GRAPPA—A hybrid graph neural network for predicting pure component vapor pressures

Although the pure component vapor pressure is one of the most important properties for designing chemical processes, no broadly applicable, sufficiently accurate, and open-source prediction method has been available. To overcome this, we have developed GRAPPA — a hybrid graph neural network for pred...

Full description

Saved in:
Bibliographic Details
Main Authors: Marco Hoffmann, Hans Hasse, Fabian Jirasek
Format: Article
Language:English
Published: Elsevier 2025-05-01
Series:Chemical Engineering Journal Advances
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S266682112500047X
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850142569790439424
author Marco Hoffmann
Hans Hasse
Fabian Jirasek
author_facet Marco Hoffmann
Hans Hasse
Fabian Jirasek
author_sort Marco Hoffmann
collection DOAJ
description Although the pure component vapor pressure is one of the most important properties for designing chemical processes, no broadly applicable, sufficiently accurate, and open-source prediction method has been available. To overcome this, we have developed GRAPPA — a hybrid graph neural network for predicting vapor pressures of pure components. GRAPPA enables the prediction of the vapor pressure curve of basically any organic molecule, requiring only the molecular structure as input. The new model consists of three parts: A graph attention network for the message passing step, a pooling function that captures long-range interactions, and a prediction head that yields the component-specific parameters of the Antoine equation, from which the vapor pressure can readily and consistently be calculated for any temperature. We have trained and evaluated GRAPPA on experimental vapor pressure data of almost 25,000 pure components. We found excellent prediction accuracy for unseen components, outperforming state-of-the-art group contribution methods and other machine learning approaches in applicability and accuracy. The trained model and its code are fully disclosed, and GRAPPA is directly applicable via the interactive website https://ml-prop.mv.rptu.de/.
format Article
id doaj-art-efb594021079499dbc5919e05a58c466
institution OA Journals
issn 2666-8211
language English
publishDate 2025-05-01
publisher Elsevier
record_format Article
series Chemical Engineering Journal Advances
spelling doaj-art-efb594021079499dbc5919e05a58c4662025-08-20T02:29:02ZengElsevierChemical Engineering Journal Advances2666-82112025-05-012210075010.1016/j.ceja.2025.100750GRAPPA—A hybrid graph neural network for predicting pure component vapor pressuresMarco Hoffmann0Hans Hasse1Fabian Jirasek2Laboratory of Engineering Thermodynamics, RPTU Kaiserslautern, Erwin-Schrödinger-Str. 44, 67663 Kaiserslautern, GermanyLaboratory of Engineering Thermodynamics, RPTU Kaiserslautern, Erwin-Schrödinger-Str. 44, 67663 Kaiserslautern, GermanyCorresponding author.; Laboratory of Engineering Thermodynamics, RPTU Kaiserslautern, Erwin-Schrödinger-Str. 44, 67663 Kaiserslautern, GermanyAlthough the pure component vapor pressure is one of the most important properties for designing chemical processes, no broadly applicable, sufficiently accurate, and open-source prediction method has been available. To overcome this, we have developed GRAPPA — a hybrid graph neural network for predicting vapor pressures of pure components. GRAPPA enables the prediction of the vapor pressure curve of basically any organic molecule, requiring only the molecular structure as input. The new model consists of three parts: A graph attention network for the message passing step, a pooling function that captures long-range interactions, and a prediction head that yields the component-specific parameters of the Antoine equation, from which the vapor pressure can readily and consistently be calculated for any temperature. We have trained and evaluated GRAPPA on experimental vapor pressure data of almost 25,000 pure components. We found excellent prediction accuracy for unseen components, outperforming state-of-the-art group contribution methods and other machine learning approaches in applicability and accuracy. The trained model and its code are fully disclosed, and GRAPPA is directly applicable via the interactive website https://ml-prop.mv.rptu.de/.http://www.sciencedirect.com/science/article/pii/S266682112500047XVapor pressure predictionMachine learningGraph neural networksAntoine equationPhase equilibria
spellingShingle Marco Hoffmann
Hans Hasse
Fabian Jirasek
GRAPPA—A hybrid graph neural network for predicting pure component vapor pressures
Chemical Engineering Journal Advances
Vapor pressure prediction
Machine learning
Graph neural networks
Antoine equation
Phase equilibria
title GRAPPA—A hybrid graph neural network for predicting pure component vapor pressures
title_full GRAPPA—A hybrid graph neural network for predicting pure component vapor pressures
title_fullStr GRAPPA—A hybrid graph neural network for predicting pure component vapor pressures
title_full_unstemmed GRAPPA—A hybrid graph neural network for predicting pure component vapor pressures
title_short GRAPPA—A hybrid graph neural network for predicting pure component vapor pressures
title_sort grappa a hybrid graph neural network for predicting pure component vapor pressures
topic Vapor pressure prediction
Machine learning
Graph neural networks
Antoine equation
Phase equilibria
url http://www.sciencedirect.com/science/article/pii/S266682112500047X
work_keys_str_mv AT marcohoffmann grappaahybridgraphneuralnetworkforpredictingpurecomponentvaporpressures
AT hanshasse grappaahybridgraphneuralnetworkforpredictingpurecomponentvaporpressures
AT fabianjirasek grappaahybridgraphneuralnetworkforpredictingpurecomponentvaporpressures