Otc 23025

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

O

OTC 2302
25
S
Seismic Analysis
A
of Soil-Pile-Structture Interraction off Pile Sup
pported L
LNG
T
Tank
D
Dongdong Ch
hang, Nick O'R
Riordan, Mich
hael Willford, and John Pow
well, Arup

C
Copyright 2012, Offshore Technology Confere
ence
T
This paper was prepare
ed for presentation at the Offshore Technolog
gy Conference held in Houston, Texas, USA , 30 April3 May 2012 .
mitted by the author(s)). Contents of the pape
T
This paper was selected for presentation by an
a OTC program comm
mittee following review
w of information containned in an abstract subm
er have not been
re
eviewed by the Offshore Technology Confere
ence and are subject to
o correction by the autthor(s). The material dooes not necessarily re
eflect any position of the Offshore Technologyy Conference, its
officers, or members. Electronic
E
reproduction
n, distribution, or stora
age of any part of this paper without the wriitten consent of the O
Offshore Technology C
Conference is prohibite
ed. Permission to
re
eproduce in print is res
stricted to an abstract of
o not more than 300 words;
w
illustrations mayy not be copied. The abbstract must contain co
onspicuous acknowled
dgment of OTC copyrig
ght.

A
Abstract
T
The complexity
y of dynamic soil-pile-structure interaction for pile suppported Liquefi
fied Natural Gas (LNG) tankk has been
ddemonstrated th
hrough numeriical studies using different methods,
m
whichh raises the im
mportance of thhoroughly evalluating and
ccomparing diffferent methods to provide a more
m
challengin
ng appraisal off each methodds limitations aand general acccuracy. In
thhis paper, dy
ynamic soil-pille-structure in
nteraction analyses using tw
wo different nnumeric methoods were unddertaken to
ddetermine the seismic
s
tank reesponse and fou
undation demaands of a pile ssupported LNG
G tank in high sseismic regionn. The tank
is 60m in diam
meter and supported by 480 number
n
750m
mm circular piiles. The dynaamic soil-pile--structure interraction was
aanalyzed using both direct meethod and indirrect (substructu
ure) method.
IIn the direct method, nonlineear dynamic fin
nite element (F
FE) analyses w
were performedd in the time doomain using FE software
O
OpenSees with
h a fully coupleed constitutive model. The direct
d
method F
FE model conssisted of a finitte mesh of nonnlinear soil,
nnonlinear soil-p
pile interaction
n springs, pilees, and tank co
omponents. T
The FE model was shaken w
with a series oof recorded
eearthquake mo
otions that weere spectrally matched to th
he design speectra developeed from a proobabilisticseism
mic hazard
aassessment (PS
SHA) study.
IIn the substruccture method, the
t soil-foundaation and the superstructure
s
were treated sseparately, whhere frequency--dependent
ffoundation imp
pedances were calculated and
d modeled as sp
pring and dashhpot. The two methods weree evaluated andd compared
ffor their ability
y to capture and
d present the essential
e
dynam
mic features annd loading mecchanism of thee tank-foundatiion system,
aand sensitivities and limitation
ns of the two methods
m
are disscussed.
T
The direct meth
hod is considered preferable when the inerrtia effects are significant andd the soil respoonse is significcantly nonliinear. Howev
ver, with app
propriate calibrrations and asssumptions, thhe substructurre method is also able to reasonably
aapproximate the essential feattures of seismic response of the
t tank-foundaation system.
Introduction
D
Dynamic soil-p
pile-structure in
nteraction is siignificant in th
he seismic desiign and analyssis of pile-suppported LNG tannks, due to
thhe large mass of the tanks and
a the resultin
ng large inertiaal forces transfferred from thhe tank to the ffoundation to m
modify the
sseismic site reesponse. The complexity of seismic soil-pile-structure interaction foor LNG tankss has been dem
monstrated
thhrough numerrical studies using
u
differentt methods (e.g
g., Willford ett al. 2010, K
Kausel et al. 1978), which raises the
im
mportance of thoroughly
t
evaaluating differeent dynamic numerical
n
methhods to providee a more challlenging appraissal of each
m
methods limitaations and geneeral accuracy.
This paper describes and compares two different meth
hods undertakeen to carry out seismic assesssment of a pilee supported
L
LNG tank in a high seismic zone,
z
i.e., the direct
d
method and the indirecct method (subbstructure methhod). The direect method
ssolves soil, fou
undation, and tank
t
in one step by perform
ming an equivaalent two-dimeensional (2-D) nonlinear finiite element
((FE) analyses in
i the time dom
main using soft
ftware OpenSeees. OpenSees is a validated software platfform developedd by PEER
aand NEES (http
p://opensees.beerkeley.edu/). The direct meethod FE moddel consists of a beam-colum
mn model of thhe tank and
nnonlinear piles that are attach
hed to a 2-D no
onlinear soil mesh by nonlineear soil springss representing ssoil-pile interaaction. The
ssubstructure method
m
treats taank and found
dation separately by followinng a three stepp procedure prroposed by Kaausel et al.
11978, where eq
quivalent linearr frequency dep
pendent found
dation impedannces are calculaated and modeeled as a set of spring and
ddamper in the structural
s
modeel.

OTC 23025

The two methods are used to analyze the dynamic soil-pile-structure-interaction of the pile supported tank, and compared
for their abilities to capture the essential dynamic features and loading mechanism of the tank-foundation system. Finally
sensitivities and limitations of the two methods are discussed.
Site Conditions
Site Location.
The LNG terminal site is located between two sub-parallel north-west south-east trending faults in Southeast Asia. It is a
seismic zone 3 by the ISO standards.
Based on downhole shear wave velocity measurement, the site is NEHRP site class D.
The Tank.
The LNG tank is 60m in diameter with a 88,000 m3 volume, supported by 480 number concrete piles.
Ground Conditions.
The ground at the site is generally comprised of a downward succession of soft alluvium underlain by firm clay underlain
by stiff to hard clay. Table 1 describes the ground model and soil parameters used for the soil-pile-structure interaction
analyses of the tank.
Table 1: Design ground model for the LNG tank
3

Layer

Soil Type

Layer Thickness (m)

Unit Weight (kN/m )

PI (%)

SPT Blow Count, N

Alluvium

16

33

2-10

Firm Clay

16

18

26

25-50

Stiff Clay

10

19

26

50-60

Hard Clay

70

19

26

60-85

Groundwater is at about 1.5m below ground surface.


Soil strength and stiffness parameters were determined using standard correlations from in-situ and laboratory testing.
Dynamic soil properties were developed from standard penetration test (SPT) data and downhole shear wave velocity
measurement. Three sets of simplified design lines were developed, i.e., upper bound (UB), best estimate (BE), and lower
bound (LB) soil stiffness and strength properties.
Seismic Hazard.
A probabilistic seismic hazard assessment (PSHA) was carried out for the LNG Terminal site. Two design levels of
earthquakes were defined, the OBE (Operating Basis Earthquake) with a return period of 475 years, which corresponds to
10% probability of exceedance in 50 years; and the SSE (Safe Shutdown Earthquake) with a return period of 2475 years,
which corresponds to 2% probability of exceedance in 50 years. The design spectra were generated at the bedrock level at
the depth of 100m from the PSHA study, indicating a peak ground acceleration (GPA) around 0.19g for OBE events and
0.33g for SSE events, as shown in Figure 1.
8.0

475 years

7.0

Spectral acceleration (m/s2)

2475 years
OBE_Design
6.0

SSE-Design

Sa= 7.18/T

5.0

4.0

Sa= 3.88/T
3.0

2.0

1.0

0.0
0.0

2.0

4.0

6.0

8.0

10.0

12.0

Response period (sec)

Figure 1: Proposed horizontal design spectra at depth of 100 m for soil (5% damping)

Ground Motions.
Three sets of realistic acceleration time history records (seed motions) that are compatible with the regional tectonic
environment, seismic setting, and regional geology were selected from the NGA ground motion database. The selected seed

O
OTC 23025

m
motions were all
a recorded att the bedrock level of class D sites basedd on the NEH
HRP standards with VS-30 varrying from
3300m/s to 400m
m/s.
Th
he first set of motions
m
was reccorded at the CWB
C
99999 TA
AP051 station during the ChhiChi earthquakke on Sept.
20, 1999 in Taiw
wan;
he second set of motions was recorded at the USGS 289 W
Whittier Narrow
ws Dam stationn during the Hector Mine
Th
earrthquake on Occt.16, 1999 in California;
C
Th
he third set of motions
m
was reecorded at the CDMG 244000 LA Obregoon Park duringg the Landers earthquake
on
n June 28, 1992
2 in California.
T
The selected grround motion records
r
were sp
pectrally match
hed to the 5%-ddamped designn base level speectra at the 1000m depth
ffor both OBE and
a SSE levels of spectra, and
d the matched motion
m
spectraa are plotted wiith target specttra in Figure 2.. The
sspectrally matcched motions arre used as inpu
ut base motionss for the dynam
mic soil-pile-strructure interacction analyses.

F
Figure 2: Spectrally matched horizontal
h
motio
ons at 100m dep
pth for OBE and
d SSE.

D
Dynamic Soil-Pile-Structu
ure-Interactio
on Analysis Direct Metthod
IIn the direct meethod, the tank
k system, piles, and soil media are representted in one moddel with discrette elements usiing explicit
ffinite element (FE) formulattion. The equ
uations of motiion are solvedd directly in onne solution inn the time dom
main. This
aapproach is ab
ble to treat thee non-linearity
y of structure, piles, soil, annd soil-pile intteraction sprinngs simultaneoously. The
innertial effect from
f
the struccture and kinem
matic interaction from the ffoundation are coupled in thhe analysis, theerefore, no
aadditional load combination needs
n
to be con
nsidered.
The 2-D no
onlinear dynam
mic soil-pile-stru
ucture interaction analyses w
were performedd using the FE software OpennSees. The
F
FE model is composed of a co
ontinuous soil mesh, soil spriings, piles, andd tank componeents, as shown in Figure 3.
he clay soil wass modeled as solid-fluid fully
y coupled soil qquadUP elemeent and the PresssureIndependM
MultiYield
Th
(PIIMY) soil mod
del;
So
oil pile interacttion was modelled as zero-len
ngth nonlinear soil springs (pp-y, t-z, q-z) siimulating laterral soil pile
intteraction, skin friction, and tip
p resistance;
Pilles and tank weere modeled ass vertical nonlin
near Beam Collumn elementss with distributted masses;
Th
he rocking stifffness of the tan
nk base was mo
odeled as a rottational spring with the stiffnness value derivved from a
strructural model.
For the dyn
namic soil-pile-structure-interraction analysiis, both UB annd LB soil propperties were uused for the OB
BE level of
eearthquakes, an
nd BE soil prop
perties were ussed for the SSE
E level of earthqquakes.

OTC 23025

F
Figure 3: Schem
matics of the dirrect soil-pile-strructure-interaction model in Op
penSees.

S
Soil Models.
A 2-D soil mesh of 100m
m wide by 100m tall with 10
000m in plane depth was useed in the FE m
model to simullate ground
rresponse underr earthquake sh
haking. Artificial lower (bo
ottom) and lateeral (side) trannsmitting bounddaries were pllaced. The
hhorizontal dimeension between
n the transmittiing boundariess was sufficienttly large such tthat the motionn at the lateral bboundaries
ccan be considerred identical to
o that of the freee-field.
The soils were
w
modeled as
a two-phase materials
m
that co
ouple the solidd and fluid phaases in a u-p ellement formulation. The
m
material modell is pressure-in
ndependent-mu
ulti-yield (PIM
MY) material bby Yang and E
Elgamal (20022), where the volumetric
sstress-strain ressponse is lineaar-elastic and in
ndependent of the deviatory response withh a Von Mises type yield surffaces. The
loow-strain sheaar modulus is computed
c
based
d on measured
d shear wave vvelocity at diffeerent layers, annd the moduluss reduction
ccurve, G/Gmax
x, is based on th
he Vucetic and
d Dobry (1991)) relation for cllay with differeent PI values.
P
Pile and Superrstructure Pro
operties.
The comparatively straigh
htforward grou
und conditionss enabled the 4480 foundationn piles to be ttransformed innto a single
m
mega pile of th
he equivalent EI
E value of thee pile group, an
nd modeled ass nonlinear Beaam Column ellements in the FE model.
T
The moment-cu
urvature relatio
onship of the pile
p sections was
w analyzed uusing commerccially availablee software Oassys AdSec,
w
which showed an approximaate bi-linear fleexural behavio
or, i.e., an initiial elastic reesponse at the cracked sectioon stiffness
ffollowed by a softer post yiield strain harrdening. An elastic-plastic
e
w
with strain harrdening constittutive model w
was used to
m
model the section behavior. The reduced flexural stiffness of the craccked pile sectiion was determ
mined from thee momentssecant stiffnesss analysis in AdSec based on
n the expected levels of axial load and bendding moments in the pile undder seismic
looads.
To capture the dynamic responses of the
t tank systeem, a lumped-m
masses-on-beaams model waas used, with key model
eelements as follows:
Taank base slab;
Taank wall and ro
oof;
Taank product imp
pulsive SDOF element (tuned
d at 0.18s);
Taank product con
nnective SDOF
F element (tuneed at 8.2s).
M
Masses were diistributed to the various tank nodes, and varrious EI valuess were assignedd to each beam
m column elemeent.
S
Soil Pile Intera
action.
The dynamic soil-pile inteeraction (kinem
matic interactio
on) is modeledd as nonlinear soil springs thhat connect thee mega pile
too the 2-D soil mesh, including p-y (laterall resistance), t--z (shaft frictioon) and q-z (ennd bearing) springs. Configuurations of
thhe free-field siite response, th
he nonlinear p-y
y, t-z, and q-z springs are schhematically shoown in Figure 4.

O
OTC 23025

F
Figure 4: Schem
matics of soil-pile interaction fo
or the direct me
ethod.

The nonlineear soil spring


gs include elasttic, plastic, draag and closuree components accounting forr gapping, draagging, and
rradiation damping, as describ
bed in Boulang
ger et al. (1999)) and Curras ett al. (2001). T
The capacities aand backboness of the p-y
ssprings are baseed on Matlocks (1970) recom
mmendations for
f clay. The uultimate skin frriction and endd bearing of thee t-z and qz spring in the clay were calcculated using the
t method preesented by APII (2000), and tthe backbones approximate tthe relation
pproposed by Reeese and O'Neiill (1988) for piles in clay.
F
Foundation Ro
ocking.
Foundation rocking is mo
odeled as a ro
otational spring
g attached to tthe tank base slab to accounnt for the globbal rocking
ddegree of freed
dom of the tank
k. The rotational spring stiff
ffness was derivved from a 3-D
D finite element structure model of the
ppiled LNG tank
k by applying a unit overturn
ning moment an
nd measuring tthe global rotaation of the basse slab. The annalysis was
ccarried out in Oasys
O
program GSA, commerrcially availablle structural annalysis softwaree.
D
Damping.
Soil material damping is included in th
he nonlinear soil
s model afteer Vucetic andd Dobry (19911) for PI=30 cclay curve.
R
Radiation damp
ping for the piiles are approx
ximated using the
t dashpot coomponents of tthe soil springgs. Structural ddamping is
m
modeled as freq
quency-depend
dent Rayleigh damping
d
to maatch the requireed system damp
mping values.
A
Analysis Sequeence.
Time historry analyses weere performed for the direct method modeel, where accellerations, displlacements, pilee demands,
aand tank respon
nses are record
ded simultaneou
usly in terms of
o time historiees.
Th
he soil meshes are built and co
onsolidated to achieve gravitty field;
Pille, superstructu
ure and soil sprrings are impleemented into thhe soil model;
Th
hree pairs of sp
pectrally match
hed ground mottion time historries are appliedd at the bedrocck level of the 100m deep
soiil column in th
he seismic site response analy
ysis using SIRE
REN, and the w
within soil motiions were recoorded in the
soiil body at 100m
m depth;
Th
he recorded witthin soil motion
ns are applied at the bottom oof soil body at 100m depth.
Both LB an
nd UB soil prop
perties were ussed for the OBE
E analyses to ccapture the inflluence of soil sstiffness on struuctural and
ffoundation dem
mands; BE soil properties werre used for the SSE analyses.
Nine sets off dynamic soil--pile-structure--interaction anaalyses were carrried out:
OB
BE + LB Soil Properties:
P
Chii Chi, Hector Mine,
M
& Landerrs;
OB
BE + UB Soil Properties:
P
sam
me motions as above;
a
SS
SE + BE Soil Properties: samee motions as ab
bove.
A
Analysis Resullts - OBE.
This section
n presents the dynamic resp
ponses of the soil-pile-structu
s
ure-interactionn model to thee design grounnd motions,
inncluding soil, foundation and
d structural acccelerations, design response spectra, tank aand slab displacement, founddation force
aand moment deemands, structu
ural overturning moment, etc.
All results are presented in
i terms of peak values, tim
me histories, annd design specttra. The outpuuts of this secttion can be
uused for seismic design of thee structure and foundation of the LNG tank..

OTC 23025

Table 2 sum
mmarizes the co
omputed peak accelerations at
a the foundatioon slab, tank rooof, and impullsive fluid and convective
ffluid levels for three different earthquake motions
m
at OBE
E level with a PGA of 0.19gg (with UB soill properties). T
The results
sshow that the Hector motion
n east compon
nent induces the
t largest acccelerations andd load demands for all struuctural and
ffoundation com
mponents. Theerefore, the Heector motion reecord would bee used as the bbaseline motioon for the founndation and
sstructural desig
gn.
T
Table 2: Pea
ak Accelerations for different earthqu
uake motions
s UB soil properties
Earth
hquake

Foundation
n Slab (g)

Tank Roof (g))

Im
mpulsive (g)

Convectiive (g)

Ch
hichi

0.20
07

0.236

0.228

0.07
75

He
ector

0.25
50

0.268

0.264

0.06
66

Lan
nders

0.22
27

0.252

0.241

0.10
00

Table 3 sho
ows the estimated peak accelerations for the Hector m
motion record ccalculated usinng both UB annd LB soil
pproperties. Thee results indicaate that the UB
B soil properties generate the largest structuural and foundaation acceleratiions during
eearthquakes, th
herefore, UB caase results will be used for thee tank design.
T
Table 3: Peak
k Acceleratio
ons for differrent soil prop
perties Hec
ctor motion
Soil Property
P

Foundation
n Slab (g)

Tank Roof (g))

Im pulsive (g)

Convective (g)

UB
U

0.25
50

0.268

0.264

0.06
66

LB
L

0.21
16

0.234

0.228

0.07
75

Figure 5 presents the acceeleration time histories of tan


nk roof, founddation slab, andd convective fl
fluid mode for the Hector
m
motion with UB
B soil propertiies. It shows that
t
the foundaation slab and tank roof motiions are in phaase, while the convective
m
motion has a much
m
longer perriod and is out of phase with other modes.

F
Figure 5: Comp
puted acceleratio
on time historie
es for tank roof, slab, and conv
vective mode

Figure 6 and Figure 7 illustrate the comp


puted responsee spectra (5% ddamping) at thee tanks base aand roof elevattions for all
thhree earthquak
kes with both UB
U and LB soiil properties. The
T UB case ggenerates higheer demand in sshorter periodss, while the
L
LB case generrates higher deemands in thee longer period
ds. Thereforee, the enveloope curve covvering both U
UB and LB
rresponses is thee recommended
d horizontal deesign response spectrum at taank base and rooof for an OBE
E event.

F
Figure 6: OBE Tank
T
Base Horizontal Accelerration Response Spectra

O
OTC 23025

F
Figure 7: OBE Tank
T
Roof Horrizontal Accelerration Response
e Spectra

In the direcct method, seissmic demands on the tank-fo


oundation systtem can be dirrectly recordedd from the anaalysis. The
ppeak structural and foundatio
on responses, including strucctural overturniing moment att the base slab, total shear annd bending
m
moment at the pile
p heads, aree compared bettween UB and LB soil properrties in Table 44. The peak piile and structurre demands
aare mostly co-eexistent, indicaating they are laargely in phasee.
T
Table 4: Peak
k Tank and Foundation
F
Global
G
Respo
onses - OBE
Soil Property
P

OTM at Bas
se of Slab
(MN**m)

Rotation Spring
g
Moment (MN*m
m)

Total M
Moment at Pile
He ad (MN*m)

Total Shea
ar at Pile
Head (M
MN)

UB
U

313
38

3757

533

148
8

LB
L

270
07

3173

464

128
8

The equivallent peak uni-d


directional sheaar force and bending momentt (unfactored) oon a single pilee are listed in T
Table 5.
T
Table 5: Peak
k Single Pile Demands OBE
Soil Property
P

Max. Shear (kN) at


a Pile Head

Max. Moment (kkN*m) at Pile Head

UB
U

309

1110

LB
L

267

9
967

The simultaaneous displaccement, shear and bending moment


m
profilles for a singlee pile at the ttiming of peakk pile head
m
moment are plo
otted in Figure 8 for the OBE
E Hector motion
n with UB soill properties. T
The results withh LB soil propeerties show
loower pile demands and slighttly larger pile head
h
deflection
n and are not reeported here.

F
Figure 8: Single
e pile peak response OBE Ch
hiChi motion, UB
B soil properties
s.

OTC 23025

A
Analysis Resullts Comparee OBE with SS
SE.
Table 6 com
mpares the PGA
A and computeed peak acceleerations at the ffoundation slab
ab, tank roof, im
mpulsive, and convective
ffluid levels betw
ween OBE and
d SSE. Table 6 compares thee peak single piile demand bettween OBE annd SSE. The PG
GA of SSE
is about 1.74 tim
mes of OBE, while
w
the found
dation and stru
uctural responsees of SSE is abbout 1.4 times of OBE due too the larger
ssystem nonlinearity in the hig
gher level of grround shaking.
T
Table 6: Com
mpare Peak Accelerations
A
s for OBE an
nd SSE
Motion and Soil Properties

PGA (g)

Foundation Slab (g)

Ta
ank Roof (g)

Impulsive (g
g)

Conve
ective (g)

OBE UB
U

0.19

0.250

0.268

0.264

0..066

SSE BE
B

0.33

0.360

0.375

0.371

0..123

Ratio
o

1.74

1.44
4

1.40

1.41

1
1.86

T
Table 7: Com
mpare Single Pile Demand
ds for OBE and
a SSE
Motion and Soil Properties

PGA (g)

Single Pile
P Head Max. S
Shear (kN)

Single Pile He
ead Max. Momen
nt (kN*m)

OBE UB
U

0.19

0.250

0.268

SSE BE
B

0.33

0.360

0.375

Ratio
o

1.74

1.44

1.40

D
Dynamic Soil-Pile-Structu
ure-Interactio
on Analysis Substructu
ure Method
ms into two parts in the analysis by trreating the
T
The Indirect (substructure)
(
method dividees the soil-strructure problem
ssoil/foundation
n and the structu
ure separately.
The substru
ucture method involves
i
the fo
ollowing three steps,
s
also as pplotted in Figurre 9.
Deetermine the free-field motion
ns to input at the
t base of riggid footing: this was realized by seismic sitte response
anaalysis using thee bedrock motiions in commeercial site respoonse software O
Oasys SIREN;
Deetermine the frrequency-depen
ndent foundatiion impedancees of the founddation (base sllab and piles?): this was
callculated in the frequency dom
main using com
mmercial softw
ware DYNA6;
An
nalyze the cou
upled soil-struccture system in
i a FE formuulation, comprrising the tankk mass conneccted to the
fou
undation represented as a lateral and rockin
ng spring/damppers: this was performed in F
FE software SAP2000 in
thee time domain.

F
Figure 9: Substructure method
d 3 steps procedure (Kausel et
e al. 1978).

The primary
y assumption used
u
in this thrree-step metho
od requires thatt the base of thhe structure is rigid (i.e. infinnitely stiff)
w
with respect to the soil. The foundation
f
pilee-soil system iss also assumedd to be elastic.
C
Comparison of Direct and
d Substructu
ure Methods
C
Comparisons were
w made betw
ween the direct method analysses and Substruucture methodd analyses in this section.
T
Tank and Foundation Respo
onse.
Table 8 com
mpares the com
mputed peak acccelerations at tank base and roof between tthe two methoods. Substructuure method
ggenerated higheer tank base an
nd roof acceleraation response than direct meethod by aboutt 40%.

O
OTC 23025

T
Table 8: Horizontal Peak Acceleration
ns Chichi OBE
O
with UB
B Soil Properrties
Method

Direct
D
Method

Substrructure Method

Tank Base

0.21 g

0.30 g

Tank Roof

0.24 g

0.32 g

Table 9 sum
mmarizes the computed
c
peak
k foundation demands
d
from the two approoaches. The suubstructure meethod overeestimated the pile
p demands compared
c
to direct method. Moreover, th e substructure method did nnot include thee kinematic
eeffect, which will
w require ad
dditional load combinations
c
to
t obtain the ttotal pile demaands; while thee direct methood includes
bboth inertia and
d kinematic efffects and the reesults can be diirectly used forr pile design.
T
Table 9: Peak
k Foundation
n Demands Chichi OBE
E with UB So
oil Properties
s
Method

Direct
D
Method

Substrructure Method

Foun
ndation Shear (MN)

129

175

Found
dation OTM (MN**m)

2765

3230

Table 8 and
d Table 9 demo
onstrate that the substructure method generaated about 20%
% to 40% higher peak accelerrations and
ffoundation dem
mands than the direct method,, a more conserrvative approacch overall.
D
Damping.
Figure 10 compares
c
the computed fou
undation hysterresis dampingg in the horizoontal directionn between the direct and
ssubstructure methods.
m
The damping calcu
ulated from direct method iss about 11% ffor OBE and 15% for SSE,, while the
ddamping from substructure
s
method
m
is about 16% for OBE and 24% for S
SSE.

F
Figure 10: X-Dirr foundation hys
steresis comparison - OBE ChiChi motion, UB
B soil properties
s

A
Acceleration Response
R
Specctra.
Figure 11 compares
c
the computed
c
acceeleration respon
nse spectra at tank base andd roof for OBE
E and UB soil properties.
B
Both methods show
s
amplificaation at the tank
k-foundation system
s
fundam
mental period beetween 0.5s to 1s.
Tank Base Hori zontal Acceleratiion Response Spe
ectra
OBE

Tank Roof
R
Horizontal Acceleration
A
Resp
ponse Spectra
OBE

16.0

16.0
14.0

14.0

12.0

12.0

Direct Method

Dirrect Method
10.0

Substructure Method

Acceleration (m/s2)

Acceleration (m/s2)

10.0
8.0
6.0
4.0
2.0

Substructure Method
8.0
6.0
4.0
2.0
0.0

0.0
0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

0.0

1.0

Period (s)

F
Figure 11: Comparison of horiz
zontal response
e spectra betwe
een the two met hods

2.0

3.0

4.0
Period (s))

5.0

6.0

7.0

8.0

10

OTC 23025

Sensitivity Study of Direct Method


The purpose of the sensitivity study was to understand the influence of input variables on some key output variables such as
pile demands and tank response. This is important for understanding which input parameters most influence system
responses, and therefore must be carefully characterized, and which have little effect on the analysis results.
The following input variables were each assigned a range of values for sensitivity study:
The stiffness of rotational spring;
The top 4m soft alluvium fill.
Stiffness of Rotational Spring.
The effect of including a rotational spring at the base slab was investigated by comparing the system responses with the
fixed head case. The resulting percentage change in peak acceleration and peak foundation demands are summarized in
Table 10. The baseline case is with the rotational spring in place. This shows that the assumed rotational stiffness generates
a near fixed rotation response for the piles.
Table 10: Influence of Rotational Spring on Peak System Responses ChiChi OBE
Rational Spring Stiffness

Total Shear at Base of Slab


(MN)

Total Moment at Pile Head


(MN*m)

OTM (MN*m)

Slab Acc. (g)

Baseline Case

100%

100%

100%

100%

Fixed Head

96%

92%

94%

96%

Top 4m of Soft Alluvium Fill.


The effect of the top 4m of soft alluvium soil on the system responses was studied by removing the 4m soil in the FE
model. The foundation and structure responses with and without the 4m soft alluvium is compared in Table 11. By
removing the 4m soft alluvium, the system had a softer response, and pile head had a bigger deflection. Therefore, pile head
bending moment increased by 12% while pile head shear decreased by 4%. Overall the top 4m soft soil benefits the pile
design and therefore should be retained in the final design.
Table 11: Influence of Top 4m Alluvium on Peak System Responses ChiChi OBE
Soil Profile

Total Shear at Base of Slab


(MN)

Total Moment at Pile Head


(MN*m)

OTM (MN*m)

Slab Acc. (g)

With 4m soft soil

172

626

3602

0.298

Without 4m soft soil

166

712

3484

0.227

Change

-3.6%

12.1%

-3.4%

-31.2%

Summary
The dynamic soil-pile-structure interaction analyses in OpenSees is a direct method where the entire tank system, piles, and
soil were modeled in a single model and analyzed in one step in the time domain. The 2-D FE model included solid-fluid
coupled soil elements in a u-P formulation, elastic and inelastic beam column elements, nonlinear soil springs. The rocking
effect of the pile group was modeled using a rotational spring on the base slab. The model was able to capture the seismic
responses of soil, foundation, and tank system in a time history manner during earthquake shaking.
Two major components are needed considered for seismic analysis of pile supported LNG tanks, the inertial forces due to
the accelerations of tank masses, and the kinematic interaction that represent the interaction between soil and foundations.
The inertia forces are significant due to the large tank mass, and the kinematic interaction is significant during strong ground
shaking. It is important to appropriately incorporate the two components simultaneously in dynamic analysis of LNG tanks.
The direct method is an optimal method for seismic soil-pile-structure-interaction analysis since it couples inertial effect
from the tank and kinematic interaction between soil and foundation simultaneously. The following advantages are
summarized for the direct method:
It uses a single model and analyzes the soil-pile-structure-interaction in a single solution;
It models the nonlinear soil, pile, and structure behavior directly and explicitly, removing empirical
simplification and linearization that are necessary in the substructure method;
The solution is carried out in the time domain, removing the need for complex frequency dependent impedance
functions and the associated simplification for removing the frequency-dependency of the spring and damper
constants;
It coupled inertia and kinematic effects in one analysis step, therefore no additional considerations for loading
phasing and load combinations.
Considerations when using direct method:
The location of the artificial transmitting boundary of the soil body should be deep enough, such as at the
bedrock level or at a very stiff soil layer. The shear wave velocity at the soil where the transmitting boundary is
placed should be no less than 500m/s;
The horizontal dimension between the lateral artificial transmitting boundaries should be sufficiently large to
simulate a near free-field response;

OTC 23025

11

The soil mesh should be sufficiently fine, and the vertical dimension of the solid elements should be sufficiently
small so that at least 5 elements exist for the shortest wavelength of interest.
Soil-pile interaction is an inherently three-dimensional problem that either requires adoption of p-y elements or
use of a more realistic 3-D soil continuum with small enough elements around the pile to capture the interaction
behavior. Further work on 3-D analysis would be required to fully benchmark the direct method for dynamic
soil-pile-structure interaction.
The substructure method could be an alternative to direct method with appropriate calibrations. It normally generates more
conservative results than the direct method. The substructure method also has some major limitations as listed below:
It requires linearization of a highly nonlinear seismic soil-pile interaction since the concept of impedance
functions is a linear concept applicable to cases in which the shaking level is low and soil strain is small.
Therefore, substructure method is not suitable for seismic analysis when significant nonlinear behavior of soil is
expected;
Simplifications are often required to remove the frequency-dependency of the spring and dashpot properties
because of the limitation in the FE method that only support frequency-independent properties;
It only includes inertia effects from the structure, with piled foundation, kinematic interaction from the
foundation needs to be incorporated into the inertial demands with some additional analyses, which introduces
more uncertainties involving with load combinations.
References
American Petroleum Institute (API)., 2000. Recommended Practice for Planning, Design, and Constructing Fixed Offshore Platforms, API
RP 2A - WSD, 21th edn., American Petroleum Institute.
Boulanger, R. W., Curras C. J., Kutter, B. L., and Wilson, D. W., 1999. Seismic soil-pile-structure interaction experiments and analyses., J.
Geotech. Geoenviron. Eng., Sept. 1999, 750-759.
Brandenberg, S. J., Boulanger, R. W., Kutter, B. L., and Chang, D., 2005. Behavior of pile foundations in laterally spreading ground during
centrifuge tests, J. Geotech. Geoenviron. Eng., 131(11), 1378-1391.
Chang, D., Boulanger, R. W., Kutter, B. L., and Brandenberg, S. J., 2005. Experimental observations of inertial and lateral spreading loads
on pile groups during earthquakes, Earthquake Engineering and Soil Dynamics, GSP 133, ASCE, pp.
Chang, D., Boulanger, R. W., Kutter, B. L., and Brandenberg, S. J., 2006. Dynamic analyses of soil-pile-structure interaction in laterally
spreading ground during earthquake shaking, Seismic Performance and Simulation of Pile Foundations in Liquefied and Laterally
Spreading Ground, GSP 145, ASCE, pp. 218-229.
Chopra, Anil K., 2001. Dynamics of Structures: Theory and Applications to Earthquake Engineering, 2nd ed., Prentice Hall.
Curras C. J., Boulanger, R. W., Kutter, B. L., and Wilson, D. W., 2001. Dynamic experiments and analyses of a pile-group-supported
structure, J. Geotech. Geoenviron. Eng., July 2001, 585-596.
Elgamal, A., Yang, Z., and Parra, E. (2002). Computational Modeling of Cyclic Mobility and Post-liquefaction Site Response. Soil
Dynamics and Earthquake Engineering 22, page 259-271.
Gazetas, G., and Dobry, R., 1984. Simple radiation damping model for piles and footings, J. Eng. Mechanics, 110(6), 937-956.
Kausel, E., Whitman, R.V., Morray, J.P., and Elsabee, F. (1978). The spring method for embedded foundations. Nuclear Engineering and
Design, Vol. 48, pp. 377-392.
Koyama, K., Sunasaka, Y., Morioka, Y., and Miyamoto, Y. (2000). Seismic safety evaluation of an existing tank supported by piles in
liquefiable ground. 12 WCEE, 2000.
Matlock, H., 1970. Correlations of design of laterally loaded piles in soft clay, Proc. Offshore Technology Conference, Houston, TX, Vol
1, No.1204, 577-594.
Matlock, H., Foo, S. H. C. and Bryant, L. M., 1978. Simulation of latearl pile behavior under earthquake motion, Earthquake Engineering
and Soil Dynamics, Pasadena, California, 600-619.
NGA Strong Ground Motion Database. http://peer.berkeley.edu/products/strong_ground_motion_db.html
OpenSees: http://opensees.berkeley.edu/
Vucetic, M.V., and Dobry, R. (1991). Effect of Soil Plasticity on Cyclic Response. Journal of Geotechnical Engineering, 117, 89-107.
Willford, M., Sturt, R., Huang, Y., Almufti, I., and Duan, X., 2010. Recent advances in non-linear soil-structure interaction analysis using
LS-DYNA. Workshop on soil structure interaction (SSI) knowledge and effect on the seismic assesemtn of NPPs structures and
components. Ottawa, Canada, 2010.
Yang, Z., and Elgamal, A., 2005. OpenSees manual for PDMY and PIMY materials, http://cyclic.ucsd.edu

You might also like