Platinum group elements (PGE) are mineral resources that serve as strategic economic drivers for the Republic of South Africa. Most of the known to date remaining reserves of PGM’s in South Africa are found in the UG2 chromite layer of the Bushveld Igneous Complex. Platinum concentrators experience significant losses of valuable PGE in their secondary milling circuits due to insufficient liberation of platinum-bearing particles. The interlocked texture between chromite and the valuable minerals predisposes the PGM ores to an inefficient froth flotation and thereby leads to drastic problems at the smelters. The Council for Mineral Technology (Mintek) aimed at improving the secondary ball milling of the Platinum Group Ores by optimisation of the ball milling parameters from the perspective of a preferential grinding of the non-chromite component in the UG2 ore. To this end, we looked at determining which one amongst speed, liner profile and ball size better controls the energy consumed. Moreover, this work sought at determining which combination of the above variables services the reduction of the chromite sliming of UG2 ores. Prior to the experimental work, preliminary evaluations of the load behaviour and power draw under different milling conditions were performed by use of the Discrete Element Modelling (DEM). The DEM was also used to assess the distributions of tumbling mill’s dissipated impact energy between balls and between balls and liners. The ability of Discrete Element Modelling (DEM) to match selected experimental scenarios was as well appraised. The actual ball milling test results indicated that variables, such as mill liner profile and ball size affect the milling efficiency and the size distribution of the products whereas, the mill rotational speed had little to no effect. Use of square lifters and small balls enhanced the grinding efficiency. These results agreed fairly well with the DEM simulation predications. A model describing the chromite composition as a function of density and particle size was also developed. The model was found to be reliable in the range of data tested and proved to be a strong function of the ore sample density and the grain size was acting as a correction factor.