Spatio-Temporal Data Prediction of Braking System Based on Residual Error

With the improvement of industrial intelligence, the train braking system has also ushered in a more rapid development, however, the huge consumption of funds required for the construction of the brake system experimental platform has discouraged researchers, and the low flexibility also hinders the research of this region. In the recent years, the hardware-in-the-loop (HIL) simulation platform was introduced into the field of braking system research. With the cascade strategy, small trains can simulate the working conditions of large trains. In this process, there is a very important link: predict the pressure value of subsequent trains under the same test time. We found that the air pressure values which belong to time-series data of trains at different locations in one experiment still have spatiality. Therefore, these are data with temporal and spatial characteristics. Not only that, we will connect the temporal and spatial characteristics of different test rooms as a new consideration point for testing. In this paper, we propose a long short-term memory (LSTM) network based on the idea of residual error (R-LSTM), which predicts the air pressure between trains at different locations in different trials, and the RMSE index used has reached 2.7938.