Professional Documents
Culture Documents
UBCSAND UDM Documentation PDF
UBCSAND UDM Documentation PDF
Version904aR
DocumentationReport:
UBCSANDConstitutiveModelonItascaUDMWebSite
by
MichaelH.Beaty,PhD,PE,GE
BeatyEngineeringLLC
and
Prof.PeterM.Byrne,PhD,P.Eng.
UniversityofBritishColumbia
February,2011
TableofContents
1 Introduction............................................................................................................1
2 DescriptionofUBCSANDVersion904a...................................................................2
2.1 ElasticResponse..........................................................................................2
2.2 PlasticResponse..........................................................................................3
3 DescriptionofUBCSANDVersion904aR................................................................7
3.1 Staticanalysismode..................................................................................12
4 SingleElementBehaviorofUBCSAND904aR.......................................................13
4.1 Typicalstressstrainandstresspathbehavior.........................................13
4.2 Cyclicstrengthcurve.................................................................................18
4.3 Weightingcurve........................................................................................19
4.4 Effectofinitialstaticshearstress.............................................................21
4.5 Modulusreductionanddampingbehavior..............................................25
4.6 Effectofconfiningstress...........................................................................28
4.7 EffectofKo................................................................................................30
4.8 Rateofexcessporepressuregenerationandvolumetricstrain..............33
4.9 ComparisontocyclicDSSdataonFraserRiversand................................34
5 Postearthquakeanalysis......................................................................................43
5.1 Revisedrucomputation............................................................................44
6 CaseHistoryComparison......................................................................................45
6.1 UpperSanFernandoDam.........................................................................45
7 References............................................................................................................65
8 Appendices............................................................................................................67
Appendix1:AdditionalreferencesforUBCSAND...............................................67
Appendix2:GenericinputparametersforUBCSAND904aR.............................69
UBCSANDConstitutiveModelversion904aR February2011
1 Introduction
UBCSANDisaneffectivestressplasticitymodelforuseinadvancedstressdeformation
analyses of geotechnical structures. The model was developed primarily for sandlike
soils having the potential for liquefaction under seismic loading (e.g., sands and silty
sandswitharelativedensitylessthanabout80%).Themodelpredictstheshearstress
strainbehaviorofthesoilusinganassumedhyperbolicrelationship,andestimatesthe
associatedvolumetricresponseofthesoilskeletonusingaflowrulethatisafunctionof
thecurrentstressratio.Themodelcanbeusedinafullycoupledfashionwherethe
mechanicalandgroundwaterflowcalculationsareperformedsimultaneously.
One of the first uses of UBCSAND was for predicting the behavior of the CANLEX
(CanadianLiquefactionExperiment)embankments.Thefillswererapidlyconstructedon
loosetailingdepositstocreatealiquefactionresponse(Puebla,etal.,1997).Themodel
was soon adapted to seismic evaluations and applied to the response analysis of the
WildlifeSiteandtheSuperstitionHillsEarthquakeof1987(BeatyandByrne,1998).The
modelcontinuedtoberefinedaftertheseearlyanalyses.
ThefirstversionofUBCSANDthatwaswidelyusedforseismicanalyseswascompleted
in2002andbecameknownasUBCSAND904a.Thisversionhasseenconsiderableuse
andscrutinysinceitsdevelopment,andhasbeenusedasthebaseversionforseveral
modifiedcodes.Oneofthesemodifications, version904aR,wasdevelopedduringthe
evaluation of Success Dam in California to improve the behavior of the model under
certain types of loading. Particular focus was given to the prediction of excess pore
pressureswhensignificantstaticshearstresseswerepresent.
Thisdocumentpresentsanoverviewofthe904aversionofUBCSAND,adescriptionof
thechangesmadeforthe904aRversion,andanaccountofthebehavioroftherevised
modelintermsofelementtestsimulationsandthebackanalysesofacasehistory.
Page|1
UBCSANDConstitutiveModelversion904aR February2011
Cyclic shear strains induce plastic volume compaction in granular soils. Martin et al.
(1975) presented quantitative data in their landmark paper and showed that the
amountofcompactionpercycleisproportionaltothecyclicshearstrainamplitudeand
accumulated volume compaction, and is independent of normal effective stress. They
also showed that the pore pressure generated per cycle is dependent on the plastic
volumetricstrain,thereboundmodulusofthesoil,andthestiffnessoftheporefluid.
The response of sand is controlled by the skeleton behavior. A fluid (air water mix) in
the pores of the sand acts as a volumetric constraint on the skeleton if drainage is
curtailed. It is this constraint that causes the pore pressure rise that can lead to
liquefaction.Providedtheskeletonordrainedbehaviorisappropriatelymodeledunder
monotonicandcyclicloadingconditions,andthestiffnessoftheporefluidanddrainage
are accounted for, the liquefaction response can be predicted. This is the approach
incorporatedintoUBCSAND.
UBCSAND is a constitutive model that directly estimates the response of the soil
skeletontogeneralincrementsofloading.Theresponseoftheporefluidiscoupledto
the skeleton response through the bulk modulus of the fluid. UBCSAND is based on
classicplasticitytheoryandthecharacteristicsandbehaviorobservedinlaboratorytests
undermonotonicandcyclicloadingconditions.TheUBCSANDmodelanditsuseshave
beendocumentedinmanypapers,includingthoselistedinAppendix1.
Theelasticcomponentofresponseisassumedtobeisotropicandspecifiedbyashear
modulus,Ge,andabulkmodulus,Be,asfollows:
ne
[1] G
e
K Ge Pa
Pa
[2] Be Ge
e
where K G is a shear modulus number that depends on the relativedensity
andvariesfromabout500forloosesandto2000fordensesand,
Pa isatmosphericpressureinthechosenunits,
isthemeanstressintheplaneofloadingequalto ( x y ) / 2 ,
ne variesbetween0.4and0.6,orapproximately0.5,and
dependsontheelasticPoissonsratiowhichisintherange0.0~
0.2(Hardin1978)withtheresultthat variesbetween2/3and
4/3orisapproximatelyunity.
Page|2
UBCSANDConstitutiveModelversion904aR February2011
Plastic strains are controlled by the yield surface and flow rule. The yield surface is
representedbyaradiallinefromtheorigininstressspaceasshowninFigure1.Forfirst
timeshearloading,theyieldsurfaceiscontrolledbythecurrentstressstate,pointAin
Figure1.Astheshearstressincreases,thestressratio ( / ) increasesandcauses
the stress point to move to point B. and are the shear and normal effective
stressesontheplaneofmaximumshearstress.Theyieldsurfaceisdraggedtothenew
locationpassingthroughpointBandtheorigin.Thisresultsinplasticstrains,bothshear
and volumetric. The plastic shear strain increment, d p , is related to the change in
shearstressratio, d ,asshowninFigure2andcanbeexpressedas
1
[3] d p
d
G /
p
[4] G p G ip ( 1 R f ) 2
f
istheplasticmodulusatalowlevelofstressratio ( 0 ) ,
p
where Gi
f isthestressratioatfailureandequals sin f ,
f isthepeakfrictionangle,and
R f is the failure ratio used to truncate the best fit hyperbolic
relationshipandpreventtheoverpredictionofstrengthatfailure.
R f generally varies between 0.7 and 0.98 and decreases with
increasingrelativedensity.
Shear Stress,
Yield Locus
YieldSurface
B
Page|3
UBCSANDConstitutiveModelversion904aR February2011
Stress Ratio, (= / )
Gp /
B
A
d p
Theassociatedincrementofplasticvolumetricstrain, d v ,isrelatedtotheincrement
p
ofplasticshearstrain, d p ,throughtheflowruleasfollows:
[5] d vp (sin cv ) d p
where cv istheconstantvolumefrictionangleorphasetransformationangle.Thisflow
rule can be derivedfrom energy considerations and is similar to stress dilation theory
(Rowe1962;MatsuokaandNakai1977).
Yieldlociandthecorrespondingdirectionoftheplasticstrainsresultingfromtheflow
rule are shown in Figure 3. Significant shearinduced plastic compaction occurs at low
stressratios,whilenocompactionispredictedatstressratioscorrespondingto cv .At
stressratiosgreaterthan cv ,shearinducedplasticexpansionordilationispredicted.
This simple flow rule is in close agreement with the characteristic behavior of sand
observedinlaboratoryelementtesting.Uponunloading,definedasareductioninthe
magnitude of , the sand is assumed to behave elastically and no plastic strains are
generated.
Thesignofthestressratioiscontrolledbythesignoftheshearstressonthehorizontal
plane.Thisisasimplifyingassumptionbutrecognizestheimportanceofthehorizontal
plane in many geotechnical structures (i.e., the importance of simple shear loading).
Bothpositiveandnegativevaluesof themaximumstressratioareseparatelytracked.
This allows the plastic behavior to include aspects of both kinematic and isotropic
hardening.
Page|4
UBCSANDConstitutiveModelversion904aR February2011
Theplasticshearmodulusdiscussedaboveisapplicableforvirginorfirsttimeloading.
Whenever the current load increment pushes the yield surface outside the previous
maximum stressratio limits,thisincrement ofloadingisconsideredfirst time loading.
Whenanincrementofloadingoccurswithinthepreviousmaximumstressratiolimits,
thesandisassumedtobehaveplasticallybutwithaplasticmodulusthatisseveraltimes
stiffer than for first time loading. The maximum stress ratio limits are defined as the
largestpositiveandnegativevaluesof thathaveoccurredsincethestartofloading.
Not all loading increments generate plastic strains in version 904a. Unloading occurs
whenthereisadecreaseinthemagnitudeofthestressratio.Ifthestressratioshould
thenbegintoincreaseinmagnitudebeforetherehasbeenachangeinthesignofthe
stress ratio, then this increment of loading is considered a reloading increment.
Reloading increments are assumed to respond elastically with no plastic shear or
volumetricstrains.Reloadingoccursuntilthestressratioequalsthepreviousmaximum
stress ratio that occurred during the current loading cycle. Once this stress ratio has
beenachieved,subsequentloadingincrementsgenerateplasticstrainsThisdefinitionof
reloadingisshownschematicallyonFigure4.
, dp
cv
, d vp
Figure3.Directionsofplasticstrainsassociatedwithlocationofyieldsurface.
Page|5
UBCSANDConstitutiveModelversion904aR February2011
Loading
Loading Unloading
Stress Ratio
Unloading
Loading
Reloading
Time
Loading Unloading
Figure4.Stressratiohistoryshowingloading,unloading,andreloading.
Page|6
UBCSANDConstitutiveModelversion904aR February2011
Revisionstoversion904awerepromptedbyanalysesofSuccessDaminCalifornia.The
selection of an appropriate constitutive model for these analyses followed a rigorous
evaluationmadebytheSacramentoDistrictOfficeoftheArmyCorpsofEngineers.This
evaluationprocessbenefitedfromthegeneralguidanceandinputoftheadvisorypanel
fortheSuccessDamproject.Italsoledtoseveralchangestotheconstitutivemodel.
Preliminary analyses of Success Dam using the 904a version of UBCSAND showed a
significantly smaller zone of high excess pore pressure beneath the upstream shell of
the dam than was anticipated. Concerns regarding the extent of predicted high pore
pressures were supported by several independent evaluations: 1) estimates of pore
pressure generation based on results from Quad4 analyses, 2) estimates of pore
pressuregenerationfromacyclecountingFLACmodeldevelopedbyURS,and3)andan
examinationofshearstresshistoriespredictedbytheUBCSANDmodelinkeyelements.
Theinitialstaticshearstress,orstaticbias,wasfoundtohaveanunexpectedinfluence
onthegenerationofexcessporepressure.
Version 904a includes the simplifying assumption that cycles of partial unloading and
reloadingareelastic.Thesecyclesaredefinedasoneswheretheshearstressdropsand
then increases but there is no reversal in the direction of the shear stress. For many
locations beneath the upstream shell of Success Dam, the initial static shear stress is
larger than the magnitude of most cyclic loading cycles. In other words, many of the
larger shear stress cycles were considered partial unloadreload cycles and did not
contributetotheexcessporepressure. Anexampleshearstresshistorycomputedfor
SuccessDambeneaththeupstreamshellisshowninFigure5.
UBCSANDwasmodifiedtoimproveresponsepredictionsforcaseswithsignificantstatic
bias.Thiswasaccomplishedusingpublishedrelationshipsbetweencyclicstrengthratio
andstaticbiasasaguide.Althoughtheeffectofstaticbiasonliquefactionresistanceis
somewhatuncertain,itistypicallyaddressedinsimplifiedproceduresthroughafactor
termedK.TworelationshipsforKwereproposedbyHarderandBoulanger(1997)and
IdrissandBoulanger(2003).Theserelationshipswerebasedonalimitedsetofsimple
sheartestresultswhichwereconsideredtoprovidethemostreliabledata.Theseplots
representthecurrentstateofpracticefortheevaluationofslopesandembankments.
ThemodificationstoUBCSANDforversion904aRweregovernedbytwocriteria:
1. capturethegeneraltrendsincorporatedintotheKplots,and
2. limitchangestothe structure,assumptions, inputparameters,andbehaviorof
the904aversion.
Page|7
UBCSANDConstitutiveModelversion904aR February2011
1000
Significant pore pressure generation
Shear Stress on Horizontal
500
0
Plane (psf)
-1500
-2000
-2500
0 5 10 15 20 25
Time (sec)
Figure5.ExampleofpredictedshearstressbeneathupstreamshellofSuccessDamusing
UBCSANDversion904a.
TheprimarychangesmadetoUBCSANDaresummarizedbelow:
1. Partialunloadreloadcyclesgenerateplasticvolumetricstrains
In version 904a, the location of the yield surface is not modified when an
element unloads until a shear stress reversal occurs. For partial unloadreload
cycles (i.e., no stress reversal), the response is entirely elastic until the stress
state once again reaches the yield surface. For version 904a, no plastic
volumetricstrainsaregeneratedduringthesepartialunloadreloadcycles.
In version 904aR, the location of the yield surface is now modified during
increments of unloading. The yield surface systematically drops as the
magnitude of the mobilized stress ratio decreases. The yield surface lags
slightlybehindthedecreasingstressstatesothatasmallelasticzoneseparates
the stress state and the yield surface. Adjusting the yield surface in this way
allows for the generation of plastic volumetric strains when the element is
reloaded even if a stress reversal has not occurred. The generation of pore
pressuresduringpartialunloadreloadcyclesissupportedbylaboratorytesting,
suchasthedatadiscussedinSection4.9.
Theplasticshearstiffnessassociatedwithreloadingiscontrolledbytheinternal
variable m_urstif, which directly factors the plastic shear modulus. This factor
Page|8
UBCSANDConstitutiveModelversion904aR February2011
wasneededinordertoreasonablypredictthebehaviorshownintheKcharts.
TheKfactorrelatesthecyclicresistancewithnostaticshearbias(=0)tothe
cyclic resistance at a static bias of . The static bias is defined as the initial
shearstressonthexyplanedividedbytheinitialverticaleffectivestress.
Themagnitudeofm_urstifisafunctionoftherelativechangein .m_urstifis
equal to a minimum of 1.0 at the instant of a stress reversal and smoothly
increasestoarelativelylargefactorforsmallunloadingcycles.Therelationship
used to define m_urstif was developed by matching model predictions to
expectedKbehaviorusingdirectsimpleshear(DSS)simulations.
2. Plasticshearstiffnessmodifiedfornonsymmetricloadingcycles
Inversions904aand904aR,noplasticstrainsaregeneratedduringincrements
ofunloading(i.e.,decreasingstressratio).Anyplasticstrainsthatshouldoccur
during unloading are assumed to be accounted for during the subsequent
loading cycle after a shear stress reversal. In other words, the actual plastic
behavior can be approximated if the plastic shear modulus during loading is
madesomewhatsofter.Thiswillinduceadditionalplasticstrainsduringloading
to account for any strains that were missed during the unloading increments.
This approximation is generally reasonable, although it did produce some
undesirabletrendsinthepredictedKbehavior.Theprimarymotivationforthis
modificationwastosmooththepredictedrelationshipbetweenstaticbiasand
liquefactionresistance.
The plastic shear stiffness is now modified to account for the effects of non
symmetry. Symmetry is evaluatedat stress reversals by computing the ratioof
the peak stress ratio from the previous two halfcycles of loading (i.e.,
ratio_k=peak_k1/peak_k2).If ratio_k1thenthepreviousfullcycleofloadingis
consideredtobesimilartoasymmetricloadcycleandtheplasticshearstiffness
is not adjusted. If ratio_k< 1 then the load cycle is considered nonsymmetric
andtheplasticshearmodulusisstiffenedusingtheinternalparameterm_sym.
The adjustment factor m_sym ranges between 1.0 for symmetric cycles to a
maximum of 1.3 to 1.9 for highly nonsymmetric cycles. This definition for
symmetryisillustratedinFigure6.
3. Postdilationsoftenermadeafunctionofaccumulateddilativestrains
Plastic dilative strains are used by UBCSAND to identify elements that will
experiencesignificantplasticvolumetriccontractionuponashearstressreversal.
This plastic volumetric contraction is induced in the model by significantly
softeningtheplasticshearmodulusafterastressreversal.Inversion904a,the
required amount of softening is based solely on the plastic dilative strain
Page|9
UBCSANDConstitutiveModelversion904aR February2011
experiencedduringtheprevioushalfcycle.Thiswasrevisedinversion904aRso
thatthemagnitudeofsofteningisbasedontheaccumulatedamountofplastic
dilationthattheelementhasexperiencedsincethestartofloading.Thisdilation
isaccumulatedonlywhenthestressratioisclosetothemaximumallowedstress
ratio(i.e.,m_ratf).
Thisreviseddefinitionassumesthatplasticdilativestrainscausethesoilskeleton
tosoften,andthatthisdamagetotheskeletonenduresbeyondthecurrentload
cycle. This change in defining the softener allows for a moregradual transition
into liquefied behavior. This was particularly important for simulating the
responseofdensersandsunderastaticbias.
The postdilation softener is a function of both the loading symmetry and the
accumulated dilative strains. The internal parameters m_symdil and m_dilsft address
thesetwoaspects.Thefunctionalrelationshipsforthesefactorswerederivedthrough
DSSsimulationsandcomparisontoexpectedliquefactiontriggeringandKbehavior.
peak_k-1
peak_k-2
peak_k
ratio < 1
Time
Note: ratio1 :previousfullcycleconsideredsymmetric.
ratio<1 :previousfullcycleconsiderednonsymmetric.
Figure6. llustrationofsymmetricandnonsymmetricloadingcycles.
Page|10
UBCSANDConstitutiveModelversion904aR February2011
4. Smoothtransitionbetweenprimaryandsecondaryyieldsurfaces
UBCSANDusestwoyieldsurfacestoincorporateplasticresponseduringloading.
The primarily yield surface is active for conditions of virgin or primary loading.
Primary loading is defined as an increase in stress ratio above the previous
maximum stress ratio experienced by the element. The occurrence of primary
loading is evaluated separately in both the positive and negative loading
directions.Incontrasttoprimaryloading,asecondaryyieldsurfaceisusedwhen
loadingincrementsoccurbelowthepreviousmaximumstressratio.
In version 904a, an abrupt change in response can occur when the model
switchesfromthesecondarytotheprimaryyieldsurface.Theseabruptchanges
aremostnoticeableinplotsofstresspath.Versions904aRincludesatransition
between these two yield surfaces. As the current stress ratio approach the
previous maximum stress ratio, the properties of the yield surface begin to
interpolatebetweenthoseofthesecondaryandprimarysurfaces.Thebenefitof
thistransitionistoproducesomewhatsmootherstresspathbehavior,although
theoveralleffectonmodelresponseisexpectedtobeminor.
5. RevisedrelationshipsforRfandf
Twoadjustmentsweremadetothegenericinputparameters.Thefrictionangle
at failure, m_phif, was increased for sands with (N1)60 greater than 15. For
example, the value of m_phif for (N1)60=25 was increased from 35.5 to 37.5.
This change in frictionangle wasmade inorder to stiffen the postliquefaction
responseofdensersands.Therelationshipforthehyperbolicadjustmentfactor,
m_rf, was also revised. The current equation better represents the trend
reportedinByrneetal.(1987).
6. Calibrationequationsform_hfac1
Page|11
UBCSANDConstitutiveModelversion904aR February2011
etal,2001).Theseequationsform_hfac1,providedAppendix1,areoptionaland
canbeeasilyreplacedbyprojectspecificrelationships.
Page|12
UBCSANDConstitutiveModelversion904aR February2011
ForthesakeofmakingcomparisonsbetweentheUBCSANDliquefactionresponseand
variousempiricalandlaboratoryrelationships,acleardefinitionisneededfordefining
the onset of liquefaction in the UBCSAND element. Liquefaction is assumed to occur
when eitherofthefollowingtwocriteriaissatisfied:the excessporepressureratio ru
exceeds0.85orthemaximumshearstrainexceeds3%.ruisameasureoftheincrease
inporepressurewhereruequals0iftheporepressuresdonotchange,andruequals1
when effective stress become equal to zero (see Section 5.1). These two criteria are
similartothoseoftenusedtodefinetheonsetofliquefactioninalaboratorytest.Theru
criterionwasoftenthecriticalcriterionforanalysesusinglowervaluesof(N1)60cs,while
theshearstraincriterionwasoftensatisfiedforcaseswithlarger(N1)60csvalueswherea
significantstaticbiaswaspresent.
ThesefiguresareintendedtodemonstratethetypicalbehaviorofUBCSANDinasimple
shearsimulation.AdirectcomparisontolaboratorytestsresultsisshownSection4.9.
ThedrainedmonotonicpredictionsareshowninFigure7.Thedensermaterialisseento
haveastifferstressstrainresponse,withdilativevolumetricstrainsafterashearstrain
ofabout0.3%.The(N1)60=5testshowsasmalldilativeresponseaftershearstrainsof
1
TheresultsanddescriptionofUBCSAND904aRpresentedinthereportrefertotheconstitutivecode
versionUBCSAND904aRDP.dr8.
Page|13
UBCSANDConstitutiveModelversion904aR February2011
Figure9showspredictionsfordrainedcyclicloading,whileFigure10showspredictions
for undrained cyclic loading. The applied cyclic loading in each case was equal to the
CRR15asdeterminedfromtheNCEER/NSFchart.Forthedrainedcase,15cyclesofthis
loading were applied. The trends observed from the cyclic loading are reasonable,
althoughUBCSANDisstillshowntobesensitivetotheinitialKoconditions.Thepotential
importance of Ko on the liquefiability of an element is seen most clearly in the
predictions for (N1)60=5: the volumetric strain versus shear strain plot for drained
conditions(Figure9c)andthestresspathplotforundrainedconditions(Figure10c).The
cyclicanalysisresultsareconsistentwiththemonotonicloadingpredictions.
80
Shear Stress Sxy (kPa)
60
40 (N1)60 =5 Ko = 1.0
(N1)60 =5 Ko = 0.5
20 (N1)60 =15 Ko = 1.0
(N1)60 =15 Ko = 0.5
0
0 1 2 3 4 5
Shear Strain %
a.Stressstrain
0.1
Volumetric Strain (%)
-0.1
-0.2
-0.3
0 1 2 3 4 5
Shear Strain %
b. Volumetricstrainversusshearstrain
Figure7. UBCSAND904aRpredictionsofmonotonicdrainedloading.
Page|14
UBCSANDConstitutiveModelversion904aR February2011
120
Shear Stress Sxy (kPa)
(N1)60 =5 Ko = 1.0
100 (N1)60 =5 Ko = 0.5
80 (N1)60 =15 Ko = 1.0
(N1)60 =15 Ko = 0.5
60
40
20
0
0 1 2 3 4 5
Shear Strain %
a.Stressstrain
120
Shear Stress Sxy (kPa)
100
(N1)60 =5 Ko = 1.0
80
(N1)60 =5 Ko = 0.5
60 (N1)60 =15 Ko = 1.0
(N1)60 =15 Ko = 0.5
40
20
0
0 50 100 150 200
v' kPa
b.Stresspath
Figure8. UBCSAND904aRpredictionsofmonotonicundrainedloading.
Page|15
UBCSANDConstitutiveModelversion904aR February2011
20
10
-20
-0.02 0 0.02 0.04 0.06 0.08
Shear Strain %
b.Stressstrainfor(N1)60=15
0
Volumetric Strain (%)
-0.02
-0.04
-0.1
-0.02 0 0.02 0.04 0.06 0.08
Shear Strain %
c.Volumetricstrainversusshearstrainfor(N1)60=5
0
Volumetric Strain (%)
-0.02
-0.04
-0.06
(N1)60 =15 Ko = 1.0
-0.08 (N1)60 =15 Ko = 0.5
-0.1
-0.02 0 0.02 0.04 0.06 0.08
Shear Strain %
d.Volumetricstrainversusshearstrainfor(N1)60=15
Figure9. UBCSAND904aRpredictionsofdrainedcyclicloading.
Page|16
UBCSANDConstitutiveModelversion904aR February2011
20
10
-20
-6 -4 -2 0 2 4 6
Shear Strain %
b.Stressstrainfor(N1)60=15
20
Shear Stress Sxy (kPa)
10
v' kPa
c.Stresspathfor(N1)60=5
20
Shear Stress Sxy (kPa)
10
-10
-20
0 20 40 60 80 100
(N1)60 =15 Ko = 1.0
v' kPa (N1)60 =15 Ko = 0.5
d.Stresspathfor(N1)60=15
Figure10. UBCSAND904aRpredictionsofundrainedcyclicloading.
Page|17
UBCSANDConstitutiveModelversion904aR February2011
ThegenericinputparametersAppendix2werecalibratedtoreproducetheliquefaction
triggeringbehaviorrecommendedbythe1997NCEER/NSFworkshop(Youdetal.,2001).
This was done by recognizing that the NCEER/NSF triggering chart corresponds to
earthquakes with magnitudes of about 7.5. The cyclic shear stress history induced by
earthquakesofthismagnitudecanbeapproximatedby15uniformcyclesofshearstress
withamagnitudeequaltothecyclicshearstressdeterminedfromthetriggeringchart.
Inotherwords,thecyclicresistanceratioindicatedbytheNCEER/NSFcurveforagiven
correctedSPTblowcount,or(N1)60cs,shouldjustinduceliquefactioninanelementifitis
appliedin15uniformcycles.
The results of the UBCSAND simulation are shown in Figure 11. This figure shows the
cyclic stress ratio causing liquefaction in 15 uniform cycles versus (N1)60. Each of the
plottedvaluesofcyclicresistanceratio(CRR)reflectthehighervaluedeterminedfrom
twoanalyses:onewithKoequalto0.5andthesecondwithKoequalto1.0,whereKois
theratioofhorizontaleffectivestresstoverticaleffectivestressatthestartofloading.A
discussionoftheinfluenceofKoonthetriggeringresistanceofUBCSANDisprovidedin
Section4.7.
The CRR predicted by UBCSAND is seen to increase gradually and smoothly with
increasing(N1)60cs.AdirectcomparisonoftheCRRestimatesfromUBCSANDismadeto
several current triggering curves: the NCEER/NSF triggering curve, the curve proposed
by Idriss and Boulanger (2006), and the curves proposed by Cetin et al. (2004) for a
probability of liquefaction equal to 20% and 50%. The CRR estimates generated by
UBCSAND are seen to agree closely with NCEER/NSF and the Idriss and Boulanger
relationships. The two curves developed from the Cetin et al. (2004) approach are
substantiallylower.
Page|18
UBCSANDConstitutiveModelversion904aR February2011
Weighting curves show the relative importance of stress cycles having different
magnitudes.Ittakesfewercyclesofalargeshearstresstoliquefyasandascomparedto
asmallshearstress,andthisrelationshipisreflectedintheweightingcurve.Weighting
curves are developed in the laboratory by testing a series of equivalent sand samples
withuniformcyclesofcyclicloading.Eachsampleistestedatadifferentmagnitudeof
cyclic shear stress and the corresponding number of cycles to induce liquefaction is
recorded.Theresultingdataisplottedtoproduceaweightingcurve,generallyshownas
CSRversusthelogofthenumberofcyclestoliquefaction.Itisconvenienttoplotthe
CSRdatainanormalizedfashion,whereeachCSRvalueisdividedbyCSR15,whichisthe
CSRcausingliquefactionin15cycles.
0.5
UBCSAND 904aR.dr6
1997 NCEER/NSF Workshop (Youd et al. 2001)
0.3
Cyclic Resistance Ratio
0.2
0.1
vo' = 1 atm
Mw = 7.5
0
0 5 10 15 20 25 30
Corrected Clean Sand Blowcount (N1)60cs
Figure11. ValuesofCRRpredictedbyUBCSAND904aRandcomparedtosemiempirical
relationships.
Page|19
UBCSANDConstitutiveModelversion904aR February2011
fordensersandstendedtobesomewhatsteeperthanforloosersands.Thedataalso
suggeststhatthesteepnessoftheweightingcurvecanbeaffectedbythecriteriaused
todefinetheonsetofliquefaction.
WeightingcurvesweregeneratedbyUBCSANDbyperformingasimilarseriesofcyclic
DSS simulations. The tests were performed for a range of relative densities and the
computed values are plotted on Figure 12. For comparison, the weighting curve
inherent in the magnitude correction relationship proposed by Idriss and Boulanger
(2006)isalsoshown.ThiscurvewasdevelopedfromthevaluesofKmproposedbyIdriss
and Boulanger in combination with the corresponding cycles of significant loading
assignedtoeachearthquakemagnitude.TheweightingcurvesgeneratedbyUBCSAND
are in reasonable agreement with the curve developed from Idriss and Boulanger
(2006). The weighting curves generally become steeper as the (N1)60 values increase,
exceptforthecurveestimatedfor(N1)60=2whichplotssteeperthanexpectedbased
ontheothercurves.
Theeffectofinitialconfiningstressontheweightingcurvewasevaluatedasshownin
Figure 13. The weighting curves predicted by UBCSAND are affected by this change in
confiningstress,buttoarelativelymodestdegree.Theexpectedrelationshipbetween
weightingcurveandeffectivestressisnotknown.
2.5
(N1)60 =2, CRR15 =0.053
(N1)60 =5, CRR15 =0.072
(N1)60 =10, CRR15 =0.113
2 (N1)60 =15, CRR15 =0.160
(N1)60 =20, CRR15 =0.215
CRR / CRR 15
0.5
1 10 100
Number of Cycles to Liquefaction
Figure12. CyclicstrengthcurveforUBCSAND904aRfor'vo=1atm.
Page|20
UBCSANDConstitutiveModelversion904aR February2011
The effect of the initial static shear bias on the cyclic behavior of a sand has typically
beenevaluatedthroughcyclicelementtests.Uniformcyclesofshearloadareappliedto
asampleofthesand,andthemagnitudeofthecyclicstressratiotocauseliquefaction
inasetnumberofcyclesisdetermined.Thisisrepeatedforvariouslevelsofinitialshear
stress.Thisinitialshearstressproducesaconstantbiasinthecyclicshearloadasshown
inFigure14.Themagnitudeoftheinitialshearstressistypicallyexpressedas,whichis
theinitialshearstressnormalizedbytheinitialverticaleffectivestress.
Changing the static shear stress will influence the cyclic resistance ratio CRR. In a
simplifiedliquefactiontriggeringanalysis,thecorrectionfactorKisdefinedastheCRR
forastaticbiasofdividedbytheCRRwhenequalszero(i.e.,K=CRR/CRR=0).
A recent evaluation was made by Idriss and Boulanger (2003) of a limited number of
cyclic simple shear tests performed to investigate K. This tests evaluated a range of
relative densities and initial shear stress conditions, including those having no shear
stress reversal on the horizontal plane. These tests were evaluated in terms of the
relativestateparameterconcept,andarelationshipbetween ,(N1)60,sandgraintype,
2.5
(N1)60 =5, CRR15 =0.072, Sigvo' = 1 atm
0.5
1 10 100
Number of Cycles to Liquefaction
Figure13. WeightingcurvesfromUBCSAND904aRfor'vo=1atmand4atm.
Page|21
UBCSANDConstitutiveModelversion904aR February2011
alpha = 0.2
Increasing
alpha = 0.1
Shear Stress
alpha = 0
Cycles or Loading
Figure14. ExampleshowingeffectofstaticbiasonuniformloadcycleswithsameCSR.
effectivestress,andKwasproposed.Therecommendedrelationshipforquartzsands
isshowninFigure15andcomparedagainsttheoriginallaboratorytestdata.Figure16
showstheearlierKrelationshipbyHarderandBoulanger(1997)thatwaspresentedat
the1997NCEERworkshop.
Therelationshipbetween ,(N1)60,andCRRwaspredictedbytheUBCSANDmodelby
simulatingaseriesofcyclicsimplesheartests.Allofthesesimulationswereperformed
using an initial vo = 1 atm. The CSR required to trigger liquefaction in 15 cycles was
determinedforeachcombinationofand(N1)60.Figure17comparestheKpredictions
forUBCSAND904aandUBCSAND904aR.
Page|22
UBCSANDConstitutiveModelversion904aR February2011
2
Laboratory Test Data
Vaid & Finn 1979 (N ~ 21)
N ~ 21
Boulanger et al. 1991(N ~ 14)
1.5
Vaid & Finn 1979 (N ~ 12)
Boulanger et al. 1991(N ~ 6)
N ~ 14
K 1
K
N ~ 12
0
0 0.1 0.2 0.3 0.4
Figure15. Relationship between K, and (N1)60. Data and plotted trends from Idriss
andBoulanger(2003).Laboratorydatafromtestsperformedatvo=2atm.
Figure16. KrelationshipfromHarderandBoulanger(1997).
Page|23
UBCSANDConstitutiveModelversion904aR February2011
2
UBCSAND
Predictions:
(N1)60 = 5
1.5
(N1)60 = 10
(N1)60 = 15
K
K
(N1)60 = 25
1
2 UBCSAND
Predictions:
(N1)60 = 2
1.5 (N1)60 = 5
(N1)60 = 10
(N1)60 = 15
K
K
(N1)60 = 20
1
(N1)60 = 25
(N1)60 = 30
Page|24
UBCSANDConstitutiveModelversion904aR February2011
AsecondfeatureoftheKpredictionsfromversion904aistheeventualdecreaseofK
for (N1)60=5 at > 0.15. This occurs due to the combination of a large static shear
stress and the initial pulse of the cyclic load. This initial pulse induces strain softening
andcausestheelementtofailunderamonotonicload.Theshearstrainsresultingfrom
thismonotonicloadaresignificantandsatisfytheliquefactiontriggeringcriteria.
Animprovedrelationshipbetween ,(N1)60,CSR,andliquefactionresistanceisseenin
the K predictions for UBCSAND 904aR shown on Figure17(b) and Figure 18. The
objectivesof904aRweretoeliminatetheartificialincreaseinthepredictedvaluesofK
withincreasing ,tocreateconsistencybetweenthepredictedKcurvessothatlarger
values of K would occur for larger values of (N1)60 over the full range of , and to
generally produce K estimates that were approximately equal or somewhat less than
thosesuggestedbythedataandrelationshipspresentedbyIdrissandBoulanger(2003).
A further restraint was imposed to prevent the model from predicting K values
significantlylargerthan1.Thisconstraintreflectstheuncertaintyregardingtheeffectof
staticbiasinfieldsituation,includinghowoutofplanemotionswillaffectporepressure
generationbeneathslopes.Mostoftheseobjectiveswereachievedin904aR.
The relationship between secant modulus, hysteretic damping, and the magnitude of
2
UBCSAND
Predictions:
(N1)60 = 2
1.5
(N1)60 = 5
K (N1)60 = 10
K
1 (N1)60 = 15
(N1)60 = 20
(N1)60 = 25
0.5
(N1)60 = 30
Background figure is
0
fromHarder and
0 0.1 0.2 0.3 0.4 Boulanger
(NCEER, 1997)
Figure18. ComparisonofKfromUBCSAND904aRversusHarderandBoulanger(1997).
Page|25
UBCSANDConstitutiveModelversion904aR February2011
ThereductioninshearstiffnesswithstrainthatispredictedbyUBCSANDisseentobein
reasonable agreement with the Idriss (1999) trend for sand. The amount of damping
predictedbythemodelforsymmetricloadcyclesisseentobesignificantlyhigherthan
anticipatedfromsoiltests.Forexample,theanticipateddampingatacyclicshearstrain
of 0.1% is approximately 10% to 20% of critical damping. The damping produced by
UBCSAND at this strain level is approximately 30% for the two sands that were
simulated.TheminimumdampingproducedbyUBCSANDforsymmetricloadingatsmall
strainsrangesfromnear0%to10%.Thelargerthananticipateddampingproducedby
UBCSANDisdueinlargeparttothesimplificationofelasticunloadingatthemaximum
shearmodulusGmax.Theuseoflinearelasticunloadingcreatesanextendedstiffportion
tothestressstraincurve,producinglargerloopareasthanwouldbeanticipatedfroma
laboratorytest.
AsignificantdifferenceisseenbetweentheKo=0.5andKo=1.0analysesatsmallstrain
levels.Thisdifferenceiscausedbythetendencyfortheelasticresponsetodominatefor
theKo=0.5analysisuntilthedirectionofthepeakshearstresssufficientlyrotates.
ThelargedampinginherentintheUBCSANDmodeloccursforsymmetricloadcycles.A
somewhat reduced damping is anticipated for cycles that are nonsymmetric. For
example,samplesthatpartiallyunloadandthenreloadwithastiffenedshearmodulus
should dissipate a relatively small amount of energy through hysteresis during the
unloadreloadcycle.
Page|26
UBCSANDConstitutiveModelversion904aR February2011
1
UBCSAND (N1)60=10 Ko=1.0
UBCSAND (N1)60=10 Ko=0.5
0.8 UBCSAND (N1)60=20 Ko=1.0
UBCSAND (N1)60=20 Ko=0.5
Idriss curve for sand (1999)
Gsecant / Gmax
0.6
0.4
0.2
0
0.001 0.01 0.1 1 10
30%
20%
10%
0%
0.001 0.01 0.1 1 10
Page|27
UBCSANDConstitutiveModelversion904aR February2011
Adjustmentfactorsm_hfac1weredevelopedforthe904aRversionthatreproducethe
KcurverecommendedbyYoudetal.(2001).TheexponentfintheKequationwas
estimatedbyassumingthefollowingrelationshipbetweenrelativedensityDrand(N1)60:
Dr2=(N1)60/46. The relationship between (N1)60, vo, and m_hfac1 was developed by
selectingvariouscombinationsof(N1)60andvo,applyingacyclicloadequaltoKtimes
theexpectedCRRat vo=1atm,thenadjustingm_hfac1untilthecyclicDSSsimulation
producedliquefactionin15cycles.Thisprocedureledtoanequationform_hfac1asa
functionof(N1)60andvoasdescribedinAppendix2.Itisinterestingthattheestimated
relationship for m_hfac1 is in the form of a power curve, similar to the adopted
relationshipforK.
Each combination of (N1)60 and vo were evaluated at two initial values of horizontal
effectivestresscorrespondingtoKo=0.5andKo=1.0.Thefinalcalibrationparameter
wasselectedfromtheinitialstresscasethatprovedtobethemoredifficulttoliquefy.
Ingeneral,simulationswithlowerinitialstressesorlower(N1)60valuestendedtoliquefy
more easily at Ko=1.0 conditions. Simulations with higher initial stresses or higher
(N1)60bvaluestendedtoliquefymoreeasilyatKo=0.5conditions.
The resulting adjustment factors are shown in Figure 20. This figure shows m_hfac1
values greater than 1 when vo is less than 1 atm. Although some increase in cyclic
resistance is expected at low initial confining stresses, the value of m_hfac1 would
typicallybelimitedtoamaximumofthevalueatvoequalto1atm.
Figure21showsthevaluesofKthatwereestimatedusingUBCSANDandthegeneric
input parameters from Appendix 2. These estimates are directly compared to the
NCEER/NSF recommendations. The relationship developed for m_hfac1 is shown to
produceagoodagreementwiththeKcurves.ValuesofK wereestimatedfor voless
than 1 atm both with and without the restriction on m_hfac1. Restricting m_hfac1 to
thevaluecomputedfor vo=1atmdidreducetheestimatedvaluesofKatlow vo,
althoughtheseKestimateswerestillsomewhatabove1.0.
Page|28
UBCSANDConstitutiveModelversion904aR February2011
1.5
(N1)60 = 2
(N1)60 = 5
(N1)60 = 8
1
(N1)60 = 10
m_hfac1
(N1)60 = 12
(N1)60 = 14
0.5 (N1)60 = 16
(N1)60 = 24
(N1)60 = 30
0
0 1 2 3 4 5 6 7 8 9 10
vo' (atm)
Figure20. Calibrationresultsshowingrelationshipbetweenm_hfac1,(N1)60,andvo.
1.4
m_hfac1 not NCEER Dr = 0.40, f = 0.8
restricted NCEER Dr = 0.60, f = 0.7
1.2
NCEER Dr = 0.80, f = 0.6
UBCSAND: (N1)60 = 4, Dr ~ 0.29
1.0 UBCSAND: (N1)60 = 11, Dr ~ 0.49
UBCSAND: (N1)60 = 18, Dr ~ 0.63
UBCSAND: (N1)60 = 28, Dr ~ 0.78
0.8
Ksigma
m_hfac1
0.6 restricted
0.4
0.2
0.0
0 2 4 6 8 10
Sigvo' / Patm
Figure21. KvaluesestimatedusingUBCSAND904aRandgenericinputparameters.
Page|29
UBCSANDConstitutiveModelversion904aR February2011
25
Number of Cycles to Liquefaction
20
15
10
0
0.5 1 2 3 5 8 10
SIGvo' / Patm
N= 3 N= 5 N= 9 N = 15 N = 20 N = 27
Figure22. Predictednumberofcyclestoliquefactionversus'voand(N1)60usingNCEER
Krelationshipandgenericequationsform_hfac1.
4.7 Effect of Ko
The liquefaction response predicted by UBCSAND is a function of Ko, the ratio of the
initialhorizontaleffective stresstotheinitialverticaleffectivestress.Somerelationship
isanticipatedbetweenKoandCRRsinceliquefactionresistanceshouldbeinfluencedby
theinitialmeaneffectivestress(e.g.,themeaneffectivestresshasastronginfluenceon
the shear stiffness response, which in turn relates to the pore pressure generation).
However, this relationship appears to be exaggerated in UBCSAND since the plastic
effectsofprincipalstressrotationarenotconsidered.
Page|30
UBCSANDConstitutiveModelversion904aR February2011
The rotation of principal stress can be a problem in any constitutive model based on
classical plasticity when the direction of maximum shear stress is not coincident with
thedirectionoftheappliedcyclicloading.Asashearloadisappliedtothemodeland
plasticstrainsaregenerated,thedirectionofmaximumshearstresswilltendtorotate
until it approximates the direction of the applied shear loading. This rotation should
generate plastic shear and volumetric strains, but these additional strains are not
consideredintheUBCSANDmodel.
TheeffectofKowasconsideredduringthecalibrationofUBCSAND.Severaltrendswere
noted when performing DSS simulations of undrained sands using UBCSAND with an
appliedcyclicloadequaltoCRR15.Itwasobservedthatsandswithrelativelylowvalues
of(N1)60tendedtoliquefymoreeasilyatKo=1thanatKo=0.5.Incontrast,sandswith
highvaluesof(N1)60tendedtoliquefymoreeasilyatKo=0.5thanatKo=1.0.Itwasalso
notedthatthisrelationshipwasafunctionof'vo.Atalowstressasandmightliquefy
moreeasilyatKo=0.5,whilethatsamesandwouldliquefymoreeasilyatKo=1.0ata
higherstresslevel.
TohelpevaluatethetrendofliquefiabilityversusKo,thepredictednumberofcyclesat
eachKowasconvertedintoanequivalentCRR15usingtheweightingcurvesdeveloped
forUBCSANDinSection4.3.Theresultingcurvesrevealthedirectrelationshipbetween
CRR15andKoasshowninFigure23.ThisrelationshipisseentobefairlysubtleforKo<1
and (N1)60 > 10 15. Significantly reduced values of CRR15 (i.e., more liquefiable) can
occur for smaller (N1)60 values with Ko near unity. All values of (N1)60 tended to be
significantlymoreresistanttoliquefactionathighervaluesofKo.
Page|31
UBCSANDConstitutiveModelversion904aR February2011
3
(N1)60 = 2, Sigvo' = 1 Atm
(N1)60 = 5, Sigvo' = 1 Atm
2.5
(N1)60 = 10, Sigvo' = 1 Atm
CRR_actual / CRR_15
0.5
0
0 0.5 1 1.5 2
Ko
a)'vo=1atm
3
(N1)60 = 5, Sigvo' = 1 Atm
(N1)60 = 5, Sigvo' = 4 Atm
2.5
(N1)60 = 15, Sigvo' = 1 Atm
CRR_actual / CRR_15
0.5
0
0 0.5 1 1.5 2
Ko
b)Comparisonof'vo=1atmand'vo=4atm.
Figure23. PredictedrelationshipbetweenKoandCRR15forUBCSAND904aR.
Page|32
UBCSANDConstitutiveModelversion904aR February2011
The rate of excess pore pressure generation predicted by UBCSAND was evaluated by
comparing published trends from laboratory tests with trends predicted from DSS
simulations.ThedatafromtheDSSsimulationsusedtheanalysespreviouslydescribed
inSection4.1.
Figure 24 compares the rate of excess pore pressure generation (ru) summarized by
Seed et al. (1976) to the predictions made by UBCSAND. The two sets of plots agree
well,withtwoexceptions.Theinitialrateofporepressuregenerationisrelativelyslow
for the (N1)60=5 material with Ko=0.5. This is related to the importance of the initial
stress state on the UBCSAND prediction for the low blowcount sand. The second
deviationoccurswhenliquefactionisapproached.TheSeedetal.trendshowsarather
smoothincreasetowardsafullyliquefiedstate,whiletheUBCSANDpredictionsbecome
a bit irregular as the element approaches liquefaction. Some of these fluctuations are
duetothecyclesofdilationandcontractionthatarepredictedneartheoccurrenceof
liquefaction. The UBCSAND predictions appear smoother and are in better agreement
withtheSeedtrendsifonlythemaximumrufromeachhalfcycleisplotted.
(N1)60 =5 Ko = 1.0
(N1)60 =5 Ko = 0.5
0.8 (N1)60 =15 Ko = 1.0
(N1)60 =15 Ko = 0.5
Excess Pore Pressure Ratio
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
Cyclic Ratio, N / Nliq
Figure24. RateofexcessporepressuregenerationfromUBCSAND904aRversustrend
reportedbySeedetal.(1976).
Page|33
UBCSANDConstitutiveModelversion904aR February2011
Figure 25 compares the relationship between factor of safety for liquefaction FSLIQ
versusru.ThepublishedtrendusedforthecomparisonisfromMarcusonetal.(1990).
FSLIQ is defined as the CSR that will liquefy the element in a given number of cycles
divided by the CSR that is actually applied to the element for that same number of
cycles. The corresponding ru value is the maximum value obtained during the given
number of loading cycles. The same combination of (N1)60, Ko, and CSR were
investigatedasshowninFigure24.ThevaluesofruversusFSLIQpredictedbyUBCSAND
give reasonable agreement with the published trend. UBCSAND appears to predict
somewhatlargerincreasesinporepressureduetosmallloadingcyclesthanwouldbe
expectedfromthepublishedinformation.
A direct comparison between the 904aR model and cyclic laboratory data in DSS is
showninFigure26throughFigure29.ThedataisfromtestsperformedattheUniversity
ofBritishColumbiaonFraserRiversand.Thesandwasreconstitutedbyairpluviationto
a relative density Dr of 40%. While the 904a version is not capable of simulating the
observed increase in pore pressure for the cases that do not have a shear stress
reversal,the904aRversionisseentogiveareasonablerepresentationoftheobserved
porepressureresponse.
TheinputparametersfortheUBCSANDanalysisweredevelopedbyfirstnotingthatthe
sample with no static bias and an applied CSR of 0.08 had liquefied in 17 cycles. The
UBCSAND model was then run under the same loading conditions using the generic
inputparameters.The(N1)60thatisusedtodefinetheseparameterswasthenadjusted
untiltheUBCSANDelementliquefiedin17cycles.An(N1)60valueof6.9wasrequiredto
achieve liquefaction in 17 cycles, which suggests Dr2(N1)60/43 for this sand. These
simulationswererunwithaninitialKoof0.5.
Figure 26 shows the comparison between the laboratory data and the UBCSAND
simulation for the case of no static bias. Some significant differences are noted. One
differenceistherateofporepressuregeneration,whichinitiallyincreasesataslower
rate than was observed in the laboratory. This is due, in part, to how principal stress
rotation is addressed in UBCSAND: stress rotation with no change in maximum stress
ratio produces an elastic response. This stress rotation component is relatively
significant in the early stages of the DSS simulation. This causes the initial load
increments in the simulation to generate a reduced plastic response until the internal
stresses rotate and become aligned with the applied load increments. For this
simulation, about 9 load cycles are required before the horizontal effective stress
becomes equal to the vertical effective stress. To address this simplification in the
UBCSAND model, the plastic stiffness parameters of the model have been softened
duringthecalibrationprocesssothatthefullexcessporepressureisstillachievedinthe
correctnumberofcycles.
Page|34
UBCSANDConstitutiveModelversion904aR February2011
a. TypicalrelationshipsbetweenruandFSLIQfromlaboratorydata
(fromMarcusonetal.1990)
1
(N1)60 = 5 Ko = 0.5 Nliq = 14.5
(N1)60 = 5 Ko = 1.0 Nliq = 6.0
0.8
(N1)60 = 15 Ko = 0.5 Nliq = 11.5
(N1)60 = 15 Ko = 1.0 Nliq = 15.0
Excess Pore Pressure Ratio
0.6
0.4
0.2
0
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6
FSLIQ
Figure25. RateofexcessporepressuregenerationfromUBCSAND904aRversustrend
reportedbySeedetal.(1976).
Page|35
UBCSANDConstitutiveModelversion904aR February2011
Another significant difference between the laboratory data and simulation is the
stiffness of the element prior to liquefaction. The UBCSAND model predicts a stress
strain response that is approximately 5 to 7 times stiffer than observed in the
laboratory. This is due to differences between the assumptions of the generic input
parametersand the specific properties associated with this sand. Themaximum shear
stiffness estimated from the laboratory data appears to be somewhat smaller than is
normally expected from sand with a relative density of 40%. For example, using the
TokimatsuandSeed(1987)relationshipbetweenGmaxand(N1)60,asandwithan(N1)60
of6.9andaninitialmeaneffectivestressofabout60kPawouldbeexpectedtohavea
Gmaxofabout65000kPa.Theinitialloadcyclesmeasuredinthelaboratoryandshown
on Figure 26 show a cyclic strain of approximately 0.10%. The anticipated secant
modulus atthis strain level isexpected to beabout80%of Gmax,orabout50000kPa.
The effect of a modest increase in pore pressure will tend to reduce this modulus
somewhat.Butthecorrespondingsecantmodulusdeterminedfromthelaboratorydata
is only 9000 kPa, or about 1/6 of the anticipated value. This difference between
observed and predicted response could be addressed through a materialspecific
calibration.
Toevaluatetheeffectofprincipalstressrotationonthepredictions,theanalysisshown
inFigure26wasrepeatedwithaninitialKoof1.0.UsingthisKovaluemeansthecyclic
loading will be coincident with the direction of maximum shear strain at the start of
loading. A new representative (N1)60 of 8.0 was selected to achieve liquefaction in 17
cycles.ThepredictedresultsareshownonFigure27.TheKo=1simulationisshownto
provideamuchcloserrepresentationtothelaboratorytestresults.Theincreaseinpore
pressure with load cycles is an almost identical match, and the plot of stress path is
moresimilar,particularlyintheearlierloadcycles.Thereisstilladiscrepancybetween
the stiffness revealed in the stressstrain plots, although the stiffness of the Ko = 1
simulationissomewhatsofterthanfortheKo=0.5simulation.
Figure28providesacomparisonforthesameconditionsandinputparametersexcept
for an initial static bias equivalent to =0.106. This produces a loading state with no
stress reversals on the xy plane. As with =0.0, the initial stiffness predicted by
UBCSAND and the generic input parameters is much larger than was observed in the
laboratory test, and the initial rate of pore pressure generation is much slower. As a
result,theUBCSANDsimulationrequiresadditionloadcyclesbeforeitliquefies:4cycles
forthelaboratorytestand13forthesimulation.Theloadingstiffnessofthesimulation
after several postliquefaction cycles generally agrees with that observed in the
laboratory.
Page|36
UBCSANDConstitutiveModelversion904aR February2011
1
0.8
0.6
ru
0.4 DSS data
UBCSAND
0.2
0
0 5 10 15 20
Number of Cycles
a)Rateofporepressuregeneration
10
Shear Stress Sxy (kPa)
DSS data
5 UBCSAND
-5
First 15 cycles of loading
-10
-0.2 -0.1 0 0.1 0.2
Shear Strain %
b)Stressstrainbehavior(first15cycles)
10
Shear Stress Sxy (kPa)
DSS data
5 UBCSAND
-5
-10
-10 -8 -6 -4 -2 0 2 4 6
Shear Strain %
c)Stressstrainbehavior
10
Shear Stress Sxy (kPa)
-5
-10
0 25 50 75 100
DSS data
v' kPa UBCSAND
c) Stresspath
Figure26. LaboratoryDSSandUBCSAND904aRusinggenericinputparameters
(Dr=40%,=0,Ko=0.5insimulation,andCSR=0.0826).
Page|37
UBCSANDConstitutiveModelversion904aR February2011
1
0.8
0.6
ru 0.4 DSS data
UBCSAND
0.2
0
0 5 10 15 20
Number of Cycles
a)Rateofporepressuregeneration
10
Shear Stress Sxy (kPa)
DSS data
5 UBCSAND
-5
First 15 cycles of loading
-10
-0.2 -0.1 0 0.1 0.2
Shear Strain %
b)Stressstrainbehavior(first15cycles)
10
Shear Stress Sxy (kPa)
DSS data
5 UBCSAND
-5
-10
-10 -8 -6 -4 -2 0 2 4 6
Shear Strain %
c)Stressstrainbehavior
10
Shear Stress Sxy (kPa)
-5
-10
0 25 50 75 100
DSS data
v' kPa UBCSAND
d)Stresspath
Figure27. LaboratoryDSSandUBCSAND904aRusinggenericinputparameters
(Dr=40%,=0,Ko=1.0insimulation,andCSR=0.0826).
Page|38
UBCSANDConstitutiveModelversion904aR February2011
1
0.8 DSS data
0.6 UBCSAND
ru
0.4
0.2
0
0 2 4 6 8 10 12 14 16
Number of Cycles
a)Rateofporepressuregeneration
25
Shear Stress Sxy (kPa)
15
10
5 DSS data
UBCSAND
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Shear Strain %
b)Stressstrainbehavior(first3cycles)
25
Shear Stress Sxy (kPa)
DSS data
20 UBCSAND
15
10
0
-2 0 2 4 6 8 10 12 14
Shear Strain %
c)Stressstrainbehavior
25
Shear Stress Sxy (kPa)
20
15
10
5
0
0 25 50 75 100
DSS data
v' kPa UBCSAND
d)Stresspath
Figure28. LaboratoryDSSandUBCSAND904aRusinggenericinputparameters
(Dr=40%,=0.106,Ko=0.5insimulation,andCSR=0.0867).
Page|39
UBCSANDConstitutiveModelversion904aR February2011
Figure29comparesresultsforthesameconditionsasforFigure28exceptforthiscase
theappliedCSRisonly0.06.Thesimulationandlabtestpredictasimilarresistanceto
liquefaction:22.5cyclestoliquefactionforUBCSANDversus15.5cyclesfortheDSStest.
As with the other comparisons, the initial shear stiffness in the UBCSAND analysis is
significantlystifferthanobservedinthelaboratorytest.Theothersignificantdifference
isthepostliquefactionstressstrainresponse.Theloadingstiffnessafterliquefactionin
UBCSANDisrelatedprimarilytotherateofdilationofthesoilskeletonandtheresulting
impactontheeffectivestress.Thepostliquefactionstiffnessobservedinthelabtestis
significantlylargerthantheUBCSANDprediction.
DSS tests without a static bias were performed on the same sand but at a relative
densityof80%.An(N1)60of28wasusedtodeveloptheinputparametersforUBCSAND
usingtherelationshipofDr2(N1)60/43.Figure30showscomparisonsbetweentheDSS
dataandtheUBCSANDsimulationsusingthegenericinputproperties.TheappliedCSR
wasequalto 0.29. TheUBCSAND analysisagain showsstiffer initialresponsethanthe
laboratory data. The UBCSAND element is predicted to liquefy in 21 cycles, while the
laboratory test showed liquefaction in approximately 11.5 cycles. The biggest
differencesbetweenthetestandsimulationareseeninthestresspathplotandalsoin
thestressstrainresponseafterliquefaction.However,theporepressuregenerationand
stressstrain behavior predicted by UBCSAND appear to be generally appropriate for
densesand.
Page|40
UBCSANDConstitutiveModelversion904aR February2011
1
0.8
ru 0.6
0.4 DSS data
UBCSAND
0.2
0
0 5 10 15 20 25 30
Number of Cycles
a)Rateofporepressuregeneration
20
Shear Stress Sxy (kPa)
15
10
5 DSS data
First 12 cycles of loading UBCSAND
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Shear Strain %
b)Stressstrainbehavior(first12cycles)
20
Shear Stress Sxy (kPa)
DSS data
15 UBCSAND
10
0
-5 0 5 10 15 20
Shear Strain %
c)Stressstrainbehavior
20
Shear Stress Sxy (kPa)
15
10
0
0 25 50 75 100
DSS data
v' kPa UBCSAND
d)Stresspath
Figure29. LaboratoryDSSandUBCSAND904aRusinggenericinputparameters
(Dr=40%,=0.106,Ko=0.5insimulation,andCSR=0.0662).
Page|41
UBCSANDConstitutiveModelversion904aR February2011
1
0.8
ru 0.6
0.4
DSS data
0.2 UBCSAND
0
0 5 10 15 20 25 30
Number of Cycles
a)Rateofporepressuregeneration
40
Shear Stress Sxy (kPa)
20
20
20
-20
-40
0 25 50 75 100
DSS data
v' kPa UBCSAND
d)Stresspath
Figure30. LaboratoryDSSandUBCSAND904aRusinggenericinputparameters
(Dr=80%,=0.0,Ko=0.5insimulation,andCSR=0.29).
Page|42
UBCSANDConstitutiveModelversion904aR February2011
5 Post-earthquake analysis
The framework of the UBCSAND constitutive model was derived from observations
madeonlaboratoryelementtests.Keyaspectsofthemodel,includingtherelationships
between shear stiffness/effective stress/stress ratio, and between plastic volumetric
strain/shearstrain/stressratio,wereallderivedfromthesegeneralobservations.Basing
themodelframeworkonhighqualitylaboratorytestsallowsafundamentalapproachto
themodeldevelopment.
While the ability of the model to represent laboratory behavior can be demonstrated
through the simulation of element tests, a critical requirement for the model is to
ensurethatitcansimulatethebehaviorobservedinthefield.Suchfieldbehaviorcanbe
verycomplexduetomanyfactorsthatareonlyapproximatedinthelaboratory,suchas
complex 3dimensional loading, pore pressure drainage, and stratigraphy on both a
largeandsmallscale.
Oneaspectthatmaynotbeadequatelyaddressedbyalaboratorybasedmodelisthe
prediction of residual, or postliquefaction, strengths. These strengths have been
inferred from field case histories through the back analysis of observed slumps and
slides (Seed and Harder, 1990; Olson and Stark, 2005). The low strength values
estimatedfromthesecasehistoriesarelikelyaffectedbycomplexmechanisms,suchas
pore water inflow, void ratio redistribution, and stratigraphic mixing. While UBCSAND
willpredictasignificantlysoftenedstressstrainbehaviorafterliquefaction,theresulting
mobilized strength may not be consistent with common interpretations of residual
strength.
To implement the postearthquake analysis, the input motion is terminated and the
model is allowed to run for a period of time to allow any residual motion to decay.
ZoneswithpeakruvaluesexceedingtherulimitarethenconvertedfromtheUBCSAND
model to the simpler MohrCoulomb model. An undrained strength equal to the
Page|43
UBCSANDConstitutiveModelversion904aR February2011
residualstrengthisassignedandlowvaluesofshearandbulkmoduliareused.Forthe
SuccessDamanalyses,theshearmodulusofliquefiedzoneswastakenas10timesthe
residual strength, and the bulk modulus was assigned a value equal to 100 times the
shearmodulus.ThenonliquefiedzonescontinuetousetheUBCSANDmodeltopermit
a more accurate prediction of stress strain response due to load redistribution. The
analysis is then continued in dynamic mode and the model deforms until stability is
regained.
The ability of UBCSAND to dilate significantly with strain allows large strengths to be
mobilizedintheseelements,strengthsthataresignificantlyhigherthanwouldtypically
be used in a postliquefaction stability evaluation. Although these strengths could
developinthefield,itislikelytheywoulddegradeasporewaterflowedintothedilating
zonesfromadjacentareas.Toaddressthisconcern,theabilityforUBCSANDtomobilize
strengththroughdilationaftertheearthquakewaslimitedtoamaximumofthedrained
strengthdeterminedineachelementatthestartoftheearthquake.
The excess pore pressure ratio ru in any element has traditionally been defined as
ru=(uuo)/'vo,whereuis theporepressureat thetimeruis defined,uo is theinitial
pore pressure, and 'vo is the initial vertical effective stress. This definition for ruwas
developed for simple 1D situations with horizontal motion such as represented by a
SHAKE analysis column. In these situations, the vertical total stress does not change
duringtotheearthquake.ruequalszeroatthestartoftheearthquake,andwillequal1
attheinstanttheeffectivestressesbecomezero.Thepurposeoftheruparameteristo
giveanormalizedmeasureoftheporepressureincrease,with0indicatingnoincrease
and1indicatingastateofliquefaction.
Thetraditionaldefinitionforruissomewhatproblematicinageneral2Danalysis.Total
stresses change during the earthquake due to temporary fluctuations as well as
permanent changes due to stress redistribution. The traditional definition for ru can
show large fluctuations during the earthquake that are not related to liquefaction.
Because ofpermanentchangesintotalstress,thepeakvalue ofrucorrespondingtoa
liquefiedelementmightbeverydifferentthan1,oftenwithinarangeofperhaps0.7to
1.5.
Asmallchangecanbemadetothetraditionaldefinitionofruthatmaintainstheoriginal
intent of this index. The excess pore pressure ratio can be defined as ru=1 'v/'vo,
where 'v is the vertical effective stress at the time that ruis defined. This definition
maintainsmuchofthecharacterofthetraditionaldefinitionalthoughitstillsuffersfrom
fluctuationsinrurelatedtonormalstresschanges.However,thenewrunowequals0at
the start of loading and 1 at the instant the effective stresses vanish. The improved
stability in estimating ru values near 1 is needed when ru is used as a criterion for
definingliquefiedzones.
Page|44
UBCSANDConstitutiveModelversion904aR February2011
The ability of the modified UBCSAND model to predict the behavior observed in case
histories was evaluated by analyzing the Upper and Lower San Fernando dams and
predicting their response to the 1971 San Fernando earthquake. The cross section
geometryandearthquakeloadingmodeledforthesecasehistoriesfollowstheoriginal
interpretationofSeedatal.(1973).TheUBCSANDparametersweredefinedusingthe
genericinputparameters.Theuseofbothmedianand33rdpercentileblowcountswas
investigated.
Observedseismicresponse
The magnitude 6.6 San Fernando earthquake occurred on February 9, 1971. The dam
waslocatednearthewesternedgeoftheobservedfaultrupture.Indicationsofpossible
surfacerupturewereobservedwithinthereservoiroftheLowerSanFernandodama
Rolled Fill
18 m
HF Sand Clay
Core HF Sand
Upper Alluvium
Lower Alluvium
Bedrock
Figure31. RepresentativecrosssectionofUpperSanFernandodam.
Page|45
UBCSANDConstitutiveModelversion904aR February2011
short distance below the dam. Peak ground accelerations (PGA) at the site were
estimated to be about 0.55 to 0.6 g (Seed et al. 1973). This compares well with the
medianPGAestimatedfromseveralcurrentattenuationrelationships(SSA1997).
The occurrence of liquefaction was suggested by increased water levels in the three
standpipe piezometers within the embankment. Water overflowed from two of these
instruments.Asinkholewasalsoobservedinthedownstreamshellaboveacrackinthe
outletconduit.
Seismicloading
TheinputmotionselectedforthisanalysiswasthePacoimadamrecordasmodifiedby
Seed et al. (1973) and shown on Figure 32. Although the input motion appears
reasonableforanearfieldrecord,theactualseismicloadingexperiencedbythedamis
notknown.Thisisacommonconcerninbackanalysissinceseeminglyminordifferences
in the character of the input motion may produce a pronounced effect on the
displacementresponse.
Theinputseismicmotionwasconvertedtoanequivalentshearstresshistoryandthen
appliedtoacompliantboundaryatthebase.Acompliantboundarywasusedtoreduce
unintendedreflectionsoffofthebaseofthemodel.Theresultingmotionatthebaseis
similar to the within motion that would be estimated in a SHAKE analysis, although
the FLAC motion also incorporates the twodimensional influence of the overlying
foundationandembankment.
The orientation of the input stress history (i.e., positive or negative polarity) was
selected so that the direction of the large velocity pulse in the model was reasonably
consistent with the orientation of the pulse at the Pacoima Dam recording site.
Maintaining a similar orientation was considered potentially important due to the
pronouncednearfieldcharacterofthetimehistory.
(N1)60characterization
Representative (N1)60 blowcounts for the hydraulic fill shells are given in Table 2. The
values are based on SPT tests performed during April and May 1971 and have been
correctedforconfiningstress, energyratio,and theestimateddensification causedby
the earthquake. Both 33rd percentile and median values are provided and are
Page|46
UBCSANDConstitutiveModelversion904aR February2011
designated(N1)6033and(N1)6050,respectively.The(N1)6033isintendedtogiveameasure
ofthelooserfractionofthesoilunit.
The(N1)6050valuesinTable1havebeenmodifiedfromthosepublishedbyHarderand
others(Harderetal.1989,SeedandHarder1990).Thecorrectionusedforearthquake
inducedvolumetricstrainwasrevisedtoreflectthekinematicdeformationspredictedin
finite difference analyses (Beaty 2001). In other words, the revised corrections were
based on smaller estimates of volumetric strain since a portion of the observed
settlementswereattributedtothemovementofthesoilmassratherthandensification.
Inaddition,thedistributionofblowcountswithinthelowesthydraulicfillzonebeneath
thedownstreamshellandthezonedescribedasUpperAlluviumbeneaththeupstream
shellaresimilar.Sincetherewererelativelyfewdatapointswithineachofthesezones,
andmuchoftheUpperAlluviumzonewastentativelyloggedashydraulicfillduringthe
drilling,theirblowcountswerecombinedtoproduceanaveragedistributionatthebase
oftheembankment.
Staticanalysis
ThestaticanalysiswasperformedinFLACusingahyperbolicstressstrainmodelbased
0.3
Acceleration
-0.3
-0.6
0.8
pgv = 0.77 m/s
(m/s) .
Velocity(m/s)
0.4
Velocity
-0.4
0 5 10 15 20
Time (sec)
Figure32.
ModifiedPacoimaDammotionfrom1971SanFernandoearthquake.
Page|47
UBCSANDConstitutiveModelversion904aR February2011
Table1.Cleansandcorrectedblowcountsofhydraulicfill.
Depthbelow
Zone crest(m) (N1)6050 (N1)6033
HF(upper) 7.014.6 10 7
HF(lower) 18.622.2 13 9
on the Duncan formulation (Duncan and Chang 1970). The construction and loading
sequencewasapproximatelymodeledbybuildingtheembankmentmodelinlayersand
thenraisingthereservoirinstages.Theseepagecalculationswereperformedusingthe
groundwater flow capabilities of FLAC. This process gave a reasonable if simplified
estimateofinitialeffectivestressesandseepageforces.Thematerialpropertiesusedin
thestaticanalysis,includingstiffness,density,andstrength,werebasedprimarilyonthe
testing and data evaluation performed during the 1973 study (Seed et al., 1973). The
permeabilityvalueswereapproximatedfromtheAtterberglimitsandgradationsusing
various empirical relationships, including adaptations of the KozenyCarman equation
(Carrier,2003;Aubertinetal.,2005).TheselectedvaluesareshownonTable2.
Seismicanalysis
Theseismicanalysis,includingliquefactionresponseanddeformations,wasperformed
inFLACusingvariousconstitutivemodels.TherevisedUBCSANDmodelwasusedforthe
liquefiablehydraulicfillshellzones.AhystereticmodeldevelopedprimarilyatUBCwas
usedfortheloweralluvium,clayeycore,androlledfillzones.Andalinearelasticmodel
wasassignedtotheunderlyingrock.Zonesdefiningtherockatthebaseofthemodel
arerequiredaspartofthecompliantbasedefinitioninFLAC.
Page|48
UBCSANDConstitutiveModelversion904aR February2011
Table2.Propertiesusedforstaticanalysis,USFD.
G tan gent 1 R f
K ge Patm m
failure
Patm
2 Definesrelationshipbetweenelasticbulkmodulusandmeanconfiningstress:
me
B K b Patm m
Patm
Page|49
UBCSANDConstitutiveModelversion904aR February2011
1
Clay Core
Kge = 650
0.8
Modulus Reduction Factor
0.6
0
0.0001 0.001 0.01 0.1 1 10
70
30
Lower Alluvium
20
Kge = 2400
10
0
0.0001 0.001 0.01 0.1 1 10
Page|50
UBCSANDConstitutiveModelversion904aR February2011
0.8
Lower Alluvium
Modulus Reduction Factor
Kge = 2400
0.6
0.4
Average (Seed & Idriss 1970)
Idriss 1999 (20 to 50 feet)
0.2 Idriss 1999 (50 to 120 feet)
Rf= 0.30 Esyy = 2116. psf phi = 37 deg
Rf= 0.30 Esyy = 6348. psf phi = 37 deg
0
0.0001 0.001 0.01 0.1 1 10
70
Lower Alluvium
30 Kge = 2400
20
10
0
0.0001 0.001 0.01 0.1 1 10
Page|51
UBCSANDConstitutiveModelversion904aR February2011
TheadditionalpropertiesusedtodefinetheseismicanalysisareshowninTable3.The
range in undrained strength values for the clayey core is approximate. The range was
estimated from torvane test results (Seed et al., 1973) as well as limited CPT tests
reportedbyBardet(1995).ThegenericUBCSANDpropertieswereusedinconjunction
with blowcounts corrected to clean sand conditions. The primary analyses assumed
properties based on (N1)6050 (or median) blowcounts The Idriss and Boulanger (2006)
correctionfor finescontentwasused, which added 5blowsto eachblowcount foran
average fines content of 25%. The Idriss and Boulanger (2007) curve for residual
strengthSrwasusedasshowninFigure35.Thefinescontentcorrectionfor(N1)60andSr
was2blowsforanaveragefinescontentof25%.Theresidualstrengthwaslimitedto
thedrainedstrengthinanyelement.
Table3.Propertiesusedforseismicanalysis,USFD.
Su/'vo
Case1 0.13
Case2 0.25/0.13
3
100*cos(37)+
Su psf 'mo*sin(37)
'mo*sin(37)
K2max 30 52 110
V s 3300
Kge1 650 2400 1150
ne1 0.5 0.5 0.5
Rf1 0.3 0.3 0.3
Kb2 650 2400 1150
me2 0.5 0.5 0.5
G psf 4.7e7
B psf 6.3e7
1 Defineshyperbolicrelationshipforshearstressversusstrain.
2 Definesrelationshipbetweenelasticbulkmodulusandmeanconfiningstress.
3 Peakstrengthratio=0.25.Oncepeakstrengthisreachedinanelement,available
strengthratioreducesto0.13inthatelement.
Page|52
UBCSANDConstitutiveModelversion904aR February2011
Figure35. Residualstrengthcurve(copiedfromIdrissandBoulanger(2007)).
Baseanalysispredictions
The base analysis uses median blowcounts and a strength ratio of 0.13 in the clayey
core.ThestressstatejustbeforethestartofearthquakeloadingisshowninFigure36
while the initial pore pressure distribution is presented on Figure 37. This figure also
showsthegroundwaterlevelsmeasuredinthreeobservationwellsshortlybeforethe
earthquake.ThesesimplemeasurementssuggesttheFLACseepageanalysispredictsa
Page|53
UBCSANDConstitutiveModelversion904aR February2011
reasonableifsomewhatlowestimateofporepressurewithinthedownstreamshellof
theembankment.
ResultsfromthebasedynamicanalysesarepresentedinFigure38toFigure41.These
figures provide the final response predictions at the end of the postearthquake
analysis.
Figure 38 shows the extensive areas of high excess pore pressure that have been
predictedwithintheupstreamshellandnearthebaseofthedownstreamshell.Much
ofthesaturatedhydraulicfillintheupstreamshellispredictedtoliquefy,exceptfora
fairlysubstantialzonenearthecore.
Figure 39 shows contours of maximum shear strain predicted within the dam. The
highest shear strains occur near the base of the downstream shell and are associated
withapronounceddownstreammovement.Shearstrainswithintheupstreamshellare
smallerinmagnitudeandsomewhatmoredispersed.Thestrainsintheupstreamshell
indicateashallowcircularslipaswellasamoredeepseatedmovementalongthebase
oftheshell.
Figure 40 presents the final estimate of displacement vectors. The vectors show
predominantly downstream movement of the dam, which generally agrees with the
actual observations and measurements of dam response. A pronounced movement of
the upstream shell into the reservoir is also predicted. However, limited observations
madeaftertheearthquakedonotsuggestsuchlargemovementsoftheupstreamshell.
Thecrestispredictedtosettlealmostverticallywithlittlenetlateralmovement,while
actualmeasurementsshowthecrestmovingsignificantlydownstream.
Thedifferencesbetweenobservedandpredicteddisplacementaremostclearlyseenon
Figure41.Ingeneral,themagnitudeandorientationofthepredicteddisplacementare
in reasonable agreement with the observed response. The 904aR analysis appears to
overpredictmovementsoftheupstreamshellintothereservoir,whichaffectsboththe
lateralandverticalmovementspredictedatthecrest.Movementsalongthetopofthe
downstreambermappeartobewellpredictedbythe904aRanalysis,althoughitshould
be noted that the analysis does not include settlements due to postearthquake
consolidation. Displacements along the downstream slopeof the damare significantly
overpredictedbythemodel.Theactualdisplacementmeasurements,althoughlimited,
suggestthatstrainswithinthedownstreamshellweremoreuniformlydistributedover
theheightofthefill.Thisconclusionwasdevelopedduringtheinitial1973study(Seed
et al. 1973). In contrast, the strains predicted in the 904aR analysis tend to be
concentratednearthebaseofthehydraulicfill.
Page|54
UBCSANDConstitutiveModelversion904aR February2011
a.Effectiveverticalstress(psf)
b.Shearstressxy(psf)
Ko contours
2.00E-01
3.00E-01
4.00E-01
5.00E-01
6.00E-01
7.00E-01
8.00E-01
9.00E-01
1.00E+00
1.10E+00
c.Kocontours
Figure36. Predictedstressstateatstartofearthquake(baseanalysis).
Page|55
UBCSANDConstitutiveModelversion904aR February2011
Figure37. Predictedporepressuresatstartofearthquake(baseanalysis).
Ru contours
2.00E-01
4.00E-01
Original Final 6.00E-01
boundary boundary 8.00E-01
1.00E+00
Figure38. Peakestimatesofporepressureratio(baseanalysis).
Page|56
UBCSANDConstitutiveModelversion904aR February2011
Max. shear strain increment
1.00E-01
2.00E-01
3.00E-01
4.00E-01
5.00E-01
6.00E-01
7.00E-01
8.00E-01
Figure39. Contoursofmaximumshearstrainatendofpostearthquakeanalysis
(baseanalysis).
Peakdisplacement=2.9m
Figure40. Displacedshapeanddisplacementvectorsatendofpostearthquakeanalysis
(baseanalysis).
Page|57
UBCSANDConstitutiveModelversion904aR February2011
4
Horizontal Displacement (m)
Observed data
Closed circle: Serff et al., 1976
2
-2 Observed
Base Analysis
-4
2
Observed data
Vertical Displacement (m)
-1
Observed
-2
Base Analysis
-3
-50 0 50 100
Figure41. Predictedsurfacedisplacementsversusobserved(Serff,Harderetal1990).
Page|58
UBCSANDConstitutiveModelversion904aR February2011
Thebaseanalysisdemonstratesthatwhilethe904aRmodeldoesnotcapturealldetails
oftheobservedresponse,itdoespredictdisplacementbehaviorthatisgenerallysimilar
inbothmagnitudeandpatterntotheobservedbehavior.
Parametricanalysispredictions
SeveralparametricanalyseswereperformedasdescribedinTable4.Allanalyseswere
identicaltothebaseanalysisexceptasindicated.
Figure43comparesresultsfromthe904aand904aRmodels.Twoestimatesofsurface
displacementareshown:oneattheendofshakingandthesecondattheendofthe
postearthquake analysis. The 904aR model provides a better estimate of the final
horizontaldisplacementmagnitudes,althoughthe904amodelisseentogiveabetter
predictionofverticaldisplacements.Thedifferencebetweenthedisplacementestimate
attheendofshakingandattheendofthepostearthquakeanalysisissubstantialfor
the 904aR analysis and negligible for the 904a analysis. This change in the relative
importanceofthepostearthquakeanalysisisduetothepredictedextentofhighexcess
pore pressures below the downstream slope. The peak predicted pore pressure ratios
fromthesetwoanalysesareshowninFigure44.
Table4.Summaryofparametricanalyses,USFD.
(N1)60 Permeabilityof
for Su/'vo HydraulicFill HydraulicFill
AnalysisCase triggering forcore Model Shells
AnalysisA
median 0.13 904aR Table2
(baseanalysis)
0.25(peak)
AnalysisB median 904aR Table2
0.13(res)
Page|59
UBCSANDConstitutiveModelversion904aR February2011
Analysis E investigates the influence of the hydraulic conductivity of the hydraulic fill
shells.Thisanalysisisthesameasthebaseanalysisexceptthepermeabilityvalueinthe
shellshasbeenreducedbyafactorof1/10.Predictionsofsurfacedisplacementforthis
caseareshown inFigure45.Reducingthepermeabilityoftheshellscauseda modest
decrease in the predicted displacements at locations downstream of the crest. One
possible cause is the reduced ability in Analysis E for high pore pressures to migrate
fromzonessusceptibletoliquefactiontozonesthatarelesssusceptible.Thechangein
permeabilitydoeshavesomeimpactonthedistributionofpeakexcessporepressure,
asshowninFigure44.
These limited parametric studies help to confirm the results obtained from the base
analysisbydemonstratingonlymodestvariationsinthepredictedresponseforanalyses
using version 904aR. The studies also show the importance of the postearthquake
analysis to the prediction of displacements. In addition, the version 904aR model
providesasomewhatbetterestimateofbehaviorinthiscaseascomparedversion904a.
Page|60
UBCSANDConstitutiveModelversion904aR February2011
4
Observed data
Closed circle: Serff et al., 1976
Horizontal Displacement (m)
Observed
An C: Su13_N3350_904aRdr8
-4
Observed data
Closed circle: Serff et al., 1976
1 Open circle: Harder et al, 1989
Vertical Displacement (m)
-1
Observed
An C: Su13_N3350_904aRdr8
-3
Figure42. PredictedsurfacedisplacementsforanalysesA,BandC.
Page|61
UBCSANDConstitutiveModelversion904aR February2011
4
Observed data
Closed circle: Serff et al., 1976
Horizontal Displacement (m)
0
Observed
An. A: Base Analysis (EOS)
An. A: Base Analysis
-2
An. D: Su13_N50_904a (EOS)
An. D: Su13_N50_904a
-1
Observed
An. A: Base Analysis (EOS)
An. A: Base Analysis
-2 An. D: Su13_N50_904a (EOS)
An. D: Su13_N50_904a
Page|62
UBCSANDConstitutiveModelversion904aR February2011
ru Contours
1.00E-01
2.00E-01
3.00E-01
4.00E-01
5.00E-01
6.00E-01
7.00E-01
8.00E-01
9.00E-01
1.00E+00
a.AnalysisA(version904aR)
b.AnalysisD(version904a)
c.AnalysisE(version904aRandreducedshellpermeability)
Figure44. Contoursofpeakexcessporepressureratio,ru,foranalysesA,DandE.
Page|63
UBCSANDConstitutiveModelversion904aR February2011
4
Observed data
Closed circle: Serff et al., 1976
Horizontal Displacement (m)
Observed
-4
2
Observed data
Closed circle: Serff et al., 1976
Open circle: Harder et al, 1989
1
Vertical Displacement (m)
-1
Observed
-3
Figure45. PredictedsurfacedisplacementsforanalysesAandE.
Page|64
UBCSANDConstitutiveModelversion904aR February2011
7 References
Aubertin,M.,Chapuis,R.P.,andMbonimpa,M.(2005).DiscussionofGoodbye,Hazen;
Hello, KozenyCarman, by W. David Carrier III, ASCE Journal of Geotechnical and
GeoenvironmentalEngineering,August.
Bardet, J.P., Davis, C.A. (1996). Performance of San Fernando Dams during 1994
NorthridgeEarthquake,ASCEJournalofGeotechnicalEngineering,122(7),554564.
Byrne,P.M.,Cheung,H.,andYan,L.(1987).Soilparametersfordeformationanalysisof
sandmasses,CanadianGeotechnicalJournal,24(3),366376.
Carrier(III),W.D.(2003).Goodbye,Hazen;Hello,KozenyCarman,technicalnote,ASCE
JGGE,129(11),10541056.
Harder, L.F. Jr., and Boulanger, R.W. (1997) Application of K and K Correction
Factors,Proc.oftheNCEERWorkshoponEvaluationofLiquefactionResistanceofSoils,
Report NCEER970022, National Center for Earthquake Engineering Research, SUNY
Buffalo,N.Y.,pp.167190.
Harder, L. F., Hammond, W. D., Driller, M. W., and Hollister, N. (1989). The August 1,
1975Orovilleearthquakeinvestigation,Bulletin20388,Calif.Dept.ofWaterResources.
Idriss,I.M.(1999).Personalcommunication.
Idriss, I.M. and Boulanger, R.W. (2006). Semiempirical procedures for evaluating
liquefactionpotentialduringearthquakes,SoilDynamicsandEarthquakeEngineering,
26(2006),115130.
Idriss, I.M. and Boulanger, R.W. (2003). Estimating K for use in Evaluating Cyclic
Resistance of Sloping Ground, In Proc., Eighth U.S.Japan Workshop on Earthquake
Resistant Design of Lifeline Facilities and Countermeasures Against Liquefaction,
Technical Report MCEER030003, Multidisciplinary Center for Earthquake Engineering
Research.
Idriss,I.M.andBoulanger,R.W.(2007).ResidualShearStrengthofLiquefiedSoils,In
Proc.,27thUSSDAnnualMeetingandConference,March2007.
Page|65
UBCSANDConstitutiveModelversion904aR February2011
Marcuson,W.F.,Hynes,M.E.,andFranklin,A.G.(1990).EvaluationandUseofResidual
StrengthinSeismicSafetyAnalysisofEmbankmentDams,EarthquakeSpectra,Vol6.,
No.3,pp.529572.
Martin,G.R.,Finn,W.D.L.,andSeed,H.B.(1975).Fundamentalsofliquefactionunder
cyclic loading, Journal of the Geotechnical Engineering Division, ASCE, 101(GT5),
pp.423438.
Matsuoka,H.,andNakai,T.1977.StressstrainrelationshipofsoilbasedontheSMP.
In Proceedings of the Specialty Session 9, 9th International Conference on Soil
MechanicsandFoundationEngineering,pp.153162.
Rowe,P.W.1962.Thestressdilatancyrelationforstaticequilibriumofanassemblyof
particlesincontact.InProceedingsoftheRoyalSocietyofLondon,Mathematicaland
PhysicalSciences,SeriesA,269:500557.
Seed,H.B.,Martin,P.P.andLysmer,J.(1976).PoreWaterPressureChangesduringSoil
Liquefaction,JournaloftheGeotechnicalEngineeringDivision,ASCE,Vol.102,No.GT4,
pp.323346.
Seed,H.B.,Lee,K.L.,Idriss,I.M.,&Makdisi,F.(1973).AnalysisoftheSlidesintheSan
Fernando Dams during the Earthquake of Feb. 9, 1971. Report No. EERC 732,
EarthquakeEngineeringResearchCenter,UniversityofCalifornia,Berkeley.
Serff, N., Seed, H.B., Makdisi, F.I., and Chang, C.Y. (1976). Earthquake Induced
DeformationsofEarthDams,Rep.No.EERC764,Univ.ofCalif.,Berkeley.
Youd,T. L., Idriss, I.M., et al. (2001). Liquefaction resistance of soils: summary report
from the 1996 NCEER and 1998 NCEER/NSF workshops on evaluation of liquefaction
resistanceofsoils.J.Geotech.AndGeoenvir.Engrg.,ASCE,127(10),817833.
Page|66
UBCSANDConstitutiveModelversion904aR February2011
8 Appendices
Appendix 1:
Additional references for UBCSAND
Sloping ground with silt Centrifuge Naesgaard, Byrne, SeidKarbasi, & Park
layers simulation (2005)
Centrifuge
Slopingground Byrne,Park&Beaty(2003)
simulation
Note:Selectedpaperscanbedownloadedfrom
http://www.civil.ubc.ca/liquefaction/publications.htm
References:
Byrne, P.M., Park, S.S., Beaty, M., Sharp, M.K., Gonzalez, L., & Abdoun, T. (2004).
Numerical modeling of liquefaction and comparison with centrifuge tests, Canadian
GeotechnicalJournal,Vol.41(2):193211.
Byrne, P.M., Park, S.S. & Beaty, M. (2003). Seismic liquefaction: centrifuge and
numericalmodeling,inProc.,3rdInternationalFLACSymposium,Sudbury,October.
Page|67
UBCSANDConstitutiveModelversion904aR February2011
Naesgaard, E., Byrne, P.M., SeidKarbasi, M. & Park, S.S. (2005). Modeling flow
liquefaction, its mitigation, and comparison with centrifuge tests, Geotechnical
EarthquakeEngineeringSatelliteConference,Osaka,Japan
SeidKarbasi,M.,Byrne,P.M.,Naesgaard,E.,Park,S.S.,Wijewickreme,D.&Phillips,R.
(2005).ResponseofSlopingGroundwithLiquefiableMaterialsDuringanEarthquake:A
Class A Prediction, in Proceedings, 11th International Conference, International
AssociationforComputerMethodsandAdvancesinGeomechanics,Italy.
Yang,D.,Naesgaard,E.,Byrne,P.M.,Adalier,K.&Abdoun,T.(2004).Numericalmodel
verification and calibration of George Massey Tunnel using centrifuge models,
CanadianGeotechnicalJournal41(5):921942.
OtherReferences:
Beaty, M. & Byrne, P.M. (1998). An effective stress model for predicting liquefaction
behaviour of sand. In P. Dakoulas, M. Yegian, & R. D. Holtz (Eds.), Geotechnical
EarthquakeEngineeringandSoilDynamicsIII,ASCEGeotechnicalSpecialPublicationNo.
75,Vol.1,ProceedingsofaSpecialtyConference(pp.766777).Seattle:ASCE.
Puebla, H., Byrne, P.M., & Phillips, R. (1997). Analysis of CANLEX Liquefaction
Embankments:PrototypeandCentrifugeModels.CanadianGeotechnicalJournal,34(5),
641657.
Page|68
UBCSANDConstitutiveModelversion904aR February2011
Appendix 2:
Generic input parameters for UBCSAND 904aR
water bulk = ; Generic input parameters assumed fmod = 5e5 kPa
prop m_n160 = ; Assign appropriate value of (N1)60cs
prop m_pa = ; Assign value of atmospheric pressure in model units
prop m_phicv = 33.
prop porosity = 0.5
def properties
loop i (1,izones)
loop j (1,jzones)
;ELASTIC
$N160 = z_prop(i,j,m_n160)
z_prop(i,j,m_kge) = 21.7*20.* $N160^.333 ;Shear Mod
z_prop(i,j,m_kb) = z_prop(i,j,m_kge)*.7 ;Bulk mod
z_prop(i,j,m_me) = 0.5
z_prop(i,j,m_ne) = 0.5
;
;PLASTIC PROPERTIES
z_prop(i,j,m_kgp) = z_prop(i,j,m_kge)* $N160 ^2*.003 +100.0 ;shear Mod
z_prop(i,j,m_np) = .4
z_prop(i,j,m_phif) = z_prop(i,j,m_phicv) + $N160 /10.0
z_prop(i,j,m_phif) = z_prop(i,j,m_phif) + max(0.0,( $N160 -15.)/5.)
;
;plastic modification factors
z_prop(i,j,m_hfac2) = 1.0 ;Secondary hardener
z_prop(i,j,m_hfac3) = 1.0 ;dilation "hardener"
; m_hfac1 = a(N) * (Sigvo'/Patm)^b(N)
; where
; a(N) = 1.05 -0.03*N +0.004*N^2 -0.000185*N^3 +2.92e-6*N^4
; b(N) = 1./(-0.424 -0.259*N +0.00763*N^2)
$a_N = 1.05 -0.03*$N160 +0.004*$N160 ^2
$a_N = $a_N -0.000185*$N160 ^3 +2.92e-6*$N160 ^4
$b_N = 1./(-0.424 -0.259*$N160 +0.00763*$N160 ^2)
$SigP = max((-syy(i,j)-pp(i,j))/ z_prop(i,j,m_pa),1.0)
z_prop(i,j,m_hfac1) = $a_N * ($SigP)^$b_N
;
;failure ratio --same as in Hyperbolic model
z_prop(i,j,m_rf) = 1.1*$N160 ^(-0.15)
z_prop(i,j,m_rf) = min(z_prop(i,j,m_rf),.99)
;
;plastic anisotrophy
z_prop(i,j,m_anisofac) = 1.0 ; Generic parameters do not address anisotropy
z_prop(i,j,m_static) = 1.0 ; = 1.0 for initial static setup; = 0.0 for
dynamic
end_loop
end_loop
end
Page|69