[tadeuzagallo]

Hello,

I'm meshing a half car model for external aerodynamics simulation, using a relatively fine mesh (~30M cells) and the surface mesh looks decent (apart from some cells snapping to the wrong feature line), but I'm struggling with the shape of the cells near the surface and layers collapsing.

The attached snappyHexMeshDict has the latest configuration I used, where I significantly increased the following parameters, but it made no visible difference:

nSmoothPatch: 3 -> 10
nSolveIter: 30 -> 100
nRelaxIter: 5 -> 20
nFeatureSnapIter: 10 -> 50

I attached the logs for sMH and checkMesh, my snappyHexMeshDict, blockMeshDict and meshQualityDict zipped. I also included screenshots of the problem areas after each step (castellation/snapping/addLayers) as well as the final surface mesh.

I'd really appreciate any suggestions on what I could try next, as I've spent a few months trying to improve this and going through every thread I could find here.

Thanks,
Tadeu

[tadeuzagallo]

Sorry, I could only attach five files, here are the remaining screenshots.

[tadeuzagallo]

Sorry, I'm a newbie and completely forgot to turn on "Crinkle slice" in ParaView.

In the meantime I tried both setting minTetQuality and minVol to 1e-30 in meshQualityControls and the trick of scaling the STLs up, meshing, then scaling down. No visible difference.

Here's a couple screenshots with crinkle slice enabled, the updated quality settings and scaling the mesh.

[AtoHM]

There is no general fix for the layer generation with snappyHexMesh, its notoriously unstable.
Just a few hints that might help:
1) keep the refinement level at the surface constant where you want to grow layers. This helps to insert the layers. Currently you have smaller cells at leading and trailing edges, bigger mid-chord.

2) In the ESI version, you can use the laplacian mesh shrinker (see: https://www.openfoam.com/documentati...sh-layers.html at "DisplacementMotionSolver")
3) You can try using absolute sizing instead of relative.
...