KratosMultiphysics
[StructuralMechanicsApplication] On shell formulations
Description I am trying to better understand the shell formulations available in Kratos, especially to understand the limitations of each one for my applications. I apologize in advance if my questions are a bit amateurish!
By default, my team has been using the 3D3N thin shells formulation (ShellThinElement3D3N), but we have been getting some results which differ a lot from commercial solvers.
- From the codebase, it seems like this formulation uses the triangular ANDES formulation, which (if I understand correctly). From the code I understand that this element formulation contains bending + membrane actions, and no transverse shear stress one, is that correct?
- OBS: the link in the code comments above is broken. I am guessing it refers to this article, is that correct?
-
Many of the problems we deal with (what I believe to be) geometrically large displacements, on the order of ~10 thicknesses (think about an aluminum sheet 1m x 1m wide, 1mm thickness, fixed on a side, with point loads of around ~200N applied to the other extremity. I expect deformations of the order of 6-20 mm). Can this element still deal with these cases correctly? If not: are there alternatives built-in that I could try?
-
For these cases in (2), von Mises stresses are often 1-2 orders of magnitude larger than Ansys on the same problems. Any insights on why that may be the case would be appreciated.
Thank you!
Hi, I am trying to answer your question.
- Yes,
ShellThinElement3D3Nis based on ANDES (for the membrane part) combined with DKT (for the bending part), and it does not include transverse shear stress. - We applied the co-rotational method to enhance this shell element for geometrically nonlinear problems. Therefore, it should be possible to run your cases.
- I am also curious about this. I will try to check it myself and get back to you.
@KratosMultiphysics/structural-mechanics , perhaps anyone could add something.
I have been running some big benchmarks between Kratos and LS-DYNA implicit, with intriguing results. All my meshes have pretty high mesh quality, and I am varying the element count by subdivisions.
Results
Sharing some of the figures here (since I cannot share the code explicitly) comparing Kratos formulations with different LS-DYNA ELFORM (element formulations). Check here for a reference: Developer_Forum_Shells_Haufe_2013_v14_reduced.pdf
Very thin shell problem with point load* (thickness <1% of other dimensions)
Thin shell problem with point load* (thickness ~5% of other dimensions)
Another variant:
Thick shell problem with point load* (thickness ~20% of other dimensions)
Some observations
- It seems the ThickElementCorotational3D3N formulation always fails to converge, displacement-wise
- For very thin shells, all other Kratos elements converge in displacement, but the result differs from LS-DYNA by ~4%
- For thick shells, at least to the resolution I was able to run, the thin shell formulations converge for displacement while thick shell formulations keep growing (LS-DYNA also)
- For thin shells, Kratos' thick shell formulations more often converge to the LS-DYNA formulations than thin shell formulations do.
It would be nice to get some insight into what may be happening here.
dear @takeshimg92 , thank you for your in depth study. I am now implementing a new thick 3 noded shell and running multiple validation cases. Soon it will be available in the master. Would you be willing to try this new element?
Alejandro
Hi @AlejandroCornejo , of course! What are the main differences with respect to what is already on master?