Analysis of Rail Transit Operation and Maintenance Fault Recognition Considering Bayesian Knowledge Recognition Algorithm

ABSTRACT In the context of the development of high‐speed railways, the management of rail transit operation and maintenance is an important task for the daily operation of trains. At present, train scheduling is the most common fault manifestation in rail transit systems. The detection and identification of fault sources during scheduling are uncertain and subject to interference from subjective and objective factors. The present study employs statistical analysis to examine the occurrence of fault events in the train scheduling structure and operation process. The study integrates Bayesian network structure and related algorithms to calculate the occurrence and diagnosis of faults in the operation and maintenance process. A comparative analysis of the probability analysis and identification methods of fault occurrence revealed that the posterior probability of fault events at network nodes was the highest at 90.34%, which was 32.3% higher than the prior knowledge state. In comparing fault recognition methods, the recognition accuracy of the support vector machine algorithm was 91.17%, while the proposed Bayesian knowledge recognition algorithm was as high as 95.89%, with a specificity of 97.02%. Therefore, the superiority of its method in rail transit operation and maintenance has been proven.