Add SU2-FEniCS Perpendicular Flap FSI Tutorial

[IshaanDesai]

For the perpendicular flap FSI tutorial we have the combinations:
OpenFOAM-FEniCS
OpenFOAM-CalculiX
SU2-CalculiX
To complete all combinations and show the true black-box use of preCICE and the official adapters, this PR adds the SU2-FEniCS tutorial case.

[IshaanDesai]

The 2D mesh used for Fluid participant (SU2) is constructed in GMSH and has the structure:

[IshaanDesai]

The folder initial-run gives the setup used to run a SU2 simulation with the flap as a rigid body. The resulting flow profile is used as a reference to run the coupled FSI case.

[MakisH]

Thank you very much for adding this! 👍

There have been quite some changes (about to merge) in the run scripts and configuration, see #100. Please wait for this to be merged and then port the same changes here.

It would be good to also remove a few log files that slipped through.

[IshaanDesai]

Thanks for the feedback @MakisH ! Will look out for #100 before merging this one. I am still working on this PR because currently the simulation diverges

[MakisH]

@IshaanDesai the SU2-CalculiX also diverges (at least under some systems), so this is not necessarily FEniCS-specific.

[IshaanDesai]

While tuning the SU2-FEniCS case and analysing results, it was found that the su2-adapter works with DisplacementDeltas and the fenics-adapters works with Displacements. To get a working tutorial, the fenics-adapter would need to be modified to handle DisplacementDeltas

[MakisH]

There is also a related issue on adding Displacements to SU2: precice/su2-adapter#17

[IshaanDesai]

Thanks @MakisH for pointing that out! After discussions with @uekerman and @BenjaminRueth it is decided to currently resolve this tutorial via handling DisplacementDeltas in the fenics-adapter. Reference: precice/fenics-adapter#87

[IshaanDesai]

After further discussions it was concluded that handling the absolute Displacement values or delta Displacement values is something that a FEniCS user can do in the FEniCS case file. Due to the unique way in which FEniCS is set up and the adapter is constructed it was found unworthy to implement the handling of DisplacementDeltas in the adapter.

[IshaanDesai]

Tip displacement plot:

The non-smooth behaviour is because the structure solver in FEniCS is a linear solver.

[uekerman]

Thanks for adding this 👍
Some remarks and small tweaks:

• I get the same crash as @BenjaminRueth.

• To be consistent to how we organize tutorials currently, the case should better go into flap_perp_2D. Soonish we want to re-organize such things anyway (#28), but for now.

• Could you please format the preCICE config and remove all commented out parts there.

• We should carry the case to a valid regime of the linear structural model. So, smaller displacements. For example by increasing the E module.

[IshaanDesai]

Increasing the modulus of elasticity (E) for the structure from 4E5 to 1E6 leads to a smoother tip displacement plot:

[MakisH]

The 2D mesh used for Fluid participant (SU2) is constructed in GMSH and has the structure:

Could you please describe here how you generated this mesh? Could be useful in the future.

Same for the initial flow.

[MakisH]

Excellent! All the configuration files look good, see a few minor comments. 👍

I cannot run the case at all at the moment (system-specific issue with mpi4py), but since it works for you, go ahead.

Remember to integrate this case into the wiki.

I quickly skimmed through the FEniCS solver code, but I cannot see where the recursion could be originating from. The stack trace also does not help much. Maybe this is originating from the Python bindings?

The SU2-CalculiX case is also unreliable right now, so a convergence issue would not be necessarily a FEniCS issue.

[MakisH]

It would be good to also add this case afterwards to our system tests. We can wait for #109, or we can simply test only a few seconds.