Quantitative mineralogy for improved modelling of shaking tables

Gravity separation is a key technology in the recovery of a range of mineral ores including coal, tin, tungsten and tantalum. One of the problems associated with gravity concentration is that recovery is traditionally lower than for other separation techniques such as the flotation of base metals. Any improvements in recovery will, therefore, have a significant effect on the economics of operations of this type. More accurate, robust modelling of gravity separators is one means of achieving greater recoveries. Furthermore, better modelling can lead to more informed equipment selection and process design for potential economic deposits. Previous models for gravity separation equipment have typically relied on simple metal mass balances from samples of the separated products. However, with advances in technology there is a ready availability of analytical instruments for the automated quantitative determination of mineralogy. Instruments such as the Mineral Liberation Analyser and QEMSCAN make it possible to estimate particle size and density (related to mineral liberation and association) which are the primary factors in gravity separation. Models produced using quantitative mineralogy of feed streams will be potentially superior to those produced historically. This paper describes test work undertaken on a wet shaking table using a synthetic ore. The objective of this test work was to develop a model to predict the effect of changes in feed ore characteristics on shaking table performance by measuring quantitatively the mineralogy of the feed material. Particles were divided into size/density classes based on the results of QEMSCAN analyses. It was found that it was possible to use data from QEMSCAN to predict the recovery of size/density classes and to predict the effect of changes in the relative abundances of these classes on the separation. This model can potentially be used to adjust operating variables to maximise grade and recovery by monitoring the input feed. Care was needed when utilising QEMSCAN data to ensure representivity. © 2017 TAPPI Press. All rights reserved.