Fully Data-Driven Solving of Aerofoil-Wind Interactions in a Single Training of a Single Multi-Output DeepONet

Download PDF

Sebastien Andre-Sloan, Dibyakanti Kumar, Anirbit Mukherjee

Published: 02 Jun 2024, Last Modified: 27 Apr 2024Selected for a Talk @ Advances in Data Science and Artificial Intelligence (ADSAI) Conference 2024EveryoneRevisionsCC BY 4.0
Abstract: In this note we demonstrate how a multi-output DeepONet can be trained using just Colab-Pro to simultaneously solve for the wind, pressure and viscosity fields around varying aerofoil shapes and inlet velocities, as given in the AirfRANS dataset. We get within one order of magnitude of the baselines using just 15% of the data as used by them.