[celestial]

Hi

Looking at the code in

coupledLduMatrix.C
coupledFvMatrix.C
coupledFvScalarMatrix.C

of foam-extend (4.0 or 4.1)

and all I see are block diagonal matrices which are therefore decoupled.

Maybe the files should be renamed as decoupledLdu(Fv)(FvScalar)Matrix.C

Any comments welcome

[celestial]

In general, I don't think people delve into the software at such a deep level. I tried to understand it since this last post but it is like breaking codes during war time or de-cyphering ancient writings in an unknown language. However, I'm now ready to accept there is a coupling between adjacent domains thru their interfaced patches. All I really want is a dictionary setting that will allow me to set the following boundary condition at the interface between domains 1 and 2.

c1 * [snGrad (X)]1 = c2 * [snGrad (X)]2 + sigma

where c1, c2 are constants and sigma is a given arbitrary field.

I am solving for field X and the above BC of course means that I'm talking about one-sided derivatives at the interface.

Simplifying the above BC to sigma = 0 would already help me a lot, probably.

That is the billion dollar question.

[celestial]

As I mentioned before, in another thread, forget about OpenFOAM when you encounter those boundary conditions. Just couple it to FEM software Elmer using MPI and interpolation interface EOF. Elmer can handle that kind of BC: just look at their solver StatElecSolver, for example.

[hjasak]

There are 2 different things here: coupled matrices you are looking at for multi-zone simulations, such as conjugate heat transfer. There are also block matrices for the coupled equations on the same zones, such as block p-u solution.

Hrv