[hulli]

Dear Foamers


I have a question about the simulation of flow past a non uniform obstacle in a box using inflow / outflow boundary conditions and a refinement box using the pimpleFoam slover. Basically similar to the motorbike case but with water. At this stage I intended to simulate that using a pure LES.

However, due to my bad experience with mesh refinementboxes and LES, which was already mentioned in several treads:



Mesh in LES


So I am a bit unsure now if I can use to use a pure LES approach since I observed velocity fluctuations (e.g., Wale, dynamicKe etc.) in particular in the combination with refinement boxes (i.e due to differing resolution as the mesh becomes finer).


So I was wondering by which means am I “allowed” to take a LES? Or would it be better / correct to use a hybrid LES RANS model such as the SpalartAlmaresDDES ? I am asking this because I plan to refine the background mesh with a refinement box of level 2 and a surface refinement of level 4 near the obstacle to get to a low enough Yplus values.



Any help is highly appreciated !!!


Thank`s H

[hulli]

... come on ... No ideas ? Is it saver to use a Hybrid turbulences model with mesh refinement or is a pure LES unaffected by that??? Thanks

[hulli]

Noone interesed in this

[BlnPhoenix]

No one forbids you using LES. It all depends on what you want to achieve with your simulation. Usually LES is used when RANS fails or isn't accurate enough. So what is your motivation to NOT use RANS?

Also, if you don't want mesh induced velocity fluctuations -> go for uniform fine mesh in that area, if that area is an area of interest.

[hulli]

Hi BlnPhoenix,

First of all thank you for your kind reply!!! May be I was not clear enough...

My motivation is to simulate the turbulent wake that develops behind a obstacle in a relatively small domain i.e., 2m in length only. I tried the RANS approach you suggested before but the turbulences appeared relatively smoothed out as one would expect it using RANS.

Therefore, I came to the idea to use a LES, which does make sense if one intends to simulate turbulences in high resolution. As mentioned above, when experimenting with refining the mesh I observed velocity fluctuations, in particular as the flow was transitioning form one refinement box trough another, so I end up with velocity fluctuations, which are related to the different levels in the mesh...

The point, which I am trying to make is that with snappyHexMesh my obstacle, which is relatively complex in my case, is also refined by some level. In order to get this complex structure in a the mesh SHM generates a lot of non uniform cells in the vicinity of the object by defining "refinement regions". Consequently, one would always end up with a non uniform mesh in the vicinity of the object, with SHM, right?

Therefore, I am was wondering, if I am "allowed" to use a pure LES even if I know that there are non uniform cells. Or more precisely,wouldn’t it be better to use a hybrid model such as a DES, to avoid potential fluctuations caused by a non uniform cells as the RANS approach would kick in?



Best H

[BlnPhoenix]

Hi,

yes snappy creates some non-uniform cells like tetraeaders, that can not be avoided.

Again if the areas where finer cells transit into larger cells are problematic i would use the finer grid spacing in the whole domain of interest. But maybe, and i don't know if thats true Hybrid models can do something.

Maybe someone else can help more.

Good luck

[hulli]

Hi,

well, thank you so much that helps a lot. Good to know about the SHM. Usually, I observed fluctuations only from coarser to finer cells and not the opposite.... so it should not be too problematic... so far have not found any fluctuations anymore. I just got them when having a couple of refinement boxes stacked after each other... So thanks again for the hint...