Inverses Modellieren in der numerischen Strömungsmechanik für kompressible Zweiphasenströmungen mithilfe von Machine Learning und differentieller Programmierung

mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) 2026 Thesis Open access

Read full text
Authors: Aaron B. Buhendwa
Cited by: 0

Abstract

This thesis advances the solution of inverse problems in CFD using machine learning and differentiable programming. It applies PINNs to infer two-phase flow fields, introduces a hybrid ML-based level-set interface reconstruction, and develops the JAX-Fluids framework for differentiable CFD. Finally, it extends the ODIL method for flow and shape inference from schlieren images in compressible flows.

Fields of Study & Subjects

Model Reduction and Neural Networks Computational Fluid Dynamics and Aerodynamics Fluid Dynamics and Heat Transfer

Publication details

Journal / source
mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich)
Publication date
29 January 2026
Language
EN
Document type
Thesis