Research on deformation prediction of subway tunnel structure based on LSTM network

Abstract

In order to improve the accuracy of subway tunnel settlement prediction, the role of long short-term memory network (LSTM network) in subway tunnel settlement prediction was studied. Reverse neural network (BP neural network), support vector machine (SVM) and LSTM network are used to build models, combined with the measured data of Shanghai subway tunnel, the prediction accuracy of the models is compared and analyzed. The test results show that LSTM network is better than BP neural network and support vector machine, and have high prediction accuracy. Compared with the BP neural network model and the support vector machine, the average prediction error of the LSTM network model is reduced by 72% and 75%, the average relative error is reduced by 72% and 75%, and the root mean square error value is reduced by 73% and 78%, the predicted results are closer to the actual measurement results. The research shows that the LSTM network, one of the deep learning methods, is introduced into the subway tunnel settlement monitoring, which improves the prediction accuracy.