Hei,

to validate my LES code I am simulation the flow around bluff bodies. As a first test case I take the classical example of a flow around a prismatic obstacle at Re=22000. If I do a 3D simulation everthing is ok and I obtain a Strouhal number of ~ 0.132. If I try a 2D calculation St increases to about 0.23.

Is this what I can expect from a 2D solution of an inherent 3D problem or is there something wrong with my code? Has anybody solved this test case in 2D?

Best regards, Hans

hi, now i am validating my LES code through simulating a cylinder with Re=3900, 3d, compressilbe,unsteady flow. the results are not good. the pressure coefficient at front stagnate point is about 1.2,not 1. and Cp=-0.6__ -0.8, is that right? the mesh total points are 200*200*24. if you do not mind , could you send me your prismatic obstacle mesh and the parameters in flow,for example,Ma,Re,Uinf,Pinf etc.

by the way, I'm a graduate student of BUAA in Beijing,China.may be we can help each other. yours sincerely Bin Li

Hi,

Generally, 2D LES doesn't make sense. Even if the computed configuration is somehow 2D in a mean sense (i.e. with an homogeneous flow direction), LES have to be three-dimensional. The reason for this is that turbulence is generally a 3D phenomenon, with vortex stretching along all the directions of space. Since LES resolves part of the turbulent structures, the computation should be able to reproduce the stretching of the largest vortices, and therefore a third dimension is desperately required.

The few exceptions to this rule occur for flows with real 2D turbulence (for instance stratified flows) and require specific subgrid model (TKE spectrum is really different when looking at 2D or 3D turbulence).

In conclusion, I think that the discrepancies you've observed for ``2D LES'' are meaningless.

Hope this help.

Best regards.

I thuoght that in general 3D turbulence is about the fact that large vortices break up in smaller vortices. [/quote] The modern concept of turbulence, developed in the early 1920s ( Richardson 1922), brought forth the idea that " the big whirls have little whirls that feed on their velocities, and little whirls have lesser whirls and so on to viscosity - in the molecular sense". [/unquote]

In fact in 2D it works the other way around; small vortices tend to group together to form larger structures. See for instance http://www.cfm.brown.edu/people/hall...cts/mix2d.html

Hence, comparing results of 3D and 2D LES seems bogus to me

This is what I was looking for:

http://cnls.lanl.gov/Highlights/1998-02/html/

What Lionel said doesn't mean that 2D turbulence doesn't exist. He just said that 2D turbulence is very different than 3D.

In 3D, U_mean_Z_direction equal to zero doesn't lead to u_fluctuation_Z_direction is equal to zero. So you must solve U_Z_direction in your LES to get the correct behavior.

In 2D, U_Z_direction doesn't exist.

i solve the N-S equations for 3d compressible turbulence and find the pressure coefficient is about Cp=1.2 at front stagnate point,-0.6 -- -0.8 was found at back. as you known, the Cp at front is 1.0 in theory in 2D. is this means that i could ignore the difference? by the way , the spanwise scale at least is Pi/2.(now ,i can't remember the reference) in order to contain the large scale votex. because i am improving my LES code now ,and have to validat it with some examples. i need the supersonic examples or that with large Re number. anybody could help me? or could you give me some advice about compressible LES. yours sincerely Bin Li

I didn't say it doesn't exist. It's just different behaviour so do not expect similar results as you would for laminar quasi 2D flows

I would say that 2D turbulence is not turbulence. Don't ask me what is it, because I do not know. I think in this case the name is misleading.

Turbulence is 3D