Problems with convergence

[lorenzino]

Hi,
I am a student and I am trying to simulate a quasi-static cycled test on a shear wall. After defining the wall geometry and the materials, I define the first 5 displacement-cycle peaks in terms of wall drift ratio (in total I have 9 displacement-peaks) and the model works great, giving me reasonable results. But, when I insert also the sixth displacement-peak, it fails to converge. I tried to add different kind of alghoritms to solve it and I also tried to change tolerance and the displacement increment, but it still does not work. What should I do? In my opinion the problem should not be about the model, since it works for the first 5 peaks, but should be a numerical problem.

Thanks,
kind regards

[selimgunay]

Did you check whether the wall has sufficient strength before non-convergence occurs. It could be possible that the response has severely degraded or on a negative slope. If this is not the case, you can try even smaller time steps.

Furthermore, having convergence in the previous cycles does not indicate that the experienced nonconvergence is not due to the model. It could well be because of the element that you are using. Which element do you use to model the walls?

[Jhno]

Try increase the damping and/or reduce the timesteps.

[lorenzino]

selimgunay wrote:
> Did you check whether the wall has sufficient strength before
> non-convergence occurs. It could be possible that the response has severely
> degraded or on a negative slope. If this is not the case, you can try even
> smaller time steps.
>
> Furthermore, having convergence in the previous cycles does not indicate
> that the experienced nonconvergence is not due to the model. It could well
> be because of the element that you are using. Which element do you use to
> model the walls?

I do not know if it is a matter of strength or, in general, a problem about the wall design because I am trying to realize a model of a wall that has been already designed and tested. I have the test results (load-displacement curve)and I would like to obtain similar results with my model (it is a model calibration).

Regarding the elements, I am using SFI MVLEM elements. Could them represent a problem?

Thanks for your help.

[selimgunay]

The SFI MVLEM elements can be the source of nonconvergence at the large displacements. Could you check references 1 and 4 in the below link to see how the authors handled similar situations.

http://opensees.berkeley.edu/wiki/index ... r_RC_Walls