Prediction of disc cutter wear considering ground conditions and TBM operating parameters

초록

Tunnel boring machines (TBMs) are being exceedingly used in tunnel projects to minimize noise and vibrations during excavation. The disc cutter, which is a tool mounted on the cutterhead of a TBM, continuously interacts with the ground because of the rotation of the cutterhead, resulting in inevitable wear. Because replacing the disc cutter involves substantial time and cost, it significantly reduces the operation and excavation rates of the TBM and impacts the total time and cost of the TBM project. Thus, accurate prediction of cutter wear or lifespan is crucial. In this study, we estimated the coefficient bearing rate (mm/km), which is an index for quantifying wear, considering both the ground conditions and operating parameters of the slurry TBMs. To address the issue of insufficient uniaxial compressive strength (UCS) data, we first performed UCS prediction, followed by prediction of the Coefficient Wearing rate. The estimation of the Coefficient Wearing rate utilizes machine learning techniques, such as the k-nearest neighbors, Support vector machine, random forest, gradient XGBoost, LightGBM, and CatBoost. The entire dataset was divided into five training/test sets for the purpose of five-fold cross-validation. The model performance was evaluated using the coefficient of determination (R2) and root mean square error (RMSE), with the hyperparameter optimization process conducted via five-fold cross-validation. The average R2 and RMSE values for the LightGBM on the test data were 0.624 and 0.00791, respectively. Finally, to evaluate the feature importance of the prediction models, a SHapley Additive exPlanation (SHAP) analysis was conducted. © 2024 The Author(s).

키워드