Fault Prediction Method of Railway Braking System Based on LSTM Neural Network

摘要

With the development of rail transit, the safety and reliability of its on-board system are becoming more and more important. In this paper, the braking system is used as the entry point. According to its composition and fault handling, the huge on-board sensor data and fault data are screened, which determine a clear range for the main analysis and processing in this paper. We propose a neural network model for braking system fault prediction. After the external sorting algorithm to complete the processing of a large amount of data, this model can predict future fault status based on these past sensor data. At first it didn't work very well but the performance of the model is significantly improved after data enhancement. This model is based on full connected networks and long short term memory (LSTM), and this model has a better effect than SVR, especially in terms of recall rate.