ML-SAFT: A machine learning framework for PCP-SAFT parameter prediction

ChemRxiv

Abstract

The Perturbed Chain Polar Statistical Associating Fluid Theory (PCP-SAFT) equation of state (EoS) is widely used to predict fluid-phase thermodynamics, but parameterization of PCP-SAFT for individual molecules is often challenging. We propose a machine learning framework called ML-SAFT for predicting parameters of PCP-SAFT. In order to provide data for training machine learning models, we created the largest dataset of regressed PCP-SAFT parameters in the literature. We then conducted extensive evaluation of several machine learning architectures for predicting PCP-SAFT parameters. We found that our best model provided accurate predictions for a wider range of molecules than existing predictive techniques with 39 \% average absolute deviation (AAD) in vapor pressure predictions and 9 \% AAD in density predictions.

Authors 8

1. Kobi Felton Aachen

   University of Cambridge · RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

   University of Cambridge

2. Lukas Rasßpe-Lange Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

3. Jan G. Rittig Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

4. Kai O. Leonhard Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

5. Alexander Mitsos

   BASF (Germany)

   Affiliation as printed

   BASF (Germany)

6. Julian Meyer‐Kirschner

   BASF (Germany)

   Affiliation as printed

   BASF (Germany)

7. Carsten Knösche

   BASF (Germany)

   Affiliation as printed

   BASF (Germany)

8. Alexei A. Lapkin

   University of Cambridge

   Affiliation as printed

   University of Cambridge