Chemometric method for Donnan dialysis physicochemical model simplification. Prediction of: Transport, recovery, concentration, and desalination efficiency

A method of transforming the exact physicochemical Donnan dialysis (DD) model into some chemometric regression models was developed. The thermodynamic network (TN) model of DD consisting of 40 coupled ordinary differential equations was successfully simplified into the regression models of practical interest. In the proposed method which is similar to the Response Surface Methodology (RSM), the real experiments of Donnan dialysis transport were replaced with numerical simulations, except for the model validation. The numerical simulations were planned and conducted according to the Box-Behnken and central composite face-centered designs. The physicochemical and transport operational parameters (independent variables) were correlated with four selected main features of DD transport and recovery efficiency (dependent variables calculated from the TN models). It was experimentally confirmed that the method proposed in this paper leads to developing relatively simple regression models with a high predictive power which can be used e.g. for predicting desalination effect. Moreover, the interpretation of the significant regression model parameters is in accordance with the already known factors influencing the Donnan dialysis transport, i.e. diffusion (interdiffusion) in an ion-exchange membrane.