Artificial neural networks modeling for the prediction of Pb(II) adsorption

Abstract

The work presents an artificial neural network (ANN) model predicting the efficiency of Pb(II) adsorption on polyamine-polyurea polymer modified with pyromellitic dianhydride. Adsorption percentages of Pb(II) ions, as calculated using the results of batch experiments, are used as data inputs for the ANN model. In the developed model, the contact time (5–240 min.), pH (1–7), the initial Pb(II) concentration (50–300 mg/L), amount of adsorbent (20–75 mg) and temperature (25–55 °C) values constitute the input layer, while adsorption percentage values constitute the output layer. Simulation-based development of ANN models was carried out with eight values for neurons in the hidden layer (2, 3, 5, 10, 20, 30, 50 and 100). The best results were obtained with 10 neurons. The prediction data of ANN models were statistically compared to experimental data. With the developed model’s trial period and cost savings, the adsorption ratio was estimated with an error rate of about 2%. The results show that the multilayer perception ANN model (R2 = 0.9858) justified the prediction of adsorption percentage.