Artificial neural network modeling and experimental investigation to characterize the dewatering performance of a hydrocyclone

Dewatering in mineral processing industries is of paramount importance as most wet beneficiation of minerals needs removal of water. For this purpose, we have evaluated a 50.8 mm diameter hydrocyclone in order to assess whether it can be used as a partial replacement for a thickener. A multi-layer perceptron based artificial neural network (ANN) model was developed to characterise the dewatering performance of a hydrocyclone using experimentally generated data for silica and magnetite. Parametric sensitivity analysis was undertaken by studying the influence of vortex finder diameter, spigot diameter and inlet pressure on dewatering performance. The ANN model predictions showed that solid recovery to underflow increases and water recovery to overflow decreases with increasing spigot diameter whereas solid recovery to underflow decreases and water recovery to overflow increases with increased vortex finder diameter. Both increase monotonically with increase in inlet pressure. The neural model prediction was successfully validated with the experimental data.