Energy-driven modeling and dominant factor analysis of disc cutter wear across different cutterhead regions in hard rock tunneling utilizing XGBoost

Abstract

Disc cutter wear significantly affects the efficiency and safety of TBM excavation in hard rock tunneling. To enhance wear management, region-specific wear estimation and dominant factors analysis are required for the inner, face and edge cutters of the cutterhead. Therefore, a comprehensive characterization of the cutter–rock interaction process should be considered. This study established an energy-based modeling framework to quantitatively estimate disc cutter wear and identify the dominant wear factors in each region. The cutter–rock interaction was characterized through deriving an energy‑based feature set that primarily focused on sliding friction, rolling friction and impact energy. Actual engineering data were used to develop XGBoost models for wear estimation in different cutterhead regions based on these energy features. The results showed that energy-driven models achieved higher estimation accuracy compared to models using only operational parameters. The XGBoost modeling combined with SHAP analysis revealed that sliding friction was the dominant wear factor for inner cutters, rolling friction and impact were primary for face cutters, while edge cutter wear was mainly governed by impact and sliding friction. This energy-driven modeling approach not only enabled accurate wear estimation but also clarified the dominant mechanisms of cutter wear across different cutterhead regions, offering guidance for wear management and failure prevention in TBM operations.