Artificial Neural Network Modeling of Cd(II) Ions Adsorption on Nano-porous Inorganic Sorbents

Abstract

A three-layer artificial neural network (ANN) model was developed to predict the efficiency of Cd(II) ion removal from aqueous solutions by nano-porous zeolite, based on 105 experimental sets obtained in an experimental kinetic study. The effects of operational parameters such as the initial concentration of Cd(II) ions in the aqueous solution, the mass of the nano-porous adsorbent and the contact time were studied to optimize the conditions for maximum removal of Cd(II) ions. On the basis of the results of this kinetic test, optimal operating conditions were determined to a contact time of 1,700 min and an initial concentration of Cd(II) ions in aqueous solution of 200 mg/dm3 for a constant mass of the adsorbent of 3 g. The second set of experiments resulted in an optimal contact time of 1,600 min and an adsorbent mass of 3 g for an initial concentration of Cd(II) ion of 50 mg/dm3. After supervisory backpropagation (BP) training, the ANN model was able to predict the adsorption efficiency. The ANN model consisted of a hidden layer with 4 neurons and a tangent sigmoid transfer function (tansig) and an output layer with a linear transfer function (purelin). The Levenberg–Marquardt algorithm (LMA) was found to provide a minimum mean squared error (MSE) of 0.0695 compared to other BP algorithms. The linear regression between the network outputs and the corresponding targets was proven to be satisfactory with a correlation coefficient of 0.936 for the three model variables used in this study.