Regression Analysis of XRF and Electromagnetic Sensor Responses for Improved Sorting of a Gold Ore

"Sensor based sorting faces two main challenges related to limited capacity/throughput and the ability of sensors to accurately discriminate based on estimated grades. This paper is aimed at improving sensor discrimination. Sensors have the ability to measure more than a single response. X-ray fluorescence (XRF) can detect and provide concentration information about a range of elements. Similarly, electromagnetic sensors (EMs) can generate magnetic susceptibility and conductivity responses over a range of frequency levels. The sensors thereby provide a response signature rather than a simply a single response that is a proxy for a metal grade. Furthermore, integrating sensor responses can improve the discrimination of metal bearing rock from waste. A study was conducted on a gold ore to assess sortability using XRF and EM sensors. A multi-variable regression approach was used to develop models that included a range of responses. Correlations improved significantly when compared to single response correlations. The approach showed that both positive and negative correlations contributed to the improved correlations. Stepwise regression simplified the models and showed that 2nd order interaction effects can be important. The approach also showed that combining different responses can improve discrimination accuracy. An important outcome of the study was that an ore that had been consider not to be sortable was determined to be sortable through the application of an approach that used the sensor responses in a more intelligent manner.INTRODUCTIONMining is aimed at recovering valuable minerals from their host rocks. Many methods are used to recover valuable minerals, and finding efficient and cost-effective ways is the key in research and production. . Ore sorting technology is fitting this definition. It not only discards waste rocks away from the concentrator but also recovers metal bearing ore particles that could be lost to the waste dump. The benefits of ore sorting include increased energy efficiency and reduced water and reagent usage in the downstream operations (Bamber, 2008; Gunson, Klein, Veiga, & Dunbar, 2012; Mazhary & Klein, 2015; Robben, Wotruba, Robben, von Ketelhodt, & Kowalzcyk, 2013)."