Job Offers - Research Positions

Machine Learning in Thermodynamics

In the course of the digitalization, there is currently an intensive discussion about how machine learning (ML) methods can be used in process engineering. Preliminary work at the Laboratory of Engineering Thermodynamics (LTD) has already demonstrated the potential of data-driven ML methods for predicting the thermodynamic properties of mixtures. ML methods also open up new perspectives in other fields of process engineering; the transfer from their classical fields of application (image and text recognition, autonomous driving, etc.) to applications in process engineering is, however, not trivial. Of particular importance will be the question of how data-driven ML approaches can be combined with basic thermodynamic knowledge in a favorable manner.

Within the framework of the research project, matrix completion methods - known from the so-called "Netflix Challenge" - will be specifically combined with physical knowledge to develop hybrid predictive fluid property models. These models are furthermore to be optimized by experimental measurements using methods from Active Learning (Design of Experiments). As an example, a method for the determination of diffusion coefficients using NMR spectroscopy, which is established at the LTD, can be used for this purpose.

In this project funded by the Carl-Zeiss-Stiftung, you will work in an interdisciplinary team of process engineers, computer scientists and mathematicians in close collaboration with the Chair of Machine Learning at the TU Kaiserslautern and the Fraunhofer Institute for Industrial Mathematics. Furthermore, there is a cooperation with the University of California, Irvine.

download job opening (German version)



Last Change: June 10th 2020