Uncertainty-aware slope stability assessment of opencast coal mines using PLAXIS-based numerical modelling and hybrid ANN–XGBoost prediction
Abstract
ABSTRACT Slope stability is a critical concern in open-pit coal mining, where slope integrity directly impacts operational safety and environmental sustainability. This study develops a hybrid modelling framework integrating PLAXIS-based numerical simulations with Artificial Neural Networks (ANN) and Extreme Gradient Boosting (XGBoost) to improve slope stability assessment. The study evaluates the hybrid model against traditional methods, examines the influence of geotechnical parameters, and incorporates bootstrap-based prediction uncertainty analysis. Results demonstrate enhanced predictive accuracy, with an R2 of 0.94 and RMSE of 0.10. Sensitivity analysis identified groundwater depth and slope angle as the most influential parameters affecting model predictions. Bootstrap-based uncertainty analysis further revealed an increased probability of failure for marginally stable slopes. The proposed framework provides a reliable approach for incorporating numerical modelling, machine learning, and uncertainty analysis into slope stability assessment and management in open-pit mining operations.