(180ac) Direct Calculation of PC-SAFT Parameters From Experimental LVE Data and From a GC-Plus Approach
Since several years, SAFT equations of state (EoS) are drawing the attention of many researchers in thermodynamics. The PC-SAFT (that is, the perturbed chain SAFT EoS) is one of the simplest and widely used among this family of models. Stemming from molecular theory and thus more complex than cubic EoS, PC-SAFT equations offer not yet explored possibilities in modelling fluid phase equilibria. When one wants to use them, at least three pure component parameters have to be considered. These are: the number mi of segments per chain in molecule i, the segment diameter σi in molecule i and εi, the depth of pair potential.
In order to make the PC-SAFT equations more widely used and able to represent a wide range of organic systems, the group contribution (GC) concept can be employed. Such GC-methods are nevertheless restricted to the number of groups defined by the method. Only very common groups such as methyl, ethyl, hydroxyl, carboxyl and so on, are in general defined by the GC-methods, thus excluding systems with specific atom groups which are essential for industrial applications. To avoid such limitations, an alternative consists in creating the missing groups through the connectivity index concept.
In this work, a robust and widely applicable fitting procedure has been developed for the direct calculation of the PC-SAFT parameters from experimental LVE data. Using these fitted parameters, two methods aimed at predicting the set of parameters (mi, σi and εi) for a pure component were developed. The first one is a classical GC-method with many distinct functional groups and thus applicable to a large set of pure compounds. This method covers a wider range of chemicals and leads generally to smaller deviations on the estimation of LVE properties than other published methods.
The second one relies on the atom connectivity approach and allows to create missing groups when a component cannot be described by the first GC-method.
Our predictive model has been tested on hundreds of pure component property data from organic chemicals as well as on multicomponent mixture data including polymers, with very promising results.