Fault Prediction Method of the On-board Equipment of Train Control System Based on Grey-ENN

Abstract

On-board equipment is the core component of Train Control System. It is of great significance to perform the fault prediction of on-board equipment in order to improve the safety of the train. This paper proposes a fault prediction method based on Grey-Elman neural network(Grey-ENN) for 300T on-board equipment. Firstly, through the statistics and analysis of the AE-log data of on-board equipment, the operation states evaluation and division have been completed. Secondly, the GSM-SVM (Support Vector Machine is optimized by Grid Search Method) model has been used to recognize operation states, followed by verifying the validity of the equivalent failure rate. The experiment results show that the fault states can be distinguished based on GSM-SVM with the accuracy of 93.4%. Finally, a joint fault prediction model has been employed to accomplish the complete prediction of serious and emergency faults with overall prediction accuracy of 86%, which verifies the feasibility and effectiveness of the Grey-Enn prediction method, and fault prediction result has certain guiding significance for maintenance decision.