help on convergence

[aneeman]

Hi,

I am using Template3Dep, Drucker-Prager model with Von Mises yield surface for non-associativity and Armstrong-Frederick tensorial evolution law. I am having a hard time getting it to converge (it converges fine with a linear tensorial evolution law).

I am pasting my analysis code below, any thoughts would be appreciated.

thanks!

alisa

----------------------------
wipeAnalysis

# Define analysis parameters
set Threshhold [expr 1e-6*$m];
set dLambda1 [expr 1.0/$Steps_sw];
set MaxIte 1000;
set Show 1; #show test norm

set lf1 0.01

system UmfPack

constraints Penalty 1e12 1e12

test NormDispIncr 1.0e-5 30 1
integrator LoadControl $lf1 1 $lf1 $lf1
algorithm Newton
numberer RCM
analysis Static
set NN1 25
puts "Self weight analysis for soil begins..."

analyze $NN1

[rjaeger]

Hi Alisa,

Do any steps converge for the A-F hardening?? What happens if you make the increments smaller?

Can you post the material commands you used?

Robbie

[aneeman]

[quote="rjaeger"]Hi Alisa,

Do any steps converge for the A-F hardening?? What happens if you make the increments smaller?

Can you post the material commands you used?

Robbie[/quote]

Thanks Robbie!

I was finally able to get it to work with small constants, ha 45 and Cr 15.

I started with ha 3000 and Cr 300, and the 3rd step of self weight the norms got really huge. 300 300 would run 15 step self-weight, 32 steps with point load, and I kept reducing until I could run 25 self-weight and 100 with my load.

Pasting my soil material code below.
#------------------------------------------------------
# Drucker-Prager model for sand
#------------------------------------------------------
#Yield surface von mises for non-associativity
set DPys "-VM"


#potential surface
set DPps "-DP 0.0"


# Scalar evolution law: linear hardening coef = 1.0, originally 0.0
set ES1 "-Leq 0.0"


#Armstrong-Frederick nonlinear tensorial evolution laws
# It changes alpha_ij in Drucker-Prager or Von-Mises yield surface
# ha 3000 Cr 300 very fast non-convergence for non-associativity
# 300 300 will run 15 step self-weight, 32 load for non-associativity
set ET1 "-NLEij 45 15"

# initial stress (pre-defined unit kPa)
set stressp "0.10 0 0 0 0.10 0 0 0 0.10"

# EPState
set E [expr 17400*$kPa]; #Young's Modulus or any number
set Eo [expr 17400*$kPa]; #Young's Modulus at reference pressure (1 P_atm)
set Vs 0.35; #Poisson's Ratio

set rho [expr 1478.0*$kg/pow($m, 3)]; #unit weight

set rho1 $rho

set phi [expr 37.1*$deg ]; #Friction angle
set alpha [expr 2.0*sin($phi)/( 1.7321*(3-sin($phi)) )]

#_____ ___E___Eo__v______rho__________________alpha___k
set EPS "-NOD 1 -NOS 2 $alpha 0.0 -stressp $stressp"
#
# where
#alpha = 2 sin(phi) / (3^0.5) / (3-sin(phi) ), phi is the friction angle
# and k is the cohesion

set mat1 1

# Create nDMaterial using Template Elastic-PLastic Model
nDMaterial ElasticIsotropic3D 2 $E $Vs $rho

nDMaterial Template3Dep 1 2 -YS $DPys -PS $DPps -EPS $EPS -ELS1 $ES1 -ELT1 $ET1

[aneeman]

set lf1 0.01

system UmfPack

#constraints for dafalias manzari
#constraints Penalty 1e8 1e8

#constraints for drucker prager
constraints Penalty 1e12 1e12

test NormDispIncr 1.0e-5 30 1
integrator LoadControl $lf1 1 $lf1 $lf1
algorithm Newton
numberer RCM
analysis Static

set NN1 25
puts "Self weight analysis for soil begins..."


analyze $NN1

[rjaeger]

Hi Alisa,

Thank you for posting your results and commands. I'm glad you got it working.

Robbie