fbd_inv error with pinned beams

[jackrussel]

Dear seismo support,
I found the problem with the model. when I run your model with no change there's no problem. but when I use <sort by name> in element connectivity and <BRB-A-111> is on the top of list I have the previous problem. this is unbelievable. maybe the windows doesn't work properly.

[Stelios_Antoniou]

Jackrussel, the convergence criteria that you are using are extremely loose, hence the difference that you experience between the runs. In any case, you have fractured steel, you have a 2% interstorey drift (3rd-4rth level), what you expecrience is structural collapse

[jackrussel]

Stelios_Antoniou,
I can't understand loose convergence criteria.

[seismosoft]

Dear jackrussel,

We have looked further and carefully into your model and can confirm that there seems indeed to be a problem when end-releases are used together with stiffness-proportional damping, as you signalled back at the end of September.

We are now looking into this issue, and wonder if, for the time being at least, a workaround for your analyses could be to:
- use truss elements instead of frame elements with end-releases?
- use frame elements with links at the ends instead of frame elements with end-releases?
- use frame elements with end-releases without (stiffness-proportional) damping?

Thank you very much for signalling this problem to us; as soon as we manage to resolve it, we will release a new version. In the meantime, we do hope that at least one of the workarounds suggested above can be of some help.

Seismosoft Support

PS: in models that are on the verge of numerically instability (due to structural failure, member instability, presence of strong stiffness irregularities, etc), using displacement/rotation convergence criteria larger than the default values of 1E-4 may indeed lead to erratic behaviour, as mentioned by Stelios Antoniou. As we noted above, however, the main problem with your model is not this, but rather an apparent difficulty of SeismoStruct in dealing with the combined presence of end-releases and stiffness-proportional damping..

[jackrussel]

seismosoft,
thank you very much. do you mean that there is the same problem with rayleigh damping? rayleigh damping expresses damping as a combination of mass and stiffness. in some models I have the same problem with rayleigh damping. the mass proportional damping works normally but I don't know the results of it are reliable or not(spurious energy dissipation).

[seismosoft]

Yes, this behaviour occurs for stiffness proportional damping (and Rayleigh damping is indeed proportional to stiffness and mass).

And, yes again, in general, mass proportional damping is not something that we would normally recommend.

Seismosoft Support

[jackrussel]

Hi seismosoft,
I have modeled a 2d structure with your guidance and using link elements instead of end release but there's a problem. The displacements are too large. May this problem has a relation with link elements and Rayleigh damping?

[iliya56]

Hi seismosoft
In my model in which some elements are pinned,I have tested your suggestions but it didnt work and the problem persists.Is it related to software bugs or my model has another problem?
I have sent my model to your email address.
Thanks alot

[seismosoft]

Dear iliya56,

Please try to start a new topic, rather than continuing dicussions on topics that are not related to your problem. It is not very helpful for Forum users trying to find answers to their difficulties when there is a very long post with loads of questions on answers on ever changing topics.

Regarding your model, it is not easy to spot a convergence problem in a 1500 element model. We would note the following:
(1) moment releases in frame elements cause convergence problems when the element gets in the inelastic range. use the truss elements instead
(2) check if the structural masses are correct. It seems strange that at the first step of the dynamic analysis the solution diverges, whlist it runs fine with the gravity loading
(3) Run an eigenvalue analysis to see if parts of your structure are not well connected.
(4) Assuming that there are software bugs, because a 1500 element model does not converge is rather superficial. Bugs in the program is obviously a possibility, however the model has been unstable from the beginning with parts unconnected to each other etc. Hence before looking for problems in the program try to spot possible problems in the model.

SeismoSoft Support

[iliya56]

Hi seismosoft
Thanks for your guidance.
I have run 15 IDA analysis for this model without any releases , whith different degrees of intensity up to 20g but I didnt have any problem.
If I had unconnected elements , as you told,then none of static and eigen value analysis should converge, but it is not so.
I want to know that there is sth wrong with my model or such structures cant sustain earthquake with pinned elements.
Thanks