way to speed up the pushover analysis

GunAndRose · Post by **GunAndRose** » Wed Mar 07, 2007 12:21 pm

Besides increasing the dU, any other choice to speed up the pushover analysis maintaining the same accuracy?
Thanks.

Post by **fmk** » Wed Mar 07, 2007 12:27 pm

if it's a big problem look at changing the solver. post the analysis part of the script.

silvia · Post by **silvia** » Wed Mar 07, 2007 12:58 pm

what materials and elements are you using???

GunAndRose · Post by **GunAndRose** » Wed Mar 07, 2007 2:14 pm

So, what kind of solver would you suggest to use instead.
here is the script:

constraints Plain
system ProfileSPD
numberer RCM
test NormDispIncr 1.0e-8 100 0
algorithm Newton
integrator LoadControl 0.1 1 0.1 0.1
#initialize
analysis Static

if {$PUSHOVER == "DispControl"} {
set dU 0.1;
# Displacement increment
#set maxU [expr 0.035*$heightBent];
if {$PlasticPushover=="nope"} {
set maxU [expr 0.005*$heightBent];
} else {
set maxU [expr 0.015*$heightBent];
}

# Max displacement

# node dof init Jd min max
if {$structuretype == "bents"} {
integrator DisplacementControl 10003 2 $dU 1 $dU $dU
} else {
integrator DisplacementControl 11003 2 $dU 1 $dU $dU
}
set numSteps [expr int($maxU/$dU)]
#puts "numSteps=$numSteps"
analysis Static
analyze $numSteps

}

fmk wrote:if it's a big problem look at changing the solver. post the analysis part of the script.

GunAndRose · Post by **GunAndRose** » Wed Mar 07, 2007 2:15 pm

I am using concrete02, nonlinear beam column element with fiber section.

silvia wrote:what materials and elements are you using???

silvia · Post by **silvia** » Wed Mar 07, 2007 5:18 pm

what steel material are you using?

GunAndRose · Post by **GunAndRose** » Wed Mar 07, 2007 7:44 pm

Steel01. Sorry I gforgot to mention that.

silvia wrote:what steel material are you using?

silvia · Post by **silvia** » Wed Mar 07, 2007 8:08 pm

thanx.
at least it's not a material that's holding you up.
i do recommend you use steel02, though, it's a nicer material, and just as fast.
how big is your model?
what are you doing that takes so long?

silvia · Post by **silvia** » Wed Mar 07, 2007 8:10 pm

also,
you are doing test NormDispIncr 1.0e-8 100 0
with 100 iterations.
if newton doesn't converge in 6, you should try something else.

are you having convergence problems?

GunAndRose · Post by **GunAndRose** » Wed Mar 07, 2007 9:04 pm

I am doing the fragility analysis that requires pushover analysis for the entire structure during each interation. So the speednecking is the pushover analysis on the structure.

My bridge model has 22 nonlinear beam columns with springs at the bottom of each column. EAch column only has one element since the speedy issue prevent me from using a finer mesh.

Thanks.

silvia wrote:thanx.
at least it's not a material that's holding you up.
i do recommend you use steel02, though, it's a nicer material, and just as fast.
how big is your model?
what are you doing that takes so long?

GunAndRose · Post by **GunAndRose** » Wed Mar 07, 2007 9:05 pm

This is a very good point. I will change it later.

Yes, I am having some convergence problems when the drift ratio exceeds 4%. So usually I only push upto 2.5%.

silvia wrote:also,
you are doing test NormDispIncr 1.0e-8 100 0
with 100 iterations.
if newton doesn't converge in 6, you should try something else.

are you having convergence problems?

silvia · Post by **silvia** » Thu Mar 08, 2007 5:45 am

you should be able to get beyond 2.5%.
have you tried the scripts that are in the examples manual, where modifications are made in case you don't converge with the standard analysis script?

i'd try the analysis scripts i put into the examples manual.
i think your analysis is slow because you are trying too many iterations on an analysis procedure that won't converge anyways.

let me know how it goes.