Fix cross-dyadic product: Ensure minor- and major-symmetric result #44
Conversation
|
@jfriedlein I did my best in fixing it with gfortran because currently I have no FEA code / Intel Fortran at hand to test the changes. Could you check if your models converge with these changes? Many many thanks in advance 😄 |
remove empty "example" line in section overview
|
Perfect! I checked the numeric results with my Python package adtzlr/hyperelastic but I couldn't test an actual UMAT / HYPELA2 / whatever subroutine inside a FEA package if something happened related to convergence. |
|
Hi @adtzlr, I tested the "fix-crossdyadic" branch for LS-Dyna:
For a more reliable and comprehensive test, we should think about unit tests. Probably that is a lot of work to set up, but once the framework is ready, it should be easy to add more and more tests and provide a quality assurance. Maybe a todo for the future. |
|
Yes, unit tests are indeed a great idea! Ideally would be an automated CI/CD pipeline with Github Actions. But my experience in Fortran ends with the scope of this package... I usually do some UMAT scripting but thats it. |
ah wow, very happy to hear that my document was helpful! |
make the cross-dyadic product fully symmetric, i.e.
C(i,j,k,l) = (A(i,k) B(j,l) + A(i,l) B(j,k) + B(i,k) A(j,l) + B(i,l) A(j,k))/4closes Crossed-dyadic product is wrong #40
see also Maria Holland's Hitchhikers Guide to Abaqus, p.46. Btw, thank you @mholla for this great summary on symmetry, I've overlooked that over all these years!
rebase
identity4(T)on the fixedcdya(T, T)closes Rebase hardcoded fourth-order
identity4(T)tensor oncdya(T, T)#39remove the incorrect implementation of
the inner cross-dyadic icdya(T, T)closes Inner-crossed dyadic product of two sym. tensors can't be stored as a 6x6 fourth-order tensor #37
Especially the first point was hard to detect because I've used the special case
A = Bin 99% of all applications where the implementation up tov1.1.2works without any problems.