Experimental investigation and machine learning prediction of water-weakening effects on rock breakability by conical pick

With the extension of mining and tunneling engineering into deep complex water-bearing strata, the interaction between groundwater and rock mass has become a critical factor governing mechanical excavation efficiency. The presence of water fundamentally alters the rock fragmentation characteristics, and understanding this hydro-mechanical coupling is prerequisite for optimizing conical pick performance. In this study, a comprehensive experimental framework combining macroscopic indentation tests and microscopic characterization was established to evaluate the breakability of twenty distinct lithologies under dry and saturated conditions. The variation of Peak Indentation Force (PIF) and cutting work was monitored, alongside micro-analysis using SEM and XRD to reveal the intrinsic controls of mineral composition and pore structure. The results demonstrate a lithology-dependent bifurcation: porous sedimentary rocks exhibit significant degradation in strength due to pore pressure wedging and chemical softening, whereas dense magmatic rocks remain largely insensitive to saturation. Furthermore, to bridge the gap between experimental data and field application, an Extreme Gradient Boosting (XGBoost) model was used. Feature importance analysis reveals that under water-saturated conditions, the Brittleness Index surpasses hardness as the dominant predictor for rock breakability. This study quantifies the water-weakening mechanism and provides a data-driven approach for predicting cutter performance and improving excavation efficiency in water-bearing environments.