Faults in onboard equipment within railway train control systems exhibit complexity and uncertainty,compounded by the non-textual nature of recorded data.Traditional diagnosis methods based on expert knowledge often prove inefficient and inaccurate.Bayesian network(BN)excel in handling uncertainties and related complexities.This paper focuses on CTCS3-300T onboard equipment,employing a BN model for fault diagnosis.By examining current practices in addressing typical onboard equipment malfunctions,a methodology combining expert knowledge,fault datasets,and the K2 algorithm was proposed for BN model development.Utilizing K2 algorithm and maximum likelihood estimation,structure learning and parameter learning were carried out,incrementally refining the BN diagnostic model for rapid fault localization.An optimized BN model was established,achieving a fault diagnosis accuracy of 87.1%.Compared to conventional expert-knowledge-based models,this optimal BN model enhances fault diagnosis accuracy by 37.4%.According to the results of case analysis and model validation,the model is able to ensure fault diagnosis accuracy while significantly improving diagnostic efficiency.
NETL
NSTL
万方数据
