Advanced AI-Powered Solutions for Predicting Blast-Induced Flyrock, Backbreak, and Rock Fragmentation

Abstract

Blasting operation plays a crucial role in open-pit mining projects and significantly affects the mining efficiency and operational costs. However, blasting operations are usually accompanied by several side effects, such as backbreak and flyrock hazards, which result in wasting explosive energy, damage to the surrounding environment, and poor rock fragmentation. Due to the complex nonlinear relationship between the blast pattern parameters, rock characteristics, and foregoing hazards, the conventional criteria and simple regression analysis cannot provide highly accurate and reliable predictive models. In this study, based on the compiled 152 datasets from four different open-pit mines in Iran, six machine learning (ML) algorithms, including K-nearest neighbor (KNN), random forest (RF), eXtreme gradient boosting (XGBoost), support vector machine (SVM), decision tree (DT), and linear regression (LR), were used to develop robust models for predicting backbreak, flyrock, and rock fragmentation size. Three different datasets containing different combinations of inputs were defined for model development, and the prediction performance of the models was evaluated using R2 and root mean square error (RMSE) indices. The results showed that KNN, RF, and XGBoost algorithms outperform others in predicting fragmentation, flyrock, and backbreak, respectively. Furthermore, the parameters of burden, spacing, powder factor, sub-drilling, hole depth, and uniaxial compressive strength were identified as the best set of inputs for ML-based model development. The sensitivity analyses also revealed that blast design parameters of stemming, hole diameter, and sub-drilling have the highest impact on the prediction of flyrock, rock fragmentation, and backbreak, respectively. Finally, the SHapley Additive exPlanations (SHAP) analysis significantly improved the interpretability of the developed ML models and provided more insight regarding the intricate relationships between the parameters.