Learning Forum

[arnau994]

Hello,

I ran an LS-Dyna FSI model with strong coupling in version R11.2 some months ago. I didn't set up any timestep for the ICFD solver and the Structural solver, so the CFL condition was respected in both solvers and "adecuate" timesteps were used (i.e. dt 2e-3s).

However, when running exactly the same model with R13.1, an automatic timestep of 1e-01 is used. I have checked the key cards and they are exactly the same. I don't understand why, as in the R13 manual (as well as in the R11 manual), it says that for *ICFD_CONTROL_TIME, when DT is 0, CFL condition is respected for timesteps. In this case, timesteps should be even lower due to smaller cell size.

This higher timestep in the R13 model generates an imprecision that leads to deviated results compared with the R11 model. The only difference is that the R11.2 model was run with a launcher provided by Oasys, while the R13.1 model is run with the ANSYS Mechanical APDL Product Lanucher 2023 R2. Both with double precision enabled. Below I'm providing information from the output of the solvers.

Do you know what might be the reason? I found this issue with the "FSI tutorial flap" in the LS-Dyna examples webpage, and this same timestep issue has extrapolated to my current FSI model, which is not converging (the ICFD-only does). My concern is that this divergence might be caused by the high timestep imposed by the implicit ICFD solver.

Thank you very much.
Arnau

R11.2 solver (messag file)

 

R13.1 solver (mes0000 file)

[arnau994]

Hello,

I just find out that the initial dt set in *CONTROL_IMPLICIT_GENERAL was the cause of the issue. However, this parameter was also set at 0.1s in the R11 model, so I understand that the R13 model for some reason prioritises it instead of the *ICFD_CONTROL_TIME dt.

I will set this forum entry as solved. Sorry for any inconvenience.

[Reno Genest]

Hello Arnau,

Thank you for letting us know.

I will close this topic for now.

Have a great day!

 

Reno.