Fault Diagnosis for Brake System in High-Speed Trains Using the Phased Features and Multi-layer Perceptron

Recently, High-Speed Train (HST) has been developing quickly. The brake system is the most important part in HST. Therefore its safety and reliability have attracted much attention from both academic and industrial community. Many data-driven methods are introduced to handle the fault diagnosis problem for brake system. However, the existing data-driven methods barely analyse the brake system. The used features lack of interpretability. In this paper, a fault diagnosis model using the phased features and multi-layer perceptron (MLP) is proposed for brake system. The proposed model is divided into two stages: offline and online stages. In the offline stage, after the analysis of brake system, the phased features are extracted. Then the MLP model is trained based on the phased features. In the online stage, the model learned in the offline stage is used to diagnosis the fault. The experimental platform for brake fault is constructed to validate the proposed model. The results shows the superiorities compared with other diagnosis methods in terms of precision, recall and F1-score.