You are on page 1of 4

Advances in Turbulence IX

Proceedings of the Ninth European Turbulence Conference


id:85,1] I.P. Castro & P.E. Hancock (Eds.)
c CIMNE, Barcelona 2002

The e ect of SGS models on the vortical


structures computed from Large-Eddy
Simulations
1
C. B., da Silva

1 Equipe MOST/LEGI - Institut de Mecanique de Grenoble


B.P. 53, 38041 Grenoble Cedex 09, France

Contact address: Carlos.Silva@hmg.inpg.fr

1 Introduction
One of the most impressive achievements of Large-Eddy Simulations (LES) is
the possibility of computing the large scale ow vortices which govern the large
scale transfers of mass and momentum. Since an important aspect the subgrid-
scale (SGS) models must represent is the kinetic energy exchange between the
resolved and unresolved scales, several authors have examined the inuence of
the coherent vortices on the so called \subgrid-scale dissipation" term ij Sij< ,
(e.g. Piomelli et. al 1] , O'Neil & Meneveau 2] ). In this context, a
complete account of the role played by the coherent vortices in the interscale
interactions in plane jets was recently given by da Silva & Metais 3] . Another
important issue that begins to receive increased attention is the inuence of the
SGS models on the coherent structures computed from large-eddy simulations
(e.g. Silvestrini 4] , Vreman et. al 5] , Horiuti 6] ).
The present work is focused in analyzing this problem through a-priori tests
of several SGS models. The starting point is the transport equation for (half the
square of) the resolved vorticity modulus,
@
 
1 < < + u< @ 1 < < = < < S < +  @ 2 1 < <

i i j @xj 2 i i | i {zj ij} @xj @xj 2 i i
|@t 2{z } | {z } III | {z }
I II IV
@ < < 2 kp
;  @xi @xi ; "ijk <i @x@ @x
@
 (1)
| j j
{z } | j p {z }
V VI
which is valid local and
 @uinstantaneously.
 Here, the \<"s denote a spatially
< 1 < @u< j
ltered variable, Sij = 2 @xj + @xi is the grid-scale (GS) deformation tensor
i
2 The e ect of SGS models on the vortical structures from LES

and ij = (ui uj )< ; u<i u<j is the unknown subgrid-stress tensor which must
be modeled. The eect of the SGS models in the computed vortical structures
from LES appears through term VI (SGS vorticity dissipation). Several plane
jet DNS were carried out upon which a box or top hat lter was used to separate
the grid and subgrid-scales of motion (details of the simulation are given in da
Silva & Metais 3] ).

Figure 2: PDF of terms III, V and VI


Figure 1: Proles of averaged terms III, from equation (1) at the far eld of the
V and VI from equation (1) at the far turbulent plane jet.
eld of the turbulent plane jet.

2 Results and discussion


It was observed that in the mean, terms II (ltered vorticity advection) and IV
(viscous vorticity diusion) are negligible. The evolution of the vorticity modulus
is mainly governed by its production (term III), viscous vorticity dissipation
(term V) and the SGS vorticity dissipation (term VI) (term I averages to zero).
Figure 1 shows proles of averaged terms III, V and VI from equation (1)
at the far eld of the turbulent plane jet, for two dierent lters. As expected,
the mean vorticity production (III) is balanced by the viscous vorticity dissipa-
tion (V) and the SGS vorticity dissipation (term VI). Note that, unlike what
happens when considering the transport of the ltered grid-scale kinetic energy,
the dissipation caused by the molecular viscosity cannot be neglected in LES.
Figure 1 shows also that, in the mean, the SGS vorticity dissipation (term VI)
acts as a sink of vorticity modulus (always negative), thus the coherent vortices
obtained from a LES are subjected to an additional (non-viscous) dissipation
mechanism. What is surprising from Figure 1 is that this terms is of the same
order of magnitude than the viscous dissipation term, meaning that the SGS
da Silva, C. B. 3

models will have a non negligible impact on the resolved vorticity of the ow.
Thankfully, as we shall see below all models work quite well.
Figure 2 gives some more insights into the nature of terms III, V and VI from
equation (1). The gure shows PDFs from these terms at the far eld of the
turbulent plane jet, for lter 2 = 25. The gure shows that the SGS vorticity
dissipation (term VI) can be either a sink (forward enstrophy cascade - TV I < 0)
or a source (backward enstrophy cascade - TV I > 0) of resolved vorticity, but the
forward cascade is dominates, in analogy with what happens with the resolved
kinetic energy equation.
A-priori tests were conducted using several SGS models in order to analyze
their ability to represent the SGS vorticity dissipation (term VI from equation
(1)). The models analyzed were the standard Smagorinsky (Smag), Structure
Function (SF), Selective Structure Function (SSF), Dynamic Smagorinsky (Ds-
mag) and the Gradient model (GM) (Lesieur & Metais 8] , Meneveau &
Katz 7] ).

Figure 4: PDF of real and modeled


Figure 3: Proles of real and modeled terms VI from equation (1) at the far
averaged terms VI from equation (1) at eld of the turbulent plane jet.
the far eld of the turbulent plane jet.
Figure 3 shows proles of real and modeled term VI from equation (1) at the
far eld of the turbulent plane jet. The gure shows that both the Structure
Function and the Dynamic Smagorinsky models, compare well with the exact
value of the SGS vorticity dissipation. The classical Smagorinsky model leads to
an excessive dissipation of the resolved vorticity eld whereas both the Gradient
model and the Selective Structure Function models do not provide enough vor-
ticity dissipation. One may argue that the vortices computed from LES using
these models will tend to be bigger or survive longer than the real ones. These
4 The e ect of SGS models on the vortical structures from LES

ndings are in agreement with the results from Vreman et. al 5] for the
evolution of the vorticity maximum in LES of a mixing layer.
PDFs of the real and modeled term VI are shown in gure 4. An encouraging
result comes from the forward enstrophy cascade tail (TV I < 0) which shows
that all the terms compare very well with the real PDF of term VI. Another
interesting result is that all the models provide a backward enstrophy cascade
(TV I > 0). This is in contrast with the resolved kinetic energy equation in
which the eddy-viscosity models never provide any backward energy transfer.
The Smagorinsky and Structure Function models cause the smallest backward
enstrophy cascades. The Dynamic Smagorinsky model seems to have the best
agreement with the correct values of the backward enstrophy tail, whereas the
gradient model leads to an excessive backward enstrophy cascade. This agrees
also with its well know propensity to cause numerical problems.
A deeper analysis of the topological behavior of these models is presently un-
derway. Isosurfaces of vorticity and modeled term VI show that most of the SGS
vorticity dissipation occurs right at the core of the ow vortices (the correlation
coecient between the vorticity norm and term VI modeled by the Smagorin-
sky model is C(V ISmag )  ;0:5), whereas the backward enstrophy transfer
occurs next to the vortices. Surprisingly all the models give a reasonably good
agreement with the topological behavior of the real SGS vorticity dissipation.
References
1] U. Piomelli, Y. Yu, and R. Adrian. Subgrid-scale energy transfer and near-
wall turbulence structure. Physics of Fluids A, 8(1):215{224, 1996.
2] J. O'Neil, and C. Meneveau. Subgrid-scale stresses and their modeling in a
turbulent plane wake. J. Fluid Mech., 349:253{293, 1997.
3] C. B. da Silva, and O. Metais. On the inuence of coherent structures upon
interscale interactions in turbulent plane jets. Submitted to J. Fluid Mech.
(revised form), 2002.
4] J. Silvestrini. Simulation des grandes echelles des zones de melange! ap-
plication "a la propulsion solide des lanceurs spatiaux. PhD thesis, INPG,
Grenoble, 1996.
5] B. Vreman, B. Geurts, and H. Kuerten. Large eddy simulation of the turbu-
lent mixing layer. J. Fluid Mech, 339:357{390, 1997.
6] K. Horiuti. Assessment of subgrid-scale models in dissipative vortical struc-
tures. Advances in Turbulence VIII, CIMNE, 519{522, 2000.
7] C. Meneveau, and J. Katz. Scale invariance and turbulence models for large-
eddy simulation. Annual Review of Fluid Mech., 32:1{32, 2000.
8] M. Lesieur, and O. Metais. New trends in large-eddy simulations of turbu-
lence. Annual Review of Fluid Mech., 28:48{98, 1999.

You might also like