 |
| Titel |
Direct and Large Eddy Simulation of Stratified Turbulence |
| VerfasserIn |
Sebastian Remmler, Stefan Hickel |
| Konferenz |
EGU General Assembly 2011
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 13 (2011) |
| Datensatznummer |
250053062
|
|
|
|
|
|
| Zusammenfassung |
| The dynamics of three-dimensional turbulence under the influence of density stratification
can be classified in different regimes that differ fundamentally; see Brethouwer et al. (2007)
for a good review. Up to now, strongly stratified turbulent flows were mainly studied by direct
numerical simulations (DNS). The available computer resources, however, limit the
application of DNS to Reynolds numbers that are too small for studying the regime of real
scale atmospheric problems. Large Reynolds number flows in general can be simulated
efficiently by large-eddy simulation (LES). However, most subgrid-scale (SGS) models for
LES are not suitable for stratified flows, as local isotropy of the SGS turbulence is assumed,
and therefore require ad-hoc modifications. We propose an implicit subgrid-scale model
that handles anisotropic turbulence in a straight forward way without any limiting
assumptions.
We solve the non-linear Boussinesq equations for a fluid with a constant background
stratification using a finite-volume solver. DNS are performed with a 4th order centered
scheme. The implicit sub-grid scale model is based on the Adaptive Local Deconvolution
Method (ALDM) for the incompressible Navier-Stokes equations (Hickel et al. 2006)
and the ALDM for passive scalar transport (Hickel et al. 2007). The turbulence
theoretical background of ALDM and the numerical details will be given in the full
paper.
Two generic test cases are considered. First, we computed the temporal evolution of a 3D
Taylor-Green-vortex (Brachet 1991) under the influence of stratification. We ran several fully
resolved simulations with up to 7683 cells and corresponding implicit LES with 643 cells.
The LES show good agreement with a fully resolved DNS. The total dissipation rate is
predicted correctly and also the proportions of kinetic and potential energy are
well estimated. We repeated the LES and DNS at a number of different Froude
and Reynolds numbers. In all cases we found good agreement between LES and
DNS.
As a second test case, we computed homogeneous stratified turbulence (Brethouwer et
al. 2007) at different Froude numbers. The Boussinesq equations are supplemented by a
volume force that acts on the large horizontal scales only. Three-dimensional structures with
finite vertical length scale develop only due to stratification. To ensure complete resolution of
the smallest scales, the DNS domain contained about 900 million cells. We studied the
influence of varying stratification on the averaged longitudinal kinetic energy spectra in the
vertical and horizontal directions.
We will discuss the energy budget, dissipation rates, spectra of turbulence, turbulence
length scales, and anisotropy of the DNS and draw conclusions with respect to SGS
modeling. Our computational results support that implicit LES with ALDM can be a suitable
tool for the investigation of stratified turbulence. This will enable us to compute and
characterize atmospheric turbulence beyond the reach of DNS with reasonable accuracy at
low computational costs.
References
[1]   G. Brethouwer, P. Billant, E. Lindborg, and J.-M. Chomaz. Scaling analysis
and simulation of strongly stratified turbulent flows. J. Fluid Mech., 585:343–368,
2007.
[2]   S. Hickel, N. A. Adams, and J. A. Domaradzki. An adaptive local
deconvolution method for implicit LES. J. Comput. Phys., 213:413–436, 2006.
[3]   S. Hickel, N. A. Adams, and N. N. Mansour. Implicit subgrid-scale modeling
for large-eddy simulation of passive scalar mixing. Phys. Fluids, 19:095102, 2007.
[4]Â Â Â M.E. Brachet. Direct simulation of three-dimensional turbulence in the
Taylor–Green vortex. Fluid Dynam. Res., 8(1-4):1 – 8, 1991. |
|
|
|
|
|
|