svFSIplus (C++ solver) produces different stress result from svFSI (Fortran solver)

[Viola Wu]

Hi SV developers,

Greetings! We have been trying to run the FSI simulation using svFSIplus and svFSI. In our simulation files, we found that while both solvers are able to give us stabilized results, the Von Mises stress and the Cauchy stress are 10 times lower for the results returned by svFSIplus. The difference in resultant stresses persists if we change the constitutive model from HGO to NHK. When you have time, could you look into this issue? I really appreciate your time and effort.

Link to our simulation file goes here: https://drive.google.com/drive/folders/ ... sp=sharing

Best, Viola

[Aaron Brown]

Hi Viola,

Thanks for sharing this issue. I am not super familiar with FSI simulations, but I will try to help where I can! Can I ask a few questions?

1) From the C++ and Fortran results you shared, it looks like not only von Mises and Cauchy stress are different between the two, but also other fields (e.g. displacement, velocity) are different. Can you confirm?

2) Can you share the convergence histor.dat file?

3) Could you tell us exactly which versions of svFSI and svFSIplus you are using? Links to the specific commits, like https://github.com/SimVascular/svFSIplu ... e1bac14b74, would be great.

4) I noticed that in the svFSIplus input file, you don't have boundary conditions for the mesh equation. Could you try adding boundary conditions there, like you have in the svFSI input file?

Thanks,
Aaron