Structure-based Generalized Models for Selected Pure-fluid Saturation Properties

Abstract

This study focused on developing generalized structure-based models for predicting pure-fluid surface tensions and saturation viscosities. Reliable experimental data for a wide range of molecular species were assembled from the DIPPR physical property database. The Scaled-Variable-Reduced-Coordinate (SVRC) framework was used to correlate the available data for the saturation properties under consideration. Quantitative Structure-Property Relationships (QSPR) was used to generalize the SVRC model parameters. Non-linear QSPR models involving a hybrid of Genetic Algorithms (GA) and Artificial Neural Networks (ANN) were developed for the model parameters. Specifically, the SVRC-QSPR models, in general, were found to be capable of providing generalized a priori predictions for surface tension and saturation viscosities with an absolute average deviation (AAD) of approximately 2% using end-point input data.