<<
[an error occurred while processing this directive] | [an error occurred while processing this directive] >>
Deformation Prediction and Control for Shield Tunnelling Passing under Existing Tunnels Based on BO-Adam-Bi-LSTM
(1. China Railway Development and Investment Group, Kunming 650500; 2. School of Civil and Hydraulic Engineering, Huazhong University of Science and Technology, Wuhan 430074; 3. Department of Building and Real Estate, The Hong Kong Polytechnic University, Hong Kong 999077)
Abstract To address deformation and safety control issues induced by shield tunnelling passing under existing tun? nels, a deep learning-based prediction model was designed to capture deformation development patterns. A Bi-directional Long Short-Term Memory network (Bi-LSTM) was optimized using the Adam algorithm, with parameter tuning performed via Bayesian Optimization (BO), forming the BO-Adam-Bi-LSTM model for deformation predic? tion of shield tunnelling passing under existing tunnels. The model's prediction results were compared with other neural network models, and SHAP was used to enhance interpretability and identify key construction parameters.Results show that the BO-Adam-Bi-LSTM model achieves high prediction accuracy, with R2 values of 0.935 and 0.924, RMSE values of 0.504 and 0.903, and MAE values of 0.415 and 0.824 for the test sets. SHAP analysis reveals that shield chamber pressure has a significant impact on the prediction of horizontal deformation in existing tunnels.By adjusting a few key parameters with high contributions to the model predictions, tunnel deformation values can be effectively controlled within the warning range.
Abstract：
To address deformation and safety control issues induced by shield tunnelling passing under existing tun? nels, a deep learning-based prediction model was designed to capture deformation development patterns. A Bi-directional Long Short-Term Memory network (Bi-LSTM) was optimized using the Adam algorithm, with parameter tuning performed via Bayesian Optimization (BO), forming the BO-Adam-Bi-LSTM model for deformation predic? tion of shield tunnelling passing under existing tunnels. The model's prediction results were compared with other neural network models, and SHAP was used to enhance interpretability and identify key construction parameters.Results show that the BO-Adam-Bi-LSTM model achieves high prediction accuracy, with R2 values of 0.935 and 0.924, RMSE values of 0.504 and 0.903, and MAE values of 0.415 and 0.824 for the test sets. SHAP analysis reveals that shield chamber pressure has a significant impact on the prediction of horizontal deformation in existing tunnels.By adjusting a few key parameters with high contributions to the model predictions, tunnel deformation values can be effectively controlled within the warning range.
ZHANG Mingshu1 YAO Chen1 WU Xianguo2 CHEN Hongyu3 FENG Zongbao2 YANG Sai2
.Deformation Prediction and Control for Shield Tunnelling Passing under Existing Tunnels Based on BO-Adam-Bi-LSTM[J] MODERN TUNNELLING TECHNOLOGY, 2024,V61(6): 92-99
2024, Vol. 61 