The response of diasporic-bauxite flotation to particle size based on flotation kinetic study and neural network simulation

This paper was to investigate the effect of particle size on diasporic-bauxite flotation. For this purpose, concentrate products of various size fractions were collected at different flotation times and forwarded to the testing of Al2O3 and SiO2 contents. Six flotation kinetic models were applied in the fitting of the experimental data in terms of Al2O3 and SiO2 recovery. Artificial neural network (ANN) model was also established to predict the effect of mean feed particle size on the recovery and separation efficiency of Al2O3. The results demonstrated that the classical first-order model was the most reasonable model to fit the diasporic-bauxite flotation process throughout this paper. Al2O3 and SiO2 were easier to be recovered from intermediate size fractions (-38 + 2 mu m and -54 + 38 mu m) than those from coarse size fractions (-74 + 54 mu m and -98 + 74 mu m) and/or extremely fine size fractions (-20 mu m). The separation efficiency of the intermediate size fraction was also greater than that of coarse or fine size fractions. The effective separation of coarse particles (+54 mu m) was only achieved at the initial stage of flotation process, whereas the effectual separation time of intermediate and fine particles is much longer compared with that of coarse size fractions. The prediction results using ANN model showed that the suitable mean feed size range of diasporic-bauxite flotation was from 21 to 41 pm. (c) 2017 Elsevier B.V. All rights reserved.