The Prediction of Carbon-in-Leach Gold Recovery Using Near-Infrared Spectroscopy

"A model was developed to use near-infrared (NIR) analysis of pulverized ore samples to estimate the recoverable gold content for ores from the Barrick Cortez Hills underground mine. Current routing practices use a ratio of the cyanide soluble gold content and fire assayed gold (CN/FA) to classify ores as oxide or refractory. Prediction accuracy of the NIR model was equivalent to the CN/FA ratio. Prediction accuracy was improved by combining the CN/FA ratio and the NIR spectra. Bootstrap validation estimates the root-mean-square error to be 9.92% with an estimated r-squared value of 0.893. This model will allow more accurate estimates of the variability in recoverable gold content throughout this mine. These data will aid in building an improved geo-metallurgical model or in making better informed routing decisions during the production cycle.INTRODUCTION The contained gold of an ore available for cyanidation is often related to subtle variations in the mineralogy. It is difficult and expensive to collect enough mineralogical information in order to predict how the variable mineralogy will impact the metallurgical results and project economics throughout the life of the mine or mill. Near-infrared spectroscopy may be an effective tool to rapidly and inexpensively collect discrete, quantitative mineralogical data for minerals that contain transition metal oxides, metal sulfides with semi-conduction properties, rare earth metals, and some chemical groups having di-pole covalent bonds such as carbonates, sulfates, and various types of clays (Hunt, 1977). Near-infrared spectroscopy can be a powerful tool for geo-metallurgical characterization of deposits where the presence of these minerals has direct or indirect relationships to metallurgical results such as grindability, regent consumption, or recovery. Near infrared spectroscopy is a rapid analysis that measures the intensity of reflected light in the near-infrared spectrum. The ranges for the ASD-Panalytical units used in this study extend through the visible spectrum and into part of the UV spectrum, reporting reflection by wavelength from between 350 nm to 2500 nm. Absorption peaks visible in this range are a result of interactions with outer valence elections or chemical bonds. The location and intensity of the peaks vary depending on the chemical composition and crystal structure of a solid indicating the presence and abundance certain minerals (Hunt, 1977; Clark, 1984)."