Modeling and Analysis of Copper, Iron, and Cobalt Recovery in a Hybrid Sulfuric Acid–Sodium Chloride Media Using Artificial Neural Network

The effects of operating conditions on copper, iron, and cobalt dissolution from sulfide ores in sulfuric acid–sodium chloride media are predicted using an artificial neural network (ANN) model. The artificial neural network model was developed, trained, and predicted using the feed-forward back-propagation (BP) algorithm. The sulfuric acid concentration, sodium chloride concentration, temperature, leaching time, and particle size were used as input variables to the model. A total of 204 sets of data generated from the leaching experiments were used to develop and train the model. To reach the network with a good agreement and highest generalizability and to reduce the error between the measured and predicted values, the neural networks with a various number of hidden layers (one to ten hidden layers) were investigated. In the regression analysis of the {5–10–3} architecture, the R² values were 0.998, 0.997, and 0.997, while the MSE values were 0.111, 0,148, and 0.106 for the training, validation, and testing sets, respectively. The results showed that ANN has a high potential for predicting copper, cobalt, and iron recoveries. The increase in the number of hidden layers was found to improve the performance of the ANN model. Graphical Abstract: [Figure not available: see fulltext.]