Machine learning in Computational fluid dynamics

Andre Weiner, Rolf Radespiel
TU Braunschweig, Institute of Fluid Mechanics

Outline

  1. Concept
  2. Challenges
  3. Evaluation
  4. Outlook

Concept

ML applied to CFD - new lecture/exercise with state-of-the-art topics and tools.

for2895

Lecture landing page on Github (link).

for2895

Jupyter notebooks - interactive lecture scripts.

Reveal.js slides

  • platform/device independent
  • available online
  • interactive (e.g. note-taking)

100% open-source!

  • no license costs - no strings attached
  • industry standard tools and libraries
  • Creative Commons and GPL content

$\rightarrow$ flexible and sustainable

Challenges

Mostly minor...

  • BigBlueButton vs. iPad vs. Eduroam
  • time consuming revision/creation process
  • (some) students miss essentials
  • no legal basis for guest access

Evaluation

Just a few selected numbers...

tubs_logo

Traffic created by lecture repository (two weeks).

Lecture content up-to-date?

tubs_logo

E-learning support?

tubs_logo

Outlook

  • repeat/extend lecture in WS 2023/2024
  • transfer concept to next lecture
  • solutions documented and accessible

Thank you!