notesum.ai

Published at November 13

Flow reconstruction in time-varying geometries using graph neural networks

cs.LG

cs.AI

physics.flu-dyn

Released Date: November 13, 2024

Authors: Bogdan A. Danciu¹, Vito A. Pagone¹, Benjamin Böhm², Marius Schmidt², Christos E. Frouzakis¹

Aff.: ¹CAPS Laboratory, Department of Mechanical and Process Engineering, ETH Zürich, Zürich, 8092, Switzerland; ²Technical University of Darmstadt, Department of Mechanical Engineering, Reactive Flows and Diagnostics, Otto-Berndt-Str. 3, Darmstadt, 64287, Germany

Arxiv: http://arxiv.org/abs/2411.08764v1

Graph Neural Network	MAE [ $\text{\,}\mathrm{m}\text{/}\mathrm{s}$ ]	RMSE [ $\text{\,}\mathrm{m}\text{/}\mathrm{s}$ ]
GACN	0.156	0.247
GraphSAGE	0.201	0.334
GCN	0.287	0.429