Scale down your earthquake by 90% or even more to reach a weak quake in which your structure would remain mainly elastic. Then run the analysis. Probably your analysis would be completed with no warning. If so, 

A) Change your convergence test to Norm Disp with maximum number of iteration of 100 and a reduced convergence tolerance (tol) of about 1e-5.

B) Increase integration points of the force-based element

Run the analysis with your real scale strong earthquake. Probably your analysis would be completed this time with no warning. But if your piers were so weak and near collapse, you might need to do the following modification as well,

C) Use more nodes along the Force-Based BeamColumn elements. This latter modification is seldom necessary expect for concentric braces. So your problem would be solved by considering items A and B.

Dear Amin:

Thank you for your response.

I tried with all your proposals, but the model keeps throwing bug.

I Changed the convergence test to Norm Disp with maximum number of iteration of 100 and I reduced convergence tolerance (tol) to 1e-5.

I increased the integration points of the force-based element, from 3 to 10.

I used more nodes along the Force-Based BeamColumn elements. I doubled the points number in columns.

This is the wrong message:

WARNING - ForceBeamColumn3d::update - failed to get compatible element force & deformations for element: 19

If you want, I send an email with the code and the accelerogram.

thank you in advance

Hello Javier,

You may wish to try the Krylov Newton algorithm. Is has been shown to successfully reach convergence on problems where Newton and Modified Newton algorithms have failed. It is also generally less computationally expensive than both the aforementioned algorithms, and requires less computational time. It prevents the flip-flop when converging on the tangent stiffness ("A Krylov Subspace Accelerated Newton Algorithm" by Scott and Fenves, 2003). 

Best of luck.

Did your model worked for a reduced earthquake? If you received the same warning even for a very weak earthquake, or even gravity load, your problem stems from somewhere else. Check boundary conditions, do an eigen value analysis and verify that obtained periods make sense. 

I've same problem

I trying to modeling cbf with initial camber

when earthquake load applied, get this error

WARNING UmfpackGenLinSolver::solve(void)- 4 returned in factorization UMD2FA()

WARNING NewtonRaphson::solveCurrentStep() -the LinearSysOfEqn failed in solve()

DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 0.01

LegendreBeamIntegration -- max # integration points is 10

LegendreBeamIntegration -- max # integration points is 10

OpenSees > analyze failed, returned: -3 error flag

even eigen value show this error

FullGenEigenSolver::solve() - the LAPACK dggev routine returned error code 13

WARNING DirectIntegrationAnalysis::eigen() - EigenSOE failed in solve()

Node::getEigenvectors() - eigenvectors have not been set

if somebody found any solution for this please remind me

best regards 

Hi Vahid Nazari:

I'm so sorry but I doesn't work with steel members.

Do you read: http://opensees.berkeley.edu/community/viewtopic.php?f=2&t=60681    ?

You can read 

http://peer.berkeley.edu/publications/peer_reports/reports_2008/web_PEER808_URIZMahin.pdf     too.

Best of luck

Dear Mohammad: 

I resolved the problem using krilov newton algorithm and working with a reduced earthquake.

Best wishes

Hi dear Javier

I had the same problem, I solved it by using "dispBeamColumn" command for my nonlinear elements.

Now another problem ha risen which is convergence problem which is not solve by either smaller time steps nor enlarging convergence tolerance and iteration numbers.

Thanks in advance

Hi,

In my opinion you shoud use OpenSees forum for your questions. In any case, let me try to help you based on my experience:

1. Increasing the integration section number does not increase the speed of convergence. It increases the accuracy and likely to decrease the convergence. I usually use 5 integration sections in forced based elements.

2. Too strict tolerance values may lead to convergence difficulties. On the other hand, too loose ones may not be enough to let the model behave correctly. Tolerence is not a definitive value, it depends on the type of convergence check and your model. Usually, (at least for me) tol=1e-6 to 1e-5 are good values for NormDispIncr. On the other hand, energy and unbalanced load criteria are also good, but may lead to convergence difficulties in the models having exceptionally stiff elements.

3. Decreasing dt usually works in slowly converging models.

4. Increase the number of iterations as max as possible. What I use is max 1000, if it takes more than 1000 iterations, the model is likely to be ill conditioned.

5. There is one more control in fiber section forced based elements which is the number of iterations carried out at integration sections, it could be invoked as "-iter":

i.e. element nonlinearBeamColumn 160 88 105 5 996 19 -iter $niter $tol;

default value of niter is 10, where as for tol, it is 1e-12, in my opinion one should not change the default value of tol. But especially for short length forced based elements, you may try to increase niter.

Apart from those, there are many other parameters controlling convergence. My advice would be first run the model with linear properties, to at least rule out any problem emerging from nonlinear response.

Again, those are only my opinions. You should really discuss this in OpenSees forum.

TRy to apply the answer of Ali Guney Ozcebe . I think you will solve it.

Hi

you should change solution algorithm object

The following contain information about algorithm type?

Linear Algorithm

Newton Algorithm

Newton with Line Search Algorithm

Modified Newton Algorithm

Krylov-Newton Algorithm

Secant Newton Algorithm

BFGS Algorithm

Broyden Algorithm

Hello ,

How can I define the stress, strain parameter for equivalent strut to define the concrete01 in opensees? 

Agree with Taylor C. Steele ........

Please share me the best answer might you trust...

Regards…

I faced the same problem but when I used "dispBeamColumn" elements the problem solved.