Hello,
I've checked the forum but any topic match my problem.
I have a mesh where I split cells (hexa) in 4 with Salome tool ('Split volumes') to refine some parts of my geometry.
In Salome I have only an augmentation of 3D cells, which is normal.
But when I us this new mesh in Code_Saturne, the pre-processing step indicates me some boundary cells with no family. When I export the results in paraview, I can see these few cells appeared in the middle my fluid domain. So I can run my simulation only by forcing the boundary condition to 'wall' on it.
These cells doesn't exist when I check my mesh in Salome.
Does anyone already face this kind of problem ?
Best regards,
Camille.
'Ghost' boundary cells
Forum rules
Please read the forum usage recommendations before posting.
Please read the forum usage recommendations before posting.
-
Yvan Fournier
- Posts: 4077
- Joined: Mon Feb 20, 2012 3:25 pm
Re: 'Ghost' boundary cells
Hello,
Was the refinement done with Homard ? If you chose the "nonconforming" option, you need to use mesh joining to make things conforming again.
If the geometry is not too complex, just use "all[]" for the mesh joining criteria. Otherwise, is may be better to exclude real boundary faces (using "not ()" with a list of groups), but you may still try first with "all[]" (check the joing results using visualization to make sure it does not join parts that should be left alone).
You can also use mesh refinement directly in code_saturne, if your refinement is isotropic (so hexahedra are split in 8, not 4). This has the advantage of simplicity and working with all element types. Check the user function examples for mesh modification in the Doxygen documentation for this.
Best regards,
Yvan
Was the refinement done with Homard ? If you chose the "nonconforming" option, you need to use mesh joining to make things conforming again.
If the geometry is not too complex, just use "all[]" for the mesh joining criteria. Otherwise, is may be better to exclude real boundary faces (using "not ()" with a list of groups), but you may still try first with "all[]" (check the joing results using visualization to make sure it does not join parts that should be left alone).
You can also use mesh refinement directly in code_saturne, if your refinement is isotropic (so hexahedra are split in 8, not 4). This has the advantage of simplicity and working with all element types. Check the user function examples for mesh modification in the Doxygen documentation for this.
Best regards,
Yvan