[ssa_cfd]

Hi all,

I have done a simulation, flow around buildings, using pimpleFoam and RANS (Kepsilon model). The results were good.

But when I run the same simulation with LES (Keqn model), the results are completely different.

Does anyone have idea, how to solve this.?


Thanks.

[piu58]

One of the main tasks in simulations is to ensure the correctness of the results. Or, more accurate, to get a feeling of the uncertainty.

I recommend that you first used both methods again geometries where an analytical or exactly measured solutions exists. You will find deviation to a more or lesser degree. Then you need to find out how to minimize them by changing the parameters of your model.

[ssa_cfd]

Hi Uwe Pilz,

I have experimental data from a journal, and I have compared the results from RANS simulations which are agreeable.

But in case of LES simulations, even the flow field is not matching. To be clear, according to the experiment the flow should move down in between the buildings, but in my case, the flow moves up. The LES simulations are completely not acceptable.

Thanks.

[pete20r2]

Have you tried initialising the LES case with the RANS results? Try mapFields for U and p fields at least and see if it's a issue with initial value sensitivity. Can you also post the 0 directory for each case, if not the whole case (both)?

[ssa_cfd]

Hi,

I have not tried mapping fields. I will try to do that.

In 0 folder I have not done any changes. They are same for both LES and RANS. I have just changed the turbulence properties file in case of LES.

I have always used RANS simulation and I am new to LES. What parameters should be changed.?

U:
Inlet : fixedProfile
Outlet : zeroGradient
walls : noSlip

p:
Inlet : zeroGradient
outlet : 0
walls : zeroGradient

k:
Inlet : fixedProfile
outlet: zeroGradient
walls : kqRWallFunction

nut:
Inlet : 0
outlet : zeroGradient
walls : nutkWallFunction.

all others are symmetryPlane.

[basrag]

Hi,
For comparing results, are you comparing values for U for each case?
If that is the case try comparing U from RANS and Umean from LES, that should give you some similar results since doesn't resolve eddies but LES U value do show LES.

[ssa_cfd]

Hi basrag,

Yes I am comparing the time averaged velocity field from LES, still the results are different from RANS.

[Tobi]

I just want to make a statement. Even if you are using RANS with another turbulence model, you will see that the results are different too; or change the mesh size. Are you using the same grid? Can the grid resolve the large-scaled vortexes? In general, the solution should definitely be more chaotic based. A good example can be found from Jószef Nagy. He is using the pitzDaily case and made an analysis with different turbulence models.