Metal bioleaching prediction in continuous processing of municipal sewage with Thiobacillus ferrooxidans using neural networks

Abstract

Neural network procedures were used to predict the metal (Cu, Zn and Cd) solubilization percentages in municipal sludge treated with a continuous process using Thiobacillus ferrooxidans. The neural networks consider several hydraulic retention times and FeSO 4·7H 2O concentrations in addition to pH, oxydo-reduction potential and initial metal concentrations as input variables. The resulting neural network showed adequate prediction of metal solubilization percentages with mean absolute deviations varying from 5 to 8% and good flexibility to predict solubilization percentages of observations that are not used in the training of the network (test set). A second neural network showed the ability to predict solubilization percentages for an hydraulic retention time different from those used in the building of the network keeping the mean absolute deviations at 5 to 8%. Furthermore, a simple graphical analysis of neural network results allowed us to bring out several important relations between metal solubilization and hydraulic retention time or iron sulfate concentrations.