Surrogate modelling of VLE: Integrating machine learning with thermodynamic constraints. (November 2020)

• Record Type: Journal Article
• Authors: Carranza-Abaid, Andres; Svendsen, Hallvard F.; Jakobsen, Jana P.
• Is Part Of: Chemical engineering science. Volume 8(2020)
• Publication Date: 2020-11
• Subjects: Thermodynamics -- Machine learning -- Surrogate modelling -- CO2 -- MEA; Chemical engineering; Periodicals; 660.05
• DOI: 10.1016/j.cesx.2020.100080
• Languages: English
• ISSNs: 2590-1400

Abstract

Highlights: A simple method to develop surrogate VLE models based on machine learning (ML) Physical constraints are successfully integrated into the ML model framework. The Gibbs phase rule is properly applied in the input variable selection step. Using ML for VLE calculations are ~1000× faster than with semi-empirical models. The ML surrogate method is ~10× faster and more accurate than the interpolation one. Abstract: An easy-to-implement methodology to develop accurate, fast and thermodynamically consistent surrogate machine learning (ML) models for multicomponent phase equilibria is proposed. The methodology is successfully applied to predict the vapour-liquid equilibrium (VLE) behavior of a mixture containing CO2, monoethanolamine (MEA), and water (H2 O). The accuracy of the surrogate model predictions of VLE for this system is found to be satisfactory as the results provide an average absolute relative difference of 0.50% compared to the estimates obtained with a rigorous thermodynamic model (eNRTL + Peng-Robinson). It is further demonstrated that the integration of Gibbs phase rule and physical constraints into the development of the ML models is necessary, as it ensures that the models comply with fundamental thermodynamic relationships. Finally, it is shown that the speed of ML based surrogate models can be ~10 times faster than interpolation methods and ~1000 times faster than rigorous VLE calculations.