[luca]

Hi all,
my question concerns the implementation of non conformal cyclic boundaries (called later on as arbitrary cyclic).

The idea is to calculate the geometric parameters (such as reconstructed faceCellCenters, weights and deltas) and boundary values for each field, interpolating (patchToPatchInterpolation::faceInterpolate() ) values from the twin patch (the other coupled patch).

I tried to implement such boundaries merging together Hvroje's ggiInterface and cyclic patch but what I obtain is this:

if I apply arbitrary cyclic condition on conformal cyclic patch results are almost identical to standard cyclic,

if arbitrary cyclic patches are applied on non conformal boundaries after some iteration the linear system solver diverge (usually solving the energy equation, if this may help) even if the convergence history is good and no wrong field are observed before the crash.

I thought this could be caused by the small number of nodes taken by the patchToPatchInterpolation to calculate the interpolation weights (the hit face and its neighbors) but also enlarging the set of nodes to the face hit by the vertices and their neighbor faces, the result is the same: divergence is only postponed.
If someone want to have a look, I posted the source codes and related test case meshes (conformal and non-conformal)
I used on this link:

http://htc.de.unifi.it/docpub/conf.tgz

http://htc.de.unifi.it/docpub/non-conf.tgz

http://htc.de.unifi.it/docpub/ggiCyclicTools.tgz

I have seen that there is still work going on about this subject, (Maryse Page and Håkan Nilsson: can it be a good starting point?!), since I cannot go further, I would like to hear your thoughts and comments on that. I would appreciate so much your help.

Thanks in any case.

Luca e Cosimo