Dynamic reliability analysis of traction drive system for EMU

WU Sai; LI Gang; QI Jinping; YU Qiangye

doi:10.16579/j.issn.1001.9669.2026.08.012

Abstract：[Objective]Aiming at the problem that traditional dynamic Bayesian networks cannot intuitively characterize the event correlation between nodes via conditional probability tables,and the deficiency that existing studies mostly focus on component-level analysis and ignore the dynamic failure effect of system redundancy design,a dynamic reliability analysis method adapted to electric multipe unit(EMU)traction drive systems was proposed.[Methods]Firstly,the improved continuous-time dynamic Bayesian network rules were constructed,and the calculation methods of reliability,posterior probability and importance based on the rule execution degree of node timing conditional probability table and impulse function were proposed to clarify the logical correlation of events between nodes.Secondly,based on the failure mechanism of the CRH3 EMU traction drive system,a dynamic Bayesian network model of the system was established to clarify the hierarchical failure logic.Then,the fuzzy set method and extension principle were adopted to process historical fault data,and the fuzzy interval of component failure rate was obtained to characterize data uncertainty.Finally,multi-dimensional reliability analysis was carried out to realize the identification of system weak links.[Results]The results show that compared with the traditional dynamic Bayesian network method,the maximum improvement of the proposed method in reliability and accuracy reaches 0.031104.The system reliability decreases continuously with the increase of running time,and approaches 0 at about 2 000 h.Four types of core weak components are identified,which provides reference for structural optimization and fault diagnosis of traction drive systems.