Research on Precise Identification of Rock Strength Based on Bolt Drilling Parameters

ABSTRACT During roadway excavation, the presence of weak interlayers and fractured rock masses significantly affects roof stability. To achieve timely and effective roadway support, it is crucial to identify and predict different rock types based on drilling signals from roof bolters. This study combines theoretical analysis, laboratory tests, and numerical simulation to investigate the intrinsic relationship between drilling signals and rock properties. A hybrid denoising method called “Grey Wolf Optimizer‐Variational Mode Decomposition‐Wavelet Threshold Denoising (GWO‐VMD‐WTD)” is proposed, along with a lithology identification model based on an SSA‐BP neural network. The results demonstrate that: rotational speed, torque, thrust, and vibration acceleration show significant correlations with rock properties; the GWO‐VMD‐WTD method effectively restores authentic signals with notable noise reduction; the SSA‐BP neural network model, trained using time‐domain and frequency‐domain feature parameters, exhibits strong generalization capability, robustness, and reliability, achieving 95.8333% accuracy in rock classification. This provides critical references for real‐time monitoring and control of roof deterioration zones.