X-ray computed tomography for geometallurgical ore texture analysis: Implications for comminution response

Abstract

This study provides an automated workflow for downhole ore texture analysis from drill-core X-ray computed tomography (XCT) and evaluates its relevance to comminution responses. A total of 186 m of intact drill core from four drill holes in the Kalkaroo copper-gold deposit in South Australia were analysed, including approximately 20 m covered by 21 geometallurgical test intervals with Axb measurements. For robustness and scalability over continuous drill-core datasets, stationary texture was quantified using Grey-Level Co-occurrence Matrix (GLCM) descriptors (contrast, homogeneity and entropy) extracted from depth-registered transverse XCT slices at 1 mm spacing. A Gaussian Hidden Markov Model (HMM) was then applied to the continuous multivariate descriptor profiles to delineate four persistent, geologically interpretable texture types, cross-validated against XCT volumes, optical imagery and core observations. Texture types were aggregated within each comminution test interval to compute majority- and mixture-based texture assignments, purity, which were compared with Axb values using non-parametric statistics and composition-based modelling. Systematic differences between majority texture groups (Kruskal-Wallis p=0.0015, ε2=0.729) and strong rank correlations between Axb and texture proportions indicate that XCT-derived stationary texture metrics capture variability relevant to the impact breakage response. Unlike approaches restricted to bulk composites or small, isolated specimens, the method operates directly on continuous drill core to deliver reproducible, depth-resolved texture characterisation with clear geometallurgical relevance. The workflow supports rapid screening of downhole texture variability for mine-to-mill decision-making, and the results motivate expanded comminution calibration across different deposit types to strengthen the general applicability of the proposed workflow.