(76c) Accuracy of Predictive Thermodynamic Models for Modeling Difficult Separation Processes

Reliable process design and capital cost estimation of separation processes in refinery and other petrochemical processes requires accurate thermodynamic models for process simulation. In many cases phase equilibrium data for potential processes are not available and predictive thermodynamic models must be used to estimate the phase equilibria in the system.  A popular traditional predictive model is the UNIFAC model which is based on empirical molecular group contributions. A newer approach is the COSMO model which is based on molecular surface properties derived from quantum chemistry. This study examines the accuracy of these models for modeling two types of mixtures: 1) hydrogen-bonding phenolic-rich aqueous mixtures containing hydrocarbon and typical of the downstream processing of biofuels derived from pyrolysis oil; and 2) mixtures of polar solvents and close-boiling C₄-C₆ hydrocarbon alkane/olefin isomers typical of refinery process extractive distillation alkane/olefin isomer separations. The accuracy of the models is evaluated by comparing the predictions of the models with vapor- and liquid-liquid equilibria data available in the literature. The results indicate that, depending on the situation, these models may not yield predictions of sufficient accuracy  for reliable process  design.