Reliability analysis of EMU brake system based on hyper-ellipsoid CTBN

YU Qiangye; QI Jinping; JIA Cunxiao; YU Liye

doi:10.16579/j.issn.1001.9669.2026.08.013

Abstract：[Objective]Aiming at the problems of complex and diverse fault data information sources of electric multiple unit(EMU)brake systems,difficulty in accurately describing component fault probability,as well as difficult modeling and cumbersome integral solution process of the traditional continuous-time Bayesian network(CTBN),a reliability analysis method combining hyper-ellipsoid model and improved continuous-time Bayesian network was proposed.[Methods]Firstly,the evidence interval of system bottom event faults was calculated via evidence reasoning method and constrained by the hyper-ellipsoid model,so as to avoid the simultaneous occurrence of multi-dimensional box corner values for the failure probabilities of all bottom events.Secondly,different logic gates were constructed using binary conditional probability tables,and the reliability of the top event was directly calculated through the tables,so as to eliminate the dependence on integral solution and reduce the solution difficulty of traditional methods.Then,the proposed method was compared with the discrete-time Markov model and the traditional continuous-time Bayesian network to verify its correctness and superiority.Finally,reliability analysis was conducted taking the brake system of CRH5 EMU on the Lanzhou-Xinjiang High-speed Railway as the research object.[Results]The results show that compared with the other two methods,the failure probability interval obtained by the proposed model is more accurate,with the solution time reduced by 36.86％and the calculation workload decreased by 33.82％.The posterior probability of the system can identify the weak links of the brake system and provide reference for the formulation of component maintenance plans.