I am simulating a simple uniaxial tension by pulling for a displacement of 0.4 in the Z-direction on a quarter symmetry cube with side lengths of 0.8. I am using an elastoplastic self-consistent UMAT and I noticed that when I mesh the cube into two elements by dividing it in the two directions as shown in the attached image, the simulation will fail by not converging at the minimum specified time step for the left mesh but will successfully complete for the right mesh. The status and message files are attached for reference.
I am going through the working variables in the UMAT but haven't found anything abnormal yet at the first non-converged step. Has anyone experienced this before and are there any UMAT input/output variables that I should take a closer look at?
Just some additional information: the simulation failed with 8 elements by seeding each edge with 2 elements, succeeded with 4 elements similar to the mesh on the right by making a cross on the pull face, and failed with 4 elements similar to the mesh on the left, which prompted me to try with 2-elements and it has matched the pattern so far.
Thanks!
Jesse
use equation constraints to define a constraint that all nodes on a free face will move equally in the face perpendicular direction. this is exactly the reason why the right mesh is converging but not the left mesh (hourglass shape)
Abhimanyu Kumar Thanks for the response! By using equation constraints, do you mean creating a set for free face on x axis, create a set for a reference point 0.1 units away from the free surface, then use equation constraint with coefficients 1 and -1 for DOF 1?
you can use one node on the face as master and remaining as slave... no need to create a ref pt
- Badges
- Science topic