Modeling and optimization of galena dissolution in a binary solution of nitric acid and ferric chloride using artificial neural network coupled with genetic algorithm and response surface methodology

Abstract

In this research, the optimization of lead recovery from galena with a binary solution of nitric acid and ferric chloride using response surface methodology (RSM), artificial neural network (ANN) and genetic algorithm (GA) was carried out. The galena mineral was examined for mineralogical properties with X-ray diffraction spectroscopic system while the elemental composition was analyzed with X-ray fluorescence (XRF). The results revealed that the galena mineral exists as lead sulphide (PbS). The central composite design was employed for RSM modeling while back propagation (BP) coupled with the Levenberg-Marquardt (LM) algorithm was used to construct a feed-forward neural network (FFNN). The leaching temperature, acid concentration, stirring rate, leaching time and ferric chloride concentration were chosen as input factors, while the percentage yield of lead was the response. The multilayer perceptron with design of 5-9-1 gave the best performance. Comparison of the RSM and ANN model indicated satisfactory prediction of the response, with AAD of 0.887% and 0.377%, R2 of 0.9910 and 0.9989, and RMSE of 0.815 and 0.290, respectively. The process parameters were optimized via GA and RSM optimization tools. An optimum lead yield of 90.33% was obtained with a leaching temperature of 80.2°C, HNO3 concentration of 3.55 M, stirring rate of 498.88 rpm, leaching time of 86.91 minutes and ferric chloride concentration of 0.35 M using RSM while a yield of 87.11% was achieved via GA.

Keywords

• Galena
• Genetic algorithm
• Artificial neural networks
• RSM
• Optimization
• Leaching