Hello fellow researchers,
In my recent research I have noticed a strange issue in Abaqus FEA software. It seems to calculate the deformation gradient for shell elements incorrectly in certain cases. From what I understand, the "deformation gradient" in Abaqus is, in fact, the left stretch tensor because it is calculated after considering the rigid body rotation of each element. Thus, the "deformation gradient" should normally be positive definite (symmetric and with all positive eigenvalues). In most simulations I've run, it is the case. However, this time I got a case where it gives me an asymmetrical deformation gradient with negative eigenvalues. This is unreasonable for a stretch tensor.
I have recreated the issue using two simple simulations and provided a note containing some explanations. Case 1 is the case where Abaqus calculates the deformation gradient incorrectly. Case 2 is functionally identical to case 1 but seem to calculate the deformation correctly.
Feel free to use the provided simulation input file and the FORTRAN subroutine to recreate my results.
Any thoughts and suggestions towards working around this issue is appreciated.
After some thorough testing, I was able to conclude the following:
1) If using Abaqus built-in shell formulations, Abaqus will calculate the deformation gradient correctly, whether if an orientation is assigned or not. However, this matters little since you hardly use the deformation gradient in this case.
2) If using UGENS subroutine, Abaqus can calculate the deformation gradient correctly if you do NOT assign orientation (which will default to global coordinate).
3) If using UGENS subroutine with an orientation assigned, Abaqus may calculate a wrong deformation gradient.
I provided some documentation and examples in case anyone wants to recreate this. I will also try to report this issue to SIMULIA. Hopefully, they can look into it later.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Mechanics
- Classical Mechanics
- Continuum Mechanics