[lentschi]

Hello everyone,

I'm encountering an issue with SnappyHexMesh. When I double the number of cells in each direction using blockMesh, SnappyHexMesh generates a mesh with cells at the boundary that do not follow the STL file but align with the Cartesian coordinate system instead.

I've already tried different numbers of refinement stages (feature refinement and surface refinement) without any improvement.

Any suggestions or hints from your side?

[Yann]

Hello,

It looks like castellatedMesh. If you're not using snappyHexMesh with the -overwrite option, make sure you are visualizing the last timestep in ParaView.


Yann

[lentschi]

Great hint! I'll check this out!


SnappyHexMesh created a time folder, which I converted using foamToVTK -latestTime. However, this conversion could have included a mixture of the constant/polyMesh and the time step/polyMesh.

[Yann]

When running snappy without the -overwrite option, it will write each meshing step in a new time directory so you will have:

• constant/polyMesh: initial (background) mesh created with blockMesh
• 1/polyMesh: castellated mesh
• 2/polyMesh: snapped mesh
• 3/polyMesh: layers

You don't need to use foamToVTK, but when loading your mesh in ParaView (using paraFoam or the paraview native reader with a .foam file), make sure to go to the last time step (see attached screenshot)

Yann

[lentschi]

Ich ran snappyHexMesh with the option -overwrite, but this caused the same inappropriate result. The cells are still not alligned according to the STL file.

[Yann]

Alright then it seems there is something else going on.
Could you share your files here so we can have a look?

[lentschi]

What data would you need? I need to ask my PM if it's allowed.

[Yann]

Ideally, a minimal example allowing to reproduce the problem.

If not possible, at least the exact commands you are running, with associated dict (blockMeshDict, snappyHexMeshDict, ...) and log files.

[lentschi]

log_snappy_fine_mesh.txt

log_snappy_coarse_mesh.txt

log_block.txt


Here are logs for blockMesh and snappyHexMesh. The only difference between both snappyHexMeshDicts is the maxGlobalCells to be set to 60000000 for the coarse mesh in order to avoid any refinement.


Commands:
blockMesh -dict system/blockMeshDict > log_block 2>&1 &
snappyHexMesh > log_snappy 2>&1 &
foamToVTK -latestTime
Then downloading VTKs to my local machine.


The coarse mesh is aligned to the STL at the 2nd timestep created with snappyHexMesh. The first time step looks aligned to the Catesian coordinate system such as the fine mesh does.
The fine mesh is aligned according to the Cartesian coordinate system at BOTH time steps! So there is no alignment to the STL between first and second time step.

[Yann]

Thanks for the files.
I'm not sure what is going on with your mesh.

What kind of geometry are you trying to mesh? It is surprising to start with a 126 millions cells blockMesh to end up with a 30 millions cells mesh.

By the way, why don't you mesh in parallel? 30h of meshing is insane!

[lentschi]

I am allowed to share the case. I will have acess to the data on Sunday.
Typically, I mesh in parallel, but on these days there were no free cores left. Unfortunately, I am not the only one using the cluster

[lentschi]

My case is now ready for sharing. If there is someone who is willing to support, please just PM me and I will share the case (constant + system folder) via OneDrive.


Thanks and best regards,
Markus

[lentschi]

No one out there who wanna help?

[lentschi]

OK, I solved it by myself. I just kept all refinement levels on 3 and skipped further refinement by reducing the maximum number of cells (maxGlobalCells).