Abstract
Comminution is a very important part of mineral beneficiation chain as it affects all the entire downstream processes if sufficient liberation of the valuable mineral is not achieved. The ball mill is commonly used for milling such that the mineral particles can be liberated. However, the ball mill is very inefficient and consumes about 50% of the mineral processing expenses due to excessive energy and grinding media consumption. Spherical media is the commonly used grinding media though there is ongoing research on other media shapes that may improve the ball mill efficiency. Various studies have been conducted to really understand the grinding kinetics and load behaviour inside the ball mill. Several researchers have worked on the effect of grinding media shape using cylpebs, cubes, ellipsoids and worn balls varying the critical speed and the grinding time. The current work was focused on the effect of using three different grinding media on the liberation of platinum group of metals (PGMs) in a UG2 ore using the response surface methodology (RSM). The characterization of the ore was conducted using the X-ray fluorescence (XRF) and the X-ray diffractometer (XRD) to obtain the chemical and mineralogical composition of the ore. The ore was crushed using a jaw crusher and a double roll crusher then milled afterwards. The grinding media shapes used were spheres, cubes, and truncated ellipsoids. The interstitial filling (U), ball filling volume (J) and percent solids were the operational variables varied during the study. The design of experiments (DOE), optimisation and validation of these parameters was conducted using RSM in Minitab 21 software. The DOE using central composite design (CCD) in RSM generated 20 runs for each grinding media which proffered 2 response models for each grinding media. The particle size distribution (PSD) of the milled product was determined by a stack of screens in the√2 sequence and the energy consumed was measured by a power meter. The use of surface plots, contour plots, main effect plots and the prediction tools helped in optimizing and validating the models. The significant parameters in the models were determined by the ANOVA analysis using the p values together with the Pareto charts. The R2 values from the model summary indicated whether the model fits the data well or not. Wear rate tests were also performed for three grinding media using the same parameters. Chemical analysis of the ore revealed the presence of Si (14.3%), Fe (14.8%), Mg (9.58%), Cr (9.36%), Al (4.98%), Cu (0.034%) and Zn (0.036%). The mineralogy consisted of 8 phases which were enstatite, aluminum oxide, magnesiochromite, anorthite, magnetite, millerite, djurleite and chalcocite. The results proved spheres to be the most efficient compared to truncated ellipsoids and cubes. Cubes
vii
performed the least. At optimum conditions, the % passing 75 μm was 82.58%, 80.41% and 77.07% for spheres, truncated ellipsoids, and cube respectively. The energy usage was 0.211KWh, 0.23KWh and 0.22 KWh in the same order of the grinding media. From the R2 values obtained, all the models fit the data. However, the models for the % passing 75 μm for all the grinding media had very high R2 values as compared to the energy usage models and they could be used for prediction whilst energy usage models were overfits. The most significant parameter of all the tested parameters on % passing 75 μm was U for all grinding media.
Keywords: Response surface methodology, comminution, ball mill, particle size distribution