Change C3D8 elements to C3D8I elements in perpendicular-flap solid-calculix

[AndresPedemonteFIUBA]

This modifies the CalculiX case for the perpendicular-flap tutorial to use incompatible mode eight-node brick elements (C3D8I) instead of basic linear fully integrated eight-node brick elements (C3D8).

C3D8 elements are prone to shear-locking, resulting in the structure appearing to be much stiffer than it actually is. Quoting the CCX manual:

Other elements frequently exhibit unsatisfactory behavior in certain instances,
e.g. the C3D8 element in bending states.

the element tends to be to too stiff in bending

C3D8I elements are much better for this case, where the flap bends:

The incompatible mode eight-node brick element is an improved version of the C3D8-element. In particular, shear locking is removed and volumetric locking is much reduced. (...) The C3D8I element should be used in all instances in which linear elements are subject to bending.

This "fix" just requires chaning TYPE=C3D8 to TYPE=C3D8I in all.msh and adding the DIRECT parameter to the *DYNAMIC keyword in flap.inp. This last addition is necessary to avoid CalculiX's automatic incrementation kicking in and changing the timestep size.

The change should help bring the results obtained with CalculiX more in line with those obtained with other structural solvers (FEniCS, deal.II) for this case. See #176.

Andrés

[BenjaminRodenberg]

I will take care of reviewing this PR and also check how these changes affect the results presented in the reference paper (see Fig 21).

[BenjaminRodenberg]

I applied the suggested changes and reran the examples presented in the preprint for the preCICE reference paper. The results are much better with the suggested C3D8I elements and in good agreement with the other available solid solvers FEniCS and Deal.II:

I used the preCICE vm version that is based on https://github.com/precice/tutorials/releases/tag/v202104.1.0, namely precice/vm@66ee0cd.

@AndresPedemonteFIUBA thank you again for this contribution and your help!

[BenjaminRodenberg]

Merging this PR closes #176.

[BenjaminRodenberg]

@AndresPedemonteFIUBA We would like to use the updated results in our paper and also point to this PR. If you want, we can also mention your github or real name. Can you write me a email to [email protected], in case you want to be mentioned personally? If not, we can also refer to this PR without mentioning you personally.

[precice-bot]

This pull request has been mentioned on preCICE Forum on Discourse. There might be relevant details there:

https://precice.discourse.group/t/turek-hron-fsi-with-openfoam-and-calculix/1001/5