Fingerprinting and thermodynamic modeling of poorly specified mixtures with NMR spectroscopy and machine learning

• Applicant: Professor Dr.-Ing. Fabian Jirasek
• Subject Area: Chemical and Thermal Process Engineering
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 462456621

In this research project, methods for the thermodynamic modeling of mixtures of which the composition is unknown will be developed based on NMR spectroscopy and machine learning. Such poorly specified mixtures occur frequently in practice but are difficult to handle, as classical thermodynamic models require the knowledge of the composition, i.e., can only be applied to fully specified mixtures. This challenge will be addressed in the present research project by 1) the development of automated methods to elucidate and quantify the structural groups in poorly specified mixtures based on NMR spectroscopy (1H, 13C, 13C DEPT, 1H-13C HSQC NMR experiments) and the machine-learning concept of support vector classification (SVC); 2) the development of automated methods to rationally define pseudo-components based on the obtained group-specific characterization and DOSY NMR experiments; and 3) the combination of the developed methods with thermodynamic group-contribution methods to predict fluid properties of poorly specified mixtures.

• DFG Programme: Research Grants
• International Connection: USA
• Co-Investigators: Professor Dr. Robert Bamler; Dr. Kerstin Münnemann
• Cooperation Partner: Professor Dr. Stephan Mandt