Self-Organizing Maps Analysis of Chemical–Mineralogical Gold Ore Characterization in Support of Geometallurgy

Abstract

Few studies have been published on the analysis and correlation of data from process mineralogical studies of gold ore employing artificial neural networks (ANNs). This study aimed to analyse and investigate the correlations obtained by the technological characterization of auriferous ore using an ANN called self-organizing map (SOM) to support geometallurgical studies. The SOM is a data analysis technique in which patterns and relationships within a database are internally derived and the outputs are visual, assisting in the understanding of data in the representation of 2D maps. In the representation generated, it was possible to establish that the variables of accessibility, exposed perimeter, median gold grain diameter (D50), and SiO2 and arsenic contents have strong positive correlations. Regarding geometallurgy, this study shows that SOM can identify large-scale spatial chemical–mineralogical gold ore patterns, which can help define the most relevant indicator variables for mineral processing.