子技术标准化研究院,北京100007 [3]安徽深信科创信息技术有限公司,安徽合肥230088

出　　处：《计算机工程》2025年第3期274-282,共9页Computer Engineering

基　　金：国家重点研发计划(2021ZD0110601)。

摘　　要：在分布式调度中,调度策略的协同优化已逐渐成为分布式调度研究者与分布式制造行业决策者的重点关注方向之一。针对带装配机的分布式阻塞流水车间调度问题DBFSP-A,以最大装配完成时间为优化目标,提出一种基于Q-Learning的协同优化算法QLBC。首先,建立了以最大装配完成时间为优化目标的数学模型,以增强DBFSP-A的可解释性,在算法构造初始化阶段,QLBC充分利用DBFSP-A的问题特征,分别根据加工任务的总处理时间、前置延迟时间等特定的排序规则,构造式地产生高质量的、多样性的可行调度序列作为算法的初始解;其次,在QLBC的后续迭代中,采用基于Q-Learning的协同搜索策略,自主学习地指导当前解根据其各自状态及先验反馈选择合适的搜索操作,从而实现全局搜索和局部搜索、加工过程与组装过程的协同进化与协同优化;最后,在不同实例中,测试和检验了QLBC算法的性能。实验结果表明,相较于其他6种先进算法,QLBC求解的可行调度序列时间平均节省32.09 h,在提高生产效率、节约生产成本方面更具优势。In distributed scheduling,collaborative optimization has gradually become one of the most important focus areas for researchers and decision-makers in distributed manufacturing.A Q-Learning-Based Collaborative(QLBC)optimization algorithm is proposed to solve the distributed blocking flow-shop scheduling problem with an assembly machine DBFSP-A,considering the maximum assembly completion time.First,a mathematical model is established with the objective of maximizing the assembly completion time to enhance the interpretability of DBFSP-A.The problem characteristics of DBFSP-A are utilized to constructively generate high-quality,high-diversity scheduling sequences based on specific ordering rules according to the total processing time and front delay.Subsequently,during QLBC iterations,a Q-Learning-based collaborative search strategy is used to achieve co-evolution and co-optimization by selecting appropriate search operations based on the respective states and prior feedback from the solutions.Finally,the performance of QLBC algorithm has been tested and verified in different instances.The experimental results and statistical analysis indicated that compared with the other six advanced algorithms,the scheduling sequence solved by QLBC saves an average of 32.09 h,which is advantageous in terms of improving productivity and saving production costs.

关 键 词：协同优化 分布式制造 阻塞约束 产品装配 Q-Learning算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]