Predicting energy consumption SAG mills through Bayesian generalized linear model and random forest

The mining industry consumes about 1.7% of the energy generated worldwide, which is expected to increase in the coming decades. Milling is the most energy-intensive process of a typical mining operation. Many variables (e.g., rock characteristics, mineral matrix, and equipment properties) affect energy consumption. This paper proposes that Random Forests and the Generalised Linear Model (GLM) be used to predict the energy consumption of the SAG mill, which significantly contributes to the energy consumption of mining operations. To show the performance of the proposed approach, a case study was applied to a copper mine dataset from South America. The proposed approaches were applied to forecast the SAG mill energy consumption. The outcomes demonstrated that these methods could be used to predict energy consumption. Random Forest can have a high prediction accuracy of 95% but lacks explanatory ability, as shown in R2 at 50%. GLM provided additional insights by showing the feature importances and their relationships with SAG mill energy consumption, along with considering the potential uncertainties and generating posterior probability distributions for the model outcomes. Both models identified key variables as significant predictors identically, with the GLM offering a more comprehensive view of best-case and worst-case energy consumption scenarios.