Shell test for singularity in shell model

[Xiangyu-Hu]

@DongWuTUM we probably update this draft pull request to shown at least 2D is working.

[DongWuTUM]

@SchrodingersAlphaGo @BenceVirtonomy @FabienPean-Virtonomy @Xiangyu-Hu

• We finally found out and solved the singularity problem in the shell model after a series of efforts. I will continue to conduct some tests to further check the accuracy, stability, and so on. The related code will be available soon.

https://github.com/Xiangyu-Hu/SPHinXsys/assets/100197018/b93fb80d-074b-46fc-8949-1ff80a7fdf41

[Xiangyu-Hu]

Great!

[BenceVirtonomy]

Very cool!

[DongWuTUM]

@BenceVirtonomy Hi Bence, thank you for sharing the petal deformation case! The new simulation result is shown in the video below. Could you please elaborate on how you obtain the initial particle positions and normals?

https://github.com/Xiangyu-Hu/SPHinXsys/assets/100197018/82b9af79-849a-4352-95f8-b7b7970bb8c1

[BenceVirtonomy]

Awesome:) It's from our CVT particle generator.

[DongWuTUM]

Awesome:) It's from our CVT particle generator.

Will it be open source?

[BenceVirtonomy]

I think we need a discussion there, because it would involve adding more 3rd party dependencies.

[Xiangyu-Hu]

What is that 3rd party lib? Actually we are considering add OCCT already for more generalized geometric modeling.

[DongWuTUM]

@SchrodingersAlphaGo @BenceVirtonomy @FabienPean-Virtonomy @Xiangyu-Hu

I found a problem in the shell dynamic case.

• This problem only occurs at high resolution. I only tested it at low resolution before and it was fine. With high resolution, the plate appears to show high frequency oscillations.
• The results by the original code show the same high frequency oscillations. In other words, it's not caused by the bug fix of this time.

2D plate with low resoluton:

https://github.com/Xiangyu-Hu/SPHinXsys/assets/100197018/e144c2a9-a5a7-47b1-84ec-9d1fc63c49a2

2D plate with high resoluton:

https://github.com/Xiangyu-Hu/SPHinXsys/assets/100197018/1f026a92-b663-4092-b306-44b8f616bee0

3D plate with high resoluton:

https://github.com/Xiangyu-Hu/SPHinXsys/assets/100197018/a2519a4e-5d20-4214-a176-586604b5e557

[Xiangyu-Hu]

@DongWuTUM you need pull the master so that the conflict will disappear.

[BenceVirtonomy]

OCCT

Geogram. Please discuss with Fabien before deciding on adding major dependencies, he has a lot of experience here. @FabienPean-Virtonomy

[DongWuTUM]

Updates:

• Fixed bugs:
  1. Integrate Cauchy stress along thickness now. Otherwise, the error of calculation results is large when large deformation;
  2. Correct conversion from dpseudo_nd2t to dangular_veldt for avoiding 90 degree singularity;
  3. Correct scaling of numerical damping to solve the incorrect high frequency vibration of dynamic shell cases.
• Optimizations:
  1. Simplify the calculation of dpseudo_ndt;
  2. A dynamic shell case, named test_3d_dynamic_plate, is added;
  3. A shell test with rigid rotation, named test_thin_structure_dynamics, is added in the unit tests;
  4. Two test cases, test_2d_shell and test_3d_arch, are applied very large deformation;
  5. Output the von Mises stress of shell mid-surface.

[FabienPean-Virtonomy]

OCCT looks like a huge beast, if you need only a small part of it, I would prefer you find a smaller alternative. Besides, it is LGPL, so it must necessarily be built as a shared library to not propagate its license to consumers of the lib. In any case, it must be an optional dependency.