A mechanics-based stress-energy framework for pillar-burst hazard classification in underground marble room-and-pillar mines

Abstract
Pillar-burst hazards in underground marble room-and-pillar mines pose a severe risk to worker safety and operational continuity. Their systematic characterisation remains limited by sparse field monitoring records and the absence of reproducible datasets. This study addresses this gap by developing a mechanics-based stress-energy framework for pillar-burst hazard classification that integrates stress-proximity-to-failure with energy-release potential.
A physics-consistent parametric database was constructed by sampling empirically justified ranges of depth of cover, excavation geometry, stress regime, and rock mass quality derived from documented underground stone and marble mining operations. Pillar stability was evaluated using five established hard-rock pillar strength formulations to produce conservative and optimistic capacity bounds, representing epistemic uncertainty in pillar strength estimation. Burst potential was quantified through strain energy density and a releasable energy proxy that captures energy accumulation and the spatial extent of potentially unstable rock. Stress-based and energy-based indicators were fused conservatively into five ordinal hazard classes using a maximum-evidence rule.

Statistical analysis on the corrected dataset confirmed strong separation among all five hazard classes (epsilon-squared = 0.928 for the normalised stress exceedance index,
p
 < 0.001 by Kruskal–Wallis test). Partial rank correlation analysis identified rock mass quality (GSI) and intact strength (UCS) as the dominant parametric controls, with the geometry-dependent stress concentration factor Sc ranking third. Epistemic uncertainty in pillar strength estimation was found to be statistically meaningful and relevant to hazard classification, supporting the use of multi-model strength envelopes over single-formula approaches. Physics-assisted machine learning models achieved high ordinal agreement (QWK up to 0.998), with misclassification concentrated between adjacent hazard classes.

The framework provides a practical four-step screening workflow enabling engineers to evaluate pillar-burst hazard from standard mine design inputs without requiring site-specific burst records. Parameter ranges are grounded in published case studies from operating marble and dimension stone mines.