LES
 

HI. I need a professor has told me that I seem to hold misconceptions of LES, and I am open mindedly trying to correct myself. However, I don't see where I am may be confused. If someone has a good understanding of LES, its equations, how to model a problem with LES, etc.. please contact me. First, could someone tell me what is wrong with the following: "The Kolmogrov length scale is the smallest length scale of flow. For DNS, one would need to create cell dimensions smaller than these scales. But for LES, these scales should be modeled, "

Re: LES
 

The Kolmgorov Length scale is only associated with viscous dissipation. Not necessarily the smallest length scale of the flow.

Re: LES
 

So is there a length scale smaller than the Kolmogrov scale in some situations?

Re: LES
 

Batchelor scales...in case of scalar mixing for high Schimdt numbers....ref: Tennekes and Lumley "Introduction to turbulence"

Re: LES
 

I'm kinda new at cfd, but I thought the whole idea behind LES is that you don't model the Kolmogorov-scale but only the large-scale eddys. In LES the Kolmogorov-scale is time-averaged using k-e or Smagorinski or whatnot..

Correct me if I'm wrong..

Re: LES
 

Hi,

In LES, you (numerically) solve the large scales (largest eddies that should carry most of the turbulent kinetic energy) while (ideally) filtering out the scales from some wavelength into the inertial subrange of the TKE spectrum up to the Kolmogorov scale. The (instantaneous) filtered scales are then modelled.

k-E is not desirable for such a modelling for many reasons. Here are some: first, it is a RANS model, not suited to model instantaneous and local unknowns. Second, k involves the full (resolved+subgrid) TKE. Third, since the cell size acts as a characteristic length, you only need one additional variable so as to compute a turbulent viscosity if your model requires one.

Such classical models based on the idea of added (turbulent) viscosity could for instance rely on the resolved velocity gradients (Smagorinsky, Clark) or on value of k computed using test-filtering or an additional equation (multiple alternatives).

Another way is to try to obtain an estimate of the subgrid scale data either by invoking scale-similarity between the smallest resolved scales and the highest unresolved scales (Bardina) or by roughly inverting the filtering process (Taylor expansion of the filtered data or iterative use of a test filter, see for instance papers by Geurts and Adams, Stolz).

Finally note that for the first kind of models, the model constant is generally computed using turbulent identities that formally hold only for (ensemble) averaged data. One approach to overcome the problem is to use a dynamic method base on Germano's identity coming from a test filtering of the fields so as to adapt the "constant" to the local behaviour of the flow.

For further details with a rigorous mathematical background, you may look at the book by Sagaut (see the books section of this website).

Hope this helps.

Lionel

Re: LES
 

The correct title is: A First Course in Trubulence. Tennekes and Lumley

Bak_Flow

Re: LES
 

What is Trubulence? !!