[lr103476]

Hi,

I used mergeMesh and stitchMesh to successfully combine two meshes in order to get some coarsening at the end of my computational domain. Between those two meshes, some polyhedrals are created.

Earlier in another topic it is mentioned there are two kind of decompositions: cellDecomposition (cell only) and faceDecomposition (cell and face). I assume one of those (depending on the compilation) is used by the laplaceTetDecomposition motion solver.

The problem I encountered was that I wasn't able to use icoDyMFoam on the moving polyhedrals. I solved this by recompiling the complete OpenFoam code with the faceDecomposition enabled instead of cellDecomposition.
Nevertheless some other utilities are still not working on the solutions of these moved polyhedral meshes, like liftDrag.

Now my question: what is the best way to solve the flow on moving meshes which contain some polyhedrals, such that everything still works.

Thanks....

Regards, Frank

[lr103476]

So concerning cellDecomposition and faceDecomposition.

When is which method used? and why? Does the solution depend on your choice?

Can anyone give an explanation to the previous post. Thanks..

Frank

[hjasak]

There is absolutely no reason whey the kind of polyhedral decomposition you are using should interfere with the calculation of lift and drag. In fact, neither the top-level utility nor the liftDrag library depend on any of the FEM solver or utility applications. Could you please clarify.

Regarding the method and consequeices of various decomposition types, have a look at the paper Zeljko and I published (finally!) on the subject.

mesh motion paper

In any case, it is all about what kind of decomposition will give you nicer-looking tetrahedra. In cell decomposition, you split each face into triangles starting from point 0 and then build the tets using the cell centroid. For face decomposition, you add a point into the middle of the face, split it into triangles that way and build the tets using the cell centroid again.

A quick rule of thumb is:
- if you have a "standard split hex", some of terrahedra may end up with a zero or negative volume, which is not good
- as a quick reject criterion, if you've got 2 faces sharing 3 consecutive points, you should be using face decomposition.

Hrv