Size Prediction of Railway Switch Gap Based on RegARIMA Model and LSTM Network

Abstract

Railway turnout is one of the weakest elements in railway infrastructure, whose normal operation is directly related to the safety of the passing trains. The railway switch gap is an important part of the indication mechanism of the turnout. Once the size of the switch gap exceeds the standard, the indication mechanism of the turnout will fail, causing the close of the railway route and endangering the safety of the passing trains. However, at present, the maintenance of the switch gap size still follows the traditional failure based maintenance (FBM) strategy. Therefore, this paper first analyzes the causes of the change of the switch gap size, and constructs a regression and autoregressive integrated moving average (RegARIMA) model to describe the change law of the switch gap size. Then, a size prediction method for the switch gap is proposed by utilizing this model and a long short-term memory (LSTM) network. By referring to the prediction results of this method, the maintenance personnel can make maintenance plan before the switch gap size exceeds the standard, so as to avoid the failure of the turnout. Finally, the experimental results based on condition monitoring data from a turnout show that the root mean square error, the mean absolute error and the mean absolute percentage error of the proposed prediction method are 0.0732mm, 0.0568mm and 2.8670% respectively, which means the proposed model can effectively describe the change law of the switch gap size, and the proposed prediction method has good prediction performance.