基于G-AHSMM的设备剩余寿命预测研究  

Research on residual life prediction of equipment based on G-AHSMM

作　　者：张青山[1] 张思岩 肖萌[1] 徐伟[1] ZHANG Qing-shan;ZHANG Si-yan;XIAO Meng;XU Wei(School of Management,Shenyang University of Technology,Shenyang 110870,China)

机构地区：[1]沈阳工业大学管理学院,沈阳110870

出　　处：《沈阳工业大学学报（社会科学版）》2022年第2期151-158,共8页Journal of Shenyang University of Technology(Social Sciences)

基　　金：辽宁省社会科学规划基金一般项目(L19BGL029)。

摘　　要：伴随大数据技术和智能制造的快速发展,生产设备的预知维修及多台设备的联合维修决策已成为工业制造业企业备受关注和亟待解决的现实问题。而服役设备剩余寿命的精准预测,又是预知维修决策和联合维修决策的前提。对已有设备寿命预测方法进行比较分析,将隐半马尔可夫模型加以拓展,结合伽马分布,构建设备状态监测数据驱动的剩余寿命预测模型G-AHSMM,给出求解方法,并基于某涡轮发动机的状态监测数据进行验证分析。结果表明:预测模型不仅规避了以往“状态观测值之间相互独立”的不实假设,而且相比传统HSMM具有更高的现实拟合性、求解简捷性和预测精准性,可作为企业预测服役设备剩余寿命的有效工具。With the rapid development of big data technology and intelligent manufacturing,the predictive maintenance of production equipment and the joint maintenance decision of multiple pieces of equipment have become a practical problem that has attracted much attention and needs to be solved by industrial manufacturing enterprises.The accurate prediction of the residual life of service equipment is the premise of predicting maintenance decision and joint maintenance decision.The current equipment life prediction methods are compared and analyzed.The hidden semi-Markov model is extended,and combined with Gamma distribution,the residual life prediction model G-AHSMM is constructed driven by equipment condition monitoring data.The model solution method is given,and verified and analyzed based on the condition monitoring data of a turbo-engine.The results show that the prediction model not only avoids the false assumption of “mutual independence between state observations”,but also has higher realistic fitting,simple solution and prediction accuracy than the traditional HSMM.It can be used as a effective tool for enterprises to predict the residual life of in-service equipment.

关 键 词：智能制造 设备寿命 剩余寿命预测 隐半马尔可夫模型 伽马分布 前向后向算法 状态识别 

分 类 号：F270.7[经济管理—企业管理]