[Thaw Tar]

Dear Foamers,

I wanted to do FSI simulations of VIV of flexible structures (like elastic finite length cylinders).

I decided to use ElmerFoamFSI solver and successfully compiled it. It uses Foam-Extend 3.2 as fluid solver and Elmer FEM code as structure solver.

I tried to run the HronTurekFSI3 tutorial provided by that solver with default settings. As a default, the velocity boundary condition of the plate is fixedValue instead of movingWallVelocity. If I use fixedValue BC, it runs happily and I can see the plate moves under the flow (I think it is quite a good deformation) although I do believe that I have to use movingWallVelocity BC.

However, if I run the tutorial with moving wall velocity, the simulation blows up and crashes after a few time steps. I think the mesh quality is OK and the time step is small enough. I can run that very same case using foam-extend-3.2's default icoFsiElasticNonLinULSolidFoam solver using both fixedValue and movingWallVelocity BC's for the plate.

Does anybody have any experience with ElmerFoamFSI code and movingWallVelocity BC with it? Could anybody please help me figure out how to fix this?

Thanks in advance.
Tar

[Thaw Tar]

I checked the ElmerFoamFSI code and found out that it is a weakly coupled FSI code rather than a strong coupled one. It means I cannot use the coupling schemes like Aitken and relaxation factors for the displacements.

I also found out that they wrote an iterative strongly coupled solver (for OpenFoam side) in the same solver code file. But that solver is not being used. I tried to modify the FSI solver code calling that strong coupling function. However, it does not work yet. The simulation is still diverging.

The person from ElmerFoamFSI development group suggested me to use upwind schemes for the div() terms in order to add some diffusion and relax the flow variables. I will try it soon.

If it does not work, as the next step, I will try to relax the UEqn() part like in pimple algorithm. Since the fluid solver for ElmerFoamFSI is purely icoFoam based, there is no relaxation occurring like in piso or pimpleFoam. I will include UEqn().relax() and try to increase the solver iterations like pimple algorithm.

I will report soon.

[ssyadav]

Hello Tar

Have you tried the default test cases with ElmerFoamFSI? I compiled the solver and I am able to run a HironTurekFSI case, but I am getting lot of default test cases failed after the compilation. Any clue?

Thanks

Shyam

[Thaw Tar]

Dear Shyam,

As I mentioned above, I can run default HronTurek FSI3 test with default fixedValue velocity BCs but not with the more realistic movingWall BC for the flat plate.

And I believe that one of the main reason is because ElmerFoamFSI is weakly coupled FSI. Or may be fixedValue BC is more suitable, I am not sure.


But for your question, what do you mean by default test cases? I also failed a lot of tests during the compilation stage. But as far as I remember, I could run most cases in the tests directory.

Anyway, I already gave up with ElmerFoamFSI because it is a weakly coupled solver and moved on to another strongly coupled one. It is also foam-extend based but fully finite volume approach. I desperately need a strongly coupled solver and thus moved to FOAMFSI (FYI).

Regards,
Tar