LAUSR

Abstract In Part 1, a REFLUX™ Classifier, which consists of a lower fluidized bed system and upper system of parallel inclined channels, was operated in continuous mode and used to beneficiate a dense mineral ore covering a nominal particle size range up to 300 μm. The improvements in performance when the inclined channel spacing was reduced from 6 mm down to 3 mm were investigated and reported ( Galvin et al., 2020 ). Here in Part 2, samples of the feed, product and reject streams from two of the Part 1 steady-state runs were subjected to water-based batch fractionation using a REFLUX™ Classifier with 1.77 mm inclined channel spacing. The algorithm proposed by Galvin et al. (2018) was then used to extract the partition curve from the batch fractionation data, and the results compared with the partition curve obtained using the conventional heavy-liquid sink-float method. The results confirmed that the D50 values are accurately determined by the algorithm, regardless of the accuracy of the batch fractionation method. However, as expected, the algorithm under predicted the true Ep by an amount proportional to the accuracy of the batch fractionation method. A simple Ep correction was therefore applicable, making the water-based fractionation approach a viable method for determining partition curves. Comparison of the 3 mm and 6 mm performance confirmed that the variation in the separation density with particle size was higher when using the wider 6 mm channels. Moreover, with 6 mm channels the Ep varied with particle size to the power −1.02, causing a significant increase as the particle size decreased, whereas with 3 mm channels there was little variation in Ep over this size range. This led to a 20% lower composite Ep when using 3 mm channels.