Construction and application of LSTM based prediction model for tunnel surrounding rock deformation

Abstract As an important problem during tunnel construction and tunnel maintenance, tunnel surrounding rock deformation has been paid close attention by scholars at home and abroad. The traditional method to predict the tunnel surrounding rock deformation by fitting the monitoring and measuring data of the tunnel not only consumes a lot of manpower and material resources, but also has a low accuracy in predicting the data with large fluctuations. Deep learning tunnel surrounding rock deformation prediction method can reduce labor cost and improve monitoring efficiency and accuracy. Taking a tunnel project as an example, based on the cyclic neural network model in deep learning, the tunnel monitoring and measurement data were modeled and analyzed, and the LSTM network model was constructed, which was used to analyze and predict the vault settlement of the tunnel. Then the monitoring data are compared with the traditional curve prediction and LSTM network model prediction to obtain the tunnel surrounding rock deformation results. The results show that the LSTM prediction model can predict the vault settlement in the process of tunnel construction. The accuracy and stability of the prediction are better than those of the traditional curve, no matter the data with small or large fluctuation amplitude, and are consistent with the field measured data. The research results can be used to introduce the LSTM prediction method in deep learning into tunnel construction, and provide a certain reference for tunnel construction safety.