TagsArticle Source
When Artificial Intelligence meets Computational Fluid Dynamics
The Aeroseminar lecture given by Dr. Michael Bauerheim titled “When Artificial Intelligence (AI) meets Computational Fluid Dynamics (CFD)” in the faculty of Aerospace Engineering at TU Delft.
Abstract
For decades, Computational Fluid Dynamics (CFD) has relied on solving the discretized PDEs governing the fluid flows, resulting in numerous discoveries on how flows may behave, in both academic and complex industrial configurations. Recently, Artificial Intelligence (AI) has emerged in physics and engineering, including fluid mechanics, offering new paths to renew our way to predict fluid flows. This presentation will introduce several works coupling neural networks with CFD solvers, in order to accelerate computations, to provide innovative flow control laws, or to achieve optimization at low cost. A particular focus is made on space and/or time varying problems, including 2D and 3D complex flows, such as ones encountered when dealing with flow instabilities. These works will reveal the potential benefits, but also the current limitations of those AI methods when applied in CFD, as well as works and pathways to circumvent these issues.
Short bio
Dr. Michael Bauerheim received his PhD from the Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique (CERFACS) (France) in the field of thermoacoustics in 2014. Following this, he spent time as a postdoctoral researcher at the Toulouse Institute of Fluid Mechanics and at ETH Zurich. Since 2017, he is an Associate Professor at ISAE-SUPAERO (France) and Sherbrooke University (Canada), working on aero-acoustics and bio-inspired aerodynamics using both high-fidelity numerical simulations and artificial intelligence.