Prediction of ternary ion-exchange equilibrium using artificial neural networks and Law of Mass Action

Abstract

The Law of Mass Action generally models the equilibrium data from ion exchange processes. This methodology is rigorous in terms of thermodynamics and takes into consideration the non-idealities in the solid and aqueous phases. However, the artificial neural networks may also be employed in the phase equilibrium modeling. In this study, both methodologies were tested to describe the ion exchange equilibrium in the binary systems SO42--NO3-, SO42--Cl-, NO3-Cl- and in the ternary system SO42--Cl--NO3-, by AMBERLITE IRA 400 resin as ion exchanger. Datasets used in current study were generated by the application of the Law of Mass Action in the binary systems. Results showed that in the equilibrium modeling of binary systems both methodologies had a similar performance. However, in the prediction of the ternary system equilibrium, the Artificial Neural Networks were not efficient. Networks were also trained with the inclusion of ternary experimental data. The Law of Mass Action in the equilibrium modeling of the ternary system was more efficient than Artificial Neural Networks in all cases.