BM Advance Rate Advance and Interpretability Analysis Using the #br# NRBO-XGBoost Method

Advance rate is a key indicator reflecting the interaction between tunnel boring machines (TBMs) and surrounding rock, and its reliable prediction is essential for TBM performance optimization. A hybrid model combining the Newton-Raphson-Based Optimization (NRBO) algorithm with the Extreme Gradient Boosting method (XGBoost), referred to as NRBO-XGBoost, is proposed to predict TBM advance rate. A total of 350 sample data from a TBM construction project in the west wing gas control roadway of a coal mine in Huainan, Anhui Province, were collected to train and validate the model. The capability and applicability of the model were evaluated using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). The results demonstrate that the NRBO algorithm effectively optimizes the hyperparameters of XGBoost, enabling the NRBO-XGBoost model to achieve high stability and generalization performance. In the testing phase, the R², RMSE, and MAE values of the NRBO-XGBoost model reached 0.981, 0.792, and 0.512, respectively, outperforming the standalone XGBoost model. Additionally, stable prediction results were also obtained for new data. Based on the SHapley Additive exPlanations (SHAP) method, the importance of input parameters was analyzed from both global and individual perspectives, showing that the top six influential variables were penetration index, cutterhead rotational speed, rock abrasion value, thrust, torque, and compressive strength of the rock.