From Exploration towards Predictive Geometallurgy - The Role of SEM-based Automated Mineralogy and Statistical Assessment

Geometallurgical models are constructed to quantitatively predict how ores will behave during extraction and beneficiation. Depending on data availability, complexity of data and operational stage different classes of geometallurgical models can be distinguished: 1) resource potential, 2) recoverable resources, 3) first order predictive models, 4) predictive models, and 5) real-time mining models. Here, two case studies are presented where modal mineralogy and microstructural data obtained from SEM-based image analysis are combined with complementary analytical data and statistically assessed in order to predict raw material behaviour during mineral processing. For both case studies, the necessary steps to develop existing models into truly predictive geometallurgical models are outlined. The first case study concerns the recovery of Sn from a historic flotation tailings storage facility. The second case study centres on the recovery of PGE as by-products from a chromite ore deposit as a first order predictive geometallurgical model.