[CRI_CFD]

Hello to everyone,

I am new in snappyHexMesh and I have some basic questions I guess somebody could resolve. I have worked some time on them but I canot find an answer.

I am using snappyHexMesh and I have mainly three questions:

1. The snap phase is not being well. The geometry at the inlet and outlet has changed after snapping. I attach an image of the result. I have tried surfaceFeatureExtract with an angle of 180 but this didn solve the problem (an with this utility the cell count grows in an unknown way). I need to reproduce the original geometry the best I can. What parameters are the key of a good snapping process? The resulting meshes give good results in checkMesh test but sadly this is not the right geometry (CFD simulations behave strange in this zones...

2. What is better for a good snap: a small blockMesh (I mean low density of cells in blockMesh phase that produces high cell count after running snappy) and more refinment levels or a bigger blockMesh (high cell count in blockMes phase and small cell count in snappy) with less refinment? Is it better to fit the bounding box to the geometry or can I make a bigger blockMesh so the cells that are outside the geometry are going to be deleted? I am doing internal aerodynamics.

3. When I run addLayers phase, I achieve a good clustering of prism cells close to the wall, but in some places this layers are collapsed (attached files for more details). Which parameters have I to modify for obtain a complete layer mesh? This problem implies that y+ values are higher in turblent simulations and I don't want to use wall functions. This also implies high aspect ratios (around 20) that makes that laminar simulatons does not converge in these meshes, but I expect that with turbulent simulations all will be OK when I have a complete layer mesh.

I cannot attach the geometry, sorry for that.

Oh, the blockMesh size for the attached case files and images produces 960000 cells and I use snappyHexMesh in parallel (I guess this not will be the problem but who knows, when I reconstruct it it gives a warning, like this tool is not enough tested...)

Thanks in advance for all the hints. It's great to became part of this big community. I apologise for my poor written English, hope that somebody can understand what I mean.

Best,

CRI_CFD.

[Artur]

Hi CRI and welcome to OpenFOAM

There is no unique answer to such generic snappyHexMesh problems as yours I'm afraid. Judging from the part of the geometry that may be seen in your pictures I deduce it's fairly complex. In case of ship hulls which I deal with the most I have found that it's vital to ensure a well snapped mesh for the extruded layer cells to behave nicely.

For a beginner I'd recommend reading these slides:

http://openfoamwiki.net/images/f/f0/...SlidesOFW7.pdf

Generally, what you're trying to achieve (the way I look at it) is balance the accuracy of of snapping (how well your surface is represented) and the quality of hexahedrals that your mesh will consist of. This is controlled by the mesh quality parameters at the end of your snappyHexMeshDict. Try lifting some of the tolerances (increase non-orthogonality of cells, etc.) if you feel you want your mesh to deform more. But bear in mind that if you push it too far your solution may not converge.

Another thing is the snapping criteria, of course. For complex geometries it's worth increasing the number of solver iterations (nSolveIter) to something sensible, say 200 or 300. I'd also try pushing the nFeatureSnapIter up, even to 20 or 30 if required. The meshing will take much much longer but generally I've found the results to yield much better layer cells in the cases I've been through.

Hopefully this will give you a few hints to get started.

Peace,

A

[CRI_CFD]

Thanks Artur for your fast and detailed reply.

I'll try with the settings you suggest. It's not easy to manage with snappy. The dictionary has so many parameters, so it's needed to change it one by one to understand the behaviour of each one.

Also it's clear that a balance between surface smoothness and cells quality it's needed in order to obtain a suitable mesh for simulations. And the tren of snappy to construct hexaedral meshes make it more difficult in complex geometries, where tetrahedral elemenst represents better the surface.

I'll keep trying, specially with the layers' issue. I can introduce it, but sometimes they collapse in places where I didn't expect. I guess a long trial and error phase wil give me the right value for angle and layer iterations. Maybe an option could be to create a single layer so far away from the wall and use walll functions better than place so many cell layers close to the wall to resolve boundary layer with y+<1...

Thanks again to you and to the forum for the opportunity for sharing knowledge!

[Artur]

Originally Posted by CRI_CFD

Maybe an option could be to create a single layer so far away from the wall and use walll functions better than place so many cell layers close to the wall to resolve boundary layer with y+<1...

To be fair I've had many difficulties with getting resolved meshes using snappy for complex geometries. Things like airfoils work OK but as you start going to complicated 3D shapes things can get a bit out of hand with cells collapsing, just as you're experiencing.

Perhaps it's worth to do it in a few steps: run snappy without layer addition first, get a decent mesh; when you're happy with it save it and try adding some layers in; you can do these in a few steps too to get some coarser ones into the outer BL regions and really fine ones close to the wall. This way you may find it easier to figure out which parameters affect your particular geometry the most. And you won't have to re-mesh everything when willing to just change one layer parameter.

Sounds easy but it can take a lot of time

Good luck,

A

[snowygrouch]

Hi Artur,
Are you adding in the layers later using "refineWallLayer" run from
the terminal - or are you doing the layers later using SHMesh ?

Regards

[Artur]

Hi,

I've never used the refineWallLayer function to be honest as usually I use snappy with wall functions so can get decent results with the layer addition phase on its own. This utility does seem very useful though so definitely worth looking at if you think it may help you to improve your mesh.

Sorry I couldn't give you any more feedback, I just have no experience in this matter.

Peace,

A