[Fabf]

Hi,

assuming a simple 2D-axisymmetric case (side = symmetry planes, no swirl) what effect has the chosen wedge-angle on the results? Referring to the documentation 1 to 5 degree is preferable. From OpenFOAM I know 5 deg is recommended.

Regarding my case: transient simulation of thermal stratification in a cylindrical tank. I do get "better" results (less convection, less movement in the layer) if my model is only 1deg. Mesh is quite fine and timesteps as well (1e-2s).

From commen sense it's somehow clear that, e.g. a 1 deg model introduces a smaller error than a 5deg model. But what's the detailled explanation and why is 5 deg recommended? Shouldn't the recommended angle rather be a function of mesh-density, transport variables and numerical error?

Looking forward to get some answers

[ghorrocks]

When you make the wedge angle smaller it means you more closely approximate the full circular shape and this reduces error. But you also make the quality of the mesh worse - as you now have a small angled element in the central axis. Different simulations have different sensitivities to poor mesh quality, so some simulations will run OK with a 1° wedge angle, but more sensitive ones will never converge.

So the compromise is better resolution of the full geometry against decreasing numerical stability.

[evcelica]

Axis-Symmetric is one major fallback of CFX.
One thing I do when meshing, if it is appropriate:
1.) Leave out a very small portion of the center, so that you have an annular space instead of a wedge. Place a symmetry plane on that small surface. Now you can sweep quads 1 element through the thickness.
2.) Another thing I do if I can't leave out the center, is use a mesh method similar to this: it requires more than 1 element through the thickness unfortunately, but maybe only 2,4,or 6. I haven't perfected this yet obviously, but it is better than doing a full 90 degrees.

[Fabf]

Thanks for the input! I wasn't aware of that issue.

So the recommendation for 2d-axisymmetric cases should rather be:
1) Use a 5°-Mesh and make sure the case is converging (model is allright).
2) Verify results from 1) with a 1°-Mesh.
3) If case is diverging, rsp. mesh quality is too bad, improve meshing near axis.

Luckily, the center of my geometry is already empty. So I will just change all my cases from 5 to 1 degree.

Erik, your mesh from 2) looks interesting! I keep that in mind. For another case I had quite bad elements in the center, but CFX was alright with that. Coarsen the mesh along the axis helped. Smoothing hexa-meshes made it worse (icem). Also deleting degenerating elements didn't help. Anyway, I had good experiences with the collapseEdges utility in OpenFOAM after creating axisymmetric meshes from a quasi-2D/3D-icem Mesh. But I couldn't reimport it to CFX..

[tripio2000]

To aid in the convergence, you could use the 5deg mesh case as initial condition for the 1deg mesh case.