von Karman vortex street
 

Greetings!

Though it might make some people sick, here is another question concerning vortex shedding.

Has anyone encountered a problem when vorticies move in the opposite direction and then get attached to the body?

My case: 2D, cylinder, oil, Re =100, laminar, and I'm sure about the time step. Mesh seems fine enough and it's symmetric. I use slightly asymmetric initial conditions for velocity (otherwise nothing happens, I guess due to mesh symmetry). CFX version is 12.1.

Sorry for asking about vortex street again, but I've read all the threads and found no answer. I would also very appreciate if someone gave me a link for a downloadable theory guide (it's a problem to find a paper book in English).

Thanks in advance,
Dmitry

Pictures would help.

Sure, this is a link to a movie file: http://www.youtube.com/watch?v=_FLRZgnNvHo

This is just a beat frequency effect between your vortex shedding frequency and your movie frame frequency. It's the same effect that makes videos of car wheels look like they are rotating slowly or backwards.

Will it help if I set results files to be written more freqently?

Yes, of course. Set it to be a fraction of the vortex shedding frequency, say 1/10.

What do you actually mean by asymmetric boundary conditions ??????

Statements like this always set off warning bells - most of the time the mesh is not fine enough and the user is just assuming that because the cells are really little it must be fine.

But the only way to be sure is to do a sensitivity analysis. A basic mesh size comparison is a minimum, but more advanced methods are available to give you a much better handle on the accuracy of your mesh. See http://journaltool.asme.org/Template...umAccuracy.pdf for a summary or read "Computational Fluid Dynamics" by Roache for the full story.

Thank you, Glenn! I've already read this article (thanks to one of your posts) and I understand the importance of performing sensivity analysis before accepting final results. But this time I didn't do it because I made Yplus many times smaller then recommended for such cases, just to be sure of the mesh. Besides, the case is very simple and well known. So, everything you said is right, I'm just trying not to look stupid. CFD by Roach will be my next book after I'm done with Shlichting.

Thank you, Michael! I guess your point about "results output frequency/movie frame rate" is right, but I had no time to try it yet. After I try it I'll post if it helped.

Dear Prof. Chaos! I mentioned initial conditions, not boundary conditions. I made them asymmetric to initiate vortex shedding, because otherwise I've been getting two symmetric vorticies, which stayed "attached" to the cylinder. I thought the reason was mesh symmetry that leaded to symmetric error accumulation -> no unsteadiness. I'd greatly appreciate any comments on that.

Dmitry

How did you incorporate asymmetric initial conditions ??????

To Prof. Chaos: Initial velocity = linear function of X, where X is a coordinate across the flow.

Yes, Michael was right! Output frequency was the problem. More trn results fixed it.

Thanks everybody!

I thought I'd just post my problem in this thread to avoid posting any more threads.
My model has ended up looking like this after 350 timesteps.
http://img841.imageshack.us/i/vortes.jpg/

This doesn't look right, and I'm not quite sure why.
Can anyone suggest a reason?
flow velocity = 16m/s
Timestep = 0.002s

What looks wrong about it? It looks about right to me.

The large region at the tail where the flow is diminished greatly.
I've done others with a different mesh, and that looked more similar to examples I've seen.

You mean how the vortex street does not reach the outlet boundary? You simply have not run it for long enough. Keep going and it will get there.

No, I mean how there's a big blue region at the tail of the vortex street.
Normally the vortices decay, when I look at other examples like.

What are the velocity vectors doing in that region? You never said what velocity component you showed in the image.

I still think it is just a startup thing and you just need to run longer.

The velocity component in the picture is absolute, not in any particular direction.
You're probably right, not having run it for enough timesteps is probably the reason it looks like that.
Thanks for your help.