Answers are grounded in this paper and cite their sources.
Excavation of steeply inclined coal seams produces asymmetric stress redistribution, yet contours extracted from numerical stress fields are criterion-dependent and cannot be equated with observed roof fractures. This study evaluates PG-IBO, a coordinate-conditioned boundary-reconstruction network, using contemporaneous vertical-stress sections generated by Fast Lagrangian Analysis of Continua in Three Dimensions (FLAC3D) and case descriptors for the C5103 working face. The network reconstructs signed-distance fields for the boundaries of regions defined by reference-normalised compression ratios κ ≤ 0.8 and κ ≥ 1.5, with contour smoothness and prescribed zone ordering as regularisers. The benchmark comprises 77 records, including 48 interpolation-derived records, expanded to 942 images; these derivatives are not independent numerical solutions. Against simulation-derived pseudo-labels in the reported image-level test, PG-IBO achieved a boundary mean absolute error of 5.8 m, intersection over union of 0.86 and F1 score of 0.89. In the source simulations, lower-zone peak stress rose from 26.54 to 36.13 MPa over 20–120 m of advance. Archived prop-pressure measurements showed a consistent upper-to-lower loading trend but did not locate the predicted contours. The 72-image robustness subset derived from six numerical states. These findings support image-conditioned reconstruction of numerical stress contours within the C5103 setting; transfer to other working faces and roof-instability assessment require independent simulations and stage-matched field observations.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: