GAPSO-LightGBM-based Intelligent Prediction Method of Surrounding Rock Grade in TBM Tunnelling

The tunnel boring machine (TBM) excavation parameters are sensitive to changes in geological condi? tions. Accurately and real-time identifying the quality grade of tunnel surrounding rocks is crucial for efficient tunnelling and geological hazard prevention. To address this, pre-processing is performed on TBM excavation parameter data to obtain a high-quality database. Selected the LightGBM as the basic model, a surrounding rock grade prediction model based on GAPSO-LightGBM is constructed by introducing the GAPSO optimization algorithm for hyperparameter tuning. The model's performance is compared with PSO-LightGBM, GA-LightGBM, LightGBM, XGBoost, and Random Forest models. The results show that the GAPSO-LightGBM-based surrounding rock grade prediction model outperforms the traditional models. For the prediction of grades Ⅱ, Ⅲ, and Ⅳ surrounding rocks, the F1 scores are 0.849, 0.871, and 0.893, with an accuracy of 87.5%. The field validation demonstrate that this method can effectively predict the changes in surrounding rock quality grades and provide reference for practical engineering.