Simulation of shear banding/strain softening in ABAQUS

[Shroom6969699]

Dear all,

I am trying to simulate shear banding in single crystal (i.e. copper) using ABAQUS. I have written both UMAT and VUMAT for the single crystal plasticity, and here is what I do to introduce shear banding : first I adopted a strain-softening constitutive relation, which sets the materials hardens a bit then softens after it yields. Second, I introduced a small geometry defect on the model, (like a small notch at the edge). In this way, I expect a shear band (localization) would form along the defect area.

However when I run the simulation for UMAT, I couldn't get convergence results when the material starts to soften. I tried RIKS or stabilization but it doesn't help. When I switch to Explicit, it takes very long time for the simulation to run and it crashes due to mesh distortion at some point. So I am wondering if anyone has experience in simulation of shear banding, and how could I get converged/completed results for such situation.

Thank you very much for your time.

P.S. I attached the paper I am following in case anyone want a clear picture of what I want to simulate.

 

[IceBreakerSours]

Disclaimer: I do not know the answer to your question and I do not even have the required background to help you out. Also, and unless you are lucky, forums may not be the best place (but trying doesn't hurt).

I have written constitutive laws before so I just have general questions/comments to ask/make:

Ensure you aren't making any modeling mistake that may be causing, say, excessive negative eigenvalues (apart from material softening coded in the subroutine). A uniaxial/biaxial test on a single element would be my first step.

Try to use the closest approximate material law supported by Abaqus on a single element to make sure the trouble is, in fact, coming from the subroutine.

Have you performed single element tests (using Standard/Explicit - single and double-precision) and compared the results with an analytical solution? If not, then you should. If yes, then do you write out values for certain variables of interest (either to a file or the command prompt/terminal, etc.). You should also add clever little pieces of code to detect if something is wrong and warn you about what happened. You should also try the same on a 'multi-element' model which is *fast* to run so you can make mistakes fast, learn, and fix the code, if that is what's needed.

Finally, you may wish to give *Dynamic, application=quasi-static a try.

*********************************************************
Are you new to this forum? If so, please read these FAQs:

http://www.eng-tips.com/faqs.cfm?fid=376

http://www.eng-tips.com/faqs.cfm?fid=1083

 

[Shroom6969699]

Dear IceBreakerSours ,

Thank you very much for your reply.

Yes I have tested it on single element / 8 element (3D) cases and the results seems fine to me. There is no error in completing those tasks. And the subroutine seems working fine if I use the geometric as a cubic shape, but if I changed the geometry into some thin rectangular shape, the problem of unconvergence/ mesh distortion occurs. Does this mean that the subroutine is OK and the problem is caused by model geometry or something else ?

And yes I am trying out *Dynamic, application=quasi-static to see what happen. In this case should I use UMAT or VUMAT ?

Thank you very much.

 

[IceBreakerSours]

*Dynamic is available only for the implicit solver, so you can use your UMAT with this option.

*********************************************************
Are you new to this forum? If so, please read these FAQs:

http://www.eng-tips.com/faqs.cfm?fid=376

http://www.eng-tips.com/faqs.cfm?fid=1083