This is crazy, problem with nonlinear beam

[toggi007]

Hi Vesna, would appreciate if you can take a look at above question?

best regards

[vesna]

Why don't you try smaller steps in your pushover analysis.

[toggi007]

Hi, you are absolutely right in thinking this, but I have also tried putting extremely small steps... However I think most of my Opensees models are working as they should. And I want to thank you for all your efforts to reply to my post.
On another note, I ran into a problem when setting up a code to emulate tension tests on steel bars. Basicly using steel02 material and fiber section with just the single straight layer, then transformation and nonlinearbeamcol element between two point, on fixed the other free. Then applying axial load in direction of tension but the simulation stops at first load step?
Any ideas? much appreciated

[vesna]

why don't you post your code so I can take a look.

[toggi007]

I have fixed the problem above, seems that when one problem is fixed another one rises and this is very strange behaviour in dispbeamcolumn element:

I ran into some strange behaviour in the dispbeamcolumn element the other day. I have created a model in matlab based on curvature and linear strain distribution through the crosssection, results should be similar to the what an opensees model with concrete01 and same steel01,( In my code my concrete model is bilinear (Eurocode 2) but my steel model has strainhardening), when comparing results for one beam I find that the curvature responce and the displacement are almost identical up to a certain point...seemingly opensees goes a lot further! My model stops when concrete strains reach 0.0035 but the funny thing is that opensees does the same thing, only at too great curvature and thus too great displacement. Then I tried the nonlinearbeamcolumn element and it stops at pretty much the same point as my little nonlinear program. So my question is: Why do the element recorder for both the displacement and the recorder for the moment-curvature in momentcurvature.tcl (slightly modified by taking out the axial load) show what are obviously incorrect stresses, kurvature & displacements for the dispbeamcolumn element?

[toggi007]

looks to me that as the dispbeamcolumn is based on the moment-curvature and linear strain distribution, then the solution algorithm converges to higher curvature though there are some ...."faults" in the Moment-curvature curves which I could not see when using concrete01, only when I use more complex stress-strain models for the concrete (concrete01 seems not to stop at crushing point?). So when displacements are recorded they are recorded for higher curvatures, beyond the "faults" in the moment-curvature curve... thus larger displacements. But do you know why the stress-strain recorder shows stresses that seemingly are stretched over the process? if you catch my drift?

[vesna]

Displacement based element is accurate only in the linear range. In order to capture nonlinear behavior of an element you need to model it with several displacement based elements. This may help you understand why: http://opensees.berkeley.edu/wiki/index ... ed_Element

[toggi007]

I am using eleven elements usually. There is a fault in the program see above post, recorded values of stress and strain are wrong. This is because when the program uses the curvature to get displacements (just the area under the curve on K-axis), then in places under the M-K graph where there seems not to be convergence but still the program runs through it...to higher curvature, then when it comes down to getting the nonlinear displacements they will be to large.