Control of existing tunnel deformation caused by shield adjacent undercrossing construction using interpretable machine learning and multiobjective optimization

Abstract

A hybrid intelligent framework is proposed in this paper to reduce the existing tunnel deformation caused by shield adjacent undercrossing construction (SAUC). A Bayesian optimization natural gradient boosting (BO-NGBoost) model for existing tunnel deformation prediction is developed, and the Shapley additive explanations (SHAP) approach is used to analyze the interpretability of the prediction model. The multiobjective evolutionary algorithm based on decomposition (MOEA/D) is used to optimize the construction parameters. The applicability and validity of the proposed method are tested in a case study from the Wuhan Metro. The results indicate that (1) the established BO-NGBoost existing tunnel deformation prediction model shows high accuracy. (2) Through SHAP analysis, the importance of each input parameter to the existing tunnel deformation is identified, and the key shield optimization parameters are defined. (3) By using the developed BO-NGBoost-MOEA/D algorithm to optimize the key parameters, the existing tunnel deformation is effectively controlled.