Okay quality in Tetras, worst quality when adding prism mesh
 

Hey everyone,
I am meshing a centrifugal pump using a unstructured tetra/prism mesh in Icem. The mesh is hardly acceptable when i create the tetra mesh with robust octree meshing and after merging or replacing the mesh with the prism layers the quality becomes really bad.


what I have done:
I imported the geometry from Solidworks, build Topology with Filtering Points and Lines, created the parts and set the global Mesh scale 1, Max is now at 30 (I tried setting it to 0 and got a good tetra mesh but creating a prism mesh there were holes in my mesh). Curvature are enabled (min 1; gap 1; Ref 12), Volume is Robust Octree, Part mesh as it is in the pic.
For the "best" tetra results I got so far I did some surface mesh for the very thin areas (also tried define them as thin cuts but it had no affect on the quality).
I used the curve Mesh setup on the thin ending cuts of the blades.

The Octree Mesh has already almost 4.000.000 Elements and 700.000 nodes.

Improving quality I got an OK mesh for the Tetras (see pic).

I made some testrunning in cfx and set the first prism layer that my y+ is about over 20 and about under 200.

Creating the prism layers:

What I think:
Problematic areas are the thin end of the impeller blades and the thin gap between case and impeller.

The bad elements seem to be mostly the tetras right after the prism layers. I tried to set up the inflation with the surface setup tetra width, but that leaded to a crash so i set the max size to 0..what leaded to a good tetra mesh but a failed prism mesh

Merging the prism layer after an OK Octree leads to unacceptable meshing.

Does anybody have any suggestions what I could do or where I am wrong?
I would be very thankful for an advice because creating a mesh takes about an hour or two...

And here is a pic of the bad elements from 0-0.3 after the prism mesh (and after improving quality) :(

You can try a delaunay mesh after you generated your prisms. Or even do a delaunay on your octree, improve the surface mesh then generate your prisms and eventually redo the delaunay.
As you see Octree has kind of discrete changes in the tetrasizes. Delaunay has a more smooth continuous tetra transition.

Thank you very much!! The delauney mesh improved my Quality of the tetra mesh. The Prism Mesh keeps being in a real bad quality. But now i was able to see the problematic areas im my geometry. My blades trailing edge have a radius of only 1.5mm which appears to be problematic for the prisms. I will try only one layer of prisms and cut it into pieces. Maybe this will work

Hello again,

I still have problems with my quality, but only on the trailing edge of my blade. The Aspect ratio of some tetras and the determinant of some pentas is below is between 0.1 and 0.3 and I just cant fix it.
I tried to control the area with a density (a single Point, afterwards I tried a line).
-> I got a hole lot of smaller elements which had the same bad quality.

I tried to modify the surface mesh particular for the thin cut by increasing the tetra width and lowering the size ratio.
-> a lot of small parts with the same - or even a worse quality

I lowered the Global Seed Size and the Min size.
-> Mesh was a little better but there were holes in my mesh
I tried to combinate all these methods.
But it all fired back or had no effect at all. Maybe someone has another idea or sees where I am going wrong.
Otherwise I am going to work with this mesh now because I am running out of time.

Thanks anyway for your time and suggestions

You are right. Trailing edge always have problems. I do not know specifically your problem.

But the best way I found to approach this problem is if you are using patch independent surface meshing:

* reduce curvature proximity limit from 1 to 0.1 perhaps. (Don't know if you tried that yet)

* reduce the edge criterion in volume meshing! default is 0.2 may be to 0.1 and then 0.05?

* have a delunay mesh

* create 3-5 prism layers and split them, re order them with initial height if you have a desired y+

* smooth tetras up to 0.4 with everything else frozen.

* smooth all elements upto 0.01 to 0.05 and not higher than that;

Good luck!

Thank you very much! I havent tried to split the prisms yet! Beautiful idea!

Just in case anyone else of you clueless students stumbles in this thread having problems with a thin cut blade and a real big geometrie. The solution to my problem was creating only some prism layers and splitting them afterwards in combination with setting the prism height limit factor to 1 (very powerfull option) even I stayed with the octree mesh. And of course setting the proximity below 1.

Good Luck Ya