Gabor Mode Enrichment in Large Eddy Simulations of Turbulent Flows
Speaker: Aditya Ghate, Science and Technology Corporation/NASA Advanced Supercomputing Division
Presentation
Seminar Slide Deck (PDF-45.0MB)
Abstract
Recent advances in wall-modeled Large Eddy Simulations (WMLES) allow for accurate prediction of first order turbulence statistics (means) using very low-resolution (coarse) computational grids. However, several applications such as unsteady load prediction, surface vibration, acoustic radiation and particle laden turbulence require highly accurate representation of a large spatio-temporal range of dynamically active scales that are present in high Reynolds number turbulence. The research addresses this emerging challenge in LES, and the new model substantially extends the spatio-temporal bandwidth of the solutions computed using low-resolution simulations with unprecedented accuracy. This "enrichment" is accomplished by representing subfilter scales in terms spatially and spectrally localized Gabor modes. Such a representation has a significant advantage over conventional wavelet representation; use of WKB-asymptotics based on the notion of quasi-homogeneity of small scales together with some scale separation arguments leads to substantial theoretically rigorous simplification of the temporal dynamics in an x-k frame advected by the large scales. The enrichment model is validated using several problems involving high Reynolds number wall bounded turbulence, including a problem with density stratification and Coriolis acceleration, using both ideal SGS closures (filtered DNS) and modeled SGS closures (eddy viscosity SGS models with equilibrium wall modeling). Finally, we demonstrate that in problems that lack a low-rank projection-based reduced order representation, Gabor modes can be used to enrich truncated modal representations to recover the full spatio-temporal bandwidth with excellent accuracy.
Biography