To content
Department of Mechanical Engineering

Guest talk by Miriam Dodt

Begin: End: Location: Meeting room MB II 316

Miriam Dodt (from KU Leuven) gives a talk on "Active learning Grey-box modelling for near-real time modelling of fast processes under uncertainty".

Miriam Dodt is currently a research assistant and Ph.D. student at the KU Leuven under the supervision of Prof. David Moens. She received her B.Sc. degree in Civil and Environmental Engineering and her M.Sc. degree in Computational Methods in Engineering from the Leibniz University Hanover in 2018 and 2021, respectively.

In her talk, she will present a scheme for near-real time process control of extremely fast and complex processes, using the framework of digital twins and an adaptively calibrating grey-box model. Input parameters shall be adjusted to satisfy performance under non-deterministic process drift. Due to the complexity of the processes and their fast nature, a significant discrepancy between simulation time and process time is observed, which is countered by the use of surrogate models. In this case, Kriging meta-modelling is employed with adaptive refinement to reduce the computational burden associated to its calibration while ensuring sufficiently trustworthiness in the areas of interest. While the application in mind is resistance spot welding, the current state of method is very general.

Please contact our secretary in case you want to join the hybrid presentation via Zoom.