Quantitative image textural analysis for copper flotation recovery

A key aim of geometallurgy studies is to improve predictions of rock behaviour throughout the mining value chain by better understanding relationships between rock properties and mineral processing response. With the current industry trend of diminishing grades and the increased mineralogical complexity of ore bodies, this type of information has become critical for the accurate estimation of value in a mining project. Furthermore, geometallurgical modeling can also provide important information for decision-making and process optimisation. In particular, flotation process quality is strongly affected by the spatial attributes of minerals such as ore mineral grain size distribution and their spatial association to gangue. These textural features affect the entrainment of undesired minerals into concentrate and affect the recovery of the valuable minerals by different mechanisms. In this study, the improvement in the prediction of copper flotation recovery when using textural features was assessed in a dataset belonging to a porphyry Cu deposit in northern Chile. To account for texture, a feature extraction method named Mineral Co-Occurrence Probability Fields (MCOPF) was applied to a dataset of VNIR-SWIR hyperspectral imaging (500 μm pixel size) of the same drill cores from which the flotation samples were taken from. The MCOPF method provided a means to assess and identify textural and mineralogical traits that can be further linked to the mineral processing behaviour of certain geological units. As a result, a repeatable and robust way of incorporating texture into the geometallurgical workflows of an operating mine is proposed, generating new value from the emerging imaging tools for rock characterization that are commercially available