Modelagem e análise de estimadores lineares aplicados sobre variáveis geometalúrgicas

Abstract

Modeling activities and grade estimation are essential and routinely applied in the mining industry to evaluate mineral resources, economic and strategic planning of mining in the short, medium, and long terms, and to choose the best operations parameters. The addition of geometallurgical variables in the block model – such as recovery data and flotation tailings grades – make it a potentially deleterious material in mineral processing which, when properly programmed in an appropriate blend, can enable its use, and optimize production. As well as for classical interpolation of grade, geostatistics can be used as a tool to stamp blocks with the estimation of geometallurgical variables. However, the challenge is presented in the choice of an appropriate method that can respect the condition of non-additivity of this type of variable, since it is not possible to calculate an average value that can accurately characterize an unsampled point. Despite the limitation of linear estimation methods in the quantification of these variables, this study proposes the analysis of its use for qualitative purposes and comparison of the accuracy of the estimation between classical interpolators, widely known and used in the industry. As a case study, two distinct areas of the Morro do Ouro Mine were selected, the current property of Kinross Gold, whose attempt to develop geometallurgical models was motivated by the presence of potentially deleterious clay material. The nearest neighbor method is the one that best preserves the occurrence of extreme values and represents numerically the geometallurgical characteristic of a material, but in an ideal drilling grid.