Interesting problem: Parallel Processor VOF Fluent + Dynamic Mesh + System Coupling
 

I have an interesting problem with Fluent. Bear with me here...

I'm running a 2-way FSI (via Ansys System Coupling) tank slosh model. I'm using transient Fluent for the fluids and Transient Structural for the tank. I'm currently trying something very simple: a 1D translation by commanding a velocity profile on one of the walls of the tank in transient structural. The tank can slide in X, but can't move in Y or Z.

Settings:

Geometry:
1m cube tank

Fluent:
1g in -Y
hexahedral mesh: 64000 cells
VOF explicit scheme, implicit body forces
k-e turbulence model
water + air, surface tension, wall adhesion
Cutcell Dynamic Mesh Remeshing
Convergence criteria: 1e-5 for everything
PISO scheme + default under relaxation factors
Pressure: Presto
Momentum: 2nd order upwind
Volume fraction: geo-reconstruct
turbulence: 2nd order upwind
Transient: 1st order implicit (2nd order not available for explicit VOF)
Hybrid initialization
Patching in an initial fluid volume to the bottom 0.3m of tank.
50 iterations max

Transient Structural:
1cm thick aluminum walls of tank
Nicely made mesh
Restrict motion to X
1g gravity
simple ramp velocity profile in X

System coupling:
dt = 0.005s
correct data transfers

Hardware:
Quadcore I7 2.6ghz hyperthreaded
16 gb ram

I did lots of research, and I'm fairly certain I'm using all of the right models, solver settings, etc for this kind of simulation. Just running the fluent model with a gravity term in X gives great, grid independent results at these settings.

Here's the kicker: Everything works fine when Fluent is in serial mode! It'll run the full simulation, including dynamic meshing;it's just slow as hell. But when I try to run Fluent on multiple cores, the system will start fine, run for a few system iterations until my criteria for cutcell remeshing the Fluent cell zone are reached, Fluent will remesh, Fluent will do a set of iterations, attempt to transfer information, then the system coupling .exe crashes and the simulation aborts. It spits out an error saying there was something wrong with fluent, but it's not specific.


Anyone have any idea what's up? Since it runs fine in serial mode, I don't think it's a problem with the way I'm modeling stuff. I'm guessing it has something to do with the way cutcell operates.

Has anyone successfully used parallel fluent with cutcell with system coupling? I've seen parallel fluent successfully used with system coupling on youtube, but that was with a tetra mesh and local cell remeshing.

System coupling is crazy unstable. I've had to completely rebuild my system multiple times because it will just stop working for no apparent reason and throw errors (resets, restarts, and deleting components doesn't work).

I tried using local cell remeshing (with a tetra mesh) but then my velocity command in transient structural stopped working as evidenced by my tank not moving as much as it should have in the given time (which prevented a remesh from taking place, which was the whole point of testing this). It also slowed everything wayyy down = :mad: , which leads back to the whole system coupling being unstable/buggy/extremely not user friendly.

My plan now is to ditch system coupling. I'll have to bring it back in at a later date for some flexible stuff going in the tank, but for now I'm just going to try moving the tank back and forth in fluent. Question: Can I use Fluent to prescribe system motion (profile, UDF, etc) instead of transient structural? I assume that I can since instead of a displacement being first transmitted to fluent, fluent will transmit a fluid force first.

Thank you

It's the cut-cell elements that are likely the problem. System coupling doesn't support cut-cells and Fluent won't let you set up a case with a cut-cell zone and a system coupling interface (in version 15.0). It looks like you case is not disallowed because it starts out as a hex mesh. Restarts caused lots of problems in 14.5. Version 15 is better.

... if you're doing remeshing make sure the mesh on the System Coupling interface is not getting remeshed - that's not allowed.

Hey mate,

Just wondering if you ever resolved your issue. I'm having the same problem except parallel is causing my solution to crash. It runs perfectly fine in serial as well.

Im using 2.5D remeshing, though it has the same symptoms as your problem.

Any help would be appreciated!

Cheers

I think stumpy is right. I never got system coupling working in parallel with cutcell, and apparently it's not allowed in V15. I ended up giving up on system coupling and re-working my assumptions so that my system could be approximated as structurally rigid.