Prof. Dr. Alexander Mitsos (RWTH Aachen)

This talk gives an overview of our work in developing and using machine learning techniques for chemical engineering. We first briefly discuss how deterministic global optimization with neural networks embedded can be made tractable by reduced space formulation. We then present our work on property prediction with graph neural networks, including the usage of thermodynamic consistent models. We use these predictions for the computer-aided molecular design with generative machine learning. Finally, we discuss hardware-in-the-loop optimization using inline reaction monitoring.
20.11.2025, Zeit: 16:00
Raum: Sandtorstr. 1 (MPI), Seminarraum Prigogine