Key points are not available for this paper at this time.
Ensuring pillar stability in hardrock room and pillar mining requires accurate stress determination, particularly in complex geological environments such as the Great Dyke of Zimbabwe. Traditional empirical methods, including the Tributary Area Method (TAM) and Coates’ Method, have been widely used but often fail to capture the anisotropic and heterogeneous nature of rockmasses. This study presents an advanced framework for mine pillar stress determination, integrating analytical, numerical modelling, machine learning, and geostatistical techniques to improve predictive accuracy. The primary objectives are to critically evaluate the limitations of traditional methods, assess the efficacy of numerical models such as Finite Element Method (FEM) and Discrete Element Method (DEM), explore the predictive power of machine learning models including Gradient Boosting Machines (GBM) and Extreme Gradient Boosting (XGBoost), and analyse the implementation of geostatistical techniques for improved spatial interpolation of stress distributions. A multi-method approach is employed, leveraging our recent practical studies conducted on the Great Dyke. Numerical simulations provide high-resolution stress distribution insights, while machine learning models trained on large datasets demonstrate superior predictive performance. The results highlight the significant advantages of integrating these techniques, with machine learning models achieving high accuracy in pillar stress prediction and geostatistical methods effectively mapping stress variations. The findings contribute to mining research by providing a hybrid methodological framework for real-time stress monitoring and proactive pillar design. The study emphasises the necessity of incorporating dynamic, data-driven approaches in mine design, addressing the limitations of conventional methods. It recommends the adoption of a combined analytical-computational framework to enhance pillar stability, optimise resource extraction, and minimise failure risks. These advancements mark a significant step toward resilient and efficient mining operations in structurally complex environments.
Zvarivadza et al. (Sun,) studied this question.