Dis-continous hysteresis

[hasanozkaynak]

Hello,

I am facing some problems with the nonlinear time history analysis out put in terms of force-displacement hysteresis.The curves are not in a continous form some scattering data exists even I increase the out put solution time interval about dt/10 20.

Would you please forward me how to come up with this problem.

Regards
Hasan OZKAYNAK (PhD)
hasanozkaynak@beykent.edu.tr

[huffte]

Possibly some more details would help users of the forum offer explanations, hasanozkaynak.

What type of elements are being loaded into the inelastic range. Force - based frame elements? Displacement - based frame elements? Link elements? What material models? Which link types? What is the time step in your loading function, which I assume is an accelerogram? Do you have constraints?

Best of luck, hasanozkaynak.

[hasanozkaynak]

Thank you for your interest.

Actually the model is very simple one-bay (one storey: 5 m in lateral x 2.75 m in vertical joint to joint) two storey Reinforced concrete Frame.

There is no constrains and also no link elements.
For the steel I prefer Manegetto-Pinto (A1:18.5,A2:0.25, A3:0, A4:0 Yield Strength 200000 kPa, E:2E+08) and Mander for the concrete.

The frame members were chosen to be Inelastic Frame Displacement Base Plastic Hinge (INFRMDBPH)because the others were not succesfull to complete the full dynamic analysis.

The input accelaration record has 0.01 sec. time intervals. The out put time interval is 0.0003 sec.The peak acceleration value is 0.2 and I multiply it with 9.81 during the analysis.

And also I used some additional joints near to the real column-beam intersection joints to let the analysis to be more precise.

The out put force-displacement loops are some how resemble to the reality but it scatters and discontinue like clouds...

I would appriciate for your support
Awaiting your reply
Regards
Hasan
--------

[huffte]

Hasan, if you will e-mail your model, I will try to look at it next week. I will to be back in my office until then.

[hasanozkaynak]

Thank you for the reply

Shall I e-mail it to support@seismosoft.com ?

Regards
Hasan

[huffte]

You can e-mail to the address you say and perhaps the support team can look at the model as well. But if you will click on my id in the forum you will get to my e-mail address - Tim.Huff@tn.gov.

[seismosoft]

Can it be that you are using a very loose displacement-based convergence criterion and not a force-based one?
Have you tried to make the convergence criteria a bit more stringent?

Seismosoft Support

[hasanozkaynak]

Thank you for your comment,

I use dt/32 as oupt put, and 0.0001 as convergence displacement tolerans, What do you prefer it to be ?

Regards
Hasan

[seismosoft]

We had a look at your model and we have the following suggestions (1) increase the time step, there is no reason to have such a low dt and if you increase it you will probably get a smooth hysteretic curve (2) decrease slightly the convergence criteria to 1e-5 (3) in 2d analyses it is generally good to restrain the out-of-plane DOFs, though this is probably not the reason of the wrinkly curve you are getting

Seismosoft Support

[huffte]

Hasan,

I agree whole-heartedly to the comments from the SeismoSoft support team.

In the first place, I agree that the time step of 0.00021 is mush too small for this case given that the loading time step is 0.01 and the natural periods of interest upon running an eigenvalue analysis on your model would appear to be 0.21 and 0.07 seconds. Most of the response is likely in the first mode, which would require a time step of, say 0.21/10 = 0.02 seconds. Even if you wish to get accurate 2ndmode response, the time step would be, say 0.07/10 = 0.007 seconds. So if I were running this model, I would probably set a time step of 0.005 seconds and feel rather comfortable.

Additionally, I would say that the most erratic hysteresis was observed in the extremely lightly reinforced t-beams, so I would say that the erratic hysteresis is not necessarily unexpected.

I'll offer up the additional comment that I pasted your input accelerogram into SeismoSignal and find that is has a residual drift at the end of the record of approximately 2 meters. I would recommend that you baseline adjust / filter the records you use as input for analysis. Or else, use files from sources whose records have already been corrected to eliminate residual drift. I used SeismoSignal to baseline correct your record with no filtering and the end-of-record drift was essentially eliminated.

Best of luck Hasan.