A Submodel Based Numerical Investigation of the Segment Uplift During Tunnel Construction

Shield tunneling involves complex soil‐structure interactions, thereby leading to a normally time‐consuming and labor‐intensive finite element calculation. A submodel based method for predicting segment uplift during tunnel construction is proposed in this study, which can effectively reduce the computational cost. It initially focuses on determination of the submodel boundary, where plastic dissipation energy and its associated coefficient of variation are selected as criteria. Compared to traditional boundary determination methods based on soil disturbance degree, the proposed approach demonstrates great advantages in accuracy. Finite element simulations of tunnel excavation are thereafter performed for both homogeneous and composite geological strata models, where the applicability of the submodeling technique as well as the rationality of plastic dissipation energy criteria are verified. Building upon the computationally enhanced segment uplift dataset (optimized via submodel‐driven temporal efficiency framework), a rigorous benchmarking of machine learning architectures is finally implemented, among which the LSTM shows the best prediction performance, with an R 2 value of 0.9 and RMSE of 0.02, fully demonstrating the superiority of the proposed method in predicting tunnel segment uplift.