基于双层交互Q学习算法的轴承生产智能排程  

Intelligent scheduling of bearing production based on double-layer interactive Q-learning algorithm

作　　者：裴志杰 杨晓英[3,4] 杨欣 杨逢海[3] PEI Zhijie;YANG Xiaoying;YANG Xin;YANG Fenghai(School of Business,Henan University of Science and Technology,Luoyang 471023,China;School of Economics and Management,Huanghe Jiaotong University,Jiaozuo 454950,China;School of Mechanical Engineering,Henan University of Science and Technology,Luoyang 471003,China;Henan Collaborative Innovation Center of Advanced Manufacturing of Machinery and Equipment,Luoyang 471003,China)

机构地区：[1]河南科技大学商学院,河南洛阳471023 [2]黄河交通学院经济管理学院,河南焦作454950 [3]河南科技大学机电工程学院,河南洛阳471003 [4]机械装备先进制造河南省协同创新中心,河南洛阳471003

出　　处：《机电工程》2025年第3期451-462,共12页Journal of Mechanical & Electrical Engineering

基　　金：山东省重点研发计划项目(2020CXGCO11001);河南省重点研发专项(231111222600)。

摘　　要：针对带装配的两阶段分布式混合流水车间(TSDHFSSP)环境下的轴承排程问题,提出了一种基于双层交互Q学习算法(DIQLA)的车间调度方法,以解决轴承生产智能排程问题。首先,描述了轴承的排程问题,建立了以最小化最大完工时间为目标的数学模型;然后,引入马尔科夫决策过程(MDP),模拟了轴承生产排程过程,根据两阶段生产过程,搭建了双智能体交互的Q学习模型,接着对两阶段的的智能体进行了建模,设计了双智能体的状态变量、调度规则动作集和即时奖励函数,改进了传统的贪婪搜索策略,提出了两阶段联合排程算法;最后,利用实例数据对该算法进行了仿真验证,将其与单一智能体Q学习算法(QL)及非支配遗传算法(NSGA-II)、带精英策略的改进的鲸鱼优化算法(IWOA)等算法进行了对比,先在同一算例下验证了该算法的有效性,再通过对比不同订单算例,验证了该算法的性能,并利用实例数据再次验证了该算法在两阶段排程的应用效果。研究结果表明:两阶段联合排程算法在解决轴承排程问题时具有可行性,在优化轴承生产排程方面上具有较好的效果;在实际的应用中,与原有人工排产相比,其产品的加工周期平均缩减了17%,订单交付率平均提升了9%。该方法为轴承制造类企业生产排程提供了一种智能化的方案。Aiming at the bearing scheduling problem in a two-stage distributed hybrid flow shop scheduling problem(TSDHFSSP)with assembly environment,a dual-layer interactive Q-learning algorithm(DIQLA)was proposed.Firstly,the scheduling issue of bearing was described,a mathematical model was established with the objective of minimizing the maximum completion time.Then,the Markov decision process(MDP)was introduced to simulate the bearing production scheduling process,and a double-agent interactive Q-learning model was constructed based on the two-stage production process.Subsequently,a two-stage interactive agent model was established,the state variables,scheduling rule action sets,and immediate reward functions for the dual agents were designed,the traditional greedy search strategy was improved and a two-stage joint scheduling algorithm was proposed.Finally,the algorithm was tested using actual data through simulation experiments.It was compared with the single-agent Q-learning algorithm(QL),the non-dominated sorting genetic algorithm(NSGA-II),and the improved whale optimization algorithm with an elite strategy(IWOA).The algorithm s effectiveness was first validated under the same case scenario,followed by a performance evaluation through comparisons with different order case studies.The application effect of the algorithm in two-stage scheduling was additionally verified using example data.The research results indicate that the two-stage joint scheduling algorithm in this study is feasible for solving actual bearing scheduling issues.It shows good effectiveness in optimizing bearing production scheduling.In practical application,comparing to the original manual scheduling,it has reduced the average processing cycle of products by 17%and increased the average order delivery rate by 9%.This method provides an intelligent solution for production scheduling in bearing manufacturing enterprises.

关 键 词：轴承生产 车间调度方法 智能排程 两阶段分布式混合流水车间 Q学习 双层交互 两阶段联合排程算法 

分 类 号：TH186[机械工程—机械制造及自动化] TH133.3