Maschinelles Lernen in der Thermodynamik

Die präzise Vorhersage thermodynamischer Eigenschaften ist für die chemische Industrie von zentraler Bedeutung, da experimentelle Daten oft nur begrenzt verfügbar sind. Etablierte physikalische Modelle stoßen dabei hinsichtlich Vorhersagegenauigkeit und Anwendungsbreite zunehmend an ihre Grenzen. Methoden des maschinellen Lernens (ML) bieten hierfür vielversprechende neue Ansätze.

Besonders hybride Verfahren, die die Stärken physikalischer Modellierung mit denen des ML kombinieren, haben in den letzten Jahren bemerkenswerte Ergebnisse geliefert. In meinem Projekt untersuche ich daher das Potenzial solcher hybriden Ansätze und erforsche, wie thermodynamisches Wissen systematisch in ML-Modelle integriert werden kann, um deren Vorhersagegenauigkeit und Robustheit zu erhöhen.

Veröffentlichungen

• M. Hoffmann, T. Specht, N. Hayer, H. Hasse, F. Jirasek: MLPROP – An Interactive Web Interface for Thermophysical Property Prediction with Machine Learning, Chemie Ingenieur Technik (2025). [DOI]
• N. Hayer, H. Hasse, F. Jirasek: Modified UNIFAC 2.0 – Completing Interaction Parameter Tables with Machine Learning, Industrial & Engineering Chemistry Research, 64 (2025) 10304–10313. [DOI]
• N. Hayer, T. Specht, J. Arweiler, H. Hasse, F. Jirasek: Similarity-Informed Matrix Completion Method for Predicting Activity Coefficients, The Journal of Physical Chemistry A, 129 (2025) 3141–3147. [DOI]
• N. Hayer, T. Specht, J. Arweiler, D. Gond, H. Hasse, F. Jirasek: Prediction of activity coefficients by similarity-based imputation using quantum-chemical descriptors, Physical Chemistry Chemical Physics 27 (2025) 4307-4315. [DOI]
• N. Hayer, H. Hasse, F. Jirasek: Prediction of Temperature-Dependent Henry’s Law Constants by Matrix Completion, The Journal of Physical Chemistry B 129 (2025) 409-416. [DOI]
• N. Hayer, T. Wendel, S. Mandt, H. Hasse, F. Jirasek: Advancing thermodynamic group-contribution methods by machine learning: UNIFAC 2.0, Chemical Engineering Journal 504 (2025) 158667. [DOI]
• M. Hoffmann, N. Hayer, M. Kohns, F. Jirasek, H. Hasse: Prediction of pair interactions in mixtures by matrix completion, Physical Chemistry Chemical Physics 26 (2024) 19390-19397. [DOI]
• F. Jirasek, N. Hayer, R. Abbas, B. Schmid, H. Hasse: Prediction of Parameters of Group Contribution Models of Mixtures by Matrix Completion, Physical Chemistry Chemical Physics 25 (2023) 1054-1062. [DOI]
• O. Großmann, D. Bellaire, N. Hayer, F. Jirasek, H. Hasse: Database for liquid phase diffusion coefficients at infinite dilution at 298 K and matrix completion methods for their prediction, Digital Discovery 1 (2022) 886-897. [DOI]
• N. Hayer, F. Jirasek, H. Hasse: Prediction of Henry's law constants by matrix completion, AIChE Journal 68 (2022) e17753. [DOI]
• N. Hayer, M. Kohns: Thermodynamically Rigorous Description of the Open Circuit Voltage of Redox Flow Batteries, Journal of the Electrochemical Society 167 (2020) 110516. [DOI]

Vorträge

• L. Vollmer, R. Loubet, P. Zittlau, M. Hoffmann, N. Hayer, S. Fellenz, F. Jirasek, H. Leitte, H. Hasse: Translating Thermodynamic Knowledge to Computers, Thermodynamik-Kolloquium, Stuttgart, 25.-27.09.2024.
• N. Hayer, S. Mandt, H. Hasse, F. Jirasek: Embedding Machine Learning Methods in Physical Thermodynamic Models, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases 2024, Vilnius, Lithuania, 09.-13.09.2024.
• N. Hayer, H. Hasse, F. Jirasek: Advancing Group-Contribution Methods for Thermophysical Properties of Mixtures, 22nd Symposium on Thermophysical Properties, Boulder (CO), USA, 24.-28.06.2024.
• N. Hayer, H. Hasse, F. Jirasek: Extending the Scope of Thermodynamic Group-Contribution Methods by Machine Learning, Thermodynamik-Kolloquium, Hannover, 25.-27.09.2023.
• N. Hayer, F. Jirasek, H. Hasse: Hybrid Group-Contribution Methods for Predicting Thermodynamic Properties of Mixtures, 14th European Congress of Chemical Engineering and 7th European Congress of Applied Biotechnology, Berlin, 17.-21.09.2023.
• N. Hayer, F. Jirasek, H. Hasse: Combining Machine Learning with Thermodynamic Group-Contribution Methods, ECTP2023 – 22nd European Conference on Thermophysical Properties, Venice, Italy, 10.-13.09.2023.
• N. Hayer, T. Specht, J. Arweiler, H. Hasse, F. Jirasek: Prediction of Activity Coefficients by Similarity-Based Imputation using Quantum-Chemical Descriptors, Thermodynamik-Kolloquium, Chemnitz, 26.-28.09.2022.
• N. Hayer, F. Jirasek: Learning from Netflix – Recommender Systems for Predicting Thermodynamic Properties, Young Researchers Symposium, Kaiserslautern, 22.07.2022.
• F. Jirasek, N. Hayer, T. Specht, J. Damay, M. Bortz, R. Abbas, B. Schmid, H. Hasse: Hybrid Predictive Fluid Property Models – Integration of Physical Knowledge in Data-driven Matrix Completion Methods, 13th European Congress of Chemical Engineering and 6th European Congress of Applied Biotechnology, Web-Conference, 20.-23.09.2021.

Poster

• N. Hayer, S. Mandt, H. Hasse, F. Jirasek: Advancing Group Contribution Models by Machine Learning: UNIFAC 2.0, International Conference on Properties and Phase Equilibria for Product and Process Design, Bad Gögging, 11.-15.05.2025.
• N. Hayer, S. Mandt, H. Hasse, F. Jirasek: Advancing Group Contribution Models by Machine Learning: UNIFAC 2.0, Thermodynamik-Kolloquium, Stuttgart, 25.-27.09.2024.
• N. Hayer, S. Mandt, H. Hasse, F. Jirasek: Embedding Machine Learning Methods in Physical Thermodynamic Models, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases 2024, Vilnius, Lithuania, 09.-13.09.2024.
• N. Hayer, M. Kohns, F. Jirasek, H. Hasse: Active Learning for the Prediction of Second Virial Coefficients in Mixtures, Thermodynamik-Kolloquium, Hannover, 25.-27.09.2023.
• N. Hayer, F. Jirasek, H. Hasse: Prediction of Temperature-Dependent Henry's Law Constants by Matrix Completion, Thermodynamik-Kolloquium, Chemnitz, 26.-28.09.2022.
• O. Großmann, D. Bellaire, N. Hayer, F. Jirasek, H. Hasse: Prediction of Diffusion Coefficients at Infinite Dilution by Matrix Completion, ProcessNet and DECHEMA-BioTechNet Jahrestagungen 2022 together with 13th ESBES Symposium, Aachen, 12.-15.09.2022.
• N. Hayer, F. Jirasek, H. Hasse: Prediction of Henry's Law Constants by Matrix Completion, ProcessNet and DECHEMA-BioTechNet Jahrestagungen 2022 together with 13th ESBES Symposium, Aachen, 12.-15.09.2022.
• N. Hayer, F. Jirasek, H. Hasse: Prediction of Temperature-Dependent Henry's Law Constants by Matrix Completion, European Symposium on Applied Thermodynamics (ESAT), Graz, Austria, 17.-20.07.2022.
• O. Großmann, D. Bellaire, N. Hayer, F. Jirasek, H. Hasse: Establishment of a Database and Prediction of Diffusion Coefficients at Infinite Dilution, Thermodynamik-Kolloquium, Web-Conference, 27.-29.09.2021.
• N. Hayer, F. Jirasek, H. Hasse: Prediction of Henry's Law Constants by Matrix Completion, Thermodynamik-Kolloquium, Web-Conference, 27.-29.09.2021.
• N. Hayer, M. Kohns: Thermodynamically Rigorous Description of the Open Circuit Voltage of Redox Flow Batteries, Thermodynamik-Kolloquium, Web-Conference, 27.-29.09.2021.
• N. Hayer, F. Jirasek, H. Hasse: Prediction of Henry's Law Constants by Matrix Completion, European Symposium on Applied Thermodynamics (ESAT), Web-Conference, 05.-09.07.2021.
• N. Hayer, M. Kohns: Thermodynamically Rigorous Description of the Open Circuit Voltage of Redox Flow Batteries, Annual Meeting on Reaction Engineering, Web-Conference, 10.-12.05.2021.