基于二阶段的云制造模式下高能效机床服务匹配  

High-efficiency machine tool service matching in cloud manufacturing mode based on two-stage approach

作　　者：谢迅 周丽蓉 王黎明[2] 孔琳 陈子翰 张传伟 XIE Xun;ZHOU Lirong;WANG Liming;KONG Lin;CHEN Zihan;ZHANG Chuanwei(College of Big Data Statistics,Guizhou University of Finance and Economics,Guiyang 550000,CHN;Key Laboratory of High Efficiency and Clean Mechanical Manufacture Ministry of Education,Shandong University,Jinan 250000,CHN;College of Mechanical and Electronic Engineering,Shandong University of Science and Technology,Qingdao 266590,CHN;Faculty of Mechanical and Electrical Engineering,Qingdao University,Qingdao 266071,CHN)

机构地区：[1]贵州财经大学大数据统计学院,贵州贵阳550000 [2]山东大学高效洁净机械制造教育部重点实验室,山东济南250000 [3]山东科技大学机械电子工程学院,山东青岛266590 [4]青岛大学机电工程学院,山东青岛266071

出　　处：《制造技术与机床》2025年第2期116-128,共13页Manufacturing Technology & Machine Tool

基　　金：国家自然科学基金青年科学基金项目(52305529);国家自然科学基金青年科学基金项目(52105519)。

摘　　要：为实现云制造环境下匹配出高能效优质机床服务组合方案,建立一种基于供需知识图谱的机床资源二阶段决策模型。首先,构建考虑能耗属性的机床服务供应和零件加工主客观需求的供需知识图谱。其次,建立机床服务时间、成本、能耗和质量指标计算模型,针对能耗指标采用基于实际功率、切削比能、额定功率3种计算策略。然后,面向客户需求构建二阶段匹配模型实现机床服务决策;其中,一阶段以知识图谱检索和蕴涵关系推理初选机床服务集合,二阶段以马尔可夫决策过程表征机床节能服务组合优化问题,并采用强化学习Actor_Critic算法求解。最后,通过机床服务资源仿真池构建和箱体加工案例试验,发现Actor_Critic算法相较于DQN(deep Q-learning)、PGM(policy gradient method)和DDPG(deep deterministic policy gradient)算法具备更优收敛效果,可快速匹配出云制造环境下经济节能且高效优质的机床服务组合方案。To achieve high-energy-efficient and high-quality machine tool service combinations in cloud manufacturing,a two-stage decision model based on a supply-demand knowledge graph was developed.A knowledge graph considering energy consumption attributes and the subjective and objective demands for part processing was first constructed.Secondly,a calculation model for machine tool service time,cost,energy consumption,and quality indicators was established,using three strategies for energy consumption based on power test data,specific cutting energy,and rated power.A two-stage matching model was built to meet customer needs,with initial service selection in the first stage through knowledge graph searches and implicit relationship reasoning.The second stage used a Markov decision process to optimize energy-saving service combinations,solved with the Actor_Critic algorithm from reinforcement learning.Experiments with a machine tool service resource simulation pool and box body machining cases showed that the Actor_Critic algorithm had better convergence than deep Q-learning(DQN),policy gradient method(PGM)and deep deterministic policy gradient(DDPG),enabling rapid matching of cost-effective,energy-saving,and high-quality service combinations in cloud manufacturing.

关 键 词：云制造 机床服务组合 高效节能 知识图谱 深度强化学习 

分 类 号：TH166[机械工程—机械制造及自动化]