Predictive geometallurgical modelling

The modern mining industry faces a number of important technical challenges, such as declining ore grades, complex mineral associations, fine-grain size and increasing geological variability. To meet these challenges geometallurgical models are constructed to quantita-tively predict how ores will behave during extraction and beneficiation. Current geometal-lurgical programs carried out by industry, focus on the definition of larger spatial domains that have similar characteristics. However, a consequent application and further develop-ment of geometallurgical programs should lead to an implementation of spatially more highly resolved geometallurgical resource models that are truly predictive for each mined block (including uncertainty), in order to improve raw material quality control and the process efficiency of mining operations. Such an approach would also enable the targeted recovery of by-products, which may generate significant additional revenue and improve overall efficiency. In order to construct relevant resource models, detailed quantitative information on the spatial distribution and geometallurgical behaviour of the by-products within the deposit is crucial.
The aim of this thesis is the development and creation of a predictive geometallurgical model by means of a case study and the presentation of the resulting advantages for the extraction of ores. Based on this, a general structure for the development of predictive geo-metallurgical models is developed, which can be applied to different types of commodities, as well as by-products, is cost-efficient, able to adapt to future data, and predicts metallurgi-cal parameters. As the case study serves the Thaba Chromium Mine in the western Bushveld Complex (South Africa), which is operated by Cronimet Chrome Mining SA (Pty) Ltd. In par-ticular, this thesis focusses on four distinct chromitite seams of the Lower and Middle Groups (LG and MG) at Thaba Mine, namely the LG-6, LG-6A, MG-1 and MG-2, which are considered as target seams for an open-cast and future underground mining scenario.
In order to understand the geological and geochemical architecture of the Thaba Mine deposit and as a foundation of the predictive geometallurgical model, an extensive geo-chemical dataset as well as logging and drill core data provided by Cronimet was evaluated and a 3D geological model was developed. A statistical assessment was performed to evalu-ate the variability within and between the chromitite seams and to separate the mine lease area into distinct geochemical clusters. The distribution of the samples belonging to the dif-ferent geochemical clusters was then transposed onto the geology of the mine lease area. This allowed the definition of spatial domains. These spatial domains, recognized by the as-sessment of assay data only, are then validated by mineralogical attributes; implications for mineral beneficiation are tested and verified.
According to this assessment, the chromitites of the Thaba Mine area can be subdivided into three distinct geochemical domains, domains that constitute the suitable foundation for a geometallurgical model. An extensive supergene altered domain or weathered domain is distinguished from a domain affected by hydrothermal alteration. The latter domain occurs below the depth of modern weathering, but in obvious proximity to faults and around a prominent dunite pipe. The third domain is represented by ores least affected by post-magmatic alteration processes. This domain occupies the centre of fault blocks below the extent of modern weathering.
Furthermore, the geochemical data is used to develop a tailored and easy-to-use multi-variate classification scheme for the chromitite layers in the Thaba Mine, based on a com-prehensive classification routine for the LG and MG chromitites. This routine allows a clear attribution with known uncertainty of all relevant chromitite layers. It comprises of a hierar-chical discrimination approach relying on linear discriminant analysis and involves five dis-tinct steps. Overall classification results for unknown samples belonging to one of the layers are 81 %. The approach may, however, be extended across the entire Bushveld, provided that an appropriate geochemical data set is available.
For detailed characterization of the mineral assemblages in the chromitite ores, selected core samples of the target layers were analysed in detail by various analytical methods, such as Mineral Liberation Analysis and Electron Probe Microanalysis. Therefore, we extended the common measurement protocols for electron probe microanalysis to ensure applicability to a wider range of PGM compositions and its overall accuracy as well as consistency. Based on the results two distinct major mineral assemblages are defined: The first assemblage is rich in platinum group element-sulphides, along with variable proportions of malanite/ cu-prorhodsite and alloys of Fe and Sn. The associated base metal sulphides are dominated by chalcopyrite and pentlandite, along with pyrite and subordinate millerite/ violarite. Associ-ated silicates are mainly primary magmatic orthopyroxene and plagioclase. The second as-semblage is rich in platinum group element-sulpharsenides and -arsenides as well as -antimonides and -bismuthides, which are associated with a base metal sulphide assemblage dominated by pentlandite and Co-rich pentlandite. The assemblage is also marked by an abundance of alteration minerals, such as talc, serpentine and/or carbonates, which are closely associated with the platinum group minerals. Statistical evaluation reveals that these two mineral assemblages cannot be attributed to their derivation from different chromitite layers, but document the effects of pervasive hydrothermal alteration.
The knowledge of the detailed mineralogical investigation was transferred to a large da-taset comprising similar mineralogical data for unweathered ore of the deposit. Hence, it was possible to identify seven distinct ore types via statistical assessment, subsequently val-idated through beneficiation tests of drill core material. In addition, metallurgical test work for large batch samples of the weathered domain was carried out. Furthermore, beneficia-tion tests were aligned with process chemistry and mineralogy to monitor the results.
The predictive geometallurgical model aims to express the recoverability of PGE as by-product from the chromite processing stream. Within this context, the weathered ores were regarded as a single domain, as chromite ores from this oxidized zone were consistently found to have very low PGE recoveries. Any attempt to recover PGE by flotation from this zone appears to be challenging. For unweathered ores, the approach towards a predictive geometallurgical model needs to be somewhat more complex. The following steps were performed:
(i) Building a predictive model of the recoverability of PGE as a function of chemical composition, i.e. establish a chemical proxy for PGE recoverability;
(ii) Performing a geostatistical modelling of the geochemical dataset, i.e. interpolation through cokriging, and
(iii) Combining step (i) and (ii) to generate a spatially-resolved geometallurgical model able to predict the potential to recover PGE by flotation in terms of probabilities.
The resultant predictive, spatially-resolved geometallurgical model displays the PGE pro-cessing potential in terms of probabilities and therefore incorporates uncertainty.
Based on the work flow applied in this study, a more generic framework towards a predictive geometallurgical model can be proposed that can be applied to different commodities, is able to adapt to future data, and predicts metallurgical parameters, e.g. the recoverability of an ore as probabilities (and therefore including uncertainty). Furthermore, the model can be applied to main as well as by-product and therefore represents a holistic modelling approach. Most of the modelled parameters are derived from primary ore properties (e.g. rock or particle stream), e.g. modal mineralogy, mineral association, densities, etc., combined with a minimum of empirical test work.