Hi,

I have three domains one with moving grid and other two stationary grid. I can't model it in one domain so i have to exchange the boundary conditions at interface of these domains at each iteration and time step.

I have tried different strategies but didn't worked. If anybody has worked on it or has any related information please tell.

Devendra

could you please describe in more details.. like what is the physical system you are modeling ?? DC

Dear Devendra

Problems like your is the goal of MPI protocols (http://WWW.ERC.MsState.Edu/misc/mpi/)

Take a while to search on the web and try it. I can sure you that, with one subroutine and a few MPI lines, you can solve your interprocesses communications. You don't even need more than one machine and you have MPI even for Windows NT. You can program in Fortran 77, 90 or in C.

If you know how to relate the boundaries between domains, I estimate that the learning period to do this in a case like your is about 1 month depending on your knowledge of CFD programming and the skills you have in your operating system. Good luck

Regards

Kike

I am modelling Scavenging Process in Two Stroke IC engine where cylinder is modelled as moving boundary to accomodate piston motion while transfer ducts are stationary. Here i cant modell all three in single domain as one domain has moving boundary. So i am solving all three domains separately ans sharing the information where ducts meet cylinder. This is crucial as this act as boundary condition for that domains. My problem is to sypply these boundary conditions at each iteration and each time step. Where all conservation laws has to be satisfied. Since in one domain grid is moving grids also dont match in two domains at their common face.

In the multi-block technique, the overlapping grid is used to communicate the information at the interfaces. I am not sure if it is helpful for you.

The KIVA family of codes developed by the T3 group at Los Alamos has logic for moving pistons and valves as well as fuel injector spray models. You can check at http://gnarly.lanl.gov.

You should also check with the University of Wisconsin (search this web site for the exact address). Graduate students there have done a lot of development on KIVA.

The various publications describing KIVA coming from LANL and from UWisc will likely answer your questions.

Check one of my previous papers, it may tell you something both theoretically and computationally.

On nonconservative algorithms for grid interfaces. SIAM J. Numerical Analysis, v.37, 173-193(1999)