Application of machine learning in mitigating ore and profit losses induced by blast-induced rock movement

During open-pit bench blasting, rock fragments are thrown toward the bench face, leading to significant differences in the spatial distribution of ore and waste before and after blasting. If pre-blast ore boundaries are directly used to define dig-limits and guide loader operations, substantial ore loss and economic losses may occur. Two distinct approaches can be adopted to mitigate ore loss induced by blast-induced rock movement (BIRM): optimizing blasting to allow direct application of pre-blast boundaries, or combining rock movement prediction with dig-limit determination. However, existing research on blasting simulation technologies and dig-limit determination methods remains insufficient to meet practical engineering requirements. To address these challenges, this study makes two main contributions by incorporating machine learning techniques. First, a machine-learning-aided (ML-aided) blasted muckpile analysis method was proposed, based on an innovative ML-assisted approach for post-blast ore boundary determination developed by the author. The effectiveness of this method was validated through numerical simulations and case studies involving dividing (DOPB) and center-initiation open-pit blasting (CIOPB). Second, three dig-limit determination methods (Binary integer linear programming-based method (BILP-based), genetic algorithm￾based method (GA-based), and practical heuristic-based method (PH-based)) incorporating rock movement prediction were developed and systematically evaluated through case studies. Furthermore, by integrating BIRM and equipment size constraints into dig-limit determination, a more efficient multi-layer dig-limit determination approach was proposed, achieving improvements in both ore recovery and economic benefits. Overall, this study provides valuable theoretical insights and practical guidance for mitigating ore loss and economic degradation in open-pit mines with complex geological conditions.