Data-Driven Prediction of Stress–Strain Fields Around Interacting Mining Excavations in Jointed Rock: A Comparative Study of Surrogate Models

Assessing the stress–strain state around interacting mining excavations using the finite element method (FEM) is computationally expensive for parametric studies. This study evaluates tabular machine-learning surrogate models for the rapid prediction of full stress–strain fields in jointed rock. A dataset of 1000 parametric FEM simulations using the elastoplastic Hoek–Brown constitutive model was generated to train Random For-est, LightGBM, CatBoost, and Multilayer Perceptron (MLP) models based on geometric features. The results show that the best models achieve R2 scores of 0.96–0.97 for stress components and 0.99 for total displacements. LightGBM and CatBoost provide the op-timal balance between accuracy and computational cost, offering speed-ups of 15 to 70 times compared to FEM. While Random Forest yields slightly higher accuracy, it is re-source-intensive. Conversely, MLP is the fastest but less accurate. These findings demonstrate that data-driven surrogates can effectively replace repeated FEM simula-tions, enabling efficient parametric analysis and intelligent design optimization for mine workings.