Application of recurrence CFD (rCFD) to species transport in turbulent vortex shedding. (15th January 2020)

• Record Type: Journal Article
• Authors: Abbasi, S.; Pirker, S.; Lichtenegger, T.
• Is Part Of: Computers & fluids. Volume 196(2020)
• Publication Date: 2020-01-15
• Subjects: Large eddy simulations -- Recurrence statistics -- Multi-scale problems -- Time-extrapolation -- Recurrence CFD -- Data-assisted simulation; CFD computational fluid dynamics -- DMD dynamic mode decomposition -- DNS direct numerical simulation -- LES large eddy simulation -- PIV particle image velocimetry -- POD proper orthogonal decomposition -- RANS Reynolds-averaged Navier-Stokes -- RQA recurrence quantification analysis -- rCFD recurrence CFD; Fluid dynamics -- Data processing -- Periodicals; 532.050285
• DOI: 10.1016/j.compfluid.2019.104348
• Languages: English
• ISSNs: 0045-7930

Abstract

Highlights: We investigate the recurrent patterns of turbulent vortex shedding at Re = 3900. We time-extrapolate species transport based on the flow's reappearing structures. Recurrence CFD's resilience against mesh coarsening leads to very accurate results. We achieve speed-up factors of up to 120 compared to large-eddy simulations. Abstract: The functionality of computational fluid dynamics (CFD) for turbulent flows is limited by huge computational demands which prevent any detailed long-term studies. In this publication, we apply the recently introduced, data-assisted method "recurrence CFD" (rCFD) to turbulent vortex shedding after a circular cylinder at Reynolds number R e = 3900 . Using a database of flow fields from short, conventional simulations, we time-extrapolate their behavior to arbitrary durations and obtain promising results for passive species transport with speed-up factors of more than 120 at 1/20 of the required computer power compared to the underlying large eddy simulation (LES). Besides this massive run-time reduction, we focus on data efficiency. For cases with strong, spatial scale separation, rCFD's resilience towards grid coarsening allows us to carry out calculations at lower mesh resolution provided one retains meso-scale velocity fluctuations as a contribution to diffusivity. This reduces database size which would otherwise become a bottleneck in the methodology.