206 changes during advance and liquidation (Potvin & Hadjigeorgiou, 2010). A key contribution of this study is the calibration of the FEM model using spatial underground convergence measurements, which improves realism relative to workflows relying only on laboratory intact-rock parameters. This follows the established practice of updating predictions using field monitoring to reduce uncertainty, and the calibrated model reproduces excavation-level deformation with good agreement. The adopted approach is consistent with commonly used rock-mass parameterisation concepts that link intact-rock test results to rockmass behaviour (Hoek & Brown, 1997). The results also show that “optimal” exploitation cannot be selected solely by maximising safety indices or recovery; the deformation character and load-transfer mechanism must be considered. Abrupt post-peak behaviour in stiff systems is widely regarded as operationally unfavourable at depth due to its association with damage localisation and dynamic instability processes (Brady & Brown, 2006). A current limitation is the sparse monitoring base, which reduces parameter identifiability and forces calibration to remain strictly iterative. With more data, the workflow could move toward a more constrained updating scheme consistent with inverse-problem concepts, reducing iterations and improving predictive confidence. 5. CONCLUSIONS This study presented a FEM-based workflow to support the selection of an optimal exploitation variant under deep mining conditions. A 3D numerical model was built for the analysed panel and used to compare three pillar-size scenarios (4×4 m, 5×6 m, and 8×8 m) across successive mining stages, with particular attention to stability evolution along the reference cross-section and the transition into the liquidation zone. Model parameters were updated through an iterative calibration procedure using underground convergence observations, resulting in a calibrated model that reproduces the in-mine deformation response with a good level of agreement and therefore provides a credible basis for comparative decisionmaking. The comparative results indicate that extreme variants are constrained by different limitations: the smallest pillars may lead to unacceptable stability margins (local SF < 1), whereas the largest pillars—although stable in the early stages—may promote an overly stiff system and an unfavourable, abrupt loss of load-bearing capacity during late-stage geometry changes. The intermediate 5×6 m configuration provides the most balanced performance, maintaining an acceptable stability margin while preserving high recovery, and therefore represents a practical compromise between safety and extraction efficiency. In the next stage of work, the focus will be placed on further development of the methodology by establishing stress–strain relationships for pillars under varying geometries and mining stages. This will enable a more direct parameter updating strategy, reduce the number of calibration iterations, and significantly increase the computational efficiency of the workflow, moving from strictly iterative tuning toward a more constrained and faster calibration process.
RkJQdWJsaXNoZXIy MTM0Mzk2