[ilpaso]

I'm trying to run a case with simpleFoam and kOmegaSST.

The geometry is a cycinder (diameter 3mm) with a shrinkage (diameter 1.2mm). Length 117mm
the estimated Reynolds is about 1700. (U=4.42 m/s diameter 1.2mm)

First I've to define the k and omega. According with THIS page w=sqrt(k)/L. L (turbolence length scale) is defined L=0.038*d.
The result of all these formulas are k=4.78 and w=4.8e5!!!!!

I've read THIS and with these formulas k=0.073 and w=k/epsilon=27

I am very confused.
can anyone help me?

thank you

[alexeym]

Hi,

surely someone can help you. What's your question besides a possibility of help?

[ilpaso]

Hi,
the question is:
what are the right formulas for k-w turbolence model in order to write the right values of k and w and not random values?

thank you.

[alexeym]

Are you trying to set initial values, boundary conditions, anything else?

For example if you look into turbulentMixingLengthFrequencyInletFvPatchScalarFi eld.H and turbulentIntensityKineticEnergyInletFvPatchScalarF ield.H you find the formulas OpenFOAM is using for calculating k and w from given intensity and length scale:

k = 1.5*(UI)^2

and

w = sqrt(k)/(L*Cmu^0.25)

Intensity depends on the case (from below 0.01 upto 0.2), L can be taken equal to 0.038D.

[ilpaso]

I'm trying to set up the entire case.
After the geometry the next step is to choose the right model (I hope).
But I didn't understand the formulas and I didn't find any reference in the userguide.

At the end of THIS wiki page there is a formula about w a bit different from your. w=Cmu^0.25 * sqrt(k) / L
Is this wrong?


with the formulas you wrote, if:
I = Re^(-1/8)*0.16=0,06 (supposed Ux=4.42, Uy=0, Uz=0, x is the axis of cylinder)
L = diam * 0.038= (1.2e-3)*0.038= 4.56e-5
k = 1.5*(Ux * I)^2 = 0,105
then w=13000

are they reasonable?

[alexeym]

Formula in the wiki is Cmu^(-0.25)*sqrt(k)/L and is the same as I've cited from source code.

Here are links to FLUENT and ESI Group pages on defining turbulence parameters:

http://combust.hit.edu.cn:8080/fluen...ug/node175.htm

http://support.esi-cfd.com/esi-users/turb_parameters/

They are not much different from the wiki page.

Take a look at tutorials using kOmegaSST, check the typical values in the tutorial, compare with yours.

[ilpaso]

Thank you for the links.
My model goes to convergence in about 320 iterations. (with simpleFoam and k-w turbolence model (not komegaSST)
What do you think about this result? I'm running iterative simulations and any improvement is appreciated.
The mesh has about 160000 hexahedra elements.
Attached here you can find my case if you want take a look........
There is not the mesh because it is 28Mb but there is an image of the geometry
The main diameter is 3mm and at the shrinkage the diameter is 1.2mm

[alexeym]

As I don't know your final aim, I can't say anything about results.

Case converged? Results seems to be reasonable? OK, go to the next problem.

You've shown geometry but not the mesh itself. Maybe you need to increase density near the walls, check y+.

You'd like to decrease number of iterations/calculation time? Play with discretisation schemes and linear system solvers.

[ilpaso]

thank you Alexey!
Your suggestions are always precious!

Only for info:
the mesh is structured with increased density near the wall.
Attached here there are 2 images.
I'm learning step by step

[RodriguezFatz]

Davide, are you sure you mean "structured"? This doesn't look like a structured mesh, I guess you mean a hexa-mesh, right?

[ilpaso]

I Philipp,
thank you for your reply! Every day reading this forum I can reflect as I am ignorant about the cfd world.

Ok. My mesh is not structured. Can I say it is hybrid: structured at the boundary layer and not structured in the middle?

Thank you

[RodriguezFatz]

Structured or not just describes the way the data is stored in the memory. Having a box meshed by hexas with dx=dy=dz doesn't necessarily mean, that the mesh is structured. But it can be. I think most solvers use only unstructed meshes. Your mesh is probably just completely unstructured.