Numerical instabilities on a steel braced frame

[Giosmi]

Good morning,

I modelled a 4 floor 3D steel braced frame using infrmFB element to characterise the diagonals and elfrm elements for the horizontal and vertical elements. The beam-column joint are considered pinned and are realised with M2 releases. The diagonal-column joint is modelled with a rigid element (elfrm element) and a Nlink element which possesses only translational stiffness (in the "Curves parameters" table, M1 M2 and M3 stiffness are considered but their value is very low, compared to the translational values).

The single braced works in Seismostruct (that is the model with just the braced, and its system of beams and columns). The 3D model is symmetrical in in X and Y direction, but the capacity curves of the static pushover analysis don't, and the analysis' are very sensitive to very little variation of static loads. I think the pushover analysis in the Y direction works fine (there are not secondary horizontal beams in the Y direction). In the X direction though, at some point in the analysis (after the maximum base-shear is reached) the pushover analysis suffers numerical instabilities and divergence error: multiple nodes of the secondary horizontal beam element give the "Diverge" with "Norm of out-of-balance Deformation". The secondary beams are considered simply supported on the primary beams (M2a and M2b releases).
Is it normal to have such intabilities? Can i give you the model?

[seismosoft]

If you carry out a search in this forum with terms like convergence, divergence or numerical instability, you will find numerous posts on this issue. You will also find a section on the Users' Manual on how to solve numerical problems (Appendix A). Diverging solution is an inherent feature/problem of nonlinear analysis and are to be expected, especially in the highly inelastic range. In addition to the details specified in the manual and in this forum, note the following:
1) infrmFBPH frame elements are more stable numerically than infrmFB
2) Combining inelastic frame elements with end-releases causes numerical instabilities. Maybe you could consider modelling the pin with a linear or nonlinear link element
3) You should investigate the case of the exceedance of the member capacities. If a member fails, it cannot sustain the load that it should carry and the analysis diverges.
Seismosoft Support

[Giosmi]

I checked the forum and the manuals but there was no solution that actually worked to my specific problem. I changed maximum number of iterations in the "global iterative strategy" and "element iterative strategy", increased the number of steps of the loading phase, but still there's a big difference between X and Y static pushover responses.

Moreover, there are no InfrmFB elements with end-releases. All the end releases used to consider the pin are within elfrm elements.