基于贝叶斯网络的高速动车组运营故障分析

为研究影响高速动车组正常运行的各故障因素间的因果关系，分析其耦合强度，将故障因素分为人、机、环3类，从系统角度又将机器因素分为5个子系统。人、机、环3类共识别27个故障因素；使用K2算法生成贝叶斯网络结构，引入扩展因果效应算法确定节点优先次序作为K2算法的先验知识，采用EM算法学习贝叶斯网络参数，构建基于贝叶斯网络的高速动车组运营故障分析模型；以209个CRH详细故障报告为例，对故障因素的故障发生概率进行排序并分析因素间的影响强度和灵敏度。结果表明:牵引供电系统故障发生概率较高；车门系统故障、牵引变流器故障易由内部零件故障引起，外界异物击打对受电弓影响较大；人、环因素更易引起多故障耦合；环境因素对牵引供电系统表现出较高的灵敏度。贝叶斯网络在分析高速动车组运营系统故障问题上具有可行性，分析结果有助于提升运营单位的管控能力。

Analysis on operation fault of high speed EMU based on Bayesian network

In order to study the causal relationship between various fault factors that affect the normal operation of high speed EMUs and analyze their coupling strength,the fault factors were classified into three categories: human,machine,and environment,and from the system perspective,the machine factors were divided into five subsystems.27 fault factors from the human,machine,and environment were identified in total,and the K2 algorithm was used to generate the Bayesian network structure,then the extended causality algorithm was introduced to determine the node priority as a priori knowledge of K2 algorithm.The EM algorithm was applied to learn the parameters of Bayesian network,and a fault analysis model for the operation of high speed EMUs based on the Bayesian network was established.Finally,209 CRH detailed fault reports were used as examples to sort the fault probabilities of the fault factors and analyze the influence intensity and sensitivity among the factors.The results showed that he probability of fault of traction power supply system was high.The door system fault and the traction converter fault were easily caused by the internal component fault,and the ambient foreign matter strike had great influence on the pantograph.The human and environmental factors were more likely to cause the multi fault coupling,and the environmental factors presented higher sensitivity to the traction power supply system.The Bayesian network is feasible in analyzing the fault problem of high speed EMUs operation system,and the analysis results help to improve the management and control ability of the operating unit.