Simulation stamp does not match

[Aditya Bantwal]

Hi,

I am restarting a svFSI simulation that had stopped due to cluster time out issues.

After restarting it, I get this error:

stamp.png (7.64 KiB) Viewed 475 times

https://drive.google.com/file/d/12yb6E8 ... sp=sharing

Can you please help me tackle this error? I have attached my files just in case they are needed.

Thanks,
Aditya

[David Parker]

Hi Aditya,

It seems that the simulation stamp for processor 16 read from the stFile_last.bin file does not match the simulation stamp created at simulation initialization for this process. The svFSI error message is a bit useless in that it does not state which component of the stamp does not match (e.g. number of nodes).

You are rerunning the simulation with the same number of processors?

You could change line 617 in svFSI/Code/Source/svFSI/INITIALIZE.f to print the contents of tStamp and stamp to see what does not match.

Cheers,
Dave

[Sara Zambon]

Hello! May I ask a semi-related question?

Is it normal for the solvers from svFSI to struggle with convergence (see the lsIt column) during the first timesteps immediately after a restart? Also, is it common for the pressures and flows in the "B_FS_....txt" files to exhibit sudden, strange peaks and values? I don't know if this is related to the type of solver or the type of simulation (this one is from a "moving wall" setting).

Thank you so much in advance!
Sara

Right before stopping the simulation with "STOP_SIM" file:

Before restart

Im1.png (17.37 KiB) Viewed 255 times

After Restart

Im2.png (47.08 KiB) Viewed 255 times

Im3.png (25.54 KiB) Viewed 255 times

[David Parker]

Hi Sara,

The FSI solvers can indeed have slow linear solve convergence. This could be caused by an ill-conditioned problem (BCs, mesh, etc.) or using the wrong linear solver. FSI problems should use the NS solver. You could also try to use the Trilinos or PETSc linear algebra libraries which typically have better preconditions and better convergence behavior.

Note that it my take a thousand time steps for a simulation to attain a steady-state solution.

Cheers,
Dave

[Sara Zambon]

Hi David,

the focus of my question was about the restart, specifically. Why did it cause such peaks in the pressure and flow profiles (see red box on the plot)? The peaks wouldn't have been created if I hadn't stopped the simulation with "STOP_SIM" (I did a test on that).

And secondly, I am using the <FSI> equation (LS type: GMRES - Preconditioner: FSILS), and in the MASTER.inp file it is specified that the linear solver NS can be only used for <fluid> equation. Is this not correct?

Thanks a lot,

Sara

[David Parker]

Hi Sara,

I'm don't know what might be causing the behavior you are seeing after restarting a simulation. Please open an Issue for this here https://github.com/SimVascular/svMultiPhysics/issues and we will investigate.

Cheers,
Dave