Topic: Problem of deltaT in nldeidynamic simulation

Dear all,

I'm working on one none linear dynamic analysis.

                     nldeidynamic nsteps 400000 dumpcoef 0 deltat 0.00002 nmodules 1

In the output, following sentence came out:
                     
                     deltaT reduced to 8.514585e-10, Tmin is 2.674936e-09

My question is should I have to adjust the deltaT to 8.514585e-10, so that I can get reasonable output?
How should I select the deltaT value?

Appreicate your help.
Have a good weekend.

Re: Problem of deltaT in nldeidynamic simulation

Dear Vigoss,
the analysis you're performing is a non-linear dynamic with explicit integration. This integration scheme leads to extremely small time increments, expecially if dumpcoef is zero. Damping is often used to stabilize such kind of analyses.

You should select the initial deltaT in a way that deltaT * numberOfIncrements = final desired time (e.g. the final time of a ground motion).

To avoid the explicit integration scheme, try to use NonLinearDynamic (unfortunately it is not documented in OOFEM manual).

Regards,
Giovanni

3

Re: Problem of deltaT in nldeidynamic simulation

HI, the explicit integration is only conditionally stable, the time increment should  be less than Tn/Pi, where Tn is the smallest vibration period. Typically, you can only estimate the value of Tn.
What oofem is doing: it estimates the Tn, based on diagonal entiries of stifness and mass matrices and adjusts the time increment, if necessary. However, this is only the estimate of Tn, the actual one can be smaller, so its safe to adjust the time increment to a value below the recommended one.

Re: Problem of deltaT in nldeidynamic simulation

In the manual page for DIIDynamic here, the damping coeffs are eta and delta, not alpha and beta as written.