Lagrangian Measurements and Physics-Informed Neural Network for Rayleigh-Bénard Flow Reconstruction

Anne Sergent (1, 2) , Soufiane Mrini (1) , Elian Bernard (3) , Didier Lucor (4, 5)

1 COMET - COuplages Multiphysiques Et Transferts - LISN
2 UFR 919 - Sorbonne Université - UFR d'Ingénierie
3 Phys-ENS - Laboratoire de Physique de l'ENS Lyon
4 LISN - Laboratoire Interdisciplinaire des Sciences du Numérique
5 DATAFLOT - DAtascience, trAnsition, Fluid instability, contrOl, Turbulence LISN

The aim of this work is to reconstruct the temperature field from Lagrangian measurements in a turbulent Rayleigh-Bénard flow using a Physics Informed Neural Network. First, the spatial concentration of particles in the experimental dataset is examined using a 3D Voronoï analysis with a view to merging the experimental and numerical databases.

Domaines: Mécanique des fluides [physics.class-ph]; Modélisation et simulation; Réseau de neurones [cs.NE]

Anne Sergent, Soufiane Mrini, Elian Bernard, Didier Lucor. Lagrangian Measurements and Physics-Informed Neural Network for Rayleigh-Bénard Flow Reconstruction. 26th International Congress of Theoretical and Applied Mechanics (ICTAM2024), Aug 2024, Daegu, South Korea. ⟨hal-04924332⟩