thesis

Mineralogical characterisation of brittleness indices of platinum-bearing UG2 reefs in the eastern Bushveld Complex

The dissertation develops and compares mineralogy- and mechanics-informed brittleness indices for UG2 reef using fieldwork, microscopy, laboratory tests, simulations and predictive models.

Why brittleness matters

Brittleness influences how rock fractures during drilling, blasting and mechanical excavation. In the platinum-bearing UG2 reef, a single strength value can conceal large changes in mineral composition and grain-scale structure. This dissertation therefore investigates whether mechanical tests and mineralogical observations can be combined into more useful brittleness indices for the eastern Bushveld Complex.

A multi-evidence method

The study combines field sampling, microscopic examination, laboratory testing, numerical simulation and multivariate linear predictive models. Mechanical measurements included uniaxial compressive and tensile strength, Young’s modulus, Poisson’s ratio and peak strain. Petrographic work described mineral grains, packing and the type and nature of contacts between them.

The broad range of measured compressive strengths confirmed that UG2 should not be treated as a uniform material. New mechanical indices—labelled B1, B2, B4 and B6—were retained for further analysis. Two other candidate indices did not show sufficient correlation with the measured behaviour.

Mechanical findings

Tensile and compressive strength were the most informative mechanical predictors. Adding Young’s modulus and Poisson’s ratio increased model complexity without consistently improving engineering relevance, and peak strain was comparatively weak. This is a useful modelling result in its own right: a more elaborate equation is not necessarily a better descriptor of brittle response.

Within the evaluated dataset, the B2 and B6 predictive models produced exceptionally small standard errors. These results are promising, but they remain specific to the samples and validation procedure used in the dissertation rather than proof of universal performance.

Mineralogical findings

Grain-contact type and contact nature carried the strongest mineralogical signal. Models based on these variables achieved coefficients of determination of 0.778 and 0.727 respectively. Packing characteristics were moderately informative at 0.473, while a texture-only model performed poorly at 0.0915. Sutured, concavo-convex, longitudinal, floating and point contacts were among the microstructural features considered.

The contrast shows why mineralogical descriptions should go beyond a list of constituent minerals. The way grains meet and interlock can be more consequential for fracture than texture alone.

Practical interpretation and next steps

The proposed indices can support safer excavation planning, rock-cutting assessment and machine selection by connecting laboratory strength to visible microstructure. Future research should include a broader range of minerals and geological settings, field validation and more advanced machine-learning methods. Such work could turn the dissertation’s correlations into robust, site-transferable predictions of drillability and cutting response.