CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Proper use of LES (

Karl Kevala September 6, 2001 17:57

Proper use of LES
I am doing three-dimensional sliding mesh simulation and have used several RANS models. I want to use LES. Based on relation in Temkes and Lumley for the Taylor microscale, L,

L ~ l (15 nu) / (TKE l)

where nu is kienematic viscosity, TKE is turbulent kinetic energy, and l is characteristic length.

Using this formula and experimental data for the TKE, L is about 0.1 mm throughout significant regions for my problem. There is no way to make my mesh so small. Right now, my cell sizes are about 0.25 mm length on all sides. This is about the smallest I can make my mesh and still solve in a reasonable amount of time (i.e. on the order of > 2 months).

Might I get useful information from LES simulation using mesh cells larger than Taylor microscale?


Karl Kevala September 6, 2001 18:02

Re: Proper use of LES
For L, that should have been written:

L ~ l (15 nu) / (TKE l)^ (1/2)

Chetan Kadakia September 7, 2001 19:22

Re: Proper use of LES
I am not sure of the Taylor microscales, but I am familiar with the Kolmogrov scales. If you are attempting DNS, then your grid needs to be smaller than the Kolmogrov scales. But if you are using LES, for many flows, the grid size can be one or two magnitudes greater than the Kolmogrov scales and still obtain valid results that are close to DNS and still better than a RANS solution.

Mike Henneke September 8, 2001 12:38

Re: Proper use of LES
LES models don't require you to model the Taylor microscale. The idea behind LES is to separate the turbulent motions into 'resolved' scales (which are larger than your mesh) and 'unresolved' scales (which are smaller than your mesh). Done correctly, the resolved scales are the geometry dependent scales and the unresolved scales are universal to all turbulent flows (at least that's the hope). This is a big reason that LES is problematic for combusting flows, the small scales aren't universal, but depend on the chemistry. The subgrid scale filter is necessarily grid dependent, so you cannot have grid independent results.

You should get better than RANS results so long as your grid resolves the integral length scales well, i.e., the motions that are on the order of your geometric length scales.

I'm not an LES expert, this is just what I've heard and understood. Also, I have heard that rapid changes in grid sizes and aspect ratios are problematic.

YMMV, Mike

All times are GMT -4. The time now is 07:54.