Hard to get convergence

kishor · Post by **kishor** » Sun Feb 08, 2009 5:45 pm

Dear Dr. Silvia and All,

I am modeling non linear steel roof diaphragm using Pinching4 material. We did test of two large scale steel roof diaphragms and this was Phase I. The model I built for these specimens matches well with the test data.Now we are going to test Phase II and Phase III. These specimens have higher shear strength and shear stiffness than Phase I specimens.

The problem now is, I am not getting convergence in the model for harmonic loading signals. However for seismic signals, the model converges and gives realistic values though at small time steps. For the harmonic signals, the model converges when the loading is in elastic range. I tried as given in example manual for convergence but could not get convergence in inelastic range for harmonic signal. how can I find problem and solution?
The model is relatively large to post here. I appreciate very much for your suggestions.

silvia · Post by **silvia** » Mon Feb 09, 2009 8:40 am

congratulations on the success thus far!
have you tried reducing the analysis step significantly?

kishor · Post by **kishor** » Mon Feb 09, 2009 11:18 am

Thank you very much!

Yes I tried reducing the analysis steps significantly. Like time steps of 0.0005 and even less. Previously, I thought the material definition parameters may have affected for convergence. But When I did analysis using seismic signals, it converge. Its very confusion for me that what is causing the problem for convergence.

silvia · Post by **silvia** » Mon Feb 09, 2009 3:53 pm

are the magnitudes that much larger???
check the response at thepoint of convergence problems.

kishor · Post by **kishor** » Tue Feb 10, 2009 6:52 am

Thank you!

The model converge at 0.5 amplification factor for the harmoni signal, the response is about 50% of elastic limit. When I use higher than this amplification factor say 0.6, it does not converge for which I belive that the response will be with in elastic limit. So I belive that it is not due to high amlitude loading.
The model does not converge at the very begining of time steps and analysis shows as below where first value is fundament time period.

0.1031829847132103
WARNING: CTestEnergyIncr::test() - failed to converge
after: 20 iterations
NewtnRaphson::solveCurrentStep() -the ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 0.09
OpenSees > analyze failed, returned: -3 error flag

Node: 253
Coordinates : 0 3656
Disps: -0.000976299 -0.0115338
Velocities : -0.0380972 0.0410393
commitAccels: 6661.63 328.703
unbalanced Load: 0 -0.00785326
Mass :
0 0
0 4.61e-005

Eigenvectors:
0.132387
0.898155

ID : 267 268

Node: 53
Coordinates : 0 3656
Disps: -0.000976299 -0.0115338 -2.36448e-006
Velocities : -0.0380972 0.0410393 7.00318e-005
commitAccels: 6661.63 328.703 9.28003
unbalanced Load: 0 0 0
Eigenvectors:
0.132387
0.898155
0.000274073

On the other hand, It converge well for seismic signal in high inelastic range.

silvia · Post by **silvia** » Tue Feb 10, 2009 8:35 am

interesting. and you did use the convergence scripts from the examples manuals and nothing????

check the model response locally.

kishor · Post by **kishor** » Tue Feb 10, 2009 9:06 am

Thank you!

Yes I did check with the use of the convergence scripts from the examples but I did not get any imporvements.

The model converge at 0.5 amplification factor for the harmoni signal, the response at local element, where maximum force is expected, is about 50% of elastic limit. When I use higher than this amplification factor say 0.6, it does not converge for which I belive that the response will be with in elastic limit.

Than you for your patient.

silvia · Post by **silvia** » Tue Feb 10, 2009 10:01 am

check which elements and sections are giving you problems.

kishor · Post by **kishor** » Fri Feb 13, 2009 7:53 am

Thanks
Using the scripts avialable in example manual, the analysis shows like below:

.
..
WARNING: CTestEnergyIncr::test() - failed to converge
after: 10 iterations
ModifiedNewton::solveCurrentStep() -the ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 0.78
OpenSees > analyze failed, returned: -3 error flag
WARNING: CTestEnergyIncr::test() - failed to converge
after: 10 iterations
ModifiedNewton::solveCurrentStep() -the ConvergenceTest object failed in test()

.
...
..

Trying Newton with Initial Tangent ..
WARNING: CTestEnergyIncr::test() - failed to converge
after: 10 iterations
ModifiedNewton::solveCurrentStep() -the ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 9.98
OpenSees > analyze failed, returned: -3 error flag
WARNING: CTestEnergyIncr::test() - failed to converge
after: 10 iterations
ModifiedNewton::solveCurrentStep() -the ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 9.98
OpenSees > analyze failed, returned: -3 error flag
Trying Newton with Initial Tangent ..
WARNING: CTestEnergyIncr::test() - failed to converge
after: 10 iterations
ModifiedNewton::solveCurrentStep() -the ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 9.99
OpenSees > analyze failed, returned: -3 error flag
WARNING: CTestEnergyIncr::test() - failed to converge
after: 10 iterations
ModifiedNewton::solveCurrentStep() -the ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 9.99
OpenSees > analyze failed, returned: -3 error flag
Trying Newton with Initial Tangent ..
WARNING: CTestEnergyIncr::test() - failed to converge
after: 10 iterations
ModifiedNewton::solveCurrentStep() -the ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 10
OpenSees > analyze failed, returned: -3 error flag
WARNING: CTestEnergyIncr::test() - failed to converge
after: 10 iterations
ModifiedNewton::solveCurrentStep() -the ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 10
OpenSees > analyze failed, returned: -3 error flag
Trying Newton with Initial Tangent ..
WARNING: CTestEnergyIncr::test() - failed to converge
after: 10 iterations
ModifiedNewton::solveCurrentStep() -the ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 10.01
OpenSees > analyze failed, returned: -3 error flag
WARNING: CTestEnergyIncr::test() - failed to converge
after: 10 iterations
ModifiedNewton::solveCurrentStep() -the ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 10.01
OpenSees > analyze failed, returned: -3 error flag
Trying Newton with Initial Tangent ..
Ground Motion Done. End Time: 10.010000
Done!

Does it affetcs the results with too many convergence fail..?
why the End time is 10.01 though I specified max time of 10.0 Sec?

It seesm the results are realistic and the scripts worked. I used to closed the opensees program at middle once I got those type of conversence failure. Sorry for that.

silvia · Post by **silvia** » Fri Feb 13, 2009 9:37 am

congratulations, that worked.
the end time depends on the time steps, likely, a time step was reduced, so couldn't hit 10.0 exactly.

kishor · Post by **kishor** » Fri Feb 13, 2009 10:38 am

Thanks a lot!!

silvia · Post by **silvia** » Fri Feb 13, 2009 10:52 am

these days i add a puts command in my script telling me that it managed to converge at the steps where it hadn't converged right away.