Hybrid machine learning techniques to predict fuel consumption of dump trucks in an open-pit mine in Peru

Mining dump trucks account for 31% of energy consumption and up to 60% of operating costs in open pit mines. This study aims to predict the fuel consumption of mining dumpers in an open pit mine in Peru by applying innovative hybrid machine learning models, including artificial neural networks (ANN), decision trees (DT) and random forests (RF), optimised with genetic algorithms (GA). The models evaluated the efficiency and cost savings of replacing 24 m3 dump trucks with 26 m3 dump trucks. The data were divided into 70% for training and 30% for testing. Input variables included utilisation, operating hours and total trips. The RF + GA model achieved the highest accuracy (R2 = 0.89, RMSE = 250.05, MAE = 185.09), followed by ANN + GA with an R2 of 0.87 and DT + GA with R2 of 0.86. This novel approach significantly improves the prediction accuracy, showing that replacing dump trucks can save an average of US$1.76 per hour.