[piotr.mecht]

I'm trying to prepare some mesh from simple geometry with snappyHexMesh. Without defining feature edge refinement every sharp edge turns round after meshing. With (explicitly defined) feature edges refinements too much of cells remain on edges during castellation, so snapped mesh looks 'framed'. What parameter should i control to resolve my edges correctly?

[HakikiCanakkaleli]

surfaceFeatureExtract?

Plus, in your snappyHexMeshDict, there is a conflicting setting:

//- Detect (geometric) features by sampling the surface
implicitFeatureSnap false;

//- Use castellatedMeshControls::features
explicitFeatureSnap false;

At least one of them should be turned on.

[piotr.mecht]

That commented out features{} geometry was obtained exactly with surfaceFeatureExtract (devided into 2 files with blender). When i use this part of code, it works like this: low refinement around that features leaves rounded edges, higher refinement leads to my undesired frame along the edges/

[HakikiCanakkaleli]

It seems that I edited my answer as you have been writing your answer. Please consider it. In addition increase the number of iterations of the entry below:

nFeatureSnapIter 10;

If you have already used 'surfaceFeatureExtract', you also need to turn on 'explicitFeatureSnap' and turn off 'implicitFeatureSnap' as this dominates the explicit one.

[HakikiCanakkaleli]

OK, I have been looking at your Dict entirely now.

file "wallEdges.obj";

should be

file "wallEdges.eMesh";

[piotr.mecht]

Thanks for your answer.
I used explicitFeatureSnap as true (just unchecked this on my trial and error attempts). implicitFeatureSnap just doesn't work for me, when i use this instead of explicit[..] my output mesh is the same as my bounding box from blockMesh.
Unfortunately increasing nFeatureSnapIter to 20 didn't help much. Changing *.obj files to *.eMesh format makes no difference.

[HakikiCanakkaleli]

Have you visually checked eMesh files? What were your surfaceFeatureExtract settings? And could you please share the last snappyHexMeshDict in which no settings were in conflict? Plus, could you please share 'surfaceCheck *.stl' output, if possible?

[piotr.mecht]

There they are

[HakikiCanakkaleli]

Why do you convert ".eMesh" files to ".obj", and try to use them in "features"? I think this is not a valid action. Plus, could you please further refine the resolution of your STL file?

[piotr.mecht]

I converted eMesh to obj, because obj was valid import format for Blender (which i used to separate edges). SHM returns:
Valid extensions are : 7 ( bdf eMesh featureEdgeMesh inp nas obj vtk )

i simply followed reasoning from this tutorial https://youtu.be/n9xYN59v3po?t=915 from Tobias Holzman (worked for him).

When i try to use features directly from eMesh (without distinguishing edge groups) effect is quite the same.

Since last time i tried: refining stl, increasing initial box division, reducing distance of bounding box from boundaries, refining features much more. Still with no sharp cuboid output.
(Attachments: refined stl and snapped mesh result)

[HakikiCanakkaleli]

Can you create the box with blockMesh and the half-ellipsoid volume with snappyHex?

[piotr.mecht]

Yes i know, but it's not my point. I'm not really interested in this particular channel. I'm trying to learn how to properly prepare desired mesh in sHM from cad model. I assumed that i should run correctly some simple evaluation cases before my complex heat exchanger geometry, but i got stuck at this point.

[HakikiCanakkaleli]

I see. I have never considered snappyHexMeshing without blockMesh. One more thought: Can you make the refinement level for the box edges zero? May be the issue is the castellation level therein.

[piotr.mecht]

Edges don't look like picture frame in that case, but they are rounded. That's why i wanted to refine edges in the first place. When i used feature edges distortions are visible on castellated level, so it's probably something on this level.

[HakikiCanakkaleli]

IMHO, with any software, the user needs to understand how that software was created to work, and not simply expect it to work the way they think it should work. I believe your frustration level would be greatly diminished if you change your meshing paradigm towards blockMesh+snappyHex. Otherwise, I yield. :] Thanks anyways for raising an interesting question!

[andrestg85]

I had a similar problem a few months ago. I finally discovered that you have to make a blockmesh bigger than the stl file that you want to model, and in that case almost all the edges will be very close to the original stl file.

I read that you are not using blockmesh, but maybe you can fix the problem just making a bigger blockmesh (or similar) before using SHM.

Also you can play activating/deactivating the elements of features line on SHM:

features
    (
//        {file "canal.eMesh"; level 0;}   
//        {file "fondo.eMesh"; level 0;}
    );

in generall, when i deactivate the features i get rounded edges.

[piotr.mecht]

I think i've finally solved my problem with snappyHexMesh. After increasing tolerance field in snapControls up to 1.0 my snapped edges got sharp (pics.).

I also reconstructed my geometry with blockMesh (just for fun). I wasn't satisfied with 8 or 9 hex blocks effect, so i tried to build my case from some more hexes, but the more hex field i distinguished, the harder it was to match cell grading in each block (and much easier to make some mistakes). I'll leave blockMesh for really the simpliest cases (simple cannels with no obstacles in the middle).

@andrestg85 I actually used blockMesh to create my bounding box and it had some offset from my geometry. I tried to pick different values, but my results weren't any better (when bounding box was really close there were almost no redundant cells on the edges, but I still wasn't satisfied with this round shape in the middle).

[MathiasSonderskov]

Hi @piotr.mecht,

How did your blockMeshDict and snappyHexMeshDict end up looking? I've got same issues.

Originally Posted by piotr.mecht

I think i've finally solved my problem with snappyHexMesh. After increasing tolerance field in snapControls up to 1.0 my snapped edges got sharp (pics.).

I also reconstructed my geometry with blockMesh (just for fun). I wasn't satisfied with 8 or 9 hex blocks effect, so i tried to build my case from some more hexes, but the more hex field i distinguished, the harder it was to match cell grading in each block (and much easier to make some mistakes). I'll leave blockMesh for really the simpliest cases (simple cannels with no obstacles in the middle).

@andrestg85 I actually used blockMesh to create my bounding box and it had some offset from my geometry. I tried to pick different values, but my results weren't any better (when bounding box was really close there were almost no redundant cells on the edges, but I still wasn't satisfied with this round shape in the middle).

[piotr.mecht]

Originally Posted by MathiasSonderskov

Hi @piotr.mecht,

How did your blockMeshDict and snappyHexMeshDict end up looking? I've got same issues.

Can't remember what the problem was back then, I don't have these kind of issues today Nevertheless I've found the files on my old laptop.

[buesinaw]

Would you mind providing the stl files and obj files called in your snappyHexMeshDict file?

kanau.stl
wallEdges.obj
obstacleEdges.obj

Thank you for considering this request.