I have a notched specimen as shown in attached NotchedSpecimen.png. It is basically a bar with 4 circular cuts on four sides. As a result, the specimen has 4 sharp notched edges. It is a quarter symmetry model so imagine three symmetry faces on top, right, and back of the model shown in the image.
Seeing the model have issues with the sharp corner at the bottom, I reduced the model as shown in NotchedSpecimenReduced.png, where I was justifying this reduction will not affect reaction force results.
I am applying a displacement of 0.15 over 150 seconds to simulate a quasi-static tension experiment with minimum allowed time step of 1e-8 and maximum 0.05. When I am modeling the material with J2, the simulation completes successfully. When I am using a UMAT subroutine, the simulation keeps on getting stuck at a time step where it would fail a step, reduce time step to around 1e-6. Although the simulation continues forward, moving at a pace of 1e-6 towards 150 is pretty much stuck.
Exploring my options, I increased the tolerance on the model from 0.005 to 0.05, and I believe the result tells me why it was getting stuck. As shown in IncreasedTolerance.png, the notched corner deforms abnormally. As the umat runs normally with other simulations and the model runs normally with J2, I am not completely sure what is causing this.
So this leads me to ask the community if there is any way to help with this situation? Perhaps forcing a deformation limit on the corner so it doesn't poke out like that?
Thanks!
Jesse
Not clear why you performed a transient analysis if your problem is quasi-static. A steady-state simulation is sufficient
Hello Mr.Feng.
Various factors are effective in this issue and one of these factors is the weakness of systems in solving and large volumes of information. I think you can prevent this sudden deformation to some extent by making the mesh finer in this area. Or apply loading at a slower speed.
Denis Benasciutti Hi Denis, thank you for the response. I am not sure if I completely understood your comment but the reason I added the 150 time is that the strain-rate is calculated inside the UMAT and affects the mechanical response. When given the default steady-state simulation that goes to a time of 1, the strain rate of the simulation is too high.
Mohammad Ali Falamarz Hi Mohammad, thanks for the response. I will add a bias to the mesh to have more refined mesh at the tip to see if it addresses the problem. I will update when I get results.
- Badges
- Science topic