Convergence difficulties due to abrupt stiffness change (ekill)

  • Last Post 26 September 2019
ElCid01 posted this 22 December 2018

Hello everybody

I am running a nonlinear structural simulation with multiframe restart option in Ansys APDL of a sandwich structure which is loaded via a displacement control. After each load step I postprocess relevant data to know, if a core element during the simulation fails. If such a core element fails then I restart the simulation from the previous substep and kill the elements which failed with the ekill command. Problem is that, after deactivating the failed elements the simulation does not converge anymore. NROPT,full,,on has been set and also the time increment respectively the displacement increment has been chosen very small. Nevertheless no improvement of the convergence has been achieved. The problem of course is that the stiffness becomes very low after deactivating the failed elements. What I also tried is a simulation without the failed elements from the beginning but then the simulation converges and also a very large displacement and reaction force is achieved, which means that the before mentioned calculation should be possible, but I really do not know how to overcome the convergence difficulties.

I would be very happy for any suggestions I could try.

Thanks a lot.

Best regards

El Cid




Order By: Standard | Newest | Votes
SandeepMedikonda posted this 23 December 2018

Hi El,

How does your force convergence plot look and what do the newton-raphson residuals (NROPT) suggest? Please see this discussion. Are they high in the regions where the elements are being deleted or elsewhere? If there is contact in your model and if the NR residuals are high over there, try reducing the contact stiffness. I would also recommend you to try using the Direct solver instead of the iterative one and see if it helps.

Guidelines on the Student Community

ElCid01 posted this 23 December 2018

Hi Sandeep

Since I am working with Ansys Mechanical I am not able to see this plots or I do not know how… I do not have any nonlinear contacts in my model, I only have nonlinear joint elements (MPC184) that are linked to the elements which are deleted. Therefore I cannot change any contact stiffness. To change the solver is a good idea, do you know which the standard solver is?

Thank you for your advices, I will try them as soon as possible and comment.

Regards El Cid

SandeepMedikonda posted this 23 December 2018

Ok, The default should be the PCG or the Iterative Solver. You can check this in the solution output.

ElCid01 posted this 07 January 2019

Hi Sandeep

During these days I had time to implement what you said to me about changing the solver. I tried every solver out but without success. Attached you will find the convergences plots before the element are killed and afterwards. In the first picture I make a restart at the time 0.25 approximately, at this time the elements are killed because the stresses are too high.

Thanks, best regards El Cid


jj77 posted this 07 January 2019

It is hard to say without looking at the model. As far as I understand it needs some manual intervention, or does it automatically remove elements and continue, or do you do that manually, and then continue. I believe that there is some script that does this automatically, think it is from peteroznewman, see here:



Venugopalb posted this 26 September 2019

Hello EICid01!

Can you tell me how to change the solver in ANSYS workbench. Because my model also gives unconverged solution.

My professor asked me to change the solver.