Page 1 of 2

Very different results when modeling with different element classes.

Posted: 09 Jan 2020, 15:29
by Koloubos
Hello, I am modeling a multi-column frame using the infrmFB and infrmFBPH element type. There is no other difference in terms of geometry, materials and reinforcement between my two models. I am conducting a static push-over analysis and there is a huge difference between both the behavior and the ultimate displacements of the frames(almost double in the infrmFBPH element case). I expected a difference but not such a big one. Is there an explanation?

Re: Very different results when modeling with different element classes.

Posted: 09 Jan 2020, 17:52
by huffte
The infrmFBPH results are highly dependent upon the specified plastic hinge length.

Note the following from the SeismoStruct Help System regarding infrmFBPH elements:

'The number of section fibres used in equilibrium computations carried out at the element's end sections needs to be defined. . . only a sensitivity study carried out by the user on a case-by-case basis can unequivocally establish the optimum number of section fibres.'

Similar notes, as well as the following, are provided in the Help System regarding the infrmFB element:

'As discussed in Material Inelasticity, element infrmFB is the most accurate among the four frame element types of SeismoStruct, since it is capable of capturing the inelastic behaviour along the entire length of a structural member, even when employing a single element per member. Hence, its use allows for very high accuracy in the analytical results, while giving users the possibility of readily employing element chord-rotations output for seismic code verifications'

Re: Very different results when modeling with different element classes.

Posted: 10 Jan 2020, 10:30
by seismosoft
Note that for the default infrmFB and infrmFBPH parameters, the results should not be very different. However, if you make significant changes to the plastic hinge length, the differences are expected to be high, as huffte very correctly pointed out.
Seismosoft Support

Re: Very different results when modeling with different element classes.

Posted: 10 Jan 2020, 20:45
by Koloubos
Thank you for your responses. These extreme differences that i am referring to are with the default 16,67% plastic hinge lenght. When I lower the plastic hinge percentage to 5-6%, my results are similar and comparable. A a rule of thumb, the bigger the plastic hinge percentage the bigger the ultimate displacement of my structure. Even with KA.NE.PE suggestion of the plastic hinge length, which is close to 10% for my case, the results are still very different.

Re: Very different results when modeling with different element classes.

Posted: 11 Jan 2020, 10:39
by seismosoft
This is not reasonable. Are you sure that all the other parameters are the same? How many integration sections are you using with the infrmFB element type? Can you change the element type using the default parameters in each case and see if you get the similar results?
Seismosoft Support

Re: Very different results when modeling with different element classes.

Posted: 11 Jan 2020, 15:37
by Koloubos
All the other parameters (geometry, materials, sections, loads, fibers used) are identical. I use 5 integration points for the infrmFB element. I already did that but the anomaly remains.

Re: Very different results when modeling with different element classes.

Posted: 11 Jan 2020, 16:48
by seismosoft
This is strange. Can you please send the two files to support@seismosoft.com, so that we check the models?
Thanks.

Re: Very different results when modeling with different element classes.

Posted: 11 Jan 2020, 17:56
by Koloubos
I sent you my models for examination, thank you for your time!

Re: Very different results when modeling with different element classes.

Posted: 12 Jan 2020, 14:37
by seismosoft
The results are by no means 'very different'. The response of the two models is similar, within acceptable limits, e.g. the difference in the yield strength of the models is approximately 6%.
The main difference in the models is the fact that in the infrmFB model that 10% strain level (where fracture of the steel occurs, according to your input parameters) is reached earlier with respect to infrmFBPH. However, this is a known issue in fiber models, attributed to localisation issues.For example refer to: Localization issues in force-based frame elements [2001], J Coleman1 and Enrico Spacone.
Seismosoft Support

Re: Very different results when modeling with different element classes.

Posted: 13 Jan 2020, 10:46
by Koloubos
Thank you for your response. Is there a solution to this issue that you can recommend?