Hi,
I'm a new user, so sorry for my not-so-professional-questions!
And sorry for my not-so-good English, I’m Italian!
I've noticed some little differences in the behaviour of the elements depending
by the number of integration points choosen.
Increasing number of int.points the structure should reach results more accurate, is it so?
However with 6-7 noIP the structure (a simple oscillator) shows decay on the accuracy of results, and the convergence has reached with difficulty.
Does it exist an ideal noIP?
And does it exist an ideal number of subdivisions of the fiber section?
Thank you!
Number of Int.points
Moderators: silvia, selimgunay, Moderators
np
I also have an issue with the number of integration points specified in an element. If I specify at or above a certain number of integration points, my OpenSees interpreter will either freeze up or close out immediately. Is this a quirk of the program, or could it be a problem with my own computer? Thanks.
yes, there is an optimum number of integration points, as the objective of the integration points is to give you an accurate approximation of the curvature distribution -- you might want to study-up on Gauss Quadrature and Integration.
The number, of course, depends on what you are trying to model.
Typically, however, 4-5 integration points is the optimum place to start. Less give you a less accurate curve, more may give you numerical instability as the program is trying to figure out what you are doing.
There are issues about whether Gauss quadrature represents the actual curvature distribution you want, even though it is able to represent the actual deformation you are trying to calculate. Hence, there is the beam-with-Hinges element (with hinges at the ends, so only for column elements, and elements with no distributed tranverse loads). Supposedly, this gives you a better model of the plastic hinge length. This is a new element, and a paper on it is due out soon.
And about fiber sections, you don't want too many as it requires too much out of your computer (not the program), but you want enough to capture the proper strain distribution and associated areas.
Hope this helps!
The number, of course, depends on what you are trying to model.
Typically, however, 4-5 integration points is the optimum place to start. Less give you a less accurate curve, more may give you numerical instability as the program is trying to figure out what you are doing.
There are issues about whether Gauss quadrature represents the actual curvature distribution you want, even though it is able to represent the actual deformation you are trying to calculate. Hence, there is the beam-with-Hinges element (with hinges at the ends, so only for column elements, and elements with no distributed tranverse loads). Supposedly, this gives you a better model of the plastic hinge length. This is a new element, and a paper on it is due out soon.
And about fiber sections, you don't want too many as it requires too much out of your computer (not the program), but you want enough to capture the proper strain distribution and associated areas.
Hope this helps!
Silvia Mazzoni, PhD
Structural Consultant
Degenkolb Engineers
235 Montgomery Street, Suite 500
San Francisco, CA. 94104
Structural Consultant
Degenkolb Engineers
235 Montgomery Street, Suite 500
San Francisco, CA. 94104