Polar (s)PC-SAFT: Modelling of polar structural isomers and identification of the systematic nature of regression issues

Abstract

The systematic nature of regression problems in the polar SAFT framework is identified in this work. This is achieved by considering parameter regression and VLE predictions for structural isomers of medium length linear ketones, esters and ethers. The inability to generate a unique polar parameter set by what we define as the standard regression approach (SRP) is a frequently reported problem in the literature. These problems are attributed to the similarity of saturation properties between the polar component and those of a nonpolar equivalent. There is thus an inability to distinguish between dispersion and polar contributions to component behaviour during regression. This is systematically demonstrated by the consideration of isomers, where a systematic deterioration of prediction strength is in evidence as the distinction between the properties of polar component and the nonpolar analogue diminishes with successive isomers. The fixed polar parameter (FPP) approach formalised in this work offers the best means of combatting these problems and yielding predictive parameter sets. Predictions based on this approach are consistently superior to conventional regression approaches, whether using the Jog and Chapman (JC) or Gross and Vrabec (GV) polar terms. Finally, the identical treatment of both these terms provides a balanced comparison of their performance and points towards a superior predictive capacity for the GV polar term.