Hey all!

I am curious, because some CFD users use LES to do 2D simulations! But

1. Isn't turbulence is 3D? To what extent can 2D simulation capture the essence of turbulence?

2. Is it correct to use LES do 2D simulation? Will it be still better than RANS?

3. If so, what should I take notice if I do a 2D-LES? And How could it happen?

Regards and thanks, Daniel

In my opinion 2D LES is a joke. Its much better to run RANS and get mean results.

Regards

1) Yes turbulence is 3D. But there are cases, as metereology, where a 2D approach has still some sense; look at

" Davidson - turbulence, an introduction for scientists and engineers - Oxford"

2) No, in any case. Also the theoretical 2D turbulence caracteristics are quite different from the 3D ones and are not accounted for in any LES model (also there is no reason for them being included). This actually not means that RANS will perform better.

Due to my poor english i don't understand the last question, but i can close saying that if you need to test a 2D case you have to use a RANS model. If you have an unsteady case the answer is no more so clear depending on the kind of flow. What kind of results will give you an URANS model? Will they be physical? Maybe in a meteorological simulation an URANS model has less sense that a 2D LES model.

so, do you mean 2D-LES is meaningless, it's better to use RANS in 2D approach?

True 2-D LES simulations, in spite of what has been written in many research articles over the years, are not what one would describe as being fully-turbulent. In shear flows such as wakes, mixing layers and jets the flow remains in an unsteady laminar state and does not capture the physics of the real flow. Close inspection of the results from 2-D LES of these types of flows in the literature will show this, but researchers tend to dismiss these subtle differences as they are more interested in validating new numerical schemes against what are seen as simple flow configurations.

In terms of accuracy, I would say that a 2-D LES is more accurate than a RANS (at least for shear flows) as it captures the Kelvin-Helmholtz vortices and their evolution correctly. 3-D LES is more accurate than 2-D LES as it can, on a sufficiently fine grid, capture the salient physics of the turbulent flow.

I do not fully agree.

You can use a 2-D LES if the instantaneous velocity field is 2-D. Examples can be the initial development of a mixing layer or jet, or the wake behind a long cylinder when there are only instabilities but there is no turbulence yet. When the flow is turbulent it is just physically wrong to apply a 2-D LES. It will give you the wrong results. I have never seen a 2-D LES of a turbulent boundary layer or a channel flow. It is either a 3-D LES or a 2-D RANS in these cases (or a hybrid/DES). A 2-D RANS can be used if the mean flow is 2-D, otherwise you have to use a 3-D RANS of course. That's not so difficult to understand, is it?

Yes, being careful with the urans approach in the unsteadt cases.

I'm not aware of any direct comparison between 2D and 3D LES of the initial region of jets or mixing layers, hence I think that it it shouldn't be assumed that the flows will behave the same in each case.