An investigation into the complementary capabilities of X-ray computed tomography and hyperspectral imaging of drill core in geometallurgy

Mashaba, Dineo

An investigation into the complementary capabilities of X-ray computed tomography and hyperspectral imaging of drill core in geometallurgy

Master Thesis

2022

Abstract

The mining industry is faced with the challenge of mining and processing low grade, heterogeneous, and complex ores, a phenomenon known as ore variability. These ores need to be managed at an early operational stage, ideally during drill core exploration, to avoid risks during the project phase (such as project delays and failure) and operational phases (such as plant instabilities), ultimately affecting the cash flow. The discipline of geometallurgy has arisen to manage the risks associated with ore variability by acquiring upfront knowledge of the mineral assemblage and texture before mining and processing. As we head towards the fourth industrial revolution (4IR), machine learning, intensive and automated data derived from drill cores are becoming more common. In this case, using non-destructive, rapid, and inexpensive automated scanning techniques such as 2D hyperspectral imaging (HSI) and 3D Xray computed tomography (XCT) have the potential to be incorporated into the machine learning dataset. Hyperspectral imaging is a critical component of continuous drill core scanning in geometallurgy for identifying problematic minerals in downstream mineral processing, such as the phyllosilicates (e.g., kaolinite, serpentine and talc). However, it only provides 2D imaging of the core, and its mineral identification is limited to minerals that show a definitive spectral response. On the other hand, XCT provides 3D imaging of drill cores, but is more routinely used in research applications and does not independently give the mineral assemblage. Mineral identification and discrimination for XCT is limited and requires prior mineralogical knowledge and sufficient mineral density and attenuation coefficient variation greater than 6%. No systematic study to date appears to have explored how the results from these two techniques can be integrated using a local South African magmatic nickel-copper-platinum group element (Ni-Cu-PGE) ore case study. This opened an opportunity to couple the two techniques to address and emphasize the image scanning techniques for drill core in geometallurgy and to provide further knowledge on the practicality of the HSI and XCT in drill core from image acquisition to processing. Ultimately, the aim is to investigate how well the techniques complement each other for mineral and texture identifications and, if combined, will produce additional mineralogical and textural information. The objective of this study was achieved by moving HSI cores to smaller samples than standard practice to produce 25 mm diameter mini cores instead of standard cores (e.g., 50 mm in diameter). For accurate mineral assemblage and textural characterisation of the drill cores, manual core logging, quantitative evaluation of minerals by scanning electron microscopy (QEMSCAN) and quantitative X-ray diffraction (QXRD) were used as supporting techniques. The results showed HSI scanning on the magmatic Ni-Cu-PGE drill core to be challenging because of pervasive mineral alteration and the nature of the rock types (mafic and ultra-mafic rocks) - providing limited information on the mineral assemblage and texture due to low scanning resolution and pervasive alteration (serpentinisation and chloritization) in the rocks. The limited mineral identification includes mixed-phases (such as serpentine-olivine in visible-shortwave infrared and plagioclase-chlorite in the longwave infrared) and unclassified minerals in the core. The resultant mineral assemblage was comparable to QEMSCAN and QXRD in terms of minerals present with generally similar abundances. However, useful information on the alteration mineralogy can still be extracted, such as the presence of serpentine, chlorite and talc and their association with other silicate minerals. Other parameters such as mineral grades and grain sizes were quantified on MATLAB using specially developed scripts. The interconnected grains could not be separated due to invisible boundaries on the HSI maps. Therefore, only a small number of grains were generated with larger grain size values, likely underestimating the real grain numbers. XCT provided information on valuable high-density minerals (including possible platinum-group minerals (PGMs)) and mineral texture in the cores. Due to extensive alteration in the rocks, discrimination between grey values was, however, challenging. Grey level segmentation into the different mineral groups was also noted to be dependent on the rock type. For example, plagioclase and orthopyroxene were more easily discriminated in the less altered rocks (feldspathic pyroxenite and anorthosite) than the more altered rocks (altered harzburgite and pegmatoidal pyroxenite). The high scanning resolution allowed for the extraction of mineral texture, such as mineral association and grain size distribution (GSD). The 3D XCT derived GSD was slightly coarser than the 2D QEMSCAN derived GSD. The differences in GSD are attributed to a combination of both stereological and sampling effects. However, sufficient information on ore variability can be obtained when using the pertinent scanning parameters and careful segmentation processes. These two techniques provide variable information on the mineral assemblage and texture, such as the identification of silicate minerals (particularly alteration minerals) in HSI and high-density minerals in XCT and good textural information on XCT than HSI. With the information provided, possible image overlapping scenarios of the two techniques were identified: (1) using XCT for high-density minerals, and HSI for silicate identification, (2) using XCT data with good mineral and texture discrimination (silicate associated with sulphides) to map unclassified areas in HSI, (3) is the opposite of the second scenario. Ultimately, the two scanning techniques will likely offer complementary information, although the application of this combined technique for routine work will be limited in practicality. Additionally, more work needs to be carried out with revised scanning and processing to improve the sustainability of the techniques in geometallurgy.

Keywords

Chemical Engineering

Reference:

Collections

Masters

Full item page