Fourier transform infrared (FTIR) spectroscopy and other near infrared (NIR) tools have been used in the bauxite industry for many years. Infrared spectroscopy exploits the differences in chemical composition and lattice structure to produce a characteristic response. Spectral devices, such as those from ASD Inc and the HyLoggerTM, provide qualitative mineralogical data targeted towards hydrated minerals detected in the near and short wave infrared region. The FTIR spectrum extends into the mid and thermal infrared range and can therefore respond to the presence of silicates and oxides, in addition to hydrates and carbonates.
The key to successful utilisation of infrared spectra, however, is the interpretation methodology. In this study, FTIR spectra were calibrated against quantitative X-ray diffraction data for the determination of the mineralogy of iron ore. A full pattern profiling machine learning technique was utilised for the calibration, and the assessment of the regressions determined from an independent validation set. The abundance of key minerals – hematite, goethite, kaolinite and quartz – were determined and the results correlated against X-ray fluorescence assays and loss on ignition data. The results of the study indicate that spectral techniques using a full-pattern profiling machine learning approach and artificial neural networks can be used successfully to obtain objective and quantitative mineralogical data to support field observations and analytical results for iron ore resource modelling. A comparison of this technique to the cost, quality and timeliness of other quantitative mineralogy tools is also made.
Carter, J, Auyong, K and Dixon, L, 2017. Determination of iron ore mineralogy using Fourier transform infrared spectroscopy and machine learning, in Proceedings Iron Ore 2017, pp 519–528 (The Australasian Institute of Mining and Metallurgy: Melbourne).