[RodriguezFatz]

Hi all,

I am currently trying to get a vortex street inside a pipe but pisoFoam converges to steady state no matter what I set in fvSchemes. I resolve the boundary layer and don't see the street for the k-omega-SST and the Launder-Sharma-k-epsilon model.
This is what exactly the same grid produces in Fluent (with standard settings)
vortex_street.jpg

The question is, why does it work in Fluent and not in OpenFoam? If you need more info, just tell me what to post!

Thanks,
Philipp.

[hxaxtma]

You have to set up a disturbance, guess your mesh is fully structured?
Add some acoustic noise or time dependent velocity dist. or create a small assymetry of your Mesh.

See here around second 10!
http://www.youtube.com/watch?v=hsmT0zvKNeM

[RodriguezFatz]

Ok, how can I do that?
I got a velocity inlet with U, k and omega/epsilon profiles from a first simulation at the inlet

[RodriguezFatz]

Also, I don't think that OpenFoam can handle structured grids. If you mean that my grid is created in ICEM using blocking / hexa, then I can answer "yes". But it is converted to unstructured grid when using it in OpenFoam.

[davibarreira]

Did you figure out a way to get the vortex street? Im having the same problem.

[RodriguezFatz]

Yes, if I recall correctly, I had much too high residuals. So I actually didnt solve the equations... Can you post your log file? How many time steps per (expected) vortex cycle do you have?
In my case I had to reduce the time step strongly or increase the number of pressure corrections to get lower continuity residuals.

[davibarreira]

I figured what was wrong. The problem was the turbulence model. I was using k-epsilon and they were smearing the vortices. I ran a simulation with Re = 150 with laminar flow, and the vortex street showed up. The only thing Im trying to solve now is the values of Drag and Lift Coefficient, that Im getting above the expected.

Thanks for the reply, anyways!

[RodriguezFatz]

OK... But it should work with k-epsilon as well. If you still want to get it running, just let us know.

[davibarreira]

Really? I read people saying that k-epsilon would dissipate the vortices... So what do you think would be the problem?
Actually, right now Im more worried about the Drag Coefficient. I keep on getting a 1.44 (converging with mesh improvements), but I should be getting something closer to 1.33 (in another paper, the author got a 1.38 with OpenFOAM). My flow is a laminar flow with 150 Re ... Any ideas?

Cheers

[davibarreira]

RodriguezFatz,

I manage to get the laminar flow to produce great results... now Im trying it out with higher Reynolds number with k-epsilon (also playing with some LES)... k-epsilon is giving reasonable values of drag coefficient, but no vortex-street. Would you now how to set it up?

Regards

[RodriguezFatz]

Ok, so can you post some log output of the k-epsilon case that doesn't work?
And the schemes file.

[davibarreira]

I increased the mesh refinement and let it run for a long time. After an appreciable time (the flow looked like it was stationary already) the vortices started to form.
Sorry for bothering you, I should've tried longer before asking in the forum. But thank you very much for the reply, I really appreciate it.

Anyways, here are my fvScheme and fvSolution, if you have any tips on how to improve the simulation.

regards

[RodriguezFatz]

Do you have the log as well?

[hxaxtma]

Just an additional information regarding my mesh post,

Of course OF can handle structured or unstructured meshes, what I wanted to say is, if your mesh is fully structured it is a 100% symmetrical mesh with respect to cylinder`s center axis in flow direction. The flow always follows your mesh, so it keeps symmetrical a very long time.

For decreasing this waiting time, some people tend to create a small mesh disturbance on on side of the symmetry side. So you get a small disturbance - produced by small mesh assymmetry - which equals in assymmetric flow and vortex shedding....

[RodriguezFatz]

hxaxtam, I think OF does not handle structured meshes, just unstructured. Also, structured does not mean that a mesh is symmetrcal to anything. Are you sure you use the correct words?

[hxaxtma]

Originally Posted by RodriguezFatz

hxaxtam, I think OF does not handle structured meshes, just unstructured. Also, structured does not mean that a mesh is symmetrcal to anything. Are you sure you use the correct words?

Ok, just for clearification:
OF handles structured and or unstructured meshes by solving the NS in an unstructured way.

This is what I meant....

Regarding the test case of davibarreira:
Discretization: C-H-O Mesh Type (structured) -> Axis symmetry!
Of course not in general!

[RodriguezFatz]

I don't understand you...
Structured and unstructured is just a matter of how the software internally stores the mesh and this is unstructured for Openfoam (as for Fluent and I guess most other solvers). You can not use a structured mesh in OpenFoam. Maybe you can import structured meshes by converting them to unstructured, but I don't know about that.

Where can I see davibarreiras mesh?

[hxaxtma]

Originally Posted by RodriguezFatz

I don't understand you...
Structured and unstructured is just a matter of how the software internally stores the mesh and this is unstructured for Openfoam (as for Fluent and I guess most other solvers). You can not use a structured mesh in OpenFoam. Maybe you can import structured meshes by converting them to unstructured, but I don't know about that.

Where can I see davibarreiras mesh?

Hehe I think I was not clear enough :-). I was meaning exactly the same thing you wrote. Of course you need to convert any Mesh into OF Mesh format, which is by definition unstructured. So we are now on the same lane*g*

[davibarreira]

Sorry, I dont have a log.

I think hxaxtma meant a mesh with all orthogonal cells (all the cells are cubes, therefore, the mesh could be completely symmetric).

[RodriguezFatz]

Davi, without the log I can't help, because that's where the mistakes show up usually...