[nch]

Hi everyone,

I'm currently dealing with meshing a rather complicated geometry and I'm encountering problems with the prism layer generation: it is a classical tretra/mixed mesh realized by the octree algorithm.
With no prism layer, the elements shape is quite ordered and omogeneous, and the average elements quality is quite high, but as the prism layer is generated, a number of issues are introduced and the elements at the outlet surfaces cores are heavily distorted and the average elements quality is deteriorated, in some cases even accompanied by negative quality elements generation.

Does anybody have any advice on how to fix it by setting some specific parameters befor computing or by repairing the mesh after its generation?

Thanks
Niccolò

[Gert-Jan]

We need pictures to be able to give you advice.

[AtoHM]

W/o pictures, some general advice I could give is to use element sizes that are quite similar in the region of the prism and tetra parts. When the difference in these sizes becomes too big, heavy distortion and bad quality is created.

Also when doing smoothing operations, dont smooth all elements together right away. Try first smooting tetra, tri and penta. Then freeze those and smooth quad, pyra. Finally you can smooth all together. This procedure helped me get rid of 90% of bad quality issues.

[nch]

Hi everyone,

as required I attach an image as an example of what I meant in the previous post.

Thank you all
Niccolò

Cattura.PNG

[Gert-Jan]

When smoothing, turn of the option "Allow node merging".

[nch]

Hello everyone,

I'd love to thank those users who answered me for their precious advice.

I'm afraid I need to ask another question: prism layer generation also introduces these error messages:
Checking orientation:
4 elements at face 1003279 1294088 1505746, not all tets
PYRA_5 TETRA_4 TETRA_4 TETRA_4
cells near 50.198635 206.253274 72.821806 occupy the same volume
cells 8662 and 67354
face node numbers 1003279 1294088 3069465
opposite vertices 1505746 3069455
cells near 50.198635 206.253274 72.821806 occupy the same volume
cells 137597 and 8662
face node numbers 1003279 1505746 3069465
opposite vertices 1377426 1294088
cells near 50.002213 206.628322 73.938764 occupy the same volume
cells 8662 and 64315
face node numbers 1294088 1505746 3069455
opposite vertices 1003279 1003279
cells near 50.002213 206.628322 73.938764 occupy the same volume
cells 8662 and 67354
face node numbers 1294088 3069455 3069465
opposite vertices 1505746 1003279
4 elements at face 1003279 1294088 1505746, not all tets
PYRA_5 TETRA_4 TETRA_4 TETRA_4
cells near 50.198635 206.253274 72.821806 occupy the same volume
cells 8662 and 67354
face node numbers 1003279 1294088 3069465
opposite vertices 1505746 3069455
cells near 50.198635 206.253274 72.821806 occupy the same volume
cells 137597 and 8662
face node numbers 1003279 1505746 3069465
opposite vertices 1377426 1294088
cells near 50.002213 206.628322 73.938764 occupy the same volume
cells 8662 and 64315
face node numbers 1294088 1505746 3069455
opposite vertices 1003279 1003279
cells near 50.002213 206.628322 73.938764 occupy the same volume
cells 8662 and 67354
face node numbers 1294088 3069455 3069465
opposite vertices 1505746 1003279
some tetrahedra occupy the same volume and that couldn't be fixed
faces are missoriented

Does anybody know how to fix such problems?

Thanks
Niccolò

[Gert-Jan]

Difficult.
You should check your mesh on errors.
Go to: Edit mesh\Check mesh
(I think this is the third icon).
There, deselect all checks, except the surface orientation errors, and press OK.
This allows you to find all elements with errors. Then you should put them in a subset.

By only showing these elements you can see where the problems occur. Probably you need to change your geometry locally, or set local refinement and remesh.

Alternatively, you can delete all nodes in the subset. Then you have to recheck your mesh on uncovered faces, repair these leaving you with a hole in your mesh. Not sure if you can get away with this. I sometimes can.

[nch]

Originally Posted by Gert-Jan

Difficult.
You should check your mesh on errors.
Go to: Edit mesh\Check mesh
(I think this is the third icon).
There, deselect all checks, except the surface orientation errors, and press OK.
This allows you to find all elements with errors. Then you should put them in a subset.

By only showing these elements you can see where the problems occur. Probably you need to change your geometry locally, or set local refinement and remesh.

Alternatively, you can delete all nodes in the subset. Then you have to recheck your mesh on uncovered faces, repair these leaving you with a hole in your mesh. Not sure if you can get away with this. I sometimes can.

Hi,

I have already done so: there are 16 bad elements (on 12 millions) which present either surface orientation problems or penetrating elements problem. I can't change the geometry, but given the low number of bad elements (already put in a subset) I was thinking about deleting them, hoping it does not affect the mesh.

Thanks, Niccolò

[Gert-Jan]

After deletion of the elements, you need to check your mesh and repair the uncovered faces. So, it will affect your mesh. But if your CFD-solution will be affected depends on the location. If it aint in a critical region, then you might get away with it. It also depends on the solver. CFX is much more foregiven on bad elements than Fluent.

[nch]

Hi,

I've checked the mesh again after bad element deletion, but no any error is reported, not even about uncovered face.

The only error reported (which was already present even befor this deleting intervetion) is during my attempt to export the mesh into Fluent solver: it tells me that 4 faces are attached to more than two cells.

This seem to be the last issue to fix, but, as I'm not expert, I have no idea about how to deal with it.

Thank you,
Niccolò

[Gert-Jan]

I have never seen this error before. Does it give you coordinates so you can see where it occurs?
I would check the mesh on all possible errors and warning to see what you get.
Alternatively try CFX, to see if you can import it. As mentioned, CFX is more foregiven.