Elastic-no tension material

[linguan118]

Dear all,
I want to model an FRP tube-confined concrete column under cyclic lateral loading.
The FRP tubes bear axial load to some extent, but can not stand tension force.
I'm using fiber-section beam-column element; ENT material is assigned for the FRP tube.
However, it is very hard to converge.
When I use Reinforcing steel model, it says "bad trial strain"; when steel02 is used following errors appear:

WARNING SuperLU::solve(void)- Error 1 returned in factorization dgstrf()
WARNING NewtonRaphson::solveCurrentStep() -the LinearSysOfEqn failed in solve()
StaticAnalysis::analyze() - the Algorithm failed at iteration: 124 with domain at load factor 2.29247e+008
OpenSees > analyze failed, returned: -3 error flag

When I shift from ENT to Elastic material, it runs successfully.

I don't know why. Since the sectional response is integrated by the fiber behavior; ENT only affects tension side of the section, how does it affect all the section response?

Thank you in advance!

[fmk]

the steel02 is because the system is becoming unstable, suggest using Newton with initial stiffness iterations.

[linguan118]

Thank you fmk for your suggestion.
However, Newton with initial stiffness also fails. Actually, all the reinforcing steel models fail.
When ReinforcingSteel was used following errors appear:
bad trial strain
UniaxialMaterial::setTrial() - material failed in setTrialStrain()

I checked the SR code of ReinforcingSteel
00164 #ifdef _WIN32
00165 if(_fpclass(strain)< 8 || _fpclass(strain)==512) {
00166 opserr << "bad trial strain\n";
00167 return -1;
00168 }
00169 #endif

_fpclass returns the status word containing information about the floating-point class.

When steel01 was used:
WARNING - ForceBeamColumn2d::update - failed to get compatible element forces & deformations for element: 1(dW: << 1.#QNAN)
Domain::update - domain failed in update

I think QNAN means Not A Number.

I don't why these strange errors appear. It seems that there is some number error.

[fbeckwit]

linguan,

Your load factor is very large from the first error message you reported. I'm assuming you're using the displacement control method. I had a similar problem in one of my models which was a result of forgetting to redefine a load pattern after setting the axial load constant. If this is also the case in your model, make sure you define a new load pattern with some reference load at your control DOF. Otherwise, you'll end up with a situation where the axial load is scaled with the pseudotime and the model will be very difficult to converge.

[linguan118]

fbeckwit, thank you for your kind reply. The axial load is that large because I used SI(N,mm) unit.
Actually I am just about to post that I've solved the problem by changing the unit from SI to English unit.
I don't know why, maybe because the value of load in SI unit is too large causing some value errors, since all the examples in Opensees are in English unit.
Anyone gets the reason?

[fbeckwit]

It's very possible that you encountered roundoff error associated with machine computing. When numbers get very large or very small, the computer will keep a specified number of significant digits and toss out anything behind it, which would very well cause convergence issues.

[linguan118]

Thank you fbeckwit for your help!