A Framework for In Situ Rock UCS Assessment and Full‐Hole Strength Inversion Integrating MWD Signals and Symbolic Regression Strategies

ABSTRACT Accurate in situ estimation of uniaxial compressive strength (UCS) is essential for safe geotechnical and mining operations, yet traditional coring followed by laboratory testing rarely supports rapid strength profiling. This study develops a UCS assessment framework based on measurement while drilling (MWD) signals to provide a lightweight prediction tool. Field drilling totaling 297 m provides continuous records of penetration rate, thrust, pump pressure, torque, bit rotation speed, and depth. Subsequently, symbolic regression (SR) is employed to extract functional relationships between MWD parameters and UCS. Among the compared SR models, the sparse identification of nonlinear dynamics algorithm via sequentially thresholded‐least squares (STLSQ–SINDy) achieves the most favorable balance between simplicity and predictive accuracy, with a test coefficient of determination ( R ² ) of 0.9413 and a root mean square error (RMSE) of 5.53 MPa. Nonparametric tests and residual diagnostics further confirm its consistently superior performance under the current data conditions. To reduce the potential optimism associated with random partitioning in a small dataset, leave‐one‐hole‐out cross‐hole validation is conducted. Under this more conservative setting, STLSQ–SINDy achieves a mean test R ² of 0.8709 ± 0.0471, indicating comparatively stable cross‐hole transferability within the five boreholes. The optimal analytical expression is then coupled with filtered MWD waveforms to reconstruct rock UCS profiles for five boreholes, which capture the stratified strength hierarchy and locate critical depth windows. Overall, this work presents a systematic application of sparse SR to field MWD data and demonstrates the potential of the MWD–SR framework for low‐cost rock strength prediction.