[gimli79]

Hi Foamers,

I am working on a FSI topic and have now some kind of problems.
In this case I have to update the mesh after every iteration, but I do not know how I should manage this. Is there any possibility to do that (maybe some fields), without changing the solvers, cause I am not allowed to do that.

Thanks a lot

Andy

[deepsterblue]

Update your mesh with either a moving-mesh and/or re-meshing approach.

[lars.]

Hi Sandeep, I work on an FSI problem, where the mesh deformation approach always fails after very few iterations, because several cells collapse or become inverted.

I tried to remesh with netgen whenever the mesh quality drops, which happens every 2-5 iterations. This is horribly slow. Do you know any tools that perform fast selective remeshing only in the severely deformed regions? The goal must be that the remeshing is not orders of magnitude slower than the equation solvers.

Lasse

[deepsterblue]

Every 2-5 iterations? Seems to be quite unusual to get mesh problems that quickly. Are you using a mesh-motion solver?

[lars.]

Yes, I used the laplaceFaceDecomposition solver with quadratic diffusivity. Of course I also tried a number of different settings. The question is not whether my case is usual or unusual. I am not talking about a small deformation of elastic parts. I am talking about obstacles moving freely in a fluid domain. We have tried the MRF and GGI methods, but they are both not optimal. Please refrain from discussing these points because this is a different topic.

The question is what tools are available for an efficient remeshing approach as you suggested it. Do you have experience with this approach?

Lasse

[deepsterblue]

Yup - quite a bit, in fact. I suspect that you want something similar to this:

http://www.ecs.umass.edu/~smenon/Mov...stonMotion.avi

I have the re-meshing algorithms nailed down pretty well, but I'm currently working on accurate field remapping at the moment. So until I have that sorted out, I'm afraid this won't be available for general use. Do keep a lookout for a release on OF-1.5-dev or 1.6-dev.

[lars.]

Nice.

As you are working in this field you have probably come across some relevant publications about dynamic remeshing. I am sure quite a few people here would be grateful if you share some links about the algorithms or even better about implementations that could be used until you are ready to contribute your work to OF1.x-dev.

Lasse

[deepsterblue]

The only other mesh-topology modification algorithms in place (that I'm aware of) are Hrv's layering / hex-refinement stuff. While this has been sufficient for a lot of cases, it isn't well suited for general mesh topologies.

My approach has been to use Shewchuk's edge-removal algorithm (can't remember the reference), with refinement features on tri-prism / tet-meshes. I can tell you that it's not easy to implement, given the complexity of the project, but it's been done.

You could temporarily hook-up your solvers to netgen, have it re-mesh, perform a mesh-to-mesh interpolation of fields, and re-start your solution. This is a global re-meshing approach that nobody is fond of, for obvious reasons. But unfortunately, that's the only solution I have for you at the moment.

[gimli79]

Thanks for help,
I am a beginner with openfoam, so I am happy about any kind of help.
But I have to move the single points which corresponds to the deformed structure of the transient CSD computation. How can I do that, without changing the code? The reason is that I am not allowed to compile.
I tried several dictionaries, but with them I can just move complete boundaries.
So, which dictionaries or maybe solvers should I use for that case.

The next problem is, that I have to update the mesh every timestep but I have absolutly no idea how I should achieve that. So, should I define a file for movement of all timesteps at the start in 0 directory.Or can I update the mesh every timestep. But here I dont know how to stop and restart the computation.

Thanks for your suggestions

Best regards
Andy

[vonboett]

Dear deepsterblue,

I simulate VOF-based freeface mudflows with herrschel-bulkley rheology and LES-turbulence modeling in OpenFOAM 1.6.x, and now I have to connect that to my landslide barrier code that catches the mud with steel nets. Since I have some implementations regarding the turbulence subgridscale modeling from the LTTE-Rostock, I really want to avoid changing to OpenFOAM 1.5.dev, but it seems not trivial to get laplaceFaceDecomposition solver implemented in OpenFOAM 1.6. Do you know who I could ask, whether implementing laplaceFaceDecomposition to OpenFOAM 1.6.x is a reallistic opportunity?

Thanks,

Albrecht

[deepsterblue]

The tetDecomposition solvers are probably non-trivial to port to OF-1.6, but I've never really tried, so I can't say for certain. Are you dealing with a tet-mesh?

[vonboett]

Ok, I see. I use hexahedral blocks created with blockMesh.