[RodriguezFatz]

Dear all,

I am trying to get the same results in Fluent and OpenFoam for a simple pipe flow (periodic). Unfortunately, the results are quite different and I am wondering if that is just normal for two solvers or if I am doing something wrong.
My settings are: DN25 pipe, standard k-e-model, mean velocity = 7 m/s, water, y+=1. I use "enhanced wall treatment" in Fluent and lowRe-wall functions in OpenFoam.
These are some results:
U-comparison-ske.png

k-comparison-ske.png

epsilon-comparison-ske.png
My feeling says something is wrong...

[-mAx-]

which Turbulence Model did you use in OpenFOAM?
If it can help you: http://www.cfd-online.com/Forums/ope...striction.html

[RodriguezFatz]

It's both the standard k-epsilon without any addition.

[-mAx-]

As far as I know there is no "enhanced wall treatment" option in OF like in Fluent for standard k-epsilon turbulence model.
But you can use k-Omega SST model with nutUSpaldingWallFunction.
Then set k and omega with uniform value 1e-10 instead of zeroGradient, especially if you have low y+ (y+<=1)

[flotus1]

I can confirm your results in fluent. Is the "standard" k-epsilon model exactly the same in both solvers, i.e. are the same values used for the coefficients?
Messing with the coefficients in fluent, one can approximately reproduce the results from OpenFoam

[RodriguezFatz]

Hi, I think the coefficients are the same for the standard k-epsilon. Most likely the boundary conditions are different, also I can't find exactly what Fluent does with all different turbulence values at the walls.

[RodriguezFatz]

Max, I also wrote in the thread you linked: I don't understand why you would set omega to zero for y+<=1 case. It should be some very high value instead.

[RodriguezFatz]

K-Omega-SST comparison:

U-comparison-komegasst.jpg

k-comparison-komegasst.jpg

omega-comparison-komegasst.jpg

[-mAx-]

Originally Posted by RodriguezFatz

Max, I also wrote in the thread you linked: I don't understand why you would set omega to zero for y+<=1 case. It should be some very high value instead.

Thanks for correcting me.

[RodriguezFatz]

U-comparison-SA.png

nuTilda-comparison-SA.png

[Geon-Hong]

Although you applied low Re BC on the wall, the standard k-epsilon model can not resolve appropriate turbulent field near the wall since it does not include a damping function.

Setting y+ to be close to the unity near the wall for the "standard" k-epsilon model is quite nonsense and I don't expect the solution to be correct with such settings.

It seems that you are trying to tighten a screw using pliers instead of screwdrivers. You can tighten screw somewhat using the pliers but it will not be tight enough.

Meanwhile the k-omega sst model can inherently resolve the viscous sub layer, and it is natural to estimate reasonable solution with y+ ~ 1.

If you want to maintain the y+ near the wall and use a k-epsilon family model, I'd recommend you to apply low Re k-epsilon model with low Re BC's, such as Launder-Sharma k-epsilon model.

[RodriguezFatz]

Hey Geon-Hong. Great! Thank you for the answer. That sounds pretty logical. I thought that maybe the wall function will patch all near boundary cells to apply the damping. But I didn't have a look at the source code, which should have revealed what you wrote... Thanks again!