UNIFAC BASED ARTIFICIAL NEURAL NETWORK MODELING OF ETHANOL - WATER PHASE EQUILIBRIA

An ANN-UNIFAC hybrid model was used to estimate subgroups binary interactions for the ethanol – water mixture and then predict the vapor phase molar composition of the components as well as the mixture pressure. The results obtained in this work were compared to the ones from a standard UNIFAC model. Model validation showed that the ANN-UNIFAC hybrid model was able to accurately fit the experimental data at 423.15, 473.15 and 523.15 K. The error evaluation through coefficient of determination and mean relative error showed that the proposed model had overall better output predictions than the standard UNIFAC model.