Efficiency of automated sorter performance based on particle proximity information

• Published in: Minerals engineering 2010-09, Vol.23 (10), p.806-812
• Main Authors: Pascoe, R.D., Udoudo, O.B., Glass, H.J.
• Format: Article
• Language: English
• Subjects: Mineral processing; Modelling; Sorting methods
• ISSN: 0892-6875; 1872-9444
• DOI: 10.1016/j.mineng.2010.05.021

The accurate prediction of automated sorter performance is important for the future application of this technology in the minerals industry. While it is known that sorter capacity increases with particle size, a method for predicting separation efficiency is not currently available in the literature. The separation efficiency depends on material properties, feed presentation and machine properties. In this study, we have undertaken colour separation of two materials chosen for their ease of identification using a TiTech Combisense® (BSM 063) automated sorter. Sorter performance was compared to particle proximity data obtained from the on-board counter of the automated sorter. Given the choice of feed materials, any inefficiency in sorter operation is related to the efficiency of selective air ejection which, in turn, is affected by feed presentation, feed size, compressed air valves characteristics as well as the sensitivity of the sorter. Deflection efficiency can be influenced by particle trajectories, precision of the air ejection system and the sensitivity of the sensors. The precision of ejection was also found to be an important factor with the unintentional removal of “accept” particles caused by their close proximity to particles being rejected. It was found that this loss of efficiency could be explained by “touching” particles forming aggregated groups at the identification point. An increase in throughput and a decrease in particle thickness were shown to increase the probability of finding these aggregates. As the fraction of particles to be deflected increases there is an increased probability of finding both “accept” and “deflect” particles within an aggregate and therefore an increased probability of co-deflection. A simple model for predicting the efficiency of our automated sorting system (assuming perfect identification) is presented which has only two variables, % belt loading and the fraction to be deflected.