Room-and-pillar mining refers to an underground extraction technique wherein the mineral deposits are mined horizontally and where natural pillars of rocks are left to support the weight of the upper layers of rock formation. As room-and-pillar mining is conducted further into the deeper levels of the mining field, the level of stress that is placed on the ground becomes much higher, making it highly likely that the pillar could fail—a structure breakage which could lead to a chain reaction collapse.
In the past, the assessment of pillar stability for mining engineering was based largely on empirical approaches. The major approach that was used was the Safety Factor (SF), in which the ratio between the pillar’s capacity and the load acting on the pillar is determined. However, classical empirical formulas and tributary area methods have shown their shortcomings with respect to accounting for the complex non-linear stress behavior at very deep levels (Kamran et al., 2024). Therefore, modern studies in geomechanics are becoming more dynamic and are based on the use of field-dependent parameters.
One important advancement is the use of complex 3D numerical modeling in order to analyze the time-dependent deformation and stress distribution. The current research uses non-linear numerical models based on elastoplastic failure criterion and analyzes the effects of variables like working depth, uniaxial compressive strength, and size of the excavation in deep mines. In one of the recent simulations, for instance, it is observed that working depth has a significant effect on both the major and minor stress concentration factors, making the use of proper parameters essential to avoid yield propagation (Mehra & Budi, 2024).
However, apart from merely modeling, the use of AI in geomechanics represents a complete paradigm shift in the prediction of pillars’ failure. Considering the high number of dimensions of the geological parameters, ML algorithms are currently being applied. Predictive models developed using the techniques of gradient boosting decision trees and extreme gradient boosting (XGBoost) algorithms have proved to be highly accurate in the analysis of pillar’s stability using historical data sets. Such algorithms are able to detect the key predictors, for instance, ratio of pillar’s width and height, as well as average pillar’s stress (Liang et al., 2020).
More recently, computing approaches have evolved towards decision intelligence-driven prediction models. With the incorporation of predictive algorithm and optimization approach, geomechanical studies have become highly robust. For example, the amalgamation of K-Nearest Neighbor (KNN) algorithm and meta-heuristic optimization algorithms like grey wolf optimizer helps in processing the complicated and nonlinear geological data (Kamran et al., 2024). Similarly, least squares support vector machine models, which have been optimized using subtraction average optimization algorithms, have been used to effectively model the nonlinear relation between deep rock mass strength and pillar size (Xie & Zhang, 2024).
In summary, modern research into geomechanics clearly demonstrates that the prediction of failures of pillars in deep room and pillar mines cannot be based on static formulae alone. In addition to the use of numerical modeling based on three-dimensional elastoplastic theory, machine learning algorithms offer the means for a more sophisticated and accurate method of structural analysis. The deeper the mining, the more crucial it is to use such intelligent predictive models.
References
Kamran, M., Chaudhry, W., Taiwo, B. O., Hosseini, S., & Rehman, H. (2024). Decision Intelligence-Based Predictive Modelling of Hard Rock Pillar Stability Using K-Nearest Neighbour Coupled with Grey Wolf Optimization Algorithm. Processes, 12(4), 783. https://doi.org/10.3390/pr12040783
Liang, W., Luo, S., Zhao, G., & Wu, H. (2020). Predicting Hard Rock Pillar Stability Using GBDT, XGBoost, and LightGBM Algorithms. Mathematics, 8(5), 765. https://doi.org/10.3390/math8050765
Mehra, A., & Budi, G. (2024). 3D Modelling approach to identify parametric configurations for pillar stability in underground metal mine: a case study. Geomatics, Natural Hazards and Risk, 15. https://doi.org/10.1080/19475705.2024.2367630
Xie, X., & Zhang, H. (2024). Research on Hard Rock Pillar Stability Prediction Based on SABO-LSSVM Model. Applied Sciences, 14(17), 7733. https://doi.org/10.3390/app14177733
