Quantification of mineral textures for geometallurgical predictive modelling

The demand for metals and minerals is increasing exponentially with time. Moreover, ore deposit grades are diminishing while ore complexity is increasing, leading to a net increase in the consumption of resources (e.g., energy and water) for the extraction and processing of metals and minerals, mainly at the concentration stage. Knowledge of mineral textures is important for predicting the efficiency of metallurgical and environmental processes, as they are linked to the liberation of metal species, energy consumption in comminution, tailings rheology, among others. Recent advances in rock imaging characterisation techniques, such as hyperspectral core scanning, offer the acquisition of spatially-arranged compositional data that enables the quantitative analysis of mineral textures.
This research has developed and applied new methods for rock characterisation using novel computational science techniques, which are used to generate highly detailed quantitative mineral texture information. Firstly, the new Mineral Co-Occurrence Probability Fields (MCOPF) method is described and applied to the acquisition of quantitative measures of mineral texture from hyperspectral core imaging database for a South American copper porphyry deposit. These mineral textures are then divided into a manageable number of textural groups using spectral clustering , providing a new layer of data for further geological and geometallurgical modelling. Secondly, MCOPF and textural cluster outputs were combined with geochemical data and machine learning techniques to generate predictive models of comminution energy consumption and copper recovery by flotation. These results quantify the impact of the MCOPF textural information, suggesting a potential increase in the accuracy and precision of the predictions. Finally, the MCOPF and mineral texture clustering method was applied to analyse the results of a particle flotation simulation developed using Scanning Electron Microscope (SEM) images of rock samples, providing insight on the causes of low copper recoveries yielded by certain rock units.
This thesis contributes to the challenges of the minerals industry by developing and testing methodologies that quantify mineral textures that can be inserted into the workflows of mining and metals concentrating and processing operations. The overall results indicate that the combined MCOPF and mineral texture clustering method is a robust and repeatable approach to quantifying mineral textural traits that have demonstrated benefit for optimisation of minerals processing. The workflow provides valuable information that may lead to the improvement of the value estimation of an ore deposit at the exploration stage, and to a reduction in the net energy and water consumption, and a better waste management at extraction and concentration stages.