reduce number of cells being flagged as offset

[kchinembiri]

Hi,

In my mesh I have an appreciable number of cells which are being flagged as offset. Visualising the check mesh result it appears the reason is due to the pyramid cells which act as a transition between the hexahedral near wall mesh to the fully tetrahedral internal domain.
I wanted to ask for some advice regarding how best to reduce the number of pyramid cells being flagged as offset, one way which appears to work is refining the mesh in the stream-wise direction but this has limited effect.

Another issue I have is I get a warning stating xxxx elements of type seg 2 have been ignored by the preprocessor.

Please find pictures of the mesh and listing file which might help to better explain the problem. I should also mention that changing the mesh to prismatic (first meshing the base-face and performing a mesh extrusion ) would not work in my case as later on I would like to deform the fuel pins thus they would be criss-crossing interior of the domain.

I would appreciate any advice you can give and Thank you in advance

[Luciano Garelli]

Hello,

In the theory guide of CS you can find de definition of the Cell Offset quality criteria and the geometrical parameters involved.
http://code-saturne.org/cms/sites/defau ... theory.pdf

Are you sure that your internal mesh is fully tetrahedral? Because it appears to be an 3D extrusion along the z direction.

Did you try refining the 2D triangular mesh?

If you want, upload the salome file in order to test the mesh

Regards,

Luciano

[kchinembiri]

Thank you for the reply, the answer to your first question is yes, it appears the internal mesh is fully tetrahedral. I performed the check by creating the group of all tetrahedral elements inside the mesh.

Refining the tetramesh inadvertently refines hexahedral mesh and thus does not seem to improve the problem. Another quirk worth mentioning is when I run the mesh on the local hpc cluster which uses CS 4.0.0 there are 0% cells flagged as offset but the other mesh quality criterion() are the same between the two version of CS. My workstation uses CS 4.0.6 and I thinks its probably the right one and the issue is with the approach I took.

I have attached my python script for generating the mesh, it should run straight away (apologies its a bit messy) and thank you again.

[Luciano Garelli]

Hello,

I run your script and did the check mesh in CS 4.0.2 and get 0% of bad cells with the offset criterion and 2% (21000 cells) of bad cells with Least-Squares Gradient criterion over 1.33 Mcells. But, if you change Netgen 3D by 3D Extrusion in the solid meshes you will get a mesh of 650 Kcell with 0% of bad cell.

I attach a picture with with both type of meshes, upper netgen 3D and lower 3D extrusion.

Regards,

Luciano

[kchinembiri]

many thanks for your help. One more quick question if you don't mind. Is there any particular reason why CS 4.0.6 is returning a different percentage for the offset criterion?

Regards,

Kenneth

[Luciano Garelli]

Hello,

I was checking the source code (cs_mesh_bad_cells.c) of the offset criterion in CS 4.0.0 and CS 4.0.6. This criteria has changed from
CS 4.0.0

Code: Select all

of_n = fabs(_DOT_PRODUCT_3D(v_of, v_n));

to
CS 4.0.6

Code: Select all

 of_n = _MODULE_3D(v_of) * _MODULE_3D(v_n);

Also, you can view the change in the formula in the CS theory guide

Pag. 395 http://code-saturne.org/cms/sites/defau ... theory.pdf
Pag. 395 http://code-saturne.org/cms/sites/defau ... theory.pdf

Regards,

Luciano