Rapid assessment of the sorting potential of copper porphyry ores through modelling of textures and grade distributions

Abstract

The research objective of this PhD project was to develop mathematical models that will predict the liberation of barren gangue through size reduction of the ore to assess the ores amenability to mineral processing methods such as sorting. An hypothesis was formulated that related ore sortability with gangue liberation through size reduction mechanisms such that the fundamental driver for sortability was the proportion of liberated gangue that could be generated at any given particle size. Since the aim of ore sorting is to reject mass out of an ore stream at the lowest possible metal loss, sorting can only be exploited on ores that generate sufficient quantities of barren gangue at particle sizes suitable for ore sorting technologies e.g. greater than 10 to 20 mm. Three complimentary mathematical models were developed, (one from first principles and two from the existing literature) that used easily measured grade and texture parameters obtained from crushed particles. These models were: 1. A refined log-normal distribution model for estimating the particle grade distribution of an ore from measurements of the mean sulphide mineral grade and mode sulphide mineral grade acquired from coarse particles. 2. A new mathematical model for estimating the sulphide mineral texture dimension from measurements of the mean mineral sulphide grade and mean sulphide mineral grain size from measurements acquired from coarse particles. 3. A refined Gaudin Random Liberation Model (GRLM) that predicts matrix liberation through size reduction on low grade ores from estimates of the mean sulphide mineral grain size and modelled texture dimension of the ore acquired from coarse particles. Having established the three mathematical models, they were then tested and validated with two different data sets. The first data set consisted of mass and assay values for a population of particles obtained from four size fractions sampled from four different ore types. This data was used to model the particle grade distribution of each ore type size fraction. In addition to lognormal modelling, a new sorting potential index was developed that predicts the sortability of an ore based upon the mean grade and mode grade of the particle grade distribution using the refined lognormal modelling methodology outlined in the thesis. The conclusions drawn from this data set may be summarised as follows: 1. Lognormal modelling showed that with the exception of two size fractions, the measured particle grade distributions closely followed lognormal statistics which validated the lognormal modelling process that was developed in the thesis. 2. Sortability is a direct attribute of the shape of the lognormal distribution with the mode grade determining how many liberated and near liberated gangue particles were present in the distribution whilst the position of the mean grade determined how much inhomogeneity was present in the distribution. 3. The distribution with the best properties for sorting had the mode grade positioned near the first grade class within the distribution whilst the mean grade was significantly larger than the mode grade. This ensured that there was abundant liberated or near liberated gangue particle to reject as well as sufficient inhomogeneity to separate higher grade particles from lower grade particles. The second dataset consisted of quantitative mineralogy data obtained from a size by size analysis of a quartz monzonite sampled from a Cu porphyry ore body. The conclusions drawn from this dataset may be summarised as follows: 1. Results from the three modelling procedures developed in this thesis (lognormal modelling, texture dimension modelling and matrix liberation through size reduction modelling) all confirm that the quartz monzonite ore under test was un-sortable at particle sizes greater than 4 mm. 2. The experimental design developed in this thesis can be used as a routine methodology by any quantitative mineralogical laboratory for assessing the sortability potential without the need for expensive metallurgical test work. 3. More complicated particle grade distributions such as those showing multi-mode compound distributions can be modelled using lognormal forms for the constituent textural populations that contribute to the distribution.