Mechanics Based Simulation and Physics Informed Surrogate Modelling of Blast Induced Damage in Deep Coal Roof Strata

Pre-splitting blasting is commonly used to manage hard roof conditions and reduce excessive mine pressure in deep coal mining. Despite its widespread application, the quantitative link between blasting parameters and the resulting damage characteristics has not been clearly established. In this study, a physics informed surrogate modelling approach is developed to assess blast induced damage in coal roof strata using a comprehensive numerical dataset. A total of 1,248 simulation cases were examined, covering variations in peak pressure, explosive energy, burden, and key rock properties, including tensile strength, elastic modulus, and density. In order to better reflect the underlying mechanics, three derived variables were introduced: the pressure to tensile strength ratio, energy per burden, and stiffness to density ratio. Random Forest and XGBoost models were then trained to predict fracture radius, while SHAP analysis was applied to evaluate the contribution and interaction of each parameter. The analysis shows that the pressure to tensile strength ratio has the strongest influence on damage development, with energy related and geometric factors also playing important roles. The surrogate models demonstrated strong predictive capability and were able to represent the nonlinear behaviour of the system. Compared with conventional trial and error design methods, this approach offers improved computational efficiency without sacrificing physical insight. The configuration therefore provides a practical basis for rapid assessment of blasting performance and for optimizing pre-splitting strategies in deep mining environments.