Abstract

The current challenges of Chilean mining include improving the performance of metallurgical processes utilizing few-resources. Some authors have proposed strategically modification the operating variables of processes for achieving this aim. This procedure is usually quite expensive because it is performed via experimental trials. This way, the trials to perform must be skillfully selected. The literature proposes many techniques for this purpose; nevertheless, among these highlights the response surface methodology (RSM). One advantage of the RSM is that it needs a fewer number of experiments, which means it is cheaper and requires less time. The objective of this work is determining optimal flotation conditions of a copper ore with high clay-bearing. RSM was used for this purpose. First, the experiment design utilized was Box-Behnken. Second, the response surfaces were obtained utilizing polynomials and artificial neural networks (ANN), comparing both techniques for determining the best alternative. Third, the optimal conditions for maximizing the copper recovery and inhibit the pyrite recovery were determined using swarm intelligence. The input variables studied were air flow, p80, collector dosages, and frother dosages. The solution used was seawater, and it was not used reagent for inhibiting the pyrite flotation. The results indicate that polynomials cannot describe the metallurgical behavior of copper ore, i.e., adjustment error was large. On the contrary, the ANNs are capable. The optimal operating conditions determined for achieving the metallurgical aims earlier mentioned were following air flow: 0.538 cm/s, p80: 190.7 μm, collector dosages: 10.5 ppm and frother dosages: 18.9 ppm. The copper and iron recoveries were 89.2% and 35.12%, respectively, being a significant improvement compared to standard tests. These metallurgical results were validated experimentally.