(190q) Phase Equilibria Modeling: Structure-Based Generalized Models for Activity Coefficients

Structure-based generalized models were developed for a priori predictions of vapor-liquid equilibrium (VLE) of binary mixtures. Specifically, quantitative structure-property relationship (QSPR) modeling was used to provide structure-based parameters for the Non-Random Two-Liquid (NRTL) and the Universal Quasi-Chemical (UNIQUAC) activity coefficient models. Although this model development is similar to our previous efforts where separate models were used to predict interaction parameters, Aij and Aji, an important motivation in carrying out this work was the use of a single QSPR model to estimate the interaction parameters simultaneously. A representative database comprised of diverse molecular species was utilized for these generalizations.

The hypothesis for this work was the utilization of an approach, which uses cause-and-effect to determine the significance of a given descriptor accounting for variations in molecular volume, area, shape, polarity, association (VASPA), etc. of a binary mixture. The quality of the predictions obtained for a diverse group of molecules demonstrates the validity of this integrated approach and provides credible evidence to support the above hypothesis. The superiority of non-linear techniques over linear techniques is also investigated. The artificial neural network QSPR models were found to be capable of providing generalized a priori VLE predictions within twice the absolute average deviation of the data regressions. The results of this study demonstrate the efficacy of using theory framed QSPR modeling for generalizing saturation property and phase equilibrium models.