Optimization of Copper (II) leaching process using machine learning approaches

Using machine learning approaches such as the artificial neural network (ANN) model, the factors influencing the leaching of copper from contaminated soil using sulphuric acid were determined. The feed-forward back-propagation (BP) algorithm was used for the development, training, and prediction of the artificial neural network model. The pH of the solution, acid concentration, soil-to-liquid ratio, and stirring speed were used as input variables, while the amount of copper (II) leached was used as the output. To build and train the model, 21 datasets were taken from the leaching experiments. We looked at neural networks with one to nine hidden layers to find the one with the best agreement and to find the one that could reduce the discrepancy between the predicted and measured values. A proportion of 70% of data were used for training, 15% for testing, and 15% for validation. During the regression analysis of the four inputs, nine hidden layers, and one output design, the R² value for training was 0.997, that for validation was 0.996, and that for testing was 0.997. The algorithm used was Levenberg–Marquardt with membership 11-11-11-11. The corresponding MSE values were 0.121, 0.133, and 0.105. The findings indicate that the artificial neural network holds great promise for predicting the leaching of copper (II). The results showed that the ANN model's performance improved as the number of hidden layers increased.