Convergence Issues

ddalgleish · Post by **ddalgleish** » Wed Jul 21, 2010 4:38 pm

I'm having problems with my model not converging. I'm using displacement beam columns and I've been reading around to see where convergence is an issue. I'm fairly new to finite modeling so I would like to know what are some basic ways to get a model to converge better. Can there be a convergence issue if I put too small of a tolerance on my beam columns or if I don't have a high max iterations on the beam columns? If this usually isn't the common issue then where is it?

Post by **fmk** » Fri Jul 23, 2010 3:51 pm

the dispBeamColumn does not have the convergence issues associated with the forceBeamColumn .. the convergence problem you are having is a global one .. have you tried changing using an analysis procedure which attempts multiple algorithms and integration schemes if a step fails?

ddalgleish · Post by **ddalgleish** » Sat Jul 24, 2010 10:56 am

No I haven't yet. I'm building my models through opensees navigator so I can see that my elements, ect. are all correct. I've noticed though, that navigator writes some analysis incorrectly like the newmark integrator so I have to go through and change what it writes.

Post by **fmk** » Mon Jul 26, 2010 11:39 am

you need to rewrite the navigator analysis file to try different things if the analysis fails. the navigator authors have been made aware of this and there should be a means of specifying your own analysis option file.

ddalgleish · Post by **ddalgleish** » Mon Jul 26, 2010 8:05 pm

Yeah what happened is that I wrote a program that automatically made a section defined in patches based off of input of the dimensions from AISC. It rounded widths to set the number of fibers in a direction so if the shape was thinner than half an inch it would make 0 fibers in that direction. I fixed it though. I really appreciate the help thought!