Convergence Difficulties with Interface Delamination

  • Last Post 20 April 2019
Afrah posted this 10 April 2019


I have a 3D DCB specimen and am modeling the delamination using the Bi-linear CZM Law. I am trying to match the load-displacement graph obtained from ANSYS with that from experiment. In trying to do so, I am trying to find the correct cohesive zone parameters. When I use a high value of the "normal displacement jump at the completion of debonding" (C2), which is anything greater than or equal to .1 mm, the solution converges, but the load-displacement curve is all wrong. However, when I lower the value to something like .095 mm or lower, the solution doesn't converge. I do, however, get partial results that looks much better. Here are the errors I receive:

I have tried the following things to make my solution converge:

-increased the sub-steps, here are my settings:


-Refined the mesh at contact regions using contact sizing (I'm not really sure how much I should refine my mesh, it's possible I just haven't done it enough)

- lowered the stiffness of the contacts 

- reduced the penetration tolerance

-used the normal lagrange formulation

What else can I do?

Note: I've tried to find the "contact destabilization factor" option under the Advanced contact settings, but haven't been able to find it. Is this option no longer available in ANSYS 19.2?






Order By: Standard | Newest | Votes
Afrah posted this 12 April 2019

Would it help to attach my workbench file? I would really appreciate the help. 

TAPTAPTAP posted this 12 April 2019

yeah you should attract your model so people can help you find the solution easier.

  • Liked by
  • Afrah
Afrah posted this 12 April 2019

I've attached my v19.2 file. I've attached a simulation that has a simulation time of 60 s and it converges, anything more than and it doesn't. I actually need to run it for 600 s. I had to clear generated data off modeler cause the file size was too big. 



Attached Files

Afrah posted this 20 April 2019


I've tried some more things to get the solution to converge;

-use MPC formulation

-enabled weak springs 

-increased number of iterations per substeps using the NEQIT command

any other suggestions?