Texture-based analysis of liberation behaviour using Voronoi tessellations

Abstract

A liberation simulation is outlined based on image analysis of mineral texture images such as those produced by automated mineralogical analysis. The method relies on a freeware image analysis programme, Fiji, and uses Voronoi tessellations to represent a fragmentation pattern. This pattern is superimposed onto a mineral texture image and the mineralogical composition of each tile is analysed separately to produce a liberation spectrum.A comparison is presented between actual and simulated data in terms of liberation, shape and size distribution. The correlation (r2) of simulated to measured liberation data exceeded 0.98 for all minerals assessed but for optimum confidence further validation is required over a larger size range. Comparing real and simulated particle shape gave a correlation exceeding 0.95, and it is shown the particle size distribution of Voronoi patterns can accurately reproduce that of scalped feeds (i.e. a narrow size fraction) of comminution products. Repeatability of the process is shown to be dependent on particle size, but overall is very good. To demonstrate potential applications of this analysis method, quartz liberation spectra for three granites of different grain sizes are included and discussed, and a simulated grade–recovery curves for an Au-bearing pyrite is demonstrated.