Force Based Element - Base Shear "Spikes"

[vsaggelidakis]

Hello to everyone!

My questions are:
1. Can too many FBE inside the plastic hinge length cause Numerical Instabilities?
2. Aren't FBE way more efficient than DBE? (I've read Vesna's presentation more than once!)

My case (Not for the Faint-Hearted):
I study the Inelastic Seismic Response of a Concrete Bridge (spans: 75-120-75 meters) with two Tall Piers (42 & 50 meters) and I model each Pier with 15 Force Based Elements. Although so many FBE are not needed, I want to have a satisfactory mass distribution along the Piers' height to investigate the effect of higher frequency modes implementing Non-Linear Dynamic Analyses.

While the Moment-Curvature Hysteretic Loops seem logical, when I plot the Shear Force for any element of the Piers to the analysis time, I get extreme Spikes that seem Unrealistic. Same for the Base Shear. When I model the Piers with the same number of DBE, I don't get these spikes, so I guess they do not happen due to higher frequency oscillations, but by the spread of plasticity along the Piers' Height. I tried changing almost every other parameter, from the test and tolerance, to the solution algorithm, the Time Step, the Constraints' Handler and the Number of Integration Points. So, could too many FBE along a member harm my results?

Sorry for the long post.
Thank you for your support (and your patience)!

[fmk]

what convergence test are you using? if not NormUnbalance try that one to ensure you are actually in equi and these spikes are not just due to unbalance in the resisting force

[vsaggelidakis]

Professor thank you for your reply!

I was using NormDispIncr so far (EnergyIncr gave me the same results) with Transformation constraints and Newmark integrator (γ=0.5 & β=0.25).
NormUnbalance gives my analyses a really hard time to converge compared to the tests I was using, so I think you have spotted my problem.
What would be an appropriate value of Tolerance for this test? When using a tolerance of [1.0e-8 to 1.0e-2] with the latter seeming too large, I can't
achieve convergence. With tolerance larger than 1.0e-2 I get convergence, but with the same results as before. Any inputs on about why I have this unbalance or how I can escape this problem?

Thank you very much!

[fmk]

look at the corresponding displacement increment norms as well .. put a 2 flag at the end of the convergence test .. the 1.0e-2 might be all right as disp increments could be 1e-10 and the convergence issue is due to v.v. small displ increments .. i like the norm unblance or energt increment in transient analysis as sometmies the unbalance is still larhe eve for v small disp