An Adaptive Response Surface Methodology for Determining Coal Pillar Width in Geologically Variable Longwall Panels

Simulation‐based approaches are increasingly vital for addressing complex design challenges in underground coal mining, a domain where longwall pillar design is critically challenged by geological variability, and conventional methods lack adaptive capacity. This study develops a responsive design framework integrating response surface methodology (RSM) and numerical simulation to determine optimal pillar width under variable burial depth, dip angle, and mining height. Based on a Box–Behnken experimental design and FLAC 3D simulations, quadratic models were established for roadway deformation and pillar stress concentration, showing excellent fit ( R 2 > 0.94). Pillar width is the most influential factor for both responses. Significant interaction effects were identified: widening pillars reduces depth‐induced deformation but may cause stress concentration in deep zones; a threshold width of approximately 7–8 m exists, beyond which increasing mining height promotes stress dissipation; and simultaneous increases in mining height and burial depth synergistically amplify roof displacement. Field validation under challenging geological transitions (seam thickness: 7–15 m; dip: 18°–0°) confirmed the model’s reliability. Adjusting pillar width from 6 to 8 m reduced average roof deformation by approximately 50% while keeping stress concentration within safe limits. The proposed RSM‐based approach provides a computationally efficient and practically adaptable alternative to conventional pillar design in geologically variable panels.