SAP2000 vs. OpenSees Pushover Curve

[Luisa]

Dear all,

I have modeled identical RC frames in SAP2000 and OpenSees and performed pushover analysis on both models. Unfortunately, the pushover curves obtained from these software don't match together in "nonlinear sections". I firstly thought the reason might be because of inherent differences between SAP2000 and OpenSees in nonlinear behavior modeling (since SAP2000 applies concentrated plastic hinges while in OpenSees I implement nonlinearBeamColumn elements which simulate distributed plasticity). However, the results are completely different and I think there might be a problem! I have checked my models several times, and really don't know what the cause of this difference is. I wonder if you could provide any help. Here are the graphs, and thanks a lot in advance.

http://www.4shared.com/photo/bWTcZ9Qyce ... e__1_.html

[andrett7]

Hi Luisa,

such huge difference is quite strange, even if the nonlinear behavior is different. Try to include in OpenSees a concentrated plasticity model (beamWithHinges element) for a better comparison.

Andrea

[Luisa]

Thanks for your suggestion. I used different elements in OpenSees. Not only non of them match to SAP2000 results, but also they differ with each other as well! The only thing I did was changing elements. This is worth mentioning that number of int. points in both force-based and disp-based elements are seven. I wonder if you could tell me what the cause of such a big difference is. Here are the graphs, and thanks a lot.

http://www.4shared.com/photo/HQAR17p5ce ... s__2_.html

[andrett7]

When you consider Force-based element, do you use one element (and at least 5 integration points) or more? It's ok to use only 1 finite element, while for Displacement-based element the key parameter is the mesh (so each element should be divided into more elements). Consequently, the results should be close one to each other.

Andrea

[Luisa]

What about other graphs?! Why non of them fit together? Forget about dispBeamColumn, how do you explain the others?

[andrett7]

Actually I cannot explain such huge difference, looks for example to this case study:

http://opensees.berkeley.edu/wiki/index ... ment_Frame

Andrea

[Luisa]

Thanks anyway dear Andrea. I had seen the link you have posted before. I really don't understand what the problem is with my model. I also modeled the frame in ETABS and the output completely matched with those from SAP2000. I don't know why OpenSees doesn't adapt with these two software in my case. Another issue which is strange to me is that changing elements in OpenSees results in a vast difference in pushover curves as well. After checking my model again and again ... I'm seriously getting frustrated! I hope Prof. Scott or Dr. McKenna provides a clue for my problem!

[andrett7]

Yes,

their response could be very useful.

[fmk]

The results are very different because the formulations are different (SAP and ETABS being basically the same - they are produced by the same company!) and you are modeling a strength-degrading system. As you played with the different element types and number of integration points within even OpenSees for your reinforced concrete structure while driving yourself insane you are observing the sensitivity of the solution to the details of the discretization in strength degrading/strain softening systems. Such systems do not converge to a solution by say making the discretization finer. Instead it requires you the user to understand that there will be a problem and to account for that in your modeling if you ever hope to get an answer that is a reasonable prediction of what is going to happen in a physical system. Now that you have seen the sensitivity it is time to step back and do a little research and educate yourself some more before you can begin to evaluate the results critically. Sadly few people actually understand the pitfalls of doing nonlinear analysis.

I suggest you read up on the matter. You could start looking at the documentation given for the force beam column element
http://opensees.berkeley.edu/wiki/index ... es_Element paricularly the first paper by Scott and Fenves (Plastic Hinge Integration Methods for Force-Based Beam–Column Elements) in which they talk the consequences of the loss of objectivity in modeling. Googlling the subject will brinig more than a few scholarly articles on the subject especially in the area of continuum finite elements.

[Luisa]

Thank you soooo... much dear Dr. McKenna! I greatly appreciate your generosity in supporting students like me.

[Luisa]

Dear Dr. McKenna,

Hi again!

As I understood, I should make my analytical models in an impeccable manner and according to accurate prediction of what is going to take place in a real physical system. In my M.Sc. thesis, however, I have 3D Reinforced Concrete Moment Frame models (6- 9- 12- and 15-story buildings indeed) and unfortunately I don't have any idea about their true behavior in real circumstances. I know this is not an OpenSees question somehow, but could you tell me please which of the experimental works (researches, reports, theses, ...) fulfilled on systems such as mine are highly reliable and include enough information for verification of my analyses results? Actually I need them as a reference to my study in order to gain a better understanding of the real behavior of my models, so that I can better decide on how to make my analytical models in OpenSees (elements, NIPs, ...).

Many thanks for your invaluable help in advance.

[arbarbosa]

For large structures, since there are only a few large-scale tests that have been done, the validation of the numerical response should be done at a component level. Then you should get back to some engineering fundamentals to check global and local responses.

For component validation, there is a lot of data out there. NEEShub has some data, PEER has many reports, and there are things in the literature for most common cases and failure modes.

I can provide more specifics off this post. Let me know.

Andre
--
http://web.engr.oregonstate.edu/~barbosa/


Luisa wrote:
> Dear Dr. McKenna,
>
> Hi again!
>
> As I understood, I should make my analytical models in an impeccable manner
> and according to accurate prediction of what is going to take place in a
> real physical system. In my M.Sc. thesis, however, I have 3D Reinforced
> Concrete Moment Frame models (6- 9- 12- and 15-story buildings indeed) and
> unfortunately I don't have any idea about their true behavior in real
> circumstances. I know this is not an OpenSees question somehow, but could
> you tell me please which of the experimental works (researches, reports,
> theses, ...) fulfilled on systems such as mine are highly reliable and
> include enough information for verification of my analyses results?
> Actually I need them as a reference to my study in order to gain a better
> understanding of the real behavior of my models, so that I can better
> decide on how to make my analytical models in OpenSees (elements, NIPs,
> ...).
>
> Many thanks for your invaluable help in advance.

[Luisa]

Dear Dr. Barbosa,

I greatly appreciate your consideration. I have sent you an email whose topic is "OpenSees Model Verification". Again, thank you so much for your help.

Warm regards