Experimental modelling and plant simulation of spiral concentrators: Comparing response surface methodology and extended Holland-Batt models

Abstract

Spiral concentrator and spiral plant modelling of chromite bearing ore is investigated in this paper. Spiral product stream responses (grade and recovery) were modelled according to variations in feed conditions (grade, density, flow rate) and primary splitter settings. The two modelling approaches constituted a quadratic response surface and an extension (to accommodate feed conditions) on the original Holland-Batt (1990) spline model. Additional validation of the extended Holland-Batt model behaviour was performed via regression on a second data set, based on hematite separation using spiral concentrators (Sadeghi, 2015). The quadratic model showed higher performance values on the training set; however, the extended Holland-Batt model showed better performance on the test data. Finally, both models were compared in terms of spiral concentrator plant modelling ability (based on an industrial spiral concentration plant). Quadratic response surface models were unable to produce acceptable grade-recovery curve simulations. The extended Holland-Batt model proved to be the superior spiral plant simulation model due to its ability to produce valid grade-recovery curves. The extended Holland-Batt spline-based model shows promise as an improvement over RSM for spiral modelling (and spiral plant simulation). The extended Holland-Batt model can provide the same screening, characterization and optimization functionality of RSM models for single-spiral experimental modelling, with the added benefit of simple extension to spiral plant simulation.