Enhancing flotation of oxidized copper ores through the integration of artificial neural network and the design of experiments approach for process optimization

Abstract

This study tackles the challenge of low copper recovery rates in supergene zones where copper oxides are associated with iron oxides. An artificial neural network (ANN) model was developed, achieving high accuracy (R² = 0.866) to optimize flotation processes for oxidized copper ores. Shapley values ranked sulfidizing agent (NaHS) and collector dosage (PAX) as the most influential factors, with NaHS and iron negatively affecting recovery, while PAX and copper oxide content had positive effects. Optimal conditions were validated on an industrial scale, achieving 75.66 % copper recovery, confirming the effectiveness of the optimized parameters through mineralogical analysis.