基于贝叶斯网络的地铁车辆牵引系统故障诊断  

Fault Diagnosis of Subway Vehicle Traction System Based on Bayesian Network

作　　者：陈亚峰 文静 周峰 肖惠杰 柴晓冬[1] 郑树彬[1] CHEN Yafeng;WEN Jing;ZHOU Feng;XIAO Huijie;CHAI Xiaodong;ZHENG Shubin(School of Urban Rail Transportation,Shanghai University of Engineering Science,Shanghai 201620;Shanghai Shentong Metro Group Co.,Ltd.,Shanghai 200233)

机构地区：[1]上海工程技术大学城市轨道交通学院,上海201620 [2]上海申通地铁集团有限公司,上海200233

出　　处：《计算机与数字工程》2025年第3期660-665,共6页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:51975347)资助。

摘　　要：基于上海地铁车辆牵引系统历史故障数据,提出一种车辆部件故障诊断方法。使用结构主题模型提取故障数据中的故障特征,并对特征数量进行寻优,引入文档协变量,提升故障特征的关键性;使用贝叶斯网络解决故障特征与故障原因之间的不确定性,结合专家知识和故障数据,通过动态规划-马尔可夫蒙特卡罗联合结构学习算法优化贝叶斯网络结构提升故障诊断模型精度。使用实际故障数据对诊断模型进行验证,结果表明所建立的故障诊断模型具有较高的诊断精度及优越性,诊断结果可为地铁车辆牵引系统故障快速诊断提供参考。Based on the historical fault data of the Shanghai metro vehicle traction system,a vehicle component fault diagnosis method is proposed.The fault features in the fault data are extracted using a structural topic model,the number of features is searched for,and document covariates are introduced to improve the criticality of the fault features.The uncertainty between the fault features and the cause of the faults is solved using Bayesian networks,and the fault diagnosis model is improved by combining expert knowledge and fault data and optimizing the Bayesian network structure through a dynamic programming-Markov Monte Carlo joint structure learning algorithm accuracy.The results show that the established fault diagnosis model has high diagnostic accuracy and superiority,and the diagnostic results can provide a reference for the rapid diagnosis of subway vehicle traction system faults.

关 键 词：轨道交通车辆 牵引系统 故障诊断 结构主题模型 贝叶斯网络 

分 类 号：U279.3[机械工程—车辆工程]