On the optimization of froth flotation by the use of an artificial neural network

Abstract

A multi layered, feed forward Artificial Neural Network (ANN) was used to study the effect of feed mean size, collector dosage and impeller speed on flotation recovery and grade. The results of 30 flotation experiments conducted on Jordanian siliceous phosphate were used for training the network while another 10 experiments were used for validation. Simulation results showed that a four layer network with a [9 11 5 9 2] architecture was the one that gave the least mean squared error (MSE). Using this ANN to optimize the flotation process showed that the optimum flotation parameters were 321.28 μm for the feed mean size, 0.7354 kg/TOF for the collector dosage and 1225.25 RPM for the impeller speed. Studying the effect of these parameters on flotation recovery and grade was done by analysis of variance, ANOVA. The results showed that grade was more sensitive to changes in flotation parameters than was recovery. They also showed that changes in collector dosage had a more significant effect on flotation grade and recovery than did changes in feed mean size or impeller speed.