Get rid of snap artifacts?
 

Hi Foamers,


I am using snappyHexMesh on a very simple, repetitive geometry - essentially just an array of boxes. The snapping process causes these tetrahedral areas that sort of propagate out from the snapped geometry, as seen in the before and after photos. I have been adjusting a lot of settings in an effort to get rid of them, but it generally just moves them around and I can't really justify the changes I made. Does anyone know the mechanism behind this, and how to fix it without massive refinement? I was considering making a really complicated blockMesh to get around things, but would rather just get better at general mesh design. The mesh is for DES, so I would rather opt for layers than more refinement levels, and I think the small tetrahedra outside the boundary layers are really jacking up my Co.


Thanks

Did you find the solution for it?

Sorry for the delay - I never found a solution. CFmesh doesn't seem to have the issue, but I don't know how to use it...

Before moving on to learning CFmesh... Has anyone found a solution for these artifacts yet? :confused:

So I have come to learn that these are actually not a problem. What is happening is that when you snap or add layers, it morphs the mesh around the object. So when you slice it in paraview, it will cut through different cell boundaries, because the corners of the cells are no longer at 90 degree angles. Paraview automatically triangulates things, so if you look at those cells in the spreadsheet view, they might be tetrahedral in paraview, but really are hexes in the mesh, which is confusing.



So the take home is that you don't need to worry about those artifacts. You may choose to invest your time in CFMesh for a few other reasons though...

Abe is right about the artifacts being just a representation matter in paraView.

I will just add a tip: when slicing your mesh in paraView, you can tick the "crinkle slice" option to display the whole cells in the slicing plane and get rid of the artifacts.

Cheers,
Yann

Abe, thanks a lot for the very clear explanation, it is always great to know exactly what is happening!


Yann, thanks for the additional tip, the "Crinkle slice" option really helps to visualize the issue.



Abe, about CFMesh, could you please quickly comment on its benefits over snappyHexMesh? It'd be great to get an experience-based perspective about when a given tool may be more suitable than the other.

Honestly, I just started using CFmesh, but it generally requires less imput from the user to make a usable mesh. It can be very difficult to get a decent mesh out of snappy, especially when it comes to layers. I am rarely successful in keeping layers from collapsing for complex geometries. Run a few of the tutorials and see what you think. You might also have a look at Gmsh. I have not learned it, but it seems like it might be the strongest option for a free and open-source mesher. I just haven't learned it yet because I have no time...


Maybe someone else with more experience can comment on the state of meshing softaware for those of us who cannot afford software...

Hi KTG!
What about SALOME? today I was experimenting with layer generation in SALOME for a simple L-duct geometry, I've managed to create up to 50 layer , of course checkMesh complained about the aspect ratio but I was just pushing SALOME's ability to generate a decent tetrahedral mesh with lots of layers, on the other hand I was struggling to not collapse only 3 layers in a sharp corner with snappyHexMesh (while also being very slow).

Check the attached picture, a 10 layer generated using SALOME, and passed checkMesh.

Planning to check out cfMesh though!