Shield tunneling efficiency and stability enhancement based on interpretable machine learning and multi-objective optimization

Adequate control of shield machine parameters to ensure the safety and efficiency of shield construction is a difficult and complex problem. To address this problem, this paper proposes a hybrid intelligent optimization framework that combines interpretable machine learning, intelligent optimization algorithms, and multi-objective optimization and decision-making methods. The nonlinear relationship between the input parameters and ground settlement (GS) is fitted based on the light gradient boosting machine (LGBM), and the effect of the input parameters on GS is analysed based on SHapley additive exPlanation for further feature selection. Subsequently, the hyperparameters of LGBM were determined based on the sparrow search algorithm (SSA) to better fit the input–output relationship. On this basis, a multi-objective intelligent optimization model is established to solve the optimized operating parameters of shield machine by non-dominated sorting genetic algorithm II and technique for order preference by similarity to ideal solution to reduce GS and improve drilling efficiency. The results demonstrate that the SSA-LGBM model predicts GS with high accuracy, exhibiting an RMSE of 4.775, a VAF of 0.930 and an R2 of 0.931. These metrics collectively reflect the model’s excellent performance in prediction accuracy, ability to explain data variability, and control of prediction bias. The multi-objective optimization model is effective in optimizing two objectives, and the improvement can reach up to 39.38%; at the same time, the model has high scalability and can also be applied to three or more objectives. The intelligent optimization framework for shield construction parameters proposed in this paper can generate the optimal parameter combinations for shield machine manipulation, and provide reference and guidance when there are conflicting optimization objectives.