Geometallurgical model of the characterization of clays for flotation efficiency using machine learning methodology

Abstract

The geometallurgical model is the result of the integration of the disciplines of geology, mining and metallurgy in order to add value and reduce risk in a process such as the flotation efficiency of chalcopyrite. The objective of this study was to calibrate models to quantify the content of clays and gangas of a hydrothermal deposit of the porphyry copper-gold type by NIR spectroscopy. The selected methodology was to design a chemometric model based on 173 diamond drill composites to which X-ray diffraction tests and near infrared spectroscopy were performed, validating the results with a cross-validation through a machine learning methodology. The creation of the models was carried out by means of a regularized non-linear regression by the Ridge method. Low linearity models were obtained for calcite and plagioclase minerals, with R2 values (0.51 and 0.78, respectively). The regression model presents a linearity for smectite, quartz, orthoclase and muscovite minerals showed a high R2 (0.95, 0.93, 0.64 and 0.59, respectively). The results found for the content of clays and gangue indicate that X-ray diffraction analyzes can be largely replaced by spectral models. In the case of calcite and plagioclase, it would be convenient to carry out a characterization campaign, in order to improve the model and to be able to replace the x-ray diffraction analyzes for these species, which will allow to generate a geometallurgical model in a quick and easy way. efficient with a semi-quantitative method.