John Doyle
Ansys Employee

It is not clear to me from your description, and the pictures, if the convergence failure is indicative of a numerical instability or a physical instability.

What is your expectation for this application?

Sometimes it is helpful to post process reaction force vs displacement to ascertain the state of structural stiffness at the point of non-convergence.

If the structure is going completely plastic thru an entire cross section and CZM is beginning, perhaps the structure lacks sufficient stiffness to resist further loading and the non-convergence is indicative of a physical failure.

If there is plenty of stiffness in the structure and the convergence failure is exclusively a numerical instability, you could try a number of different things in addition to what you already mentioned about contact stiffness factor:

  • Use many more substeps
  • Add stabilization damping
  • Try dropping midside nodes (in the first pictures you shared, there appeared to be some local hour glassing in a few of the elements at the corner)
  • Sometimes, it helps to add a frictional contact pair on top of the bonded pair to help stabilize the elements at the CZM surface after debonding begins.
  • Force Full Unsymmetric solver (for enhanced solver robustness after debonding)
  • Consider adding local damping to the CZM material model (refer to Section 4.20.5 on ‘Viscous Regulation’ of CZM in the Material Reference Guide.