There is a need to develop a model to predict the liberation of minerals in breakage products as this is a key step in linking comminution and separation models in an integrated mineral processing simulator. Such a model would allow metallurgists to optimise grinding operations in response to changes in ore characteristics. This thesis aims to predict mineral liberation using a model based on random breakage of a measured texture, and to validate the model prediction against actual experimental breakage tests. A published, texture-based model of mineral liberation that uses the assumption of random breakage of the mineral texture to generate product particles called the JK Gaudin Random Liberation Model (JK-GRLM) (Evans et al. 2013) has been used as the basis of this study.
In earlier empirical work a number of researchers (Berube and Marchand (1984), Manlapig et al. (1985), Wightman et al. (2008) and Vizcarra (2010)) observed the heuristic model that the degree of liberation in a given size fraction is the same regardless of where in the comminution circuit a sample is collected. This behaviour was observed for various ore types and was found to be independent of the mode of breakage. However, the published research does not explicitly consider the effect of amount of energy applied to break the ore on the distribution of the mineral across liberation and particle size classes.
Two ore types were used in this study, namely ores from Kanowna Belle (KB) and Kennecott Utah Copper Corporation (KUCC). The original JK-GRLM was applied to simulate the product particle liberation distribution from breakage of the two ores. Experimental validation of the JK-GRLM was conducted by comparing the simulated and measured liberation distribution results. In this study, three different impact breakage energies were applied to samples of each ore to investigate whether the amount of breakage energy input affects the liberation behaviour of the ore. The particulate products of these three impact breakage energies were analysed using the MLA to measure the cross-sectional area liberation that was then compared to equivalent areal liberation data from the simulated results.
Comparison of the simulated and measured results for the KB and KUCC ores showed that the simulated cross-sectional area particle composition distributions using the original JK-GRLM did not predict the measured values accurately. While non-random breakage of the ore was recognised as one possible cause of the differences between the simulated and measured liberation, it was postulated that another potential reason for the inaccurate predictions is the way in which texture is represented in the original JK-GRLM. Specifically the representation of texture using an average grain size distribution, cubic grains and random spatial distribution of the mineral grains to create the feed ore texture in the breakage simulation rather than in the actual spatial arrangement and grain shape of the measured ore. Therefore, the JK-GRLM was extended to assess the effect of using measured 3D ore textures as inputs to the model.
Selected samples of KB ore textures were measured using X-ray tomography to provide inputs to the extended JK-GRLM that allowed the actual 3D ore textures to be used. These texture samples were broken using impact breakage and the product particles analysed using X-ray tomography to measure their volumetric liberation distribution. Comparison of the simulated and measured results of these samples did not show a conclusive trend. Analysis of the results indicated that the number of texture samples which were able to be analysed was not sufficient to allow for statistically valid conclusions to be drawn. Further work is required to conclude whether using measured ore textures in the extended JK-GRLM to predict mineral liberation produces more accurate estimates of liberation than the original JK-GRLM.
Occurrences of non-random breakage, specifically grain boundary breakage, were detected in the KB ore. This has been postulated as one of the reasons that the extended JK-GRLM did not predict the liberation distribution accurately.
A key finding of this research is that the amount of impact energy applied to the KB and KUCC ores has no significant effect on the degree of liberation of a mineral in a given size fraction measured using particle sections. This provides sound experimental evidence to support the heuristic model.
One important outcome of this work is an alternative method for quantifying the mineral liberation of the breakage products. This method is a combination of the heuristic model and a new model to predict mineral distribution across breakage product size fractions proposed in this study, based on extending the “map of relative ore breakage” approach of Narayanan and Whiten (1983) to minerals. The new model of mineral distribution predicts the mineral distributions of the sulphide and gangue minerals in the KB and KUCC ores well with the predicted values, showing good correspondence to the measured values for a range of input breakage energies.