Hello there,

I am using a primitive variable NS FEM formulation for exterior problems. What should be the outflow conditions? Is there a difference for interior problems? I am trying to catch the vortex street shed behind a cylinder in 2d. What should be the size of the computational area?

Anybody with experience on FLOTRAN or PHOENICS?

Thanks a lot.

Achilleas Tsompanos

Hi,

there are many papers on the Von Karman vortex street simulation and you can find answers to some of your questions there. Just check the Journal of Computational Physics, Journal of Fluid Mechanics, etc...

I am not using commercial codes (flotran, phoenics or whatever) but the size of the computational domain for a vortex street behind a cylinder in 2D depends a little bit on the Reynolds number and what you want to check. Say if you have really nice vortices there that you want to resolve (high Reynodls number), then take the outer radial boundary to be (say) 5 to 10 times the radius of the cylinder (this depends also and how much resolution you can afford, etc...). On the other side if you have a low Reynodls number then you can take the outer radial boundary to be located at about 20-25 radii of the cylinder. Say on one side you have inflowing boundary conditions and on the other you have ouflowing boundary conditions (for best results impose them on the inflowing characteristics of the flow, other wise you might have some additional frequencies because of reflection of waves).

The only reference I have here under my hand is:

Don & Gottlieb (1990), Computer Methods in Applied Mechanics and Engineering, vol. 80, p. 39.(low Reynodls number, Re=80, Mach=0.4).

and see also the references therein.

I hope this helps,

PG

There is a classic trick for generating the von Karman vorticies.

For many codes, you must create some perturbation to get the vorticies started. Once started, the vorticies should grow and the perturbation can then be eliminated.

The most common perturbation is applied at the inlet BC. For example, put a 0.1% stronger flow on one half of the inlet than the other, then after 10 timesteps switch the perturbation to the other half of the inlet, then turn the perturbation off. It may take more than 10 timesteps. It was a 10 years ago when I read about this and experimented with it.

Hope this helps. Clifford

Dear Achilleas Tsompanos,

Please, see the following useful web site!

http://ourworld.compuserve.com/homep...ima/karman.htm

Please, do not hasitate to write me, if I could have helped you.

Sincerely,

L.Könözsy

Hi,

I have simulated the Karman vortex street with Star-CD. You can see the result in the adress below (sorry that the text only is in swedish):

http://www.smhi.se/sgn0104/miljo/cfd/index.htm

Sincerely

Lennart Wern