PDelta geomTransf on leaning columns

ca493 · Post by **ca493** » Tue May 13, 2008 10:58 am

Hello,

Are there any known issues of using the PDelta geomTransf on leaning columns that connected to the main frain via rigid links?

The PDelta and Corotational geomTransf work fine on the rest of the frame except the leaning columns (the Linear geomTransf, however, worked fine on the leaning cols).

I have tried elastic leaning cols, nonlinear, and beamWith Hinges; all of which have convergence issues with PDelta.

The leaning cols are pin supported and there is one on each side of the main frame.

Any suggestions?

Thanks!

ca493 · Post by **ca493** » Tue May 13, 2008 11:50 am

and i have also tried linking the leaning columns to the main frame using equalDOF but there are still convergence issues.

silvia · Post by **silvia** » Tue May 13, 2008 12:29 pm

interesting problem.
you might want to try reducing step size as a first measure.

ca493 · Post by **ca493** » Thu May 15, 2008 10:01 am

Hi Silvia, thanks for the help! The solution to the problem was actually due to a silly typo i had in the application of the gravity loads so i think it works just fine.

However, i did run into another error. I am modeling buckling restrained braces and giving them a Corotational geomTransf. They work just fine in a static pushover, however, i am observing some very odd behavior during dynamic analysis that does not occur when using a Linear geomTransf for them.

They BRBs are modeled such that there are 5 nonlinearbeamcolumn elements connected in series. The middle element is the smaller area core of the BRB which is meant to yield while the outer 2 elements on each side have larger areas and wont typically yield. The ends of the BRB are given a moment release.

The problem is that the BRBs start to develop large moment and shear during the dynamic analysis. The bending occurs between the yielding core element and the adjacent non-yielding elements.

Is there a possible reason why this happens for the Corotational (and also PDelta) and not the Linear geomTransf?

thank you very much!

pejman_opensees · Post by **pejman_opensees** » Thu May 15, 2008 10:21 am

I do not think its the matter of corotational or pdeltatransformation. Basically force-beam-column elements are so sophisticated but a very complicated nonlinear procedure has been implemented in their formulation. So many local and element-based iterations are performed intrinsically and possible non-convergence or inaccuracies may take place if iterations are not enough or tolerance is too tight. These problems do not encounter normally in static analyses but is possible in dynamic oscilatory analyses ( such as IDA or..) and attention should be paid working with FBCs. In order to find out what the real problem is, I recpmmend you that use displacement Beam column instead and see if there is any change in results or not. But you should increase the number of elements in order to get the same accuracy as FBCs.

ca493 · Post by **ca493** » Fri May 23, 2008 11:47 am

Hello, thank you very much for your informative response. After some adjustments to the frame, I am still running into the same problems where the BRB develops significant bending.

I am under the impression that it is a continuity issue where the nodes that are defined along the length of the diagonal BRB are not 100% aligned. Although I used sines and cosines to do this, there may be some rounding error. Considering that the Corotational transf considers p-del effects, these very small offsets in the BRB nodes become amplified at very large demands and eventually develop noticeable bending, which seems to be what i am observing.

I am still curious as to why this only occurs in dynamic analysis and not static; however, for now i will keep the BRBs with a linear transformation.