Investigation of using different thermodynamic models on prediction ability of mutual diffusion coefficient model

The aim of this study is to improve the new model for prediction of mutual diffusion coefficient in non-ideal, concentrated liquid solutions by applying two thermodynamic models (NRTL and Wilson) to estimate the derivative of activity coefficient logarithm versus the mole fraction (dlnγ1dlnx1). The model which is used in this paper is based on cluster diffusion theory. Unlike, the previous model which was proposed by the authors (Kamgar et al., 2017), this improved one is a predictive model and it does not require any vapor-liquid equilibrium data. It is just function of temperature, mole fraction and viscosity. An optimization approach has been carried out to measure the adjustable parameters of the models. Besides, the results of the improved model are compared with the model in which VLE are used for calculation of dlnγ1dlnx1. Furthermore, to validate the improved model, eleven binary system with more than 130 experimental diffusion coefficient data are used. An acceptable average relative error (ARD %) has been achieved for all the cases. Among investigated thermodynamic models, Wilson gave the best results and the least errors.