(1. Urban Construction Group Co., Ltd., Jinan 250014, Shandong, China;
2. Key Laboratory of Safety for Geotechnical and Structural Engineering of Hubei Province, School of Civil and Architecture Engineering, Wuhan University， Wuhan 430072，Hubei， China;
3. China Railway Siyuan Survey and Design Group Co., Ltd., Wuhan 430063, Hubei, China;
4. School of Civil Engineering and Architecture, Guangxi University, Nanning 530004, Guangxi, China;
5. Large Diameter Shield Engineering Co., Ltd., China Railway 14th Bureau Group , Nanjing 211899, Jiangsu, China)
Abstract: To accurately predict surface deformation induced by super-large-diameter shield tunneling in silty clay, an intelligent prediction model based on machine learning was developed. The physical and mechanical parameters of the soil were corrected by considering stratum thickness and tunnel burial depth. Three surface-deformation prediction models, namely, a Back-Propagation (BP) neural network, a Support Vector Machine (SVM), and a Random Forest (RF), were established in MATLAB. The optimum hyperparameter combinations were determined by cross-validation, grid search, and empirical formulas. Using data from the Jiluo Road Yellow River Tunnel in Jinan, the surface deformation induced by shield tunneling was predicted.
Model accuracy was evaluated using Mean Square Error (EMS), Root Mean Square Error (ERMS), Mean Absolute Error (EMA), and the Coefficient of Determination (R2). The results showed that the RF model produced the smallest prediction error, followed by the BP neural-network model, whereas the SVM model produced the largest error and had relatively weak generalization ability. The proposed model provided a reference for predicting surface deformation induced by super-large-diameter shield tunneling under similar engineering-geological conditions.
