nonlinear elements and fiber section

[ingLuca]

Hi all,
I would like to share an analysis about the behaviour of different kinds of nonlinear element.
I have read among posts that I'm not the first to find these problems, but I think it would be useful to the community to try to understand.

I can't post an image, but i can add the url to a graph wich represents my results.

http://byfiles.storage.msn.com/x1pN1mp8 ... ttM83HCv2A


The structure is a column, clamped to the bottom, 3 metres height, with a c.a. section of 0.2x0.4 m (the strong side in the direction of load), subjected to pushover analysis (Ux imposed at the top). The section is reinforced with 3 bars d=0.016 m on each side.

In the graph you can see the pushover curves obtained with OpenSEES and the one obtained with SAP.

The SAP analysis uses hinges, the OS ones different nonlinear elements:

nonlinearBeamColumn
dispBeamColumn
beamWithHinges

don't care of differences in elastic stiffness, it's due to the non-linearity of Concrete01, while SAP uses a linear elasticity before yielding. The problem is the strong difference in yielding moment for dispBeamColumn.

Has anyone experienced the same problem?
Does anyone know the reason of that?

The section is a fiber section, with Concrete01 and Steeel01, and the behaviour don't change changing fiber discetization and element points of integration.

Tell me something guys, developers or users

Luca

[silvia]

first of all, if you are interested in the yield moment, i wouldn't use concrete01, i'd use concrete02. The tensile strength of the concrete is important for that level of damage.


how many integration points are you using?

[ingLuca]

the graph you see was obtained with 10 integration points, but with also with 2 points the results are quite the same, 1% difference.

For the Concrete model, well, I'm afraid I didn't use the right term: I'm interested in the ultimate moment, not the one where the section stops being elastic linear.

I think concrete tension is not so important for my scope; otherwise, if I use the same material with 3 beams, I expect the same result, without thinking if material is the best one or not.

I tried, just to verify, and Concrete02 gives the same max moment for the dispBeamColumn, but with convergence problems wich make the curve sawtooth.

thanks to all for help

[silvia]

then something is wrong with your model, i believe.

[ingLuca]

Maybe I'm wrong, but where's the problem? I use the same script, with 3 command lines for the beam element, and in turn I put # before two lines to leave only one of them and run the same analysis. The section is the same, all the analysis parameter are the same. The beams lines are the subsequent:

Code: Select all

#element nonlinearBeamColumn 1 1 2 10 1 1
element dispBeamColumn 1 1 2 10 1 1
#element elasticBeamColumn 1 1 2 0.08 2.5e10 0.00107 1
#element beamWithHinges 1 1 2 1 0.5 1 0.5 2.5e10 0.08 0.00085 1

There is also an elastic beam I put in to simply test the problem at initial stiffness.

The parameters are the same for the beams, IDtag 1, nodes 1-2, 10 integration points, section 1, geometric transformation 1 (the problem is 2D).
The elastic parameters of the interior of beamWithHinges refer to an integer section of reinforced concrete.
The hinge lenght 0.5 is reasonable for a column depth of 0.4, and its results seem good.

What can I do?

[ingLuca]

please, nobody can help me?
It's a real problem for me, I have to understand what I'm doing wrong with nonlinear elements..

Well I'll be waiting..

Thank you all anyway, since I have subscirbed I found a lot of useful information.

Luca

[silvia]

can you please send me the input files with a detailed description of what you are doing at opensees-support @berkeley.edu

[ingLuca]

Sure
Let me the time to translate the comment lines in some acceptable manner (I write in Italian when I work )
I'll write you soon

Thanks a lot

[ingLuca]

I took some information at University, and they explained me that the problem is in the shape function.

a force based element use a force distribution as start point, and for a structure like this, it's very good, 'cause force distribution (bending moment) is simple

a displacement based element use a deformation distribution as start point, and this is more difficult than force distribution

They told me to try do subdivide the structure in more pieces, so that real deformation distribution is more accurately described.

Using 6 elements of 0.5 m of length the result of dispBeamColumn is quite the same of the one of nonlinearBeamColumn


I think now that is not more necessary I send you my files, but I really like to know what do you think about the things they said to me, do you think is this the problem?

There's something I can't understand: dispBased element starts from a deformation distribution then obtains a force distribution, right? If deformation is wrong force will be wrong.. Now, how forces come out from deformation? I think by using the constituive model (Force vs Deformation). Then, how is it possible that I get a force that section can't express? I tried also to use an Uniaxial Section based on Steel01 with softening: with this section-material there's no deformation wich can lead to a bending moment bigger than the limit one I imposed, but when I use dispBeamColumn it happens!

I'm a little confused...


Thank you very much for all your help, I really don't know how you all of the staff find the time to answer us

[silvia]

you can leave the comments in Italian, I am italian!

[silvia]

the displacement-based element is the typical element you use in an elastic analysis. because of its assumptions, yes, you have to define many, many elements along the length of a beam or column. The nonlinearbeamcolumn element (and the beam with hinges, which is just a modification of the nbc element), is a force-based element. With this element, you only need one element for each beam or column, with about 5 integration points (more than that may lead to numerical instabilities). In a strengthening system, you should get similar force-deformation results for all three elements. local deformations, such as curvatures, will be different, though.
When you have a softening system, negative tangent, the beam-with-hinges is the most reliable one.

[ingLuca]

Beh, lo avevo supposto che il nome non fosse molto inglese

Well, now that we have found the bug it's not necessary to send the files, but can you answer my last question? I would speak about that in the study I'm carring on:

How is it possible that at a section there is a bending moment bigger than the maximum that section can have? i understand the problems of wrong curvature in a section after the use of disp Based or force Based elements, but how can the program give me a force that is beyond the limit established by the section constitutive model?

Maybe that's not a matter of interest for users, but it's so for me, being a student.

Thanks

[silvia]

it is not possible, unless the axial force increases, which increases the moment strength.
something's up in the model.

[ingLuca]

Exactly wath I was thinking.. but:

1) I measured axial force, and it's not so large to explain the bending resistance increment
2) I noticed the same behaviour with Uniaxial section, without any connection between P and Mz.

Ok, I'll send you the model.

Luca

[silvia]

then, somethig's up with your model.