Wavelet adaptive POD for large scale flow data

Philipp Krah
TU Berlin
https://scholar.google.com/citations?user=FSv1I7gAAAAJ&hl=de

Date(s) : 20/11/2020   iCal
14 h 30 min - 15 h 30 min

The proper orthogonal decomposition (POD) is a powerful classi-cal tool in fluid mechanics, for instance used in model reduction andextraction of coherent flow features. However, its applicability to highresolution 3D DNS data is limited due to its computational complex-ity. Here, we propose a wavelet-based adaptive POD, called wPOD,which overcomes this limitation. The size of the analyzed data is re-duced by exploiting the compression properties of wavelets with errorcontrol, which yields a sparse flow representation. Numerical analysiswill illustrate the influence of wavelet compression and POD trunca-tion errors. Examples will be presented for 2D and 3D high resolutionDNS data. A comparison with the randomized singular value decom-position will be provided to compare the efficiency and the precisionof the wPOD method.

Authors : Philipp Krah (speaker), Thomas Engels, Kai Schneider, Julius Reiss

Slides: https://www.i2m.univ-amu.fr/seminaires_signal_apprentissage/Slides/20102020_KrahEngelsSchneiderReiss_wPOD.pdf

Emplacement
Site Nord, CMI, Salle de Séminaire R164 (1er étage)

Catégories



Retour en haut 

Secured By miniOrange