Combining Machine Learning with Physical Knowledge in Thermodynamic Modeling of Fluid Mixtures

Abstract

Thermophysical properties of fluid mixtures are important in many fields of science and engineering. However, experimental data are scarce in this field, so prediction methods are vital. Different types of physical prediction methods are available, ranging from molecular models over equations of state to models of excess properties. These well-established methods are currently being complemented by new methods from the field of machine learning (ML). This review focuses on the rapidly developing interface between these two approaches and gives a structured overview of how physical modeling and ML can be combined to yield hybrid models. We illustrate the different options with examples from recent research and give an outlook on future developments. Expected final online publication date for the Annual Review of Chemical and Biomolecular Engineering, Volume 14 is June 2023. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

CALiSol-23: Experimental electrolyte conductivity data for various Li-salts and solvent combinations

“…However, the validity of the model in the neighbourhood chemical space of molecules present in the dataset, will depend on the continuity and smoothness of the representation function in the chemical space. Identification of such appropriate descriptors requires exploring 58 , 59 a broad range machine learned and cheminformatics based representations 60 , 61 in combination with a wide variety of predictive classical 62 , 63 and machine learning models 57 , 64 – 68 and performing exhaustive testing. Molecular representations used by these models from the provided SMILES strings or after SMILES those to other datatypes like InChi, atomic graphs, or atomic position-based descriptions using cheminformatics tools like RDKit.…”

Gibbs–Duhem-informed neural networks for binary activity coefficient prediction

“…, the reviews in ref. 22 and 23). For example, Focke 24 proposed a hybrid neural network structure that embeds the Wilson model.…”

Mixtures Recomposition by Neural Nets: A Multidisciplinary Overview

“…To our knowledge, this is the first overview of ANNs for chemical mixtures despite some recent contributions in mixture thermodynamics and alloys . The focus being on mixtures, the reader can refer to an abundant collection of complementary reviews − to get a more exhaustive picture of ML application to chemical processes.…”