Multi-sensor Data Mining of Single Train Braking System Based on Long Short-term Memory Network

In this article, the data collected by multiple air pressure sensors during train braking is divided into training set and test set to carried out temporal data prediction by Long Short-Term Memory (LSTM) algorithm, which is an optimized algorithm of Recurrent Neural Network algorithm, and is suitable for analyzing long time data series. Due to the frequency instability caused by hardware sampling interference, a variety of data preprocessing methods and algorithm parameters configuration are used to equalize and normalize the original data to ensure the accuracy of modeling. The test results of the test set show that the model has a high accuracy rate for the train tube pressure prediction, and finally achieves information fusion between the five sensors. This has greatly contributed to the research of multi-sensor information fusion of long-large group train brake systems.