Optimized Neural Network Model to Characterize the Effects of Process Parameters on the Separation Efficiency of Iron Ore by a High Intensity Magnetic Separator

Abstract

An improved and optimized multi-input-multi-output (MIMO) neural network model has been developed to predict the output parameters e.g grade and recovery to characterize the separation behavior of a high intensity magnetic separator for processing iron ore in the particle size range of 75~300 m. The input parameters in the Neural model comprises of feed composition, % Fe, % SiO2, %Al2O3 and process parameters such as particle size, pulp density and magnetic field intensity. The effect of process parameters on the separation efficiency was characterized by conducting a sensitivity analysis. The neural network architecture has been optimized using an efficient gradient based network optimization algorithm to minimize the training error rapidly. The model is based on the data generated from WHIMS experimental investigations. There has been an excellent agreement between the optimized model predictions with the measured values pertaining to recovery and grade for magnetic separation. This is depicted by the regression fit generated between the predicted and measured values.