IRTG Modern Inverse Problems (MIP)

Multi-Scale Modeling, Model-Order Reduction and Uncertainty Quantification for Transpiration Cooling

06. April 2022 | von

Prof. Michael Herty

In this project we focus on constrained Bayesian inversion for problems governed by multi–scale partial differential equations, particularly for applications in transpiration cooling. A direct resolution of the inverse problem using existing approaches is prohibitive due to the need to resolve different scales numerically. We build upon the PIs‘ expertise on Bayesian inversion and machine learning methods for large-scale systems, asymptotic analysis for effective models  and generalized polynomial chaos expansion (gPC) and optimization  to develop novel constrained Bayesian inversion methods for multi–scale inverse problems.

We are interested in the analysis and development of optimization strategies for multi–scale fluid flow models. To obtain a successful method several challenges have to be addressed:

1) fast and reliable surrogate model to address the multi–scale aspect,

2) quantification of the propagation of uncertainty in model hierarchy and in the inverse problem and

3) development of a Bayesian framework under constraints.


Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht.