Seismosoft | Earthquake Engineering Solutions

I am trying to run a dynamic time history analysis on a bridge with isolators modelled using zero-length "plst" links. I defined the time history input at supported nodes and defined no permanent loads, letting SeismoStruct auto-convert masses to loads.

When I try to run the model, i receive the following message:

"Unable to apply the entire permanent load. Analysis terminated"

So I interpret this to mean that the model cannot even be analyzed for the self-weight. Support conditions seem to be more than adequate for stability.

Any advice on this error message will be welcomed.

As a first step, you could run eigen-value analysis with the same model, in order to make sure that there are no some sort of local instabilities that prevent the model from converging with static loads.

Thank you Stelios. The Eigenvalue Analysis worked fine - the periods were not what I expected (much too low) - but that is something I'll tweak once I get everyhting running fine. I'm still puzzled at the failure of a successful Permanent Load execution.

Hi,

Given the fact that your periods are much lower that you expect, it might be the case that your masses are very high and/or that your structure is too slender. If so, the software might not be able to reach equilibrium for the permanent load.
Verify if you define correctly the sections, materials, or masses, and if they are compatible with the units defined.

Cheers,
Romain

Huffte,
The fact that the periods are low means that the masses are low or the stiffness is very large. Hence, the large masses/gravity loads are not the problem. However, I believe that the low period values are where you should start from, it obviously means that the model is not 100% correct. Check the eigenvalue deformed shapes to see if they look reasonable.

Another way to move forward is to try to run static analysis (simply the initial loads) with your model and very small initial loads, i.e. (1) materials' specific weights, (2) sections' mass/length and (3) initial loads almost zero. By 'experimenting' (i.e. trying different loading at different locartions on the model) you could find out what is wrong with the model. For example, if the analysis diverges with very small values, it obviously means that there is something fundamentally wrong: (1) the structural configuation or (2)the convergence settings, e.g force-based convergence criterion with rigid links/constraints)

Hi,

Sorry for my previous erroneous message regarding the mass/stiffness values.
Obviously it should be as Stelios_Antoniou stated.

Thanks for the correction,
Romain

Thanks for the tips. I have actually switched from the "plst" link to a bi-linear link and the Permanent Load analysis runs in a snap. The time history results are incorrect however - displacements are much too small, and I'm starting to wonder if there are some conversion problems internal to SeismoStruct - I have created the model in English units. I am considering creating an entirely new model in SI units to see if I get the same problem - then I'll know for cetain if there is some flaw in my input or in my strategy.

Once again, thanks very much for the advice to both Stelios and Romain.

After some tweaking of parameters - I am relatively new to SeismoStruct - it would appear that the Newmark integration scheme works much better for models which incorporate non-linear links as the sole source of inelastic behavior (this is compared to the default integration scheme).

My suspicion of an internal problem with units was unfounded.

Anyone aware of any good example problems - and where I could get example seismostruct files - of structures modelled with bi-linear kinetic links to represent isolators? I am trying to model bridge structures with all inelastic behavior occurring in the isolators and using dynamic time history analysis.

Dear huffte,

Your observations on the higher stability of the Newmark integration scheme, with respect to the HHT one, are interesting, since in general the opposite has been proven to hold instead. Have you tried running such dynamic analyses on a very simple model featuring one spring+mass only, to check if the same happens?

On the issue of example models, we do hope that at least a couple of the 5000+ users registered in this Forum will respond to your request, notwithstanding the somewhat "shy spirit" of this community (with some noticeable exceptions).

If you have access to the textbook of Priestley et al (http://tinyurl.com/3f3tjrd), you may check the SeismoStruct examples contained in the enclosed CD. We know that bridges were modelled, but are not sure if they feature the isolators you are interested in.

Seismosoft Support

hi,
I have the textbook by Priestely et al. and the CD as well. I tried to run those models in seismostruct. But they are modeled in version 4.3 and when I open these models it shows list index out of bound (3). why this message is coming? when I open these models there is no nodes defined and no element connectivity. can u inform me what is the problem?

Thanks