CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   LES idea (

B. Saunders. February 20, 2002 15:26

LES idea
Hello to you all,

I am a student who is new to LES and I wish to understand if there are any similarities between LES and RANS? Thank you,


Jongdae Kim February 20, 2002 16:20

Re: LES idea
1. LES

LES uses (space) filtered Navier-Stokes and continuity equations. So the solutions are (space) filtered properties, U,V,W,P,etc.


Analysis with

RANS uses (time) filtered Navier-Stokes and continuity equations. So the solutions are (time) filtered properties, U,V,W,P,etc.

Depends on your purpose, you can choose one of those methods. In my case, turbulent flow around bluff body, LES with standard Smagorinsky model gives good results compared to experimental data. The RANS and k-epsilon model estimate reasonably but ....


sylvain February 25, 2002 05:39

Re: LES idea
I appologize for that, but RANS is NOT time filtered NS equations. To get the RANS equations, the NS equations are averaged over realizations. This allows time dependance computations when the time variation are driven by moving mesh or time dependent BCs.



Jongdae Kim February 25, 2002 11:55

Re: LES idea
I agree with you. Ensemble averaging is necessary to get RANS equations. However in computational method using RANS, how to get ensemble averaged N-S equation? To make problems simple, time filtering is a kind of Reynolds averaging. This was what I mean.



sylvain February 25, 2002 12:29

Re: LES idea
I agree, for steady state turbulence (statistically steady flow), time average values tend to ensemble average values, as an exemple :

Experimentally, for the following channel flow :

width = 0.2m; velocity = 10m/s; turbulence intensity = 20%.

one needs to average over 8s to get a correct (<1%) mean velocity value.

one needs to average over 200s to get a correct (<1%) rms value of the fluctuating velocity.

RANS solver give these average values directly.



Jongdae Kim February 25, 2002 13:56

Re: LES idea
There is numerical error which is growing as the simulation continues. Sometimes, the numerical error can be small due to positive and negative numbers. But anyway, after long time simulation, the accuracy of numerical simulation decreases. So we are careful on the total computation time to get the results from unsteady simulation. (To discuss this problem, the size of grids, the spatial and temporal discretization,etc. are to be touched.)

In my case(LES of flow analysis around 2-dim. square prism, Re=22000), I use periodic condition in spanwise direction with the assumption that the flow is homogeneous in spanwise direction. (But this is not true. My collegue measured the flow field and got some correlation in spanwise direction.) To get velocity and turb. intensity profiles considering reasonable computational time and accuracy, sometimes, I use the average of the whole spanwise direction data.

This is a kind of ensemble averaging.

(One of my problem is how to compare LES data and the data from RANS equations.)


All times are GMT -4. The time now is 15:32.