Professional Documents
Culture Documents
LateralAnalysisforSeismicLoading
ofPileFoundations
RogerHartandVarun
ItascaConsultingGroup,Inc.
Minneapolis,Minnesota,USA
16October2012
Topics
1. A brief introduction to FLAC3D What is the
applicability of FLAC3D for dynamic SSI analysis of
pile foundations?
2. A recommended procedure for seismic analysis of
soil-pile interaction in liquefying soils.
3. Lateral load test calibration of pile-soil coupling
spring properties for a FLAC3D
simulation of
lateral/seismic loading of a pile foundation
What is FLAC3D?
FLAC3D is a general-purpose code that can simulate a full range of
nonlinear static & dynamic problems, with coupled fluid flow, heat transfer
and structural interaction. Any geometry can be represented, and the
boundary conditions are quite general.
FLAC3D simulates the behavior of nonlinear continua by the generalized
finite difference method (arbitrary element shapes), also known as the
finite volume method.
FLAC3D solves the full dynamic equations of motion even for quasi-static
problems. This has advantages for problems that involve physical
instability, such as collapse, as will be explained later. To model the
static response of a system, damping is used (dynamic relaxation) to
absorb kinetic energy.
FLAC3D contains an embedded language, FISH, that gives the user access
to all internal variables and allows custom-written functions.
FLAC3D
1.
2.
3.
4.
5.
6.
7.
8.
Basis of FLAC3D
FLAC3D uses an explicit, dynamic solution scheme to solve the
full dynamic equations of motion even for quasi-static problems.
This has advantages for problems that involve physical instability,
such as collapse.
To model the static response of a system, a relaxation scheme
is used in which damping absorbs kinetic energy. This approach
can model collapse problems in a more realistic and efficient
manner than other schemes, e.g., matrix-solution methods.
FLAC3D treats interactions between separate objects (e.g., zones
or structural elements) as boundary conditions; there is no
concept of a joint element. The Distinct Element Method
(DEM) is used for interactions.
Physical instability
Path dependence
Implementation of strongly nonlinear constitutive models for
example, strain-softening models or models that exhibit volumetric
collapse.
Strainrates are
fixed during this
calculation
x min
Cp
The input strain is fixed, and there are no influences from other
elements, during the step. No incremental stress/strain matrix is
needed the constitutive relations are used directly. Plastic
behavior requires no iteration (or return algorithm) during the step.
ConstitutiveModelsfor FLACandFLAC3D
Elasticitymodels:
Isotropic
Transverselyisotropic
Orthotropic
BuiltinModels
Plasticitymodels:
DruckerPrager
MohrCoulomb
Ubiquitousjoint
Strainhardening/softening
Bilinearstrainhardening/softening/ubiquitousjoint
Doubleyield
ModifiedCamclay
HoekBrown
Cysoil frictionhardening,withellipticalcap
Chsoil simplifiedCysoil (alternativetoDuncanChang)
DynamicLiquefactionmodels:
Finn(Martinetal.,1975)model
Bryne,1991model
Creepmodels:
Viscoelastic
Burgerssubstanceviscoelastic
Twocomponentpowerlaw
Referencecreepformulation(WIPP)
Burgercreep/MohrCoulombviscoplastic
Twocomponentpowerlaw/MohrCoulombviscoplastic
WIPPcreep/DruckerPrager viscoplastic
Crushedsalt
UserdefinedModels*
Elasticitymodels:
Hyperbolicelastic
DuncanChang,1980
Plasticitymodels:
NorSand
Jardine etal.,1986
ManzariDafalias,1997
Kleine etal.,2006
Concretehydration
vonWolffersdorff hypoplastic
DynamicLiquefactionmodels:
UBCSAND
UBCTOT
Wang,1990
Rothetal.,2001
Andrianopoulos,2005
Creepmodels:
Minkley viscoplastic
Heincrushedsalt
Salzer creep
Lubby2creep
*partiallistofmodelscreatedby
(ordevelopedfor)codeusers
Deformation Properties
for an Effective Stress Analysis
Material
Soil 1
moist unit wt. sat. unit wt. porosity dry unit wt. dry dens.
(pcf)
(pcf)
(pcf)
(slugs/ft3)
126
132
0.35
110.0
3.421
Vs
(ft/sec)
856
Vp
(ft/sec)
2138
G
(psf)
2.510E+06
K
(psf)
1.230E+07
.
.
.
Strength Properties
for an Effective Stress Analysis
Material
Soil 1
.
.
.
drained cohesion
(psf)
0
drained friction
(degrees)
30
(N1)60
c1
c2
10
0.4892
8.176E-01
residual strength
(psf)
400
To simulate liquefaction:
e.g., Finn-Byrne model
RF
KW / n
K 43 G
Where KW is the water bulk modulus, n is the porosity and K & G are the drained
elastic bulk and shear moduli.
There is a temptation to decrease the water modulus, in order to increase the time step
(and reduce the simulation time). There are two cases to consider:
1.
Using the correct water modulus ( 4.1 10 7 psf for pure water), if RF 20 then it may
be decreased such that RF 20 without affecting the results significantly.
2.
If RF 20 using the correct fluid modulus, then that value should be used.
As an example, K w 4.1 10 6 psf is considered representative of saturated sandy soils
(Chaney, 1978).
In the case of models with a large range of elastic moduli, the fluid modulus can be made
to depend on local values of moduli, provided that the above conditions are respected.
Elastic/plastic models
The built-in models in FLAC3D consist of various elastic/perfectlyplastic relations. There is only hysteresis for cyclic excursions that
involve yielding.
stress
strain
Rayleighdamping
RayleighdampingmaybeusedinFLAC3D asanapproximationtohysteretic
(frequencyindependent)damping.Twoviscouselementsareusedtomakeup
thedampingmatrix:
Themassproportionaltermislikeadashpotconnectingeachgridpointto
ground.Thestiffnessproportionaltermislikeadashpotconnectedacross
eachzone(respondingtostrainrate).
Althoughbothdashpotsarefrequencydependent,anapproximately
frequencyindependentresponsecanbeobtainedoveralimitedfrequency
range,bytheappropriatechoiceofcoefficients.
Rayleighdamping cont.
Combinedcurvereaches
minimumat:
ratioofdampingtocritical
combined
stiffnessproportionalonly
massproportionalonly
frequency
Note3:1frequencyrangeoverwhich
combineddampingisalmostconstant
10
Rayleighdamping cont.
ThedrawbackswithRayleighdampingarethat:
1.
Thecenterfrequencymustbechosen fromsometimes
conflictingdata(e.g.,thesiteresonanceortheearthquakeaverage
frequency)
2.
Thestiffnessproportionaltermcausesthetimesteptobereduced
asthedampingratio(lambda),atthehighestnaturalfrequency,is
increased:
HystereticDampinginFLAC3D
FLAC3Dprovides anoptionalhystereticdampingfunctionfordynamic
simulations.Thedampingisindependentofthematerialmodels,andconsists
ofastraindependentmultiplieronthetangentshearmodulus.
M s ( )
d
dM s
Mt
Ms
d
d
/ Go
Go = smallstrain
shearmodulus
M s secant modulus
M t tangent modulus
1.2
1
Modulus reduction factor
Ifthesecantmodulusisgivenbya
degradationcurve,thenthetangent
moduluscanbederived:
0.8
0.6
0.4
0.2
0
0.0001
0.001
0.01
0.1
10
Cyclic strain %
FromSeed&Idriss (1970)
shear stress
shear strain
11
Givenaparticularmodulusdegradationfunction,theresultingtangent
modulusisusedtomultiplytheapparentshearmodulus(Go)providedbythe
constitutivemodel:
G M tG
JOB TITLE :
Theapparentstrainisthe
deviatoricstrainaccumulated
sincethepreviousreversal
point.Suchreversalpointsare
keptinastacksothat
embeddedcycleswithina
maincyclemaybefollowed.
LEGEND
12-Feb-03 15:39
step
3700
+04
2.000
HISTORY PLOT
Y-axis :
Ave. SXY
( 1, 1)
1.000
X-axis :
X displacement( 1, 2)
0.000
-1.000
-2.000
-40
-20
20
40
(10 -05 )
elasticmodelwithhystereticdamping
Thus,energyisdissipatedforminiloopsaswellasthemain
hysteresisloop.
Thehystereticdampingformulationhasthreeadvantages.
1.
StandardG/Gmax degradationcurvesusedinequivalentlinear
analysesmaybeuseddirectlyinFLAC &FLAC3D,toperformfully
nonlinearsimulationswiththesamematerialresponse.
2.
Thedampingdoesnotaffectthetimestep(incontrasttoRayleigh
damping,whichmayprofoundlyreducethetimestep).
3.
Thedampingmaybeusedwithanymaterialmodel,andwithanyof
theotherdampingschemes(optionally)active.
Onedisadvantageisthatpublisheddegradationcurvesseemtobe
inconsistent i.e.,ahystereticmodelthatconformstotheG/Gmax curve
doesnotnecessarilyconformtotheassociateddampingcurve
1.2
60
50
40
0.8
Seed data
FLAC - Sig3 fit
0.6
20
0.2
10
0.001
0.01
0.1
0
0.0001
Seed data
30
0.4
10
0.0001
0.001
0.01
0.1
10
12
(10
-01
2.000
1.500
1.000
SHAKEanalysis
0.500
0.000
-0.500
-1.000
-1.500
-2.000
-2.500
10
15
20
25
30
35
1.2
72ft
1
0.8
0.6
57.4ft
0.4
0.2
0
0.0001
0.001
0.01
0.1
Cs 630 ft /sec
G / Gmax 0.85
Cs 610 ft /sec
G / Gmax 0.80
107 pcf
113 pcf
10
strain - %
36ft
Cs 1122 ft/sec
ModulusReductionCurve
125 pcf
Dr 2.5%
Dr 4.0%
G / Gmax 0.90
Dr 2.0%
30
25
Damping ratio
20
15
10
5
0
0.0001
0.001
0.01
0.1
10
strain %
DampingRatioCurve
13
30
25
0.8
20
Damping ratio
1.2
0.6
15
0.4
10
0.2
0
0.0001
0.001
0.01
0.1
0
0.0001
10
0.001
0.01
(10
03
10
shear strain vs
shear stress for soil
with Rayleigh
damping:
G/Gmax = 0.8
Dratio = 4% and
freq. = 1.0 Hz
1.000
0.000
-1.000
-2.000
-20
-15
-10
-5
10
15
20
(10
-04
Selection of parameters
for hysteretic damping
(10
Select hysteretic
damping curve to
approximately fit
modulus reduction and
damping ratio curves
over expected strain
range
3.000
2.000
0.1
strain %
strain - %
03
shear strain vs
shear stress for
soil with default
hysteretic
damping model:
L1 = -3.156
L2 = 1.904
2.000
1.000
0.000
-1.000
-2.000
-20
-15
-10
-5
10
15
20
(10
-04
14
Liquefaction - principles
expansion
compaction
15
CyclicStressRatio(CSR)
(Byrne,P.M.,Naesgaard,E.,andSeidKarbasi,M.,2007)
CyclicSimpleShearTest
Roth et al Model
i
D i = 0.5 / N L
SchematicshowingbasicprincipalsofRothetalmodel
Roth, W.H., Bureau, G., Brodt, G., (1991) Pleasant Valley Dam: An Approach
to Quantifying the Effect of Foundation Liquefaction, 17th International
Congress on Large Dams, Vienna, 11991223.
16
Finn-Byrne Model
Forsimpleshearloading(dryconditions):
1.0
0.8
B
v
0.4
C1 exp
C1
C1 8.7 N1 60
B
v
1.25
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
Amplitudeofcyclicshearstrain
-1.0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
x10^2
Volumetricstrainincreases
withlevelofcyclicshearstrain
vB
C1
Forgiven,rateofaccumulation
decreaseswiththenumberofcycles
Volumetricstrainincreaseswhen
SPT
( N1 )60 decreases
Numberofcycles
Byrne, P. M. (1991) A Cyclic ShearVolume Coupling and PorePressure Model for Sand, in
Proceedings : Second International Conference on Recent Advances in Geotechnical Earthquake
Engineering and Soil Dynamics (St. Louis, Missouri, March 1991), Paper No. 1.24, 4755, 1991
Pile/SoilInterface(defaultbehavior)
shearstiffness,cohesion,friction,
normalstiffness,cohesion,friction,
perimeter
17
SeismicInput
Mejia&Dawson (seeproceedingsofthe4th InternationalFLAC Symposium
paper0410)presentaverycleardescriptionofthewaysinwhichseismicinput
maybeappliedtoamodel.(Thefiguresinthissectionarereproducedfromtheir
paper,withpermission).
Therearetwomainoptions
1.
2.
Rigid base(velocityoraccelerationhistoryapplieddirectly)
Flexible base(velocityhistoryconvertedtoappliedstresshistory)
Ifthetargetmotionisprovidedforanylocationexceptforthebaseofthe
model,thendeconvolution isnecessary,todevelopatimehistorytobe
appliedatthemodelbasesuchthatthesimulation wouldreproducethe
targetmotionatthespecifiedlocation,underfreefieldconditions(e.g.,no
structures).
NormallytheprogramSHAKEisusedfordeconvolution.SHAKEisan
equivalentlinearprogram,andisthusunabletofollownonlinearity
directly;itadjuststhesecantshearmodulusanddampingofeachlayer
iterativelytoobtaintheapproximateeffectofnonlinearity,averagedover
thewholetimehistory.
18
EarthquakeDeconvolution
Seismicinput
SHAKEworksinthefrequencydomain,
usingthesumoftheupward and
downwardpropagatingwaves.Ateach
interfacebetweenlayers,thereisan
analyticalsolutionforthereflected&
transmittedportionsofeachwave.By
solvingtheresultingsystemofequations,
transmissionbetweenanytwolocations
(e.g.,betweenbase&surface)maybe
computed.
(AfterMejia&Dawson,2006)
SHAKEinput&outputisavailableeither
1.
2.
19
(AfterMejia&Dawson,2006)
Notethatthewithin motion(at60m)istheactualmotionatthat
depth thesumofupward anddownwardpropagatingwaves.
TheuseofSHAKEtocomputetherequiredinputmotionfortherigidbase
ofaFLAC3D modelleadstoagoodmatchbetweenthetargetsurface
motionandthesurfacemotioncomputedbyFLAC3D,foramodelthat
exhibitsalowlevelofnonlinearity.(Theinputmotionalreadycontainsthe
effectofallthelayersabovethebase,becauseitcontainsthedownward
propagatingwave).
AdifferentapproachmustbetakenifFLAC3D istomodelmorerealistic
systems,suchas
1.
sitesthatexhibitstrongnonlinearity;or
2.
theeffectofasurfaceorembeddedstructure.
Inthefirstcase,therealnonlinearresponseisnotaccountedforbySHAKEin
itsestimateofthebasemotion.
Inthesecondcase,secondarywavesfromthestructurewillbereflected
fromtherigidbase,causingartificialresonanceeffects.
20
Formostsitesencounteredinpractice(exceptthosewheretheexistenceofaverystiffbedrock
justifiesarigidbase)aflexible basetotheFLAC3D modelshouldbeused.Inthiscase,thequiet
baseconditionisselected,andtheupwardpropagatingwaveonly fromSHAKEusedtocompute
theinputstresshistory.(Thisisderivedastheoutcrop velocityhistory,convertedtoastress
historybyusingtheformula
CS v ).
(AfterMejia&Dawson,2006)
Note that
FLAC3D choosesitstimestepfornumericalstabilitybasedonthepwavespeed
x
t min
CP
where
CP
K 43 G
Thus,small,stiffzones(elements)determinethetimestep.
Zonesizesshouldbechosensmallenoughtoresolvethesmallestwavelength:e.g.,
x min / 10
( C s / f )
21
time
time
x min / 10
x min /10
Checkfrequencycontentofinputrecord
JOB TITLE : Acceleration Record
-01
LEGEND
2.000
7-Mar-04 23:32
step
0
1.500
Table Plot
Table 1
1.000
0.500
0.000
-0.500
-1.000
-1.500
(10
-04
-2.000
-2.500
7-Mar-04 23:32
step
0
1.000
5
Table Plot
Table 2
0.800
10
15
20
25
30
35
time
0.600
0.400
Inputaccelerationrecord
0.200
10
15
20
25
frequency
Powerspectrum(fromFFT.FIS)
22
2D Extrusion Tool
Construction View
23
2D Extrusion Tool
ExtrusionView
FLAC3D Grid
24
FLAC3D Grid
with Wharf and Piles
pile element
25
6000(psf)
5000
4000
3000
2000
1000
0
Factor of Safety*
26
Apply free-field boundaries along the sides of the model after all
other grid conditions are set.
2.
3.
4.
27
vs
0.75
0.60
0.45
0.30
0.15
0.0
-0.15
-0.30
-0.45
-0.60
-0.75
28
vs
0.90
0.75
0.60
0.45
0.30
0.15
0.0
-0.15
-0.30
-0.45
-0.60
x-velocity at base
x-velocity at crest
-0.75
-0.90
29
2400(psf)
2000
1600
1200
800
400
0
30
PileDisplacementVectors at 20 sec
(with Finn-Byrne material)
max. displacement (at 20 sec.) = 3.65 ft.
PileMomentsat 20 sec
(with Finn-Byrne material)
PileMoments(kft)
1200
1000
800
600
400
200
0
200
400
600
31
General Comments
1. This model is composed of 156,000 hexahedral zones,
400 pile elements and 6480 shell elements.
2. A dynamic simulation for 10 seconds of seismic
loading requires approximately 1 day to complete on
an Intel four-core i7 (2.67GHz) computer. (FLAC3D is
multi-threaded, so faster runtimes can be expected on
computers with more cores.)
LateralLoadingofASinglePile
NumericalLoadTests
32
LateralLoadingofSinglePile
PileDesign:Responseofasinglepiletolateral
loading
Developcasespecificpycurvesaccountingfor
theeffectof
Soillayering
Rotationfixityconditionathead
Pilebendingstiffness
Pilecrosssection
GroundSlope
Axialload
Interactionwithotherloadsnearby
FLAC3DModel
ConcretePile:0.6m(2ft)diameterand6m
(16.4ft)long
Axialloadof100kN (22500lbs)
Embeddedinsoilwith=35 andc=1kPa(42
psf)
Freetorotateatthetop
33
SoilPileInterface
Interfaceneededbetweensoilandpile
elementstocapture
Notensionatpilesoilinterface(developmentof
gapbehindthepileindirectionoppositeto
loading)
Slipatpilesoilinterface
Properties:Normalstiffness,Shearstiffness,
Shearstrength(CohesionandFrictionangle),
TensilestrengthandDilationangle
InterfaceElement
34
GuidelinesforInterfaceProperties
Stiffness:
Notveryimportantasweareinterestedonlyin
slipandseparation
Usevaluesstiffenoughnottoaffectresultsbut
nottoostifftopenalizetimesteptoomuch.
4
K 3G
kn , k s 10 max
zmin
Cohesionandfriction:About2/3timesthatof
soilunlessbetterdataavailablefromtesting
FLAC3DGrid
35
Step1:StressInitialization(PileWeight)
Units:Pa
Step2:VerticalLoading
Units:Pa
36
Step3:LateralLoading
YieldingZones
Step3:LateralLoading(final)
37
Pycurves(Freehead)
Pycurves(Noheadrotation)
38
Slopingground(upslope)
Slopingground(downslope)
39
Pycurves(upslope)
Pycurves(downslope)
40
General Comments
1.This model is composed of 3276 hexahedral
zones
2.A dynamic simulation for 0.8 inches of lateral
movement at head requires approximately 30
minutes to complete on an Intel four-core i7
(2.67GHz) computer.
Final Comments
1. Know the limitations and capabilities of the numerical
analysis program you are planning to use. Does the
code have sufficient capabilities to simulate the
important conditions of the problem?
2. Start as simple as possible.
complexity as necessary.
It is better to add
41
42