Prediction of hydrocyclone performance using artificial neural networks - Synopsis

Artificial neural networks (ANNs) have found their applications in the modelling of unit operations of mineral processing plants. In this research, laboratory-scale tests were conducted, using a three-inch diameter Mozley hydrocyclone. Main parameters included pressure drop at inlet, solid per cent, vortex and apex diameter were adjusted. The corrected cut size (d50c) and the flow rates of underflow and overflow were determined. Multi layers perceptron (MLP) feed forward network architectures were designed to predict the responses. The results showed a good correlation between experimental and network output, for corrected cut size and flow rates. Keywords: hydrocyclone, artificial neural network, corrected cut size, flow rates.