Machine learning techniques for optimizing the efficiency and costs of drill steel in sandstone and granodiorite: a case study in Peru

This research aims to optimize the efficiency and costs of drilling steel in sandstone and granodiorite rocks using machine learning techniques in a Peruvian mine. Predictive models, including random forest (RF), XGBoost (XGB), decision trees (DT), and artificial neural networks (ANN), were applied, along with optimization algorithms such as genetic algorithm (GA), particle swarm optimization (PSO), ant colony optimization (ACO), and simulated annealing (SA). A dataset of 705 entries was analyzed, focusing on drill bit wear, percussion and rotation pressures, and cost per meter drilled. Model performance was evaluated using R2, RMSE, MAE, and MAPE. The ANN model demonstrated the highest accuracy, with an R2 of 0.64, RMSE of 0.05 and MAE of 0.04, for efficiency, and an R2 of 0.96, RMSE of 0.03 and MAE of 0.03 for costs. Optimization with PSO offered the best performance, increasing the efficiency of drill steel by 360% in sandstone and 423.33% in granodiorite, with a slight cost increase of 8.82% in sandstone and a cost reduction of 32.71% in granodiorite. In conclusion, machine learning techniques proved effective in optimizing the efficiency and drilling cost in different types of rocks in an underground mine.