Part II: Prediction of the dialysis process performance using Artificial Neural Network (ANN)

Abstract

In the present research, an experimental based Artificial Neural Network (ANN) model is developed to describe the dialysis process performance under different operating conditions in the charged and neutral states of dialysis. The transfer behavior of charged micelles through the charged and neutral conductive membranes was investigated using this model. The parameters are highly interconnected in this system. Moreover, using the available deterministic models for tracking the process performance and the switch of the diffusion mechanisms is not completely realizable. Therefore, using neural networks is highlighted as a recommended model for this type of situations. The concentration gradient, absolute feed concentration and membrane electrical charge are the main parameters which affect the process performance. The experimental system consists of aqueous sodium dodecyl sulphate (SDS) solution above critical micelle concentration (CMC) and commercial micro-filtration membrane (GVHP) coated by conductive poly-pyrrol (PP) or silver. The amounts and the mechanisms of mass transfer were analyzed for the two types of membranes in a variety of operating conditions. Moreover, the study over the competitive state of the process was performed where the parameters are manipulated simultaneously. The developed ANN is able to predict the process performance under individual manipulation and simultaneous manipulation of the parameters. Therefore it is able to track the governing mechanisms any time. The experimental data and the developed model show that in low concentration and concentration gradients, the diffusion mechanism and value, are different in comparison to the state with high values of concentration and concentration gradient. The neural network approach was found to be capable of modeling this complex process accurately.