Artificial neural network modeling of the hexavalent uranium sorption onto chemically activated bentonite

This study aims to study the hexavalent uranium U(VI) adsorption onto chemically activated Algerian bentonite BMP-Na in order to guarantee a clean and environmentally friendly method of the elimination of waste in nuclear energy applications. The activation of the bentonite was done chemically using hydrochloric acid HCl and saturated with Na+ cations using NaCl solution. The obtained bentonite has been characterized by using various methods such as XRD, SEM–EDX, FTIR, XRF and BET. Those characterizations have shown that the activation of bentonite has led to a significant increase in the surface area of the bentonite from 17.19 to 83.06 m2/g, which can improve its efficiency in the adsorption application. After characterization, the activated bentonite was subjected to the U(VI) adsorption application. An artificial neural network (ANN) method was used to model and optimize the U(VI) adsorption. Experiments at various pH levels, different absorbent doses and uranium (VI) initial concentrations were done and then, used to train the ANN network. The network was optimized by measuring the mean square error (MSE) with various numbers of neurons in the hidden layer. The optimal number of neurons on the hidden layer was discovered to be 12 neurons, which provided the highest coefficient of correlation R2 of 0.991. The ANN model's predictions matched very well with the experimental results. The optimization after modelization has shown that the optimal conditions for uranium adsorption are pH = 5.6, initial concentration of 67 mg/L and adsorbent dose of 5 g/L with a complete adsorption rate within only 6 h. Finally, to determine the relative importance of the various operating inputs, a Garson formula was established and the pH of the solution was found to be the most influential parameter.