The ConvergenceTest object failed

[AnnaKowal]

Hello,

I haven't been used OpenSees for a long time and would appreciate some experienced users advice.
I am trying to run Total Stress Analysis of a Layered Soil Column (Total Stress Analysis) - 2D problem. All layers defined as elastic isotropic.


I got warning on Dynamic Analysis:
After first

WARNING: CTestNormDispIncr::test() - failed to converge
after: 35 iterations current Norm: 0.000212074 (max: 0.0001, Norm deltaR: 4030.48)
AcceleratedNewton::solveCurrentStep() -The ConvergenceTest object failed in test()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 0.10625
OpenSees > analyze failed, returned: -3 error flag
Finished with dynamic analysis...
Analysis execution time: 313411 seconds

I would be really grateful for your suggestion and ideas what I should look at and how to fix that issue.


The parametrisation I am using: after defining soil nodes, layers parameters, soil materials (all layers elastic isotropic), quad elements:

#-----------------------------------------------------------------------------------------
# 7. DEFINE ANALYSIS PARAMETERS
#-----------------------------------------------------------------------------------------

#---GROUND MOTION PARAMETERS
# time step in ground motion record
set motiondT 0.005
# number of steps in ground motion record
set nSteps 24000

#---RAYLEIGH DAMPING PARAMETERS
set pi 3.141592654
# damping ratio
set damp 0.01
# lower frequency
set omega1 [expr 2*$pi*0.0]
# upper frequency
set omega2 [expr 2*$pi*15]
# damping coefficients
set a0 [expr 2*$damp*$omega1*$omega2/($omega1 + $omega2)]
set a1 [expr 2*$damp/($omega1 + $omega2)]
puts "damping coefficients: a_0 = $a0; a_1 = $a1"


set dT $motiondT

#---ANALYSIS PARAMETERS
# Newmark parameters
set gamma 0.5
set beta 0.25

#puts "Finished defyning analysis parameters..."



#-----------------------------------------------------------------------------------------
# 10. DYNAMIC ANALYSIS
#-----------------------------------------------------------------------------------------

# define constant scaling factor for applied velocity
set cFactorH [expr 1.00*$baseArea*$dashpotCoeff_s]

# define velocity time history file
set velocityFileH /Users/annakowal/Documents/2D_site_analysis/deconvolution/Kettle_park_beach/VelocityHistory_4_OpenSees/01_Akatore_centr_NS_deconv.out

# timeseries object for force history
set mSeriesH "Path -dt $motiondT -filePath $velocityFileH -factor $cFactorH"

# loading object
pattern Plain 10 $mSeriesH {
load 18 0.50 0.00
load 172 1.00 0.00
#……
puts "Dynamic loading created..."


constraints Penalty 1.e20 1.e20
test NormDispIncr 1e-4 35 1
algorithm ModifiedNewton
numberer RCM
system ProfileSPD
integrator Newmark $gamma $beta
rayleigh $a0 $a1 0.0 0.0
analysis Transient

# perform analysis with timestep reduction loop
set ok [analyze $nSteps $dT]

# if analysis fails, reduce timestep and continue with analysis
if {$ok != 0} {
puts "did not converge, reducing time step"
set curTime [getTime]
puts "curTime: $curTime"
set curStep [expr $curTime/$motiondT ]
puts "curStep: $curStep"
set remStep [expr int(($nSteps-$curStep)*2.0)]
puts "remStep: $remStep"
set motiondT [expr $motiondT /2.0]
puts "motiondT : $motiondT "

set ok [analyze $remStep $motiondT ]

# if analysis fails again, reduce timestep and continue with analysis
if {$ok != 0} {
puts "did not converge, reducing time step"
set curTime [getTime]
puts "curTime: $curTime"
set curStep [expr $curTime/$motiondT ]
puts "curStep: $curStep"
set remStep [expr int(($remStep-$curStep)*2.0)]
puts "remStep: $remStep"
set motiondT [expr $motiondT 2.0]
puts "motiondT : $motiondT "

analyze $remStep $motiondT
}
}
set endT [clock seconds]
puts "Finished with dynamic analysis..."
puts "Analysis execution time: [expr $endT-$startT] seconds"

wipe
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Many thanks,
Anna

[selimgunay]

You can try Transformation constraints and the EnergyIncr test. The best way would be to use an adaptive solution that tries other integration methods and nonlinear solution methods as well.

[surojitmondal]

Hi dear,

Changing your algorithm to Krylov-Newton is one of the ways of solving the convergence problem.

Also visit this articles:https://www.lyrics2world.com/2020/06/un ... logic.html