[Diro7]

Hello everyone!

I'm trying to mesh the attached geometry (Fig. 1).

fig1.jpg

Basically the geometry features a 1/16 cyclic symmetry, such that the bodies to be actually meshed are only the ones in Fig. 2, where the green-highlighted boundary faces (and the corresponding ones behind) are symmetric.All the other boundary faces are external and therefore treated as walls.

fig2.png

As you can see, I divided the part into different bodies to try to exploit further symmetries and the capabilities of the MultiZone method.
Everything just fine so far, and I'm able to get good quality meshes with approx. 100-200k elements. Selective body meshing and sweepable bodies work like a charm (Fig. 3).

fig3.png

I use the mesh for some custom multiphysics calculations with OpenFOAM. Results are pretty good in terms of stability and convergence, but I'd like to add some inflation on external boundaries to enforce a better turbulence boundary layer treatment without a dramatic increase of mesh size.
When I try to apply inflation on all the external boundaries, the MultiZone method works no more, due to the complex intersection pattern of inflated faces of different bodies (on the egdes highlighted in Fig. 4).

fig4.png

The only feasibile workaround I found so far is to mesh all bodies with MultiZone, except two which are meshed with Automatic/Tetrahedrons.
The resulting meshes (Fig. 5) gives no error (except a warning on stair-step inflation layers, which is reasonably expected due to the particular geometry).
Unfortunately, the quality is significantly poorer (in the tet-meshed bodies, of course) and I'm not able to properly simulate to convergence the model.
I've also tried the CutCell method, but the overall quality is poor.

fig5.png

I've been struggling with this for days.
Does anyone have some advice for tackling the problem?
Some possible alternatives I thought of are:

1. Use the original mesh as it is, with no inflation. The things "work", but screw to any possibility to properly model boundary layers.
2. Refine the original mesh globally to a stupid size. This might work, but since I'm actively working on model development I'd strongly prefer some more suitable mesh for everyday use on a limited number of CPUs
3. Divide the mesh in 2 parts (inner cylinder and outer pipes), mesh separately with the MultiZone and use a non-conforming mesh with some mapped-cyclic coupling on the 2 interface surfaces. I know OpenFOAM allows for something of this kind, but I have literally no idea on how it would perform in my case.

Does anyone have some ideas?
Perhaps some smart further splitting of the geometry, or maybe some other tool in ANSYS Meshing to add inflation while keeping the structure of the original mesh using MultiZone?

Any contribution will be much appreciated!
Thank you in advance,

Andrea

[AtoHM]

Hi, since nobody else wrote anything yet, I give you some feedback. Firstly, props for getting that kind of mesh from the multizone method, I never came so far with it.



Secondly, I wonder if it is possible to mesh this all with one blocking approach nevertheless. I would go with 4x3 blocks for a start, create an ogrid for the inner block and one at the outer "ring". Then, the innermost block can be deleted. Then the blocking can be adjusted to fit your ring properly. The parts standing out of the plane can be just added with the insert block tools. The part where the edges converge for the "center" is a bit trickier, but manageable by collapsing a block edge to get a triangular block (viewn from above). Then the only thing left would be to increase the refinement level in the blocks further away from the symmetry axis and finally resolve refinements after creating the mesh.


There might be some obstacle I have overlooked, but this might work. Good luck.

[Diro7]

Hi M!

first of all, thank you for the reply and the valuable suggestions.

I'm trying to figure out the precise meaning of what you suggested and how to apply it to my case.
I'm no meshing expert and, for instance, I've never heard of such thing as "o-grid" before (my bad, actually). Though after a quick search I realized what you are referring to and that I simply didn't know it had a name of its own

Anyway, I'm afraid you may be talking about some kind of meshing functionalities (o-grids, blocking etc) which are not available in ANSYS Meshing, or at least not in the v19 Student version I'm using. Unfortunately my institution doesn't provide full license for PhD students, and it seems for instance that ICEM is not included in Student licenses.

Am I correct, or am I just ignoring some functionalities of ANSYS Meshing?

Andrea

[AtoHM]

Uh, my bad ... I was talking about ICEM actually, sorry I just now see you tagged it with ANSYS Meshing. Are you sure you cannot access it? We are using ICEM at my company and have ANSYS Meshing available as well, but I do not know if that works vice-versa as well.

[Diro7]

Don't worry! I'll try to figure out if I can somehow access some fully licensed version of ANSYS.
Or I'll try to make use of your comments with alternative meshing tools.

Thank you for the help!
Best,

Andrea