[kalle]

Hi!

For a while I have been trying to run some axisymmetric 2D-cases. For this I have used the wedge patchtype (5 deg angle). Mostly I have set a tangential velocity to the outer cylindrical surface... much like a driven cavity. However, no case has been successful as the velocity normal to the moving wall has developed a wavy pattern, see figures. If I plot velocity vector magnitude in paraview, things looks nice, because the erroneous wall normal velocity is small compared to the tangential velocity... however, I think that the solution is not reliable, especially since the wall normal velocity is important for my cases (heat transfer). I have tried different schemes, high tolerances. Linear convergence is also OK. I suspect that the way OpenFOAM ensures continuity may be the root cause, but I have not figured it out. I have not seen this for planar cases, so it might relate to the axisymmetric geometry. I have also tried a 3D-sector mesh, using cyclic, but no improvement. For all cases I have had to export the mesh to CFX where things have gone smooth... however I am sure OpenFOAM should be able to handle this?! Anyone have any ideas?

In attached test cases I have used icoFoam and rhoPimpleFoam (just run blockMesh and then the relevant solver)

Regards,
Kalle

[olivierG]

hello,

You run a 2D axi case, and in your settings, set Uz=100 m/s , in the symmetry direction !
I guess you don"t want axis-symmetry case here !

regards,
olivier

[kalle]

Hi!

Thanks for your reply. Well, the test case simulates the situation where the outer surface rotates (100 m/s) and at the same time moves slowly in the the axial direction (1 m/s). As such the mesh may be two dimensional, but momentum for all 3-directions should be solved. Axisymmetric does not nessessarily mean that the tangential velocity component is zero.

K

[olivierG]

hello,

You say that all 3 dimension should be solver, and try to reduce to only 2.
What you need is 3D mesh with cyclic. Check the incompressible/simpleFoam/pipeCyclic for that.

regards,
olivier

[kalle]

Hmm... the solver solves for all three directions, but I will give cyclic a try. I see that the tutorial you mention has some cells also in the tangential direction, which I guess is needed for cyclic.

As I mentioned I have seen the same pattern also for a sector mesh, where I used cyclic and I had some 30 cells in the tangential direction. The radial pressure gradient is really strong for many of these cases... might cause trouble. I will look carefully on the tutorial on Monday when I am back in office!

Thanks!
K

[kalle]

Hi again!

I've had a look at the pipeCylic tutorial. It's aim appears to be to show cyclicAMI capabilities. It calculates nicely, simulating swirling flow in a straight pipe. However, my problem relates to the situation where the swirl is generated by a moving wall, rather than as swirl coming from an inlet.

As for the question if 3D swirling flow can be calculated on a 2D-wedge domain, Hrv suggests that it can be done in FOAM (just like I did it in CFX)

I have done a few of those simulations (2-D wedge with swirl) and it's OK.

from http://www.cfd-online.com/Forums/ope...tml#post188553

http://www.cfd-online.com/Forums/ope...tml#post185824


I suspect that the strong radial pressure gradient causes trouble here, or that some general problem with disctretization or solution procedure exists... but I have not been able to point it out...

Kalle