[vaina74]

Hi all,
I'm trying to use snappyHexMesh - I'm a beginner - in order to simulate an internal flow with simpleFoam. I'd like to share my preliminary case (no care for the correct refinement or other) because I'm afraid that something is not correct. I'm not sure if something is wrong with dictionaries or tolerances, maybe the geometry model is not perfectly closed and the mesh 'leaks' out.
This is my procedure:

blockMesh
decomposePar
foamJob -parallel -screen snappyHexMesh -overwrite
reconstructParMesh - constant

If I check the resulting mesh with ParaView, I see that the internal mesh includes the outside background mesh and the space inside the geometry - and I can see the geometry outline. I already checked the locationInMesh vector.
This is the link to download my preliminary case and the 3D model (the STL file is too large fo the the forum limit). I hope that somone can show me the way to fix the problem.

[chegdan]

What does your snappyHexmeshDict look like? if you post your case i bet people will help debug it

[vaina74]

Hi Dan, in my thread I included a link to my Dropbox account to download my preliminary case (also with geometry model, so the archive is too large for the forum attachment limit).
Anyway I attach my dictionaries again. If you can take a look, thank you :-)

[chegdan]

Had a look at the case and it seems as though the base mesh size was longer in one direction than the other two. This makes the base mesh not cubic, thus making the meshing process behave badly. Attached is a picture of the domain snapping (not refining near the edges) and the snappyHexMeshDict i used. You will need to do some more to capture those edges and use surfaceFeatureExtract to get the featureLines of domain.

testmesh.jpg

snappyHexMeshDict.txt

I also made the base mesh size 0.2 and did not insert any layers. I hope this helps.

[vaina74]

Thanks for your remark, Daniel. I already knew that a 'cubic' background mesh is required, but I didn't realize that I didn't update the previous z-grading setting :-(
I made some tests yesterday and I encountered some problems, related to refining and snapping features. I think I'll also use Helyx and SwiftSnap in the future, but I want to learn snappyHexMesh first.

Refining

Outwardly, the refinement level in snappyHexMeshDict are not applied and I obtain only a 0 level (0.1 m). I can't understand why the refinementSurface settings for all STL surfaces don't work (level (2 3)).

Snapping

I made a first test with the implicitFeatureSnap option but the result was not so good. I enabled the explicitFeatureSnap, but I think that features are not detected. I also set a level 3 for features refinement in order to evidence the edges meshing - am I correct? - but nothing seems to happen.

This leads to a bad resolution of openings shape because the fillets are not resolved (see img1).
Moreover, the geometry shows a critical point where curved surfaces are tangent to plane surfaces (see img2). I can fix the problem with a more simplified 'chamfered' geometry, but I'd like to know a possible solution for a similar case. I already played a little with maxNonOrtho, maxBoundarySkewness, maxInternalSkewness and maxConcave parameters. Maybe could a different resolveFeatureAngle (I used 30) help?
This is the link to my updated case (and geometry).

UPDATE: maybe it's just a problem of maxGlobalCells? I'm going to check now.

[chegdan]

I think I'll also use Helyx and SwiftSnap in the future

I only employed this because it eliminated simple errors. you can always use these applications to setup cases easily and then tweak them afterwards...which is how i work. (disclaimer, i work for the company that makes HELYX-OS)

Your number of cells grows to a number larger than the a maxGlobalCell count so you may want to raise that a bit.

Because those edges are collapsing, I would put a refinement region around that feature or refine according to an emesh file.

The most common reason for snapping and underresolved regions is the base mesh is not small enough. Raise your maxGlobalCells (along with max local Cells too), lower your base mesh size, and try again.Furthermore you may want to try adding an emesh

[vaina74]

Hi everybody, just some feedbacks and questions - thank you, Daniel!
I made a lot of tests to improve the refinement and snapping quality of my preliminary case. I also must take account of the limits of my hardware configuration and seek a compromise about the global cells number.
As i posted before, I create a (very little) chamfer between the surfaces with collapsing edges. I was able to obtain a good resolution in the critical area and other zones, but 15 M cells were too much (and almost 4 hours for meshing). So I create a larger chamfer (that doesn't affect the simulation) and bring down the number of cells (see img1). A background mesh spacing of 0.2 m is the minimum to resolve the internal structures (see img2) - the block is 30 m x 25 m x 30 m.
About

or refine according to an emesh file.

what do you mean Daniel? Do you mean the features refinement level?

Key values of my snappyHexMeshDict are:
- maxLocalCells 1 000 000
- maxGlobalCells 5 000 000
- nCellsBetweenLevels 1
- refinementSurfaces level (1 2) for inlet and outlet openings and (2 3) for the domain walls
- resolve feature angle 30
- refinementRegions levels (1E15 2) - more searchableBoxes close to inlet and outlet openings
- explicitFeatureSnap true (includedAngle 150)
- features refinement level 3
- maxNonOrtho 80 / maxBoundarySkewness 20 / maxInternalSkewness 5 / maxConcave 80

The checkMesh tool complains only about

Max skewness = 5.061746, 1 highly skew faces detected which may impair the quality of the results

I'd like to understand all snappyHexMeshDict features, something is not clear to me.

1. I define all boundaries in the snappyHexMesh dictionary and I set

            patchInfo
            {
                type patch;
            }

for inlet and outlet openings and

            patchInfo
            {
                type wall;
            }
        }

but in ParaView the domain walls are defined as patch in the Mesh Parts list (and I see a proper wall - group anyway). Is it correct?

2. What does it mean planarAngle 30? I found this setting in the original dictionary I edited.

3. I set explicitFeatureSnap true and implicitFeatureSnap false because I used the explicit feature edge handling method. Is possible to activate both and improving the snapping? Does a method exclude the other one?

4. I set

    features
    (
        {
            file "domain.eMesh";
            level 3;
        }
    );

and domain walls refinement level is 3 too. Is a feature level 3 consistent? Maybe a level 0 is enough.

5. What does it mean

writeFlags
(
    layerFields     // write volScalarField for layer coverage
);

Thanks for your help, guys.

PS maybe this thread requires now a suited title, could a moderator change that?

[chegdan]

or refine according to an emesh file.

Refine your mesh in terms of distance from an eMesh.

features
(
    {
        file "myFile.eMesh";
        levels ( (0.0 3) (0.1 2) (0.7 1) );
    }
)

[vaina74]

Thanks again, Daniel, now I got it. Anyway I didn't know that it's possibile to set a different feature refinement level for each carthesian direction. I hope that you or someone else can answer to the other questions in my previous post, I can't find anything in the forum.
Have a good day.

[chegdan]

Originally Posted by vaina74

...Anyway I didn't know that it's possible to set a different feature refinement level for each carthesian direction.

No, that is not possible, though that would be an interesting feature . This is in the form of

features
(
    {
        file "myFile.eMesh";
        levels ( (<distance1> <desiredlevel>) (<distance2> <desiredlevel2>) (<distance3> <desiredlevel3>) );
    }
)

this needs to be in order of distance.

[vaina74]

Sorry, I'm an idiot. Too much enthusiasm