Confusing error message

[mpnewcombe]

Hi there,

I get a whole bunch of:

WARNING BandGenLinLapackSolver: : solve <> - LAPACK routine returned 6
WARNING ModifiedNewton: : solveCurrentStep<> -the LinearSysOfEqn failed in solve<>
.... Algorithm failed at time 0.01
... -3 error flag...
wrong # args: should be "set varName ?newValue?"

..etc etc.

Does this simply mean I have stuffed up my fixity/constraints? I have been checking for hours now and can't find an error.

Regards,
Mike Newcombe

[fmk]

the algorithm is ailing because the solver is failing .. so yes there is a problem with the model and if this is the first step, the boundary conditions are more than likely the problem.

[mpnewcombe]

Great, thank you. I will keep looking

[mpnewcombe]

I have narrowed down the problem.

I am trying to use Elastic MultiLinear rotational springs. If I replace the Elastic MultiLinear material with a Elastic material the analysis runs fine... hence the problem seems to be my material definision.

Please see the key extracts below:

uniaxialMaterial ElasticMultiLinear 15 -strain -0.024849513 -0.013475006 -0.004199947 -0.000524993 0 0.000524993 0.004199947 0.013475006 0.024849513 -stress -1.63E+08 -1.58E+08 -1.49E+08 -1.40E+08 0 1.40E+08 1.49E+08 1.58E+08 1.63E+08 ;

element zeroLength 193 1 7 -mat 15 -dir 6

Please note that I have constrained the node 1 and 7 in x and y.

Any ideas, I am so close...

Regards,
Mike Newcombe

[fmk]

do you have a recorder on the material to see the stress-strain history.

from what you have the tangent from this material should always be positive and as a consequence not the direct cause of the error you are seeing. i am guessing it is the material properties that is causing some other materials to go negative and that is the source of the error message.

[mpnewcombe]

Yes... Also I have changed to the Krylov-Newton algorithum, which seems to be much more stable. I manage to get about 7s of run time before it crashes. The recorder data for the moment-rotation response looks good... it achieved 3 points of stiffness change, but not the fourth... I am unsure that that is the problem however.

My understanding is that after the stiffness of the final segment is extrapolated beyond the final stress-strain point; has there ever been issues once that point is reached?

Many thanks for all your help with this,

Mike N

[mpnewcombe]

I just got. I traced it back to a MultiLinear Elastic rotational spring that simulates column-base connections in my frame model. These have a significantly higher initial stiffness than the beam-column connections, and a sharp transition from the initial stiffness to the second stiffness. This transition appears to have caused the numerical instability...

[yas]

WARNING BandGenLinLapackSolver::solve() -LAPACK routine returned 115

I ,ve checked all of the boundary conditions and the material but I couldn,t find any problem for the reason of this error.what can I do.would you please help me.thanks in advance.

[fmk]

the lapack solver is failing .. you could try another solver and get a solution, but the real problem is that this error message shows is that the matrix is probably poorly conditioned. that second stiffness may be too low or the initial one too high

[yas]

hello dear all

WARNING BandGenLinLapackSolver::solve() -LAPACK routine returned 1
does the number of this warning at the end of that have specified meaning?

thanks

[fmk]

your matrix is singlular, i.e. rigid body mode or 0 stiffness. here is a link to the doc for the routine we are using. the number returned is info variable.

http://www.netlib.org/lapack/explore-ht ... solve.html