building-foundation-soil interaction effects - curee · pdf filekajima - curee research...
TRANSCRIPT
Kajima - CUREe Research Project
Building-Foundation-Soil Interaction Effects
By
Dr. Kenji Miura
Pof. G. R. Maftii
Dr. Yuji Mi.yarnoto
Prof. Ronaldo I. Borfa
Ivir. Ariyoshi Yamacia
Prof. H.Allison Smith
Mr. lviasayuki Nagano
Dr. i.P. Bardet
Mr. Yuji Sako
Dr. S.W.Chi
Mr. Yó Hyodo Dr. Q. 1-luang
Mr. Kiyoshi Masuda
lVii. Tatsuya Maeda
Mr. Eiji Kitamura
Mr. Kiichi Suzuki
Mr. YasutsugL Suzuki
Mr. AtsunobuFukiioka
1992.8
Kajima Corporation
CUREe
Kajima Corporation
CUREe;
California Universities for Research
in Earthquake Engineering
* Kajima Institute of Construction Technology
* Information Processing Center
* Structual Department, Architectual Design Division
* Civil Engineering Design Division
* Kobori Research Complex
* The California Institute of Technology
* Stanford University
* The University of California, Berkeley
* The University of California, Davis
* The University of California, Irvine
* The University of California, Los Angeles
* The University of California, San Diego
* The University of Southern California
SUMMARY REPORT
CUREe-Kajima Research Project
Nonlinear Soil-Structure Interaction
Geoffrey R. Martin, Team Leader
PROJECT OBJECTIVES
Many high rise buildings are under construction or being planned on the reclaimed coastal area
around Tokyo and San Francisco bays. These buildings usually have a pile group foundation or a
I) composite foundation, consisting of an exterior wall foundation and internal pile group. Because
of bearing or lateral capacity concerns, the behavior of structures built on such foundations is
greatly influenced by nonlinear soil-foundation-structure interaction during strong earthquakes. In
cases where liquefiable sand deposits occur in foundation soils, further complications arise during
soil-foundation interaction due to pore pressure increases in the saturated sands. An improved
understanding of the effects of nonlinear soil-foundation interaction is needed in order to develop
improved design methods for these types of foundations.
The objectives of this project were to evaluate the ability of nonlinear numerical models to simulate
the complex soil-foundation-structure interaction behavior during strong earthquake shaking. The
collaborative research program was formulated as having three main task areas:
1. A Kajima team project examining the earthquake response of buildings on a pile or
composite foundation system using two-dimensional finite element methods or beam/Winkler
spring models. A specific building (the K-Building) was used for several of the analyses. This
building has 34 stories and a basement supported on an exterior wall foundation and internal piles.
Fckmdation soils comprised liquefiable sand layers and soft clay deposits to depths of about 23
rneters"overlying dense sands. The exterior wall foundation and piles were founded in the
1
underlying dense sand stratum. An earthquake time history with a maximum acceleration of about
0.4g was used to simulate strong earthquake motion.
The second task, undertaken by the University of Southern California (USC) team, also
focused on the use of nonlinear finite element analyses of the K-Building. In these studies, the
numerical simulation of nonlinear soil behavior was accomplished through the use of a simplified
plasticity-based bounding surface theory incorporating a Drucker-type failure surface (BDP
model). The effective stress formulation used simu1ated earthquake-induced pore pressure
increases and allowed simultaneous diffusion.
The third task, undertaken by the Stanford team, focused on the development of a
methodology for evaluating nonlinear soil-structure interaction effects in the time domain for rigid
foundations on soft cohesive soils. In this study, the dynamic response of rigid foundations on an
elasto-viscoplastic-space was investigated using a linear combination of isotropic and kinematic
hardening to model the soil constitutive response.
I) (I) &VLS1t1I1
1. Kajima Team Project
The studies undertaken by the Kajima research team (1) included the following:
A verification study of a simplified numerical model for a soil-pile-foundation system using
beam-Winkler spring models. In this study, centrifuge test results for a 1/50 scale model
consisting of a rigid pile cap and a four pile group embedded in dry or saturated sand was
used to verify the model. The centrifuge test results conducted at the California Institute of
Technology were reported in a previous Kajima/CUREe project report (2). The mod4used
2
to simulate pore pressure increases in the saturated sand under earthquake shaking was that
used in the computer program DESRA.
A numerical simulation of the K-Building response to strong earthquake ground motion
utilizing a simplified stick model to simulate the superstructure and pile foundations, and a
base rotational spring to simulate the rotational stiffness of the assumed rigid exterior wall
foundation. The foundation system was driven by free field earthquake ground motions
(including the effects of pore pressure increases) computed using the DESRA program.
Displacement time histories from free field response were applied to nonlinear Winkler
springs coupled to the simplified foundation model.
A study using two-dimensional nonlinear finite element models to simulate the results of the
centrifuge model tests comprising the rigid pile cap and four pile group in dry sands.
2. USC Team Project
The studies undertaken by the USC research team (3) included the following:
Calibration of the BDP constitutive model by comparison with the centrifuge test results (2)
previously described above. These tests modeled a foundation supported by four piles in
saturated Nevada sand. Pore pressures and pile cap accelerations were monitored during
earthquake excitation in the centrifuge tests. The BDP model soil constants were obtained
from existing laboratory test data on the Nevada sand. Results of analyses were compared to
both free field and pile cap response as measured during centrifuge tests in a similar manner
to the calibration tests conducted by the Kajima team.
The application of a two-dimensional nonlinear finite element analysis to model the
earthquake response of the K-Building. In these analyses, the foundation system comprising
3
the wall/pile basement system was modeled using the foundation parameters supplied by the
Kajima team. The building structure was modeled as a stick model, again using the data
supplied by the Kajima team. The soil pore pressure response during strong ground shaking
was assessed using the BDP model.
3. Stanford Research Team
The studies undertaken by the Stanford team included the following:
Analyses of the nonlinear response of vertically oscillating circular, square and rectangular
foundations to apply harmonic loads using large scale nonlinear finite element analyses. The
constitutive model used for the supporting soil was a deviatoric elasto-viscoplastic model capable
of replicating closed ioop steady state cyclic stress-strain response. The supporting soil medium
was assumed to be a half space that could undergo plastic deformation in the vicinity of the
foundation. The large finite element mesh used was sufficient to simulate radiation damping
without the use of an artificial boundary operator.
Additional studies were performed of the horizontal rocking and torsional vibration modes
of strip and square foundations.
SUMMARY OF RESULTS AND CONCLUSIONS
The results of team research studies may be summarized as follows:
Kajima Team
1. The numerical model comprising beam elements and nonlinear lateral Winkler springs
(taking into account the changing effective stress) successfully predicted the centrifuge test results
4
of the pile foundation model in dry or saturated sand. The model effectively represented the
response characteristics of a pile foundation in nonlinear liquefied soil.
Analyses of the earthquake response of the K-Building supported on composite foundation
showed that response was strongly affected by nonlinear liquefied soil-foundation interaction
during strong earthquake shaking. Response of the superstructure changed depending on the
magnitude of input motion and nonlinearity of surrounding soil. The composite foundation
response was significantly changed by large deformation and degraded resistance of the soil due to
liquefaction.
The proposed two-dimensional finite element model incorporating soil nonlinearity, where
the soil domain is divided into multi-regions according to the degree of soil-pile interaction effects,
was successfully verified by correlation analyses using the centrifuge test results of a pile
foundation in dry sand.
USC Team
The two-dimensional finite element simulation of the centrifuge pile model, utilizing the
BDP model to describe the saturated sand response, was reasonably successful in the case of the
dry sand experiments. However, the simulation of the pile group response in saturated sands
showed lesser agreement, as computed pore pressures diffused less rapidly than measured ones
since the two-dimensional drainage simulation was more constrained than the experimental three-
dimensional drainage. Consequently, computed model response was characterized by stiffer soil-
pile interaction than observed experimentally.
The application of the two-dimensional finite element model to predict the dynamic
response of the K-Building produced results in general agreement with those obtained by the
Kajima team, who used different numerical approaches. However, the limitations of two-
dimensional simulation of pore pressure response were less apparent than in the centrifuge test
simulations as the structural response of the K-Building is strongly controlled by the assumption
that the external foundation wall and piles are elastic and are founded on elastic soil layers.
5
Stanford Team
Analyses of the nonlinear response of vertically oscillating circular, square and rectangular
foundations to harmonic loads showed that for low frequency excitations, resonance amplifies the
motions of the foundation to levels above those obtained at the zero frequency level. Horizontal
rocking and torsional vibration modes of strip and square foundations are characterized by
increased vibrational amplitudes due to material stiffness degradation.
Nonlinear soil effects are shown to be dominant over a wide range of excitation frequencies
for foundations vibrating in torsional and horizontal modes, but over a much narrower range of
frequencies for vertical and rocking modes.
RESEARCH NEEDS
The following research needs were identified as a result of team discussions following the
conclusion of the research program:
The need for improved modeling of local nonlinearity around piles (Winkler spring
simulation) during earthquake induced pore pressure increases, including the effects of pore
pressure diffusion.
The need for improved one-dimensional free field response models to include simulation of
post-liquefaction site response.
The need for improved constitutive models for three-dimensional saturated cohesionless
soil behavior capable of simulating pre-liquefaction and post-liquefaction behavior for use in finite
element models.
11
4. The need for improved two-dimensional finite element hybrid models to more correctly
simulate three-dimensional behavior of pile foundations.
REFERENCES
Miura, Kenji, et. al. "Building-Foundation-Soil Interaction Effects." Final Project Report,
Kajima-CUREe Project, February, 1992.
Scott, R.F. and Hushmand, B. "Soil-Pile Response in the Santa Cruz (Loma Prieta)
Earthquake, October 17, 1989." Kajima-CUREe Report, California Institute of Technology,
University of California, Davis, 1991.
Bardet, J.P., et. al. "A Study of S oil-Pile- Structure Interaction in Liquefiable Soils During
Strong Earthquakes." Kajima-CUREe Project Report, University of Southern California, August,
1992.
Borja, R.I. and Allison-Smith, H. "A Methodology for Nonlinear Soil-Structure
Interaction Effects Using Time Domain Analysis Techniques." CUREe-Kajima Research Project
Report, Stanford University, 1992.
Kajima-CUREe Project
Building-Foundation-S oil Interaction Effects
Topic3-1 ;A Study of Nonlinear Soil-Pile-Structure Interaction in Soft Soils
during Strong Earthquakes
Topic3-2;A Methodology for Nonlinear Soil-Structure Interaction Effects
Using Time-Domain Analysis Techniques
Final Project Report
Kenjl Miura Kiyoshi Masuda
Yuji Miyamoto Tatsuya Maeda
Ariyoshi Yamada Eiji Kitamura
Masayuki Nagano Kiichi Suzuki
Yuji Sako Yasutsugu Suzuki
Yo Hyodo Atsunobu Fukuoka
(Kaj ima Corporation)
February 1992
Contents
Introduction . 1
Simplified numerical model of a pile foundation and verification study ..................2
2.1 Numerical model of a pile foundation in dry or saturated soil .....................2
2.2 Results of correlation analysis .....................................................................12
Earthquake response of a building on composite foundation ..................................23
3.1 Description of the building ...........................................................................23
3.2 Input ground motion for earthquake response analysis ................................23.
3.3 Free ground response by the effective stress analysis ..................................28
3.4 Earthquake response of K-Building .............................................................41
2-Dimensional finite element model and verification study .....................................91
4.1 Numerical model of a pile foundation in dry soil ........................................91
4.2 Results of correlation analysis .....................................................................98
Conclusions .................................................................................................................106
References ......................................................................................................................107
1. Introduction
Many high rise buildings are under construction or being planned on the reclaimed
coastal area around Tokyo and San Francisco Bays. These buildings usually have a pile
group foundation or a composite foundation which consists of wall foundation and pile
group because of increasing bearing capacity and stability of a foundation. The behavior of
structures built on such foundations is greatly affected by nonlinear soil-foundation
interaction during strong earthquakes. Particularly in liquefiable sand deposits, further
complicated interaction between foundation and the surrounding saturated soil will occur
because of pore water pressure generation and large displacement. An improved aseismic
design method is then required to take into account nonlinear soil-foundations interaction
on earthquake response.
The earthquake response of buildings on a pile/composite foundation during strong
earthquakes is usually studied using a 2-dimensional finite element method or a beam-
Winkler spring model. In this study, two analytical methods are employed to investigate
earthquake responses supported on a pile/composite foundation. In Chapter 2, a practical
simplified numerical model of building supported on a pile/composite foundation in non-
linear liquefiable soil deposit is proposed and verified by comparison with centrifuge tests.
In Chapter 3, earthquake response analyses for representative building on a composite
foundation are performed and discussed to develop improved understandings of the
behavior during earthquake. Further, in Chapter 4, 2-dimensional finite element technique
incorpotating soil nonlinearity is proposed and verified by comparison with centrifuge
tests.
- 1 -
2. Simplified numerical model of a pile foundation and verification study 41 Numerical model for calculating earthquake response of a building supported on pile
group/composite foundation indicates in Fig.2.1. Correlation analyses for the centrifugal
test results of a pile foundation model as shown in Fig.2.2 were conducted to verify
the proposed numerical model. The centrifuge tests for verification studies have been
conducted on a 1/50 scale model consisting of a rigid pile cap and a four-pile group
embedded in dry or saturated fine sand in the previous Kajima/CUREe project1).
2.1 Numerical model of a pile foundation in dry or saturated soil
In this analysis procedure, the nonlinear effective stress method, computer program
DESRA proposed by Firm et a12), is utilized for the free ground response. The obtained free
ground responses, including displacements and excess pore water pressures, are applied to
the soil-pile foundation system as earthquake input. Superstructure and pile foundation are
idealized by a one-stick model with lumped masses and bending-shear elements. The
lumped masses are connected to free ground through nonlinear Winkler springs modified
at each step in accordance with the generation and dissipation of excess pore water
pressure. The linear rotational spring, related to the axial stiffness of the piles, is also
incorporated at the pile head.
Analysis of free ground
The computer program DESRA is based on the shear stress-strain relationship of the
Hardin-Dmevich model with the Masing rule expressed as
= G y/(1+G0 (2.1)
in which 'c = shear stress, tmaX = shear strength, y = shear strain, and G0 = initial shear
modulus. The initial shear modulus G0 and shear strength tmax are modified progressively
for the changing vertical effective stress of saturated sand subjected to an earthquake.
Physical constants employed in correlation analyses for the saturated sand are indicated in
— 2 —
Table 2.1. The initial shear modulus G0 employed in the Hardin-Drnevich model is
estimated from
G0 =A ((2.17-e)2/(1+e)) ('m0)1a (2.2)
in which e = void ratio, and 'mO = initial mean effective stress3 . The coefficient A,
obtained by resonant-column tests, is modified to correspond with the resonance frequency
of the free ground in the low input test. The shear strength 'CmaX is obtained by the Mohr-
Coulomb yield condition.
The pore water pressure generation Au is defined by
Au = Er AF,d (2.3)
S where
AF-,d = C1(y - C2EVd) + C3E2Vd / (Y+C4EVd) (2.4)
Er = ((Y')m1'm1C2 (')fl.m (2.5)
in which Er = one-dimensional rebound modulus of sand at an effective stress cy', AEVd =
volumetric strain increment under simple shear condition. Parameters C11 C21 C3 and C41
and K2, m and n are estimated from cyclic simple shear tests and one-dimensional loading
and unloading tests. These parameters are also modified to simulate liquefaction
resistances obtained by the cyclic simple shear test results. The values employed in the
correlation analyses are C1=0.114, C2=1.2, C3=0.17 and C4=1.36, and K2=0.001(tJm2),
m=O.1 and n=O.19. Fig.2.3 indicates the liquefaction resistance curve calculated using
these parameters in comparison with the cyclic simple shear test results of undrained
saturated sand.
Shear modulus values for evaluating initial soil-pile foundation interaction springs are
estimated by the secant modulus of the backbone curve of each layer using Eq.(2.1) at the
effective shear Straifl(O.657m ), which is obtained by the nonlinear free ground response.
M,J Super -Structure
Wink ler Spring Rotational Spring
>i R
Pile -Foundation
Fig.2.1 Lumped masses and bending-shear elements model with lateral Winkler springs for soil-pile foundation system
17.73 3.025
AH4l Laminar Box
1.58
1.44 SG7 - 0.0- - SG6— - -0.0
0.75- - I
ISG3
- -1_i -0.75 AH3 - 3.45 3.65
J
SG4 - LT3 - - - 5.30 £ I SQpp4 G2 -
S 1 h° 10.7 - 4.8-1 -.
0.48 OPP2 SG1 10 4 4. _18 ___
LT2 AH2 1.19 (Dry sand)
- (Saturated sand) - - - -fT1-8.75 - LT1
I
AH11t Base - (Unit: m) AH:Horizontal Accelerometer
PP:Pore Pressure Transducer SG:Strain Gauge LT:Displacement Transducer
Fig.2.2 Centrifuge model of 4-pile group foundation and transducer arrangements (The dimensions in this figure are described in prototype scale)
Table 2.1 Physical soil constants for correlation analyses
Soil : Nevada sand #120
Internal friction angle: ' 350
Void ratio: e 0.69
Density: y 1.98 tIm3
E Permeability: k 5.2x10 3 cm/sec
Coef. of earth pressure at rest: K0 0.45
Shear modulus: G0 570(GLOm) - 3240(GL-1 0.4m) tim2
1'
0.1
S
Gm = 6.20x1 03 tim2 m =1.97t1rn2 CY ,ax
= 5.63 tIm2
Analysis
est
I 0 50 'l'(!)c) 2(10
Number of cycles
Fig.2.3 Liquefaction resistance curve by analysis and cyclic simple shear tests of undrained saturated sand
EIJ
- 7 -
Evaluation of nonlinear lateral Winkler spring EA
The nonlinear lateral load-displacement relationship of a pile is also based on the
Hardin-Drnevich model with the Masing rule, as shown in Fig.2.4. Initial lateral Winkler
springs K 0 at each depth are evaluated by the inversion of soil flexibiities by ring loads at
the nodes, as shown in Fig.2.5. The position of the node corresponds to that of the local
lumped mass of the pile. The soil displacements can be expressed as
(u) = [d1 ][p)
(2.6)
where (u) and (p) are the vectors of lateral displacement and load at the nodes, d jj is
the flexibility matrix of the soil. The soil displacements (u) caused by lateral ring loads
(p) in a layered stratum4 are calculated at low frequency(f=O.2SHz) to form the S flexibility matrix. The soil flexibility at the i-th node is obtained by the superposition of
soil displacements at the i-th node caused by ring loads at all nodes. Then, the lateral
Winkler spring at the i-th node is approximately obtained by the inversion of the soil
flexibility expressed as
K,0= dii (2.7)
where K.Oi is the initial lateral Winkler spring at the i-th node.
The initial ultimate lateral soil resistance Pmax0i at the i-th node is assumed by Broms5)
to be three times the Raukine passive pressure expressed as
maxO = 37'0Kdl
(2.8)
where cF'O is the initial effective stress, K = (l+sin')/(l-sin'), 4' is the internal friction
angle, d is pile diameter and 1 is pile length equivalent to the i-th node.
The initial lateral Winkler spring and the ultimate lateral soil resistance accounting for
the pile group effect is obtained using the pile group efficiency CXH defined by
aH = KHI/(NKHS)
(2.9)
where N, K H N and KHS are the number of piles, the value of horizontal static impedance at
pile head of N-piles, and its value for a single pile respectively. The value of ct is obtained
as 0.56 by the three-dimensional thin layered element method6) using the degraded shear
S modulus. The equivalent lateral Winkler spring K and the ultimate lateral soil resistance
Pmax0C which take into account the pile group effect are approximately defined by
= cXHNKSO (2.10)
maxO = aHNPmaxO (2.11)
where cXHN is the equivalent number of piles. The values of K and P.,x at each time step
in the time domain are also modified according to the change of the effective stress defined
by
= K 0e(a/a 0)la (2.12)
P. = Pmaxoe(/o) (2.13)
S where c' is the effective stress at each time step, and is the initial effective stress.
The rotational spring is also evaluated under a pinned condition at the pile head by the
three-dimensional thin layered element method. The degraded shear modulus of soil is
employed taking into account the effective shear strain obtained from the ground response.
The viscous damping employed in the pile foundation, the lateral springs and the
rotational spring are 1 per cent, 5 per cent and 30 per cent respectively for the first mode
by the eigenvalue analysis.
— 9 —
Lateral load P maxO
P = Ky/(l +KoY/Pmaxo)
K50
Ks = Kso(0'/(y'o)1"2
Displacement y ,
max maxo(''o) ,
- Effective stress at each time step
Initial effective stress
{maxO
Fig.2.4 Non-linear lateral load-displacement relationship of Winkler spring
—10—
S
--- i s
- -H —'
I I 5
node I Ring load
Pi u } = { dJ { p
Ks0' = [ d J-1
Inodej I
K50 Initial lateral Winkler spring
Displacement
Fig.2.5 Evaluation of initial Winkler spring by ring load in layered stratum
—11—
2.2 Results of correlation analysis
I
Centrifuge tests for verification study were conducted with different input acceleration
levels for dry or saturated sand. The input acceleration had a spectrum similar to that of the
1940 El Centro earthquake strong motion record NS component. The maximum input
acceleration measured at the soil container base(AH1) for dry sand were about 85
gal(TEST-Di) and 203 gal(TEST-D2), and those for saturated sand were about 100
gal(TEST-Si) and 250 gal(TEST-S2).
Fig.2.6 indicates predicted acceleration time histories at GL-0.75m and the pile cap as
compared with TEST-D1,D2, and Fig.2.7 also indicates those in TEST-S i,S2.
Acceleration response spectra for 5 per cent damping for predicted and measured time
histories are indicated in Fig.2.8 for TEST-Dl,D2 and in Fig.2.9 for TEST-S i,S2. Fig.2.10
and Fig.2.1 1 indicate comparisons of predicted relative ground displacements and excess
pore water pressures for TEST-Si, S2. Fig.2.12 indicates comparisons of predicted pile
bending moments at SG3 for TEST-Di,D2 and TEST-S 1,S2.
It can be seen that predicted acceleration time histories and response spectra at the
ground and the pile cap in TEST-Di are in good agreement with the measured results, so
that the dry soil-pile foundation system for the low acceleration input is modeled properly.
In TEST-Si, predicted acceleration at GL-0.75m is somewhat smaller than the measured
result. The acceleration response spectrum at GL-4.8m is smaller than measured results
because the generation of excess pore water pressure is overestimated in the analysis
model. The predicted response at the pile cap in TEST-Si, however, shows good
agreement with measured one, and it can be seen that the generation of excess pore water
pressure is too small to effect the response at the pile cap. The predicted responses in
TEST-D2 approximately correspond to the measured results, and the frequency properties
of the measured results are represented well. However, discrepancies of the peak
amplitudes at the pile cap can be seen between the predicted and measured results. These
discrepancies result from the difference of the free ground response. In TEST-S2, the
—12—
predicted responses, except for the peak amplitudes, also agree well with the measured
results, including the relative ground displacements and the generation and dissipation
processes of excess pore water pressure. It is found that the shift in dominant frequency of
the soil-pile foundation system by soil liquefaction is represented well by the proposed
model. This is evidenced by comparing the predicted and measured acceleration response
spectra at the pile cap. The predicted pile bending moments in TEST-Dl,D2 and TEST-Si
also correspond to the measured results, but the peak amplitudes in TEST-S2 are
underestimated. The underestimated peak amplitudes in accelerations and bending
moments by analysis are mainly attributed to the discrepancies in the free ground response.
It is considered that these discrepancies are due to the dilatancy of soil during large strain,
and an advanced effective stress method for predicting ground response in liquefied soil
should be developed.
Fig.2.13 and Fig.2.14 indicate the calculated lateral load-displacement relationships of
the pile at GL-0.6m and GL-3.Om for TEST-Di, D2 and TEST-Si, S2. It can be seen that
the value of lateral springs are degraded by the pile displacement and the decrease in
effective stress due to excess pore water pressure generation in TEST-S2. The degradations
at GL-0.6m are much larger than at GL-3.Om due to large displacement and excess pore
water pressure buildup near the ground surface.
-13-
AL-jo. r-i r\7r-
;ec
-300
Gal 400 AH4: Pile cap - ANALYSIS TEST
30sec
-400 J
-200
ec
S 400 Gal AH4: Pile cap
A.
(TEST-Di)
(TEST-D2)
Fig2.6 Comparisons of predicted acceleration time histories at GL-0.75m and pile cap with TEST-Di ,D2
—14—
-300 Gal
200 AH3: GL-0.75m
Gal 300 * I I A ITh • I - - - - ANAT y:uc
ec
30sec
-200
Gal 450 AH4: Pile cap
-450
Gal 250 1 AH3: GL-0.75m
V ' • V
-250
(TEST-S2)
Fig.2.7 Comparisons of predicted acceleration time histories at GL-0.75m and pile cap with TEST-Si ,S2
30sec
-15-
OF
Gal Gal 2500 2500
AH4 Pile cap h=O.05
4 Pile cap h=O.05
Test Analysis
Test
oJ..T0 0.1 0.5 1.0 5.0 10
Hz 0.1 0.5 1.0 5.0 Hz1°
Gal Gal 2000 2500
AH3: GL-0.75m h=O. 05 AH3: GL-0.75m h=0.05
AnalyAnalysis
Sis Test f' Test
Ol 0.1 0.5 1.0
01 I I b.i 0.5 1.0 50 H 10
z 5.0 io Hz
Gal Gal 1500 1500
h=O.05 h=0.05 AH2: GL-4.8m AH2: GL-4.8m
Analysis A Test Analysis Test
01 o 0.1 0.5 1.0 5.0Hz1° 0.1 0.5 1.0 5.0 Hz
(TEST-Di) (TEST-D2)
Fig.2.8 Comparisons of predicted acceleration response spectra at GL-4.8m, GL-0.75m and pile cap with TEST-Di D2
—16—
Gal Gal 2000 2000
h=0.05 AH4: Pile cap AH4: Pile cap
h=0.05
'SIS Test Anal ,'i Test alysis ( \/
l v I '
0.1 Oji 111111
0.1 0.5 1.0 0.5 1.0 5.0 Hz1° 5.0 Hz1°
Gal Gal
1000
1000 f
AH3: GL-0.75m h0.05 AH3: GL-0.75m h=0.05
Test Test A
Analysis Analysis IV, \ Ij s
DL 0.1 0.5 1.0 5.0 Hz 10 0.1 0.5 1.0 5.0 Hz1°
Gal Gal 1500 1500
AH h=0.05
2: GL-4.8m AH2: GL-4.8m h=0.05
C'
Test Test - Analysis /\,. Analysis W1
o_ _ 0.1 0.5 1.0 5.0 Hz1° 0. 0.5 1 1.0 1 1 5.011 10
Hz
(TEST-Si) (TEST-S2)
Fig.2.9 Comparisons of predicted acceleration response spectra at GL-4.8m, GL-0.75m and pile cap with TEST-Si S2
-17-
1.5 cm
-1.5 J 1.0 -cm
-1.0
10.0 -cm
LT3: GL-0.75m
[12: GL-4.8m
- (TEST-Si)
ANALYSIS TEST -------------
3Osec
3Osec
-10.0 J 5.0 -1cm
L13: GL-0.75m , 1\: I -. "'' ...............,.-- .................................
3osec
-5.0 ----..---sec
Fig.2.10 Comparisons of predicted relative displacement time histories of ground with TEST-Si ,S2
-18-
0.2 1 Kg/cm2 FF4: GL-3.45m
ANALYSIS
L
TEST
30sec -0.2 - 0.2 - Kg/cm2
FF2: GL-5.3m
-0.2 - (TEST-Si)
30sec
0.5 .1Kg/cm2 FF4: GL-3.45m
30sec -0.5 - 0.5 1Kg/cm2
FF2: GL-5.3m
-0.5 (TEST-S2)
30sec
Fig.2.11 Comparisons of predicted excess pore water pressure time histories with TEST-S1,S2
-19-
xiO5Kg.cm SG3: Bending moment - ANALYSIS 101 TEST
-10 1
20
(TEST- Dl) xiO5Kg.cm
- SG3: Bending moment
-20 (TEST-D2)
xlO5Kg.cm SG3: Bending moment : xiO5Kg.cm
20 SG3: Bending moment
ec 20
(TEST-S2)
Fig.2.12 Comparisons of predicted pi'e bendimg moment time histories with TEST-Di ,D2 and TEST-Si $2
-20-
U
Winkler spring at GL-3.0 m r')
(TEST-Di) (I hi I-U2)
Fig.2.13 Calculated non-linear lateral load-displacement relationships of Winkler springs in TEST-Di ,TEST-D2 analyses
Winkler spring at GL-0.6 m Winkler spring at GL-0.6 rn
Winkler spring at GL-3.0 rn
Winkler spring at GL-3.0 m
Fig.2.14 Calculated non-linear lateral load-displacement relationships of Winkler springs in TEST-Si TEST-S2 analyses
3. Earthquake response of a building on composite foundation
3.1 Description of the building
The building (named K-Building) which is used for numerical prediction has thirty-four
stories and a basement supported on wall foundation and piles. The outline of the K-Building
is shown in Fig.3.1. The K-building is 104.5 meters high and has a basement of 6.0
meters deep. The pile length is 18.3 meters and 1.0 meter of the whole length is driven into
the supporting layer of 23.3 meters deep. This building has been planned to construct at
the reclaimed site of Tokyo Bay area which consists of liquefiable sand and clay deposits.
The foundation consists of wall foundation and twenty piles as shown in Fig.3.2. The
dimensions of the basemat and wall foundation are approximately 30 meters square. The
thickness of wall foundation is 1.2 meters. The piles are made of reinforced concrete, and
the diameter is 1.8 meters along the pile shaft and 3.0 meters at the pile tip. The soil
consists of sandy and clay strata, which is 23.3 meters in thickness as shown in Fig.3.3.
The supporting soil below the pile tip is the dense sand stratum.
3.2 Input ground motion for earthquake response analysis
The incident ground niotion(E0) for earthquake response of aseismic design is
determined at GL-53.5 meters by the semi-empirical method7). Acceleration response
spectra and time history for 2E0 are shown in Fig.3.4. The maximum acceleration value is
407 Gal.
—23—
S
i1lllI I UlIlF't\
- Thi
30-
-
- -__
II II
- i—fj'=j —ç----- (Basemat) (Underground wall and piles)
25— I
11 F-1 F
j I f11 11 11 111 0
1 Li 20 ____ ____
1 0 0
L ___ ___ 0 0
15- 0 0
i:Ji:::it —rtn i
ii i•r—=fl __ __ 0 0 I d=i .8
io'— PFnI:T I _
- 30 0
- I II II II fl Ill i A
(Wall-Pile foundation plan)
- II
!L11 H liirTfl( Liquefiable sand
II II 1 Il II GL 0.Om 4
I I I T'-"•I
S
Wall foundation
Clay
30.0
Gal 500 r
Unit: m -500 lOOsec
(Elevation)
Fig.3.1 Elevation and composite foundation plan of K-Building
-24-
)OO
UNIT:mm
angements)
MATERIAL PROPERTY
Concrete Fc=300kg/cm
E0=2. 57x 106kg/cm AXE OF EXCITEMENT
G0=l. lox lO 6 kg/cm 2 ( v =1/6)
p =2. 3t/i
Fig.3.2 Basemat and wall-pile foundation plan of K-Building
—25—
GLOm ----------------------------------
Vs(m/s) r (t/m 3) 1) GL Urn S:SAND,C:CLAY
BASE MAT OF THE STRUCTURE WB-10657t
GL-6.Om JB-8.65X1OtCI11
Vs=155, y1.7, P =0.48(S) -3.9m
Vs=80, Tl.G. v =0.48 (C) -6.9m
-7. 6m Vs105, r =1. 7, P =0. 48(5)
Vs=130, y=1.6, P =0.48(C) -9.5rn
-10.8m Vs=175, y=1.8, v =0.48(S)
Vs=325, 1=1.9, P =O. 48 (S) -15.
Vs=230, - =1. 7. p =0. 48(C)
L
L-24.3in PILES DIAPHRAGM WALL Vs450, r2.0, p0.40 (ELASTIC)
-29. 85m
Vs450, =1.9, p0.40 (ELASTIC)
-53. 5m
WAVE INPUT POINT
Vs600, y2.06, v =0.40 (ELASTIC)
<PROPERTY OF SOIL LAYER>
Fig.3.3 Soil profile and elevation of composite foundation
-26-
500
-500
Gal
Hz
Acceleration Response Spectra
100
SEC
MRX = Gal
Time History
Fig.3.4 Acceleration response spectra and time history of input motion
-27-
S 3.3 Free ground response by the effective stress analysis
Earthquake response analyses for the free ground were performed by using the effective
stress analysis DESRA to investigate response characteristics during earthquakes.
Physical soil constants employed for analytical studies are indicated in Table 3.1.
Parameters for excess pore water pressure generation in DESRA analyses are determined
by the simple shear tests converted from the triaxial tests for each sand layer. Fig.3.5(a)
and Fig.3.5(b) show the converted test results and the simulated liquefaction curve. The
liquefaction curve for sand-2 is estimated using the simple shear test results of sand-i
because the triaxial test for sand-2 was not conducted. 5 The earthquake response analyses for the free ground were conducted for the following
soil properties with different input acceleration levels.
The analytical conditions are as follows.
Case 1-1: Maximum input acceleration - 100 gal
Soil property - Linear
Case 1-2: Maximum input acceleration - 100 gal
Soil property - Nonlinear & Liquefaction
Case 1-3: Maximum input acceleration - 407 gal
Soil property - Nonlinear & Liquefaction
3.3.1 Results of earthquake response for free ground
Acceleration and displacement time histories for the free ground are shown in Fig.3.6
and Fig.3.7 for Case 1-1, Fig.3.8 and Fig.3.9 for Case 1-2 and Fig.3.10 and Fig.3.11 for
Case 1-3 respectively. Acceleration response spectra at the free ground surface are also
shown in Fig.3.12. Fig.3.13 shows the distributions of maximum acceleration, shear stress
and shear strain in the free ground for Case i-i, Case 1-2 and Case 1-3.
It is observed that acceleration .responses in liquefied soil for Case 1-2 are amplified in
the low frequency content compared with those in linear soil for Case i-i. The acceleration
—28—
time histories for Case 1-3 are significantly filtered in the high frequency range. The
displacement time histories for Case 1-3 are also amplified in the low frequency content.
The acceleration response spectra at the free ground surface indicate that the peak
amplitude at 1.5Hz are remarkably reduced by soil liquefaction, but the amplitude at the
low frequency range become larger in Case 1-3.
Amplification of maximum acceleration values cannot be seen in the ground for Case 1-
2 ,which is the low level acceleration input, compared with those of Case 1-1. Maximum
acceleration values for Case 1-3 ,which is the high level acceleration input, are remarkably
reduced in the liquefied sand layer. Maximum shear stress distributions indicate that
degradation of soil shear stress due to soil liquefaction occurs much larger than soil
nonlinearity. It is comfirmed taht larger soil shear strain also occurs in liquefied soil layer.
—29—
Table 3.1 Soil constants and parameters for earthquake response analysis
GL Om
-3. 9m
-6. 9
-7.6
5
8
-15.95
-23.3
-29. 85
-53.5
iui.t WILL
weight
r (tim3)
,wave
velocity
Vs(m/sec)
NdA. iiiu
modulus
Go(t/ n)
r.arui-
press.at
rest K0
rernieaui- lity
Kv(in/sec)
internai
friction
angle q5
5near
strength
v rd)
Sand-i 1.7 155 4170 0.5 6. 9x10 6 28.2 0. 04-U. 64
Clay 1.6 80 1045 1.0 0x108 -- 2.7
Sand-2 1.7 105 1910 0.5 6. 9x10 6 21.6 0.54-0. 58
Clay 1.6 130 2760 1.0 1.0x108 -- 3.4
Sand-3 1.8 175 5625 0.5 6. 9x10 6 30.3 1. 82-i. 97
Sand-4 1.9 325 20480 0.5 6x10 5 44.3 58-5. 18
Clay 1.7 230 9180 1.0 1.0x108 -- 13.3
(Elastic) 2.0 450 41330 -- -- -- --
(Elastic) 2.0 450 41330 -- -- -- --
9 ac rnn '77fl -- --
[Parameters for DESRA] Sand-i, Sand-2
c=1.0, c2=0.4, c'3=c3/c=0. 1612, c4=0.41 k20.09(psi), m0.43, n=0.62
Sand-3
c1.0, c20.4, c'3c3/c0.1612, c4=0.41 k 2=0.017(psi), m=0.43, n=0.62
Sand-4
: c=1.0, c2 0.4, c'3c3/cir0.1612, c4=0.41 k 2=0.021(psi), n=0.43, n=0.62
-30-
lilt I I I I I till I
(Sand-i)
G1IAX = 530000.0
TMRX = 272.00
SIGVO = 1080.0
Cl = 1.000
C2 = 0.400
C3 = 0.161
• 04 = 0.410
K2 = 0.09000
Il = 0.430
2=0.620 •
Analysis
lilt
Test
11111
w
CD
CD
ME
LIQUEFACTION CURVE
D 0
a: (fl t
U) 0 LU a: 1—U)
a: LU
(1)
2 5 10 2 5 100 2
NUMBER OF CYCLES
LIQUEFACTION CURVE
(Sand-2)
G)IPX = 387000.0
TIIRX = 220.00
SIGVO = 1880.0
Cl = 1.000
02 = 0.400
C3 = 0.161
C4 = 0.410
K2 = 0.09000
11 = 0.430
N = 0.620
El ej
Analysis Test
-.--
U•)
C
0
5
2 5 10 2 5 100 2
5
NUMBER OF CYCLES
Fig.3.5(a) Liquefaction curves for sand-i, sand-2 —31-
(Sand-3)
LIQUEFACTION CURVE a
GIIAX = 1030000.0
TIIRX = 652.00
SJGVO = 2300.0
Cl = 1.000
C2 = 0.400
C3 = 0.161
C4 = 0.430
K2 = 0.01700
II = 0.430
0.620
15 Analysis
N =
Test C)
4 5 IU 2 5
100 2 5
NUMBER OF CYCLES
LIQUEFACTION CURVE In
cc Cr
Cr) C) LU a:
U)
a:
LU
U)
C)
im
I I I I I I I
(Sand-4)
Test
GMRX = 49000 Analysis
ThRX = 1300.00
SIGVO = 2820.0
Cl = 1.000
C2 = 0.400
C3 = 0.161
C4 = 0.410
K2 = 0.02100
11 = 0.430
N = 0.620
I III bi 2 5 10 2 5 100 2 5
NUMBER OF CYCLES
Fig.3.5(b) Liquefaction curves for sand-3, sand-4 —32—
S
Gal 260 GLOm
.1 , i A I Lhk A L I h.AALAWa .4"; A. kh1w
-260 0.0
Y v
HRX 2.528E+2 oo-sec 260 GL -6.Om
-260 0.0 MAX 1.872E+2 60.
260 GL -6.9m
-260 0.0 1RX= 1.629E+2 60.
260 GL -7.6m
-260 0.0 MAX=-1.531E+2 60.
260 GL -8.55m
-260
-
0-0 HRX=-1.465E+2 60.
260 CL -9.5m
-260 0.0 HRX= 1.400E+2 60.
260 GL-1 0.8m
-260 0.0 HRX= 1.368E+2 60.
260 GL-13.4m
- W)W"Yetwo1w, 10
-260 0.0 MAX= 1.333E+2 60.
260 CL -1 5.95m
-- --.._
-260 0.0 MAX I .27qE+2 60.
260 GL -29.85m
-----
-260 0.0 MRX= 6.7314E+1 60-
Fig.3.6 Acceleration time histories calculated by DESRA (Linear, Input acc.=1 00 Gal)
-33-
30cm GLOm
3.0 0.0 tix= 2.567E+0 60. sec
3.0 GL -6.Om
-3.0 0.0 MRX= 2.169E+O 60.
3.0 GL-6.9m
-3.0
-..... -
0.0 MRX= 2.02E+0 60.
3.0 GL-7.6m
-3.0 0.0 MRX= 1.953E±0 60.
3.0 GL-8.55m
-3.0 0.0
-
MRX= 1.863E+0 60.
3.0 GL-9.5m
YU h A A ~n'
-3.0 - 0.0
44 -44TAvAV VA 1vv AV .0--v'~w ~"V , V- IV
v 0 'V AV e v̂AhA
MRX= 1.776E+0 60.
3.0 GL-1O.8m
0.0
- --.. -
MRX= 1.715E+0 60.
3.0 GL-13.4m
-3.0 0.0 MRX= I.676E+0 60.
3.0 GL-15.95m
-3.0- 0.0 MRX= I.630E+0 60.
3.0 GL -29.85m
v
-3.0 0.0 1x= 1.285E+0 60.
Fig.3.7 Displacement time histories calculated by DESRA (Linear, Input acc.=100 Gal)
-34-.
I
I
AN
S
120GaI GLOm
-120 0.0 MRXI .112E+2 60. sec
120 GL -6.Om
-120 0.0 MPX=-1.004E+2 60.
120 GL -6.9m
-120 0.0 rIRX=_9.302E+1 60.
120 GL-7.6m
-120 0.0
rvvr1YvTY jvqrTIPVR11M IVI VYV'Y IT I- IV JYWY I -V Y' YIY' I I V ~ IV
HRX=-1.079E+2 60.
120 GL-8.55m
-120 0.0 nX=-1.046E+2 60.
120 GL-9.5m
-120 0.0 tRX=1.013E+2 60.
120 GL-1O.8m
-120 0.0 FIAX=-1.096E+2 60.
120 GL-13.4m
-120 0.0 t1lX=-1.079E+2 60.
120 GL-15.95m
-120 0.0 11X=-1.048E+2 60.
120 GL -29.85m
-120- 0.0 ftX=-8.521E+1 60.
Fig.3.8 Acceleration time histories calculated by DESRA (Nonlinear & Liquefaction, Input acc.=100 Gal)
-35----
0cm GLOm
-14 . 0.0
11.0 GL-6.Om
-11.0 - 0.0
4.0 GL-6.9m
-.w-
-14.0 0.0
4.0 GL-7.6m
-'4.0 0.0
14.0 1 GL-8.55m
hAX 3.371E+0 o. sec
1RX= 2.827E+0 60.
MRX= 2.651E+0 60.
MRX= 1.951E+0 60.
I
0.0
4.0 GL-9.5m
-11.0 0.0
14.0 GL-10.8m
'-,
-14.0 0.0
14.0 GL-13.4m
-'4.0 0.0
GL-15.95m
-14.0 0.0
14.0 GL -29.85m
-14.0 0.0 MRX= 1.057E+0 60.
Fig.3.9 Displacement time histories calculated by DESRA (Nonlinear & Liquefaction, Input acc.=1 00 Gal)
-36-
MRX= 1.871E+U 60.
MRX= 1 .775E+Q 60.
MRX= 1 .529E+O 60.
4IRX= 1.1470Ee-C 60.
MRX= 1.402E+0 60.
MRX=-3.892E+2 60. -'400
0.0
00 Gal GLUm
-u-Do 0.0
GL -6.Om
-oa 0.0
:: H69 uao - GL -7.6m
-1400 0.0
100 GL -8.55m
-oo 0.0
1400 GL -9.5m
1yr
-1403 0.0
1400 ' " el
-'400 0.0
1400 (I .1QAm
-1400 0.0
(l -19m
ilAX=2.260E+2 oo. sec
MRX=-2.1SOE+2 60.
1RX=-2.187E+2 60.
h4X=-2.659E+2 60.
MRX=-2.702E+2 60.
1flX=-2.831E+2 50
MRX=-3.11814E+2 50.
MRX=-3.663E+2 60.
400 GL -29.85m
}-1400
V11111 I fTTI TRYIPY-TIr I I vIV.F-F-Iv - pv'ry.-U, V V-VIV -Y
0.0 HP.X=-3.835E+2 60.
Fig.3.10 Acceleration time histories calculated by DESRA (Nonlinear & Liquefaction, Input acc.=407 Gal)
-37-
Cm GLOm
0.0 IIRX= 1 .32'IE+l 60. sec GL-6.Om
0.0 t1AX= 1.076E+1 60. is. GL-6.9m
0.0 IIRX= 9.775E+0 60. 15. GL-7.6m
---.
0.0 MRX= 9.065E+0 60. 15. GL -8.55m
k
0.0 'IRX= 8.692E+0 60. 15. GL-9.5m
-15.
-. ..-
0.0 lIRX= 8.177E+0
15. GL-1O.8m
0.0 HPX= 7.t453E+O 60. 15. GL-13.4m
0.0 MAX= 6.943E+o 60. 15. GL-15.95m
-
0.0 HRX= 6.153E4-0 60.
GL-29.85rn
-15- - 0.0 MAX= 3.762E+O 60.
Fig.3.11 Displacement time histories calculated by DESRA (Nonlinear & Liquefaction, Input acc.=407 Gal)
-38-
Nonlinear & Liquefaction, Input acc.=407Gal
Nonlinear & Liquefaction, Inout acc.=lOOGal
0 L_. 0.1
0.5 1.0
50' iHz
Gal
200 h=O.02
Linear,Input acc.=1 OOGal
Fig.3.12 Comparisons of acceleration response spectra at the free ground surface
-39-
-20.Om
-23.3m
(Max. Acceleration)
(Max. Shear stress)
(Max. Shear strain)
-10.0 -10.0
20.0 20.Or
23.3r 0.0 100 200 300 400 500
Gal
00Gi
LNonlinear Liquefaction, Input acc.=lOOGaI
Nonlinear & Liquefaction, Input acc.=407Gal
I'
Linoar,Input acc.=looGal
I1- .II•• Nonlinear & Liquefaction, Input acc.100GaI /
/
I /
\ ' Nonlinear & Liquefaction, Input acc.=407GaI
.-..
'I
0.5
/ Liriear,Input acc.10OGal
> Nonlinear & Liquefaction, ,.'
.Input acc.=lOOGal
------------
Nonlinear & Liquefaction, Input acc.=407Gal
GL Orr GLO GL(
1.0 Kg/cm2 1.5 - 0.0 1.0 2.0 3.0 4:0
Fig.3.1 3 Comparisons of free ground responses calculated by DESRA
3.4 Earthquake response of K-Building
3.4.1 Numerical model of superstructure-composite foundation system8)
Superstructure on composite foundation is idealized by a one-stick model with lumped
masses and bending-shear elements as shown in Fig.3.14. In modeling of composite
foundation, shear rigidity(GA) is evaluated using the sectional area of the web part of wall
foundation. Bending rigidity(EI) is evaluated using the whole area of wall foundation.
Lumped mass is evaluated using the whole area of wall foundation and piles. Physical
constants of superstructure and composite foundation employed in this analytical study are
shown in Table 3.2.
Evaluation of nonlinear lateral Winkler spring
Initial lateral Winkler springs K0 are evaluated at each depth which corresponds to
the positions of lumped masses of wall foundation. In the evaluation of initial spring
value, degraded shear modulus is employed which is obtained at shear strain(O.65ymax) by
the nonlinear response of free ground.
The initial lateral Winkler spring values are calculated as follows.
The soil displacements at the discretized nodes corresponds to the position of wall
foundation, as shown in Fig.3. 16(a), can be expressed as
(u) =[D](p)
(3.1)
where (u) and (p) are the vectors of lateral displacement and load at the nodes, [ D ] is the
flexibility matrix of the soil. The soil displacements (u) caused by lateral point loads (p)
in a layered stratum are calculated at low frequency(f=O.25Hz) by the three-dimensional
thin layered element method. The stiffness matrix[ K] of soil-wall foundation system is
obtained by the inversion of soil flexibility matrix[ D ].
Next, assuming that soil displacements at the same depth are equal as shown in
Fig.3.16(b), the displacement vectors[ U] of wall foundation for each depth is expressed as
—41—
a (u)=[R](U) (3.2)
where [ R ] is a constraint matrix which relates between soil displacement and wall
foundation. Stiffness matrix[ S ] for displacement vectors[ U], which corresponds to the
position of lumped masses, is expressed as
[S]=[R]T{K][R] (3.3)
Stiffness matrix[ S] in Eq.(3.3) is obtained as a full matrix, the soil springs at each depth
are coupled. The coupled springs, however, are not appropriate for nonlinear analysis
with liquefaction. In this study, for simplicity, the following procedures are adopted to
convert the coupled soil springs into Winkler type springs(Ku).
Assuming that displacement mode of underground wall is rigid, Winkler type spring at
i-th node are approximately obtained using a matrix relation as follows:
(3.4)
(dij)=[D](l) (3.5)
Kuil/di (3.6)
Assuming that soil reaction along underground wall is uniform:
(ldj)[S](1) (3.7)
Kui=ki (3.8)
Inplace of rigid mode in Eq.(3.5), using displacement mode of underground wall by
loading at the wall foundation head:
(dij)=[D](u) (3.9)
where (u) : Displacement mode of underground wall
Kui= lId (3.10)
The rotational spring(K) at wall foundation tip is also evaluated by three-dimensional
thin layered element method, assuming that the bottom of underground wall is rigid in the
rotational direction. The rotational spring is assumed to be linear because the supporting
layer of wall foundation is enough hard.
The ultimate soil resistance for underground wall is estimated by Eq.(2.8). However,
—42-----
there are few studies for rational evaluation of these soil resistance. Further studies should
be performed to determine the ultimate soil resistance for underground wall.
The nonlinear lateral load-displacement relationship of Winkler spring is also based on
the Hardin-Drnevich model with the Masing rule, and the initial spring values and ultimate
lateral soil resistances at each depth are modified in accordance with the generation and
dissipation of excess pore water pressure.
Table 3.2 shows the physical constants of superstructure and composite foundation, and
the lateral Winkler springs and rotational spring which are evaluated assuming that
underground wall is rigid as shown in Eq.(3.4) to Eq.(3.6). Table 3.3 shows eigenvalue
analysis results of superstructure-composite foundation system with the initial Winkler
springs. Fig.3.17 also shows the natural modes of superstructure -comp o site foundation
system.
—43—
1-str
ucture
A Super
.
34• Ku Lateral Winkler spring
35 GL
36 Wall foundation
37 and Piles
44
45
46
Rotational spring KR
Fig.3.1 4 Numerical model of superstructure-composite foundation
-44 -
Table 3.2 Physical constants of superstructure-composite foundation ,and interaction spring values
NO. HEIGHT (cm)
YElGHT
(t) INERTIA (tcni 2 )
GA (t)
El (tcm 2 )
KH (t/cni)
KR (tcm/rad)
1 1. 045E+04 1. 470E+03 1. 078E+06 4. 509E+13 2 9. 985E+03 1. 096E+03 1. 585E+06 5. 908E+13 3 9. 695E+03 9. 980E+02 1. 713E+06 6. 758E+13 4 9. 405E+03 9. 980E+02 1. 754E+06 7. 131E+13 5 9. 115E+03 9. 980E+02 1. 769E+06 7. 383E+13 6 825E+03 980E+02 1. 798E+06 7. 569E+13 7 8. 535E+03 1. 056E+03 1. 848E+06 7. 822E+13 8 8. 245E+03 1. 002E+03 1. 862E+06 8. 086E+13 9 7. 9 5 5 E + 0 3 1. 002E+03 1. 867E+06 8. 147E+13
10 7. 665E+03 1. 002E+03 1. 877E+06 8. 193E+13 11 7. 375E+03 1. 002E+03 1. 894E+06 8. 251E+13 12 7. 085E+03 1. 056E+03 1. 948E+06 8. 137E+13 13 795E+03 1. 0 0 2 E + 0 3 2. 087E+06 8. 651E+13 14 6. 505E+03 1. 048E+03 2. 361E+06 9. 581E+13 15 6. 210E+03 1. 061E+03 2. 398E+06 9. 589E+13 16 5. 915E+03 1. 061E+03 2. 415E+06 9. 603E+13 17 5. 620E+03 1. 116E+03 2. 435E+06 9. 484E+13 18 5. 3 2 5 E + 0 3 1. 061E+03 2. 464E+06 9. 662E+13 19 5. 030E+03 1. 0 6 1 E + 0 3 2. 511E+06 1. 002E+14 20 4. 7 3 5 E + 0 3 1. 061E+03 2. 521E+06 1. 007E+14 21 4. 440E+03 1. 061E+03 2. 525E+06 1. 007E+11 22 4. 145E+03 1. 116E+03 2. 533E+06 9. 920E+13 23 3. 850E+03 1. 0 6 1 E + 0 3 2. 554E+06 1. 044E+14 24 3. 5 5 5 E + 0 3 1. 061E+03 2. 568E+06 1. 044E+14 25 3. 260E+03 1. 061E+03 2. 611E+06 1. 044E+14 26 2. 965E+03 1. 0 6 1 E + 0 3 2. 694E+06 1. 091E+14 27 2. 6 7 0 E + 0 3 1. 133E+03 2. 747E+06 1. 093E+14 28 2. 375E+03 1. 084E+03 2. 753E+06 1. 102E+14 29 2. 080E+03 1. 084E+03 2. 7 6 0 E + 0 6 1. 104E+14 30 1. 7 8 5 E + 0 3 1. 084E+03 2. 787E+06 1. 102E+14 31 1. 490E+03 1. 084E+03 2. 933E+06 1. 093E+14 32 1. 195E+03 1. 149E+03 3. 762E+06 1. 077E+14 33 9. 000E+02 2. 486E+03 5. 659E+06 1. 199E+14 34 4. 250E+02 3. 797E+03 3. 082E+07 1. 482E+14 35 0. 000E+00 5. 9 6 7 E + 0 3 4. 840E+09 1. 000E+10 1. 000E+17 1. 538E+03 36 -6. 000E+02 5. 520E+03 4. 541E+09 4. 330E+06 5. 560E+14 1. 239E+03 37 -6. 900E+02 3. 740E+02 4. 154E+08 4. 3 3 0 E + 0 6 5. 560E+14 1. 310E+03 38 -7. 600E+02 3. 860E+02 4. 283E+08 4. 330E+06 5. 560E+14 1. 454E+03 39 -8. 550E+02 4. 440E+02 4. 932E+08 4. 330E+06 5. 560E+14 1. 683E+03 40 -9. 5 0 0 E + 0 2 5. 260E+02 5. 841E+08 4. 3306+06 5. 560E+14 2. 107E+03 41 -1. 080E+03 9. 060E+02 1. 006E+09 4. 330E+06 5. 560E+14 2. 641E+03 42 -1. 337E+03 1. 204E+03 1. 337E+09 4. 330E+06 5. 560E+14 2. 916E+03 43 -1. 5 9 5 E + 0 3 1. 175E+03 1. 304E+09 4. 330E+06 5. 560E+14 3. 134E+03 44 -1. 840E+03 1. 146E+03 1. 272E+09 4. 330E+06 5. 560E+14 3. 608E+03 45 -2. 085E+03 1. 146E+03 1. 272E+09 4. 330E+06 5. 560E+14 4. 389E+03 46 -2. 3 3 0 E + 0 3 1 1. 041E+03 1. 1556+09 1 878E+03 1. I0IE+11
-'15-
foundation /
ug
le
Free ground respons€
CL
w,J
KR
Fig.3.15 Numerical modeling of composite foundation
x
(a) Discretized soil corresponding to underground wall foundation
Y
TT4_Ui
z (b) Contraction of soil flexibilities
x
Fig.3.1 6 Evaluation of wall foundation-soil spring
-46-
Table 3.3 Results of eigenvalue analysis
Natural
period
(sec)
Natural
frequency
(Hz)
Participation
coefficient
1st 2.09107 0.47822 1.49125
2nd 0.62435 1.60166 0.77741
3rd 0.35800 2.79331 0.65210
(1 st) (2 nd) (3 rd)
Fig.3.1 7 Natural modes of superstructure-composite foundation system with interaction springs
—47—
3.4.2 Results of earthquake response analysis for K-Building
Earthquake response analyses were performed to investigate the followings:
Effects of initial Winkler springs which are evaluated using the aforementioned
procedures on earthquake response of superstructure-composite foundation.
Differences of earthquake response by the different input acceleration levels.
Differences of earthquake response by equivalent linear analysis and nonlinear analysis
with liquefaction.
Effects of Winkler spring values evaluated using initial shear modulus of free ground and
degraded shear modulus for free ground response.
These analytical results are indicated and discussed below.
(1) Earthquake responses with initial Winkler springs evaluated by different procedures
As described before, coupled lateral soil springs connected to wall foundation are
obtained by Eq.(3.3). Simplified Winkler springs are also evaluated by using Eq.(3.6),
Eq.(3.8) and Eq.(3.10). In this study, effects of soil spring evaluation procedures on
earthquake response are investigated by comparison of the following linear analyses.
Case.2-1 : Analysis with lateral coupled soil springs obtained by Eq.(3.3)
Case.2-2 : Analysis with lateral Winkler springs obtained by Eq.(3.6)
Case.2-3 : Analysis with lateral Winkler springs obtained by Eq.(3.8)
Case.2-4 : Analysis with lateral Winkler springs obtained by Eq.(3. 10)
The conditions of these analyses are as follows:
Maximum input acceleration - 100 gal
Soil property for free ground response - Linear
Soil spring - Linear
Fig.3.18 and Fig.3.19 show acceleration and displacement time histories at the roof and
the first floor. Fig.3.20 shows acceleration response spectra at the roof and the first floor.
—48—
Fig.3.21 and Fig.3.22 also show distributions of maximum accelerations and displacements,
and those of maximum shear forces and overturning moments.
Fig.3.18 and Fig.3.19 indicate that acceleration and displacement responses in Case.2-2
to Case.2-4 give similar results to Case.2-1. Acceleration response spectra of Case.2-2 to
Case.2-4 in Fig.3.20 are also good agreement with Case.2-1. Two peaks in acceleration
response spectra at the roof coincide with the first and second natural frequency of
superstructure-composite foundation system, it can be seen that the second mode of
superstructure at 1.6Hz is amplified because the dominant frequency of input earthquake
correspond to the second frequency of superstructure. Fig.3.21 and Fig.3.22 indicate that
responses of Case.2-2 to Case.2-4 are underestimated in comparison with Case.2-1, but It
can be said that the simplified analysis models with Winkler springs give reasonable
results in acceleration and displacement responses of superstructure on composite
foundation.
—49—
(Gal) 500 1
RF Case.2-1
MPX=4 .215E+2
-500
0.0
60. 1100 - RF Case.2-2
MAX=3.841E+2 (sec)
-1400 J
0.0
60.
I00 1 RF Case.2-3
MPX=3 .5611E+2
-1100 0.0
4 00 1 RF Case.2-4 60. S
MPX=3 .989E+2
-1100 0.0
(Gal) 200 1 1FCase.2-1 MAX=1 .832E+2
iI
-200 J
0.0
60.
200 1 1FCase.2-2
MRX=1.701E+2 (sec)
-200 J
0.0
60
200 1 1FCase.2-3
MPX=1 .613E+2
-200
0.0 60.
200 1FCase.2-4 MPX=1.739E+2
..
-200
0.0 1 60.
Fig.3.18 Acceleration time histories (Case.2) -50-
(cm) is. 1 RFCase.2-1 MAX1.288E+i
-15.J
0.0 60.
15. RFCase.2-2 MAX=1 .267E+1 (sec)
-15.
0.0 60.
is. RFCase.2-3 MPX1 .261E+1
- - - -'IV 4 n v
A v A v
A '. V
vAvpf\\-Jov A
V 6 Q n V A V A V A V v v A
v
-15. 1
. 0.
is. 1 RFCase.2-4 MPX1.272E+1
-15.
0.0 60.
(cm) 2.5 1 1FCase.2-1 MPX2.1124E+0
-2.5
0.0 60.
2.5 1 1FCase.2-2 MAX=2 .321E+0 (sec)
-2.5
0.0 60.
2.5 1 1FCase.2-3 I MPX=2.371E+0
-2.5
0.0 60.
2.5 1FCase.2-4 MFRX=2.361E+0
-2
Fig.3.19 Displacement time histories (Case.2) -51-
Gal RF q o c
Hz
Gal iF 1 SC
Hz
Fig.3.20 Comparisons of acceleration response spectra (Case.2)
—52----
I 4GL
I I I -23.3m 300.0 400.0 500.0 0.0
(Gal)
GL+1 04.5m RF
GL iF 0.Om
30F
25F
1OF
15F
20F
5F
GL -23.3 m
0.0 100.0 200.0
GL+1 04.5m RFI
251
all
1 SF
1OF
5F
GL iF 0.Om
/Cgse.2-4
se.2-2
5.0 10.0 15.0 (cm)
(a)Max. acceleration (b)Max. displacement
Fig.3.21 Comparisons of maximum acceleration and displacement (Case.2)
-53-
GL+1 04.5m
GL+l 0 RF
RF
30F
25F
20 F
30F
25F
20F Case4
4.
15F
1OF
5F
GL iF 0.Om
GL -23.3 m
GL -23.3 m
15F
1OF
5F
GL iF 0.Om
\\ Case.2-1
II
Ca-se.
- - Case.2-4
- ( t\
--------------
\"\ - - \
0.0 1000.0 2000.0 3000.0 4000.0 5000.0 0.0 0.5 1.0 1.5 2.0 (ton) x107 (ton cm)
(a)Max. shear force (b)Max. overturning moment
Fig.3.22 Comparisons of maximum shear force and overturning moment (Case.2)
-54-
(2) Earthquake responses by different input acceleration levels
In this study, earthquake response properties of superstructure and composite
foundation are investigated by comparison of different input acceleration levels.
Earthquake response analyses are conducted with the following cases.
Case.3-1 Maximum input acceleration - 100 gal
Soil property for free ground response - Linear
Soil spring - Linear
Case.3-2: Maximum input acceleration - 100 gal
Soil property for free ground response - Nonlinear & Liquefaction
Soil spring - Nonlinear & Liquefaction
Case.3-3 : Maximum input acceleration is 407 gal
Soil property for free ground response - Nonlinear & Liquefaction
Soil spring - Nonlinear & Liquefaction
In Case.3-2 and Case.3-3, initial Winkler spring values are evaluated using degraded shear
modulus at the effective shear Strain(O.65ymax) which is obtained by the nonlinear response
of free ground for 100 gal and 407 gal input.
Fig.3.23 and Fig.3.24 show acceleration and displacement time histories at the roof
and the first floor. Fig.3.25 shows acceleration response spectra at the roof and the first
floor. Fig.3.26(a) and Fig.3.26(b) show distributions of maximum accelerations and
displacements including normalized results for 100 gal input. Fig.3.27(a) and Fig.3.27(b)
also show maximum shear forces and overturning moments including normalized results
for 100 gal input. Fig.3.28 shows distributions of maximum deformation of Winkler spring
at each depth.
Fig.3.23 and Fig.3.26(a) indicate that acceleration responses of superstructure in
Case.3-2 and Case.3-3 are reduced due to soil nonlinearity and liquefaction in comparison
with Case.3-1. Acceleration response spectra of Case.3-2 in Fig.3.25 also indicate the
—55—
lower peak amplitude at 1.6Hz compared with Case.3-1, which result from the difference
of free ground response. As for composite foundation response, it can be seen that larger
shear force and moment occur in liquefied soil of Case.3-2 and Case.3-3 compared with
Case.3-1. Superstructure responses in Case.3-3 indicate that acceleration response in
nonlinear liquefiable soil is not amplified in proportion to the input acceleration level, but
displacement response with low frequency content is amplified. It is found that
amplification of acceleration response of superstructure in nonlinear liquefiable soil during
strong earthquake is much smaller than those in linear soil, but shear force of wall
founadtion in liquefied soil become larger in comparison with that in linear soil. This is
caused by degradation of soil springs and large soil deformation as shown in Fig.3.28.
n
—56—
60 (sec)
MflY2 .21 2P+2
-1400 J 0.0
250 RF Case.3-2
-250 0.0
1000 1 RFCase.3-3
S -1000
0.0
MRX=8 .1425E-t-2
sI
(Gal) 1400 1 RFCase.3-1 MRX=3 .8Lj1E+2
1 F Case.3-1
MRX=1 .701E+2
-200 0.0
60. (sec)
120 1 1FCase.3-2
MAX=1 .109E+2
-120 0.0
500 1 1FCase.3-3
-500 0.0
MAX=LI .856E+2
sI
uI
Fig.3.23 Acceleration time histories (Case.3)
-57 -
0.0
50 . RFC2s P- MYd 22P-.i
(cm)
1 5. RF Case.3-1 MPX=1 .267E-i-1
-15. 0.0
60. (sec)
15. 1 RE Case.3-2
MAX=1 .213E+1
(cm) 2.5 1
lFCase.3-1 MRX=2.321E+0
-2.5
0.0 60. (sec)
2.5 1 1FCase.3-2 MRX=2.170E+0
-2.5
0.0 60.
15.1 1FCase.3-3 MAX=1.078E+1
-15.
0.0 50.
Fig.3.24 Displacement time histories (Case.3)
—58—
S
RF Gal
6000- h=0.02
0.1 0.5 1 1 1 1.0 5.0 10Hz
iF Gal
3000- h=0.02
Case.3-1
0 lTiñ
0.1 0.5 1.0 5.0 10 Hz
Fig.3.25 Comparisons of acceleration response spectra (Case.3)
- 59—
G L+ 104.5 m RF
30F
25F
20 F
15F
1OF
5F
GL iF 0.Om
GL -23.3 m
GL+1 04.5m RF
30F
25F
20F
15F
1 OF
5F
3L iF ).Om
23.3m
LI
0.0 250.0 500.0 750.0 1000.0 0.0 100.0 200.0 300.0 400.0 (Gal)
(Gal)
Normalized by input acceleration 100 gal
Fig.3.26(a) Comparisons of maximum acceleration (Case.3)
-60-
GL+1 04.5m RF
30F
25F
20 F
1 5F
1OF
5F
GL iF 0.Om
GL+i 04.5m RFI
/
/ ------------------- /
III
/I / "
Case.3-1 7
//
20F
--------------
1 5F
1 OF
5F
30F
25F
GL -23.3m
0.0 20.0 40.0 60.0 (cm)
23.3m 0.0
I,)
/
5.0 10.0 15.0 (cm)
Normalized by input acceleration 100 gal
Fig.3.26(b) Comparisons of maximum displacement (Case.3)
-61-
-62----
G L+ 1 RF
GL+1 04.5m RFIITT
0 5000 (ton)
0 1000 2000 3000
30F
25F
20F
15F
1OF
5F
GL iF 0.Om
30F
25F
20F
15F
1OF
5F
3L iF ).Om
3L 23.3m 0 0
-
HE - .0 5000.0 100000 i.'nUC
(ton)
UT -
U Case.3-1
1J:I: Case.3-3
GL -23.3m
0
S
NormaUzed by input acceleration 100 gal
Fig.3.27(a) Comparisons of maximum shear force (Case.3)
C L+ 1 RF
301'
25F
20 F
15F
lOf
51
CL iF D.Om
CL -23.3 m
GL+1 04.5m RF
30F
25F
20F
15F
1OF
5F
• 0.Om
GL -23.3m
0.0 2.5 5.0 7.5 0.0 0.5 1.0 1.5 2.0
X10 (ton cm) xl 07(ton cm)
Normalized by input acceleration 100 gal
Fig.3.27(b) Comparisons of maximum overturning moment (Case.3)
-63-
GLO.Om
GL 0.Om
I
G L-23 .3 m G L-23 .3 m 0.0 3.0 6.0
Fig.3.28
Case.3-1
LPs /
1-
/s II:
/
/
/ /
9.0 12.0
0.0 1.0 2.0 3.0 (cm)
(cm) Normalized by input acceleration 100 gal
Comparisons of maximum Winkler spring deformations (Case.3)
O
(3) Earthquake responses by equivalent linear analysis and nonlinear analysis with
liquefaction
In this study, equivalent linear analysis method is investigated by comparison with
analysis taking into account soil nonlinearity and liquefaction. In equivalent linear
analysis, linear Winkler spring values are employed which are obtained at the effective
shear strain(0.65ymax) by the nonlinear response of free ground.
Earthquake response analyses for 100 gal and 407 gal input are conducted with the
following cases.
Case.4.1-1 : Maximum input acceleration - 100 gal
Soil property for free ground response - Nonlinear
Soil spring - Linear degraded Winkler spring
Case.4.1-2 : Maximum input acceleration - 100 gal
Soil property for free ground response - Nonlinear & Liquefaction
Soil spring - Nonlinear & Liquefaction
As for Case.4.1-1 and Case.4.1-2, Fig.3.29 and Fig.3.30 show acceleration and
displacement time histories at the roof and the first floor. Fig.3.31 shows acceleration
response spectra at the roof and the first floor. Fig.3.32 and Fig.3.33 also show
distributions of maximum accelerations and displacements, and those of maximum shear
Is forces and overturning moments.
Case.4.2-1 : Maximum input acceleration - 407 gal
Soil property for free ground response - Nonlinear
Soil spring - Linear degraded Winkler spring
Case.4.2-2: Maximum input acceleration - 407 gal
Soil property for free ground response - Nonlinear & Liquefaction
Soil spring - Nonlinear & Liquefaction
—65—
As for Case.4.2-1 and Case.4.2-2, Fig.3.34 and Fig.3.35 show acceleration and
displacement time histories at the roof and the first floor. Fig.3.36 shows acceleration
response spectra at the roof and the first floor. Fig.3.37 and Fig.3.38 also show
distributions of maximum accelerations and displacements, and maximum shear forces and
overturning moments.
Response results in Fig.3.29 to Fig.3.33 indicate that equivalent linear analysis for 100
gal input gives almost same responses in superstructure and composite foundation as
nonlinear effective stress analysis. Thus equivalent linear analysis method is confirmed to
be useful for the low level input. However, as shown in Fig.3.34 to Fig.3.38, responses for
407 gal input indicate larger displacement of superstructure, smaller shear force and larger
moment of composite foundation in the equivalent linear analysis. It is confirmed that
response of structure during strong earthquake is strongly affected by nonlinear soil around
foundation and the changing of effective stress.
—66-----
(Gal) 250 RFCase4.1-1 MPY=22flflP+2
0.0 60. (sec)
RFCase.4.1-2 MOY- 250
60
120 1 1FCase.4.1-2 MAX=1 .109E+2
iF(p41-1 Mqx=1.nqnF±2 (Gal) 120
60 (sec)
-120 0.0
Fig.3.29 Acceleration time histories (Case.4.1)
-67-
(cm) is. RFCase.4.1-1
-15. 0.0
is. RFCase.4.1-2
0.0
MAX=1 .25L4E+l
MPX=1 .213E+l
0 (sec)
r
(cm) 2.5 1 1FCase.4.1-1 MAX=2 . 18LIE+0
-2.5 0.0
2.5 1FCase.4.1-2
-2.5 0.0
MPX=2 . 170E+0
80 (sec)
80.
Fig.3.30 Displacement time histories (Case.4.1)
Ga I
am
200
Hz
iF Ga
Hz
Fig.3.31 Comparisons of acceleration response spectra (Case.4.1)
—69—
GL+1 04.5m RF
30F
25F
20F
15F
1OF
5F
GL iF 0.Om
GL+1 04.5m RF
30F
25F
20F
15F
1 OF
5F
iF
GLL 1 JGL L/ -23.3m1 I i I-23.3m1 I I
0.0 50.0 100.0 150.0 200.0 250.0 0.0 5.0 10.0 15.0 (Gal) (cm)
(a)Max. acceleration (b) Max. displacement
Fig.3.32 Comparisons of maximum acceleration and displacement (Case.4.1)
-70-
GL+1 04.5m
GL±1 04.5m RFP1
RF
30F - -- - - - - - - - - - - - - - - - - - - - - - - I 30F
25F
25F
20F - ----------- 20F
I
Case.4.1-1 Case.4.1-2
1 5F
1 OF
1 5F
1 OF
\ _Case.4.1-1
Case.4.1-2\
5F I- ------------------ - - - - - - - - - - - - - - - I
5F
GL iF
GL iF O.Om
O.Om
GL r d-23.3m GL r
-23.3mr :1 I I t I
0.0 1000.0 2000.0 3000.0 4000.0 0.0 0.5 1.0 1.5 2.0 (ton) X10 (ton cm)
(a)Max. shear force (b)Max. overturning moment
Fig.3.33 Comparisons of maximum shear force and overturning moment (Case.4.1)
-71-
(Gal) 1 000 1
RF Case.4.2-1 MAX=8 .271E+2
-1000 J 0.0
60 (sec)
1 000 1 RF Case.4.2-2
MAX=8 .425E+2
-1000 0.0
(Gal) 1 F C 500 ase.4.2-1
..4vJVV'v #
-500 0.0
500 1 F Case.42-2
-500 - 0.0
60
MPX=4 .72E+2
MPX=4 .856E+2
60 (sec)
S
Fig.3.34 Acceleration time histories (Case.4.2)
-72-
Pt12-1 MPX=4L552E+1 (cm) 50.
60 (sec)
RFCas42-2 M11X=L .AF+1 50.
60
1 F Case.4.2-1
-15. 0.0
i • 1 F Case.4.2-2
w
0.0
MRX=1 .00LIE+1
MAX=1 .078E+1
60 (sec)
Fig.3.35 Displacement time histories (Case.4.2)
-73--.
Ga
Hz
iF
Gal
300
Hz
S
Fig.3.36 Comparisons of acceleration response spectra (Case.4.2)
—74—
--------------------------
1 OF 1OF
5F
------------------- 5F
GL+1 04.5m
GL+1 04.5m RF
RF
30F
30F
Case.4.2-1
25F
4 ase..- 25F
/ Case.4.2-1
Case.4.?-2
20 F
20F
1 5F
15F
• GL iF
O.Om
0.Om
GL L JGL -23.3m1 I I I I-23.3m
0.0 250.0 500.0 750.01000.0 0.0 (Gal)
(a)Max. acceleration
20.0 40.0 60.0 (cm)
(b)Max. displacement
Fig.3.37 Comparisons of maximum acceleration and displacement (Case.4.2)
-75-
GL+1 C RF
---------------------------------
I'
.....................................
\ Case.4.2-1
Case.4.2-2\
)F
5F I
)F
5F I
25F
20 F
15F
F
\\ \
GLI-J-23.3m
GL -23.3m1 I I!i I
1OF
5F
GL iF 0.Om
1 04.5m F I
0.0 5000.0 10000.0 15000.0 20000.0 0.0 2.5 5.0 7.5
(ton) x107 (ton cm)
(a)Max. shear force (b)Max. overturning moment
Fig.3.38 Comparisons of maximum shear force and overturning moment (Case.4.2)
-76-
(4) Earthquake responses with different initial Winkler spring values
In the aforementioned analyses, degraded shear modulus by nonlinear response of free
ground is employed in the evaluation of initial Winkler spring values. This procedure
incorporates soil nonlinearity around composite foundation by earthquake input. However it
can be considered that improved soil around composite foundation is not degraded as much
as free ground during earthquake. In this study, earthquake response results are investigated
using initial Winkler spring values which are evaluated in consideration of these
conditions.
Earthquake response analyses for 100 gal and 407 gal input are conducted with the
following cases.
Case.5.1-1 :Maximum input acceleration - 100 gal
Soil property for free ground response - Nonlinear & Liquefaction
Initial Winkler spring - Evaluated using shear modulus of free ground
Case.5.1-2 : Maximum input acceleration - 100 gal
Soil property for free ground response - Nonlinear & Liquefaction
Initial Winkler spring - Evaluated using degraded shear modulus by nonlinear free
ground analysis
As for Case.5.1-1 and Case.5.1-2, Fig.3.39 and Fig.3.40 show acceleration and
displacement time histories at the roof and the first floor. Fig.3.41 shows acceleration
response spectra at the roof and the first floor. Fig.3.42 and Fig.3.43 also show
distributions of maximum accelerations and displacements, and those of maximum shear
forces and overturning moments. Fig.3.44 show distributions of maximum deformation of
Winkler spring at each depth.
Case.5.2-1 : Maximum input acceleration - 407 gal
Soil property for free ground response - Nonlinear & Liquefaction
Initial Winkler spring - Evaluated using shear modulus of free ground
—77—
Case.5.2-2 Maximum input acceleration - 407 gal
Soil property for free ground response - Nonlinear & Liquefaction
Initial Winkler spring - Evaluated using degraded shear modulus by nonlinear free
ground analysis
As for Case.5.2-1 and Case.5.2-2, Fig.3.45 and Fig.3.46 show acceleration and
displacement time histories at the roof and the first floor. Fig.3.47 shows acceleration
response spectra at the roof and the first floor. Fig.3.48 and Fig.3.49 also show
distributions of maximum accelerations and displacements, and those of maximum shear
forces and overturning moments. Fig.3.50 show distributions of maximum deformation of
Winider spring at each depth.
Fig.3.39 to Fig.3.43 for 100 gal input indicate similar acceleration and displacement
responses of superstructure for Case.5.1-1 and Case.5.1-2, while shear force of wall
foundation at near ground surface for Case.5.1-1 is much larger than Case.5.1-2. Fig.3.45
to Fig.3.49 for 407 gal input indicate significant difference responses of superstructure and
wall foundation between Case.5.2-1 and Cse.5.2-2. It is found that the degree of soil
nonlinearity around foundation affects the response properties of superstructure and
composite foundation. This is due to degradation of soil resistance depending on the
deformation of Winider springs as shown in Fig.3.44 and Fig.3.50.
S
—78—
RFCasc51-1 MPY=2.17RF+2 (Gal) 250
60. (sec)
RFCase5l-2 MPX=2.212E-i-2 250
uI
(Gal) 120 1
1 F Case.5.1-1 MPX=1 .067E+2
-120 -'
0.0
60 (sec)
120 1 1FCase.5.1-2
MPX=1 .109E+2
-120
0.0
60
Fig.3.39 Acceleration time histories (Case.5.1)
—79---
(cm) I RFCase.5.1-1 MAX1 .229E+1
-15. 0.0
is 1 RFCase.5.1-2 MRX=1 .213E±1
-15. J
0.0
(cm) 2.51 1FCase.5.1-1 MAX=2.311E+0
-2.5 0.0
2.5 1 1FCase.5.1-2 MAX=2.170E+0
-2.5 0.0
60 (sec)
sI
sI
S
60 (sec)
Fig.3.40 Displacement time histories (Case.5.1)
Gal
RF
200
Hz
iF
Gal
I
AN
Hz
Fig.3.41 Comparisons of acceleration response spectra (Case.5.1)
—81—
GL+1 04.5m
G L+ 1 J4.bm RF
RI
30F
25F
20 F
15F
1OF
5F
- " Case.5.1-1
-
-----------------------------------
------------------ ----------------
----------------------- --- ----------
-------------------- ri II I , I , II I ,
7 ,
301
251
201
15F
1 OF
5F
6L iF O.Om
:1. ------------------------ -----------
:i_ -------------------- ------------- I
GL iF O.Om
GL
GL -23.3m •23 .3 rr
0.0 50.0 100.0 150.0 200.0 250.0 0.0 5.0 10.0 15.0 (Gal) (cm)
(a)Max. acceleration (b)Max. displacement
Fig.3.42 Comparisons of maximum acceleration and displacement (Case.5.1)
-82-
GL -23.3m
15F 15F
1 OF
5F
GL iF D.Om
3L 23.3m
1OF
5F
GL iF O.Om
GL+1 04.5m
G L+ 10 RFI I
RF
30F
30F
25F
25F
I
20F ----------------------- 20F
Case.
0.0 1000.0 2000.0 3000.0 4000.0 5000.0 0.0 0.5 1.0 1.5 2.0 (ton) X10 (ton cm)
(a)Max. shear force (b)Max. overturning moment
Fig.3.43 Comparisons of maximum shear force and overturning moment (Case.5.1)
-83-
GL 0.Om
I
---------------
GL-23.3m L
I
0.0
0.5 1.0 1.5 2.0 (cm)
Fig.3.44 Comparisons of maximum Winkler spring deformations (Case.5.1)
-84----
(Gal) 800 RFCase52-1 McY7 R 1 2F--2
60 (sec)
1 000 RF Case.5.2-2
MAX=8 .'425E+2
_i000
0.0
ip
(Gal)
oo 1FCase.5.2-1 MAX3.308E+2
-L00 J
0.0
60. (sec)
soo iF Case.5.2-2
MRX=4 .856E+2
-500
0.0
MIE
Fig.3.45 Acceleration time histories (Case.5.2)
—85—
RF ('.cc c 2_I M (1 V - II 1 fl I E (cm) 50.
60. (sec)
RFCsF2-2 Mr-iv_Tl OCO 50.
(cm) 1 5 .-1 iF Case.5.2-1 MAX=9 .635E+0
-15. J 0.0
60. (sec)
15. 1 1 F Case.5.2-2
MAX=1 .078E+1
-15. J
0.0
Fig.3.46 Displacement time histories (Case.5.2)
-86--
Ga.
RF
iTi
Hz
iF
Ga I
Fig.3.47 Comparisons of acceleration response spectra (Case.5.2)
—87 -
------------------
250.0 500.0 750.0 1000.0 0.0 (Gal)
I
20.0 40.0 60.0 (cm)
GL±1 04.5m RF
30F
25F
20F
15F
1OF
5F
GL iF 0.Om
GL -23.3m
0.0
15F
1OF
5F
31- iF ).Om
23.3 m
GL+1 04.5m RF
25F
30F
20 F
(a)Max. acceleration
(b)Max. displacement
Fig.3.48 Comparisons of maximum acceleration and displacement (Case.5.2)
-88-
GL -23.3m
5000.0 10000.0 15000.0 20000.0 0.0 (ton)
(a)Max. shear force
0_u (.0 X10 (ton cm)
(b)Max. overturning moment
GL+104.5m
GL+1 04.5m RF[T
RF
30F
25F
20F
I
1 5F
1OF
5F
GL1F 0.Om
GL -23.3 m
0.0
30F
25F
20F
15F
1OF
5F
GL iF 0.Om
Fig.3.49 Comparisons of maximum shear force and overturning moment (Case.5.2)
-89-
GL O.Om
Case.5.2-1
G L-23 .3 m 0.0 3.0 6.0 9.0 12.0
(cm)
Fig.3.50 Comparisons of maximum Winkler spring deformations (Case.5.2)
S
-90----
4. 2-Dimensional finite element model and verification study
This chapter describes the correlation analyses for the centrifugal test results of a
pile foundation model which consists of a rigid pile cap and a four-pile group embedded in
dry sana. This analysis is conducted to verify the numerical modeling technique in
considering with the nonlinearity of soil. The detail of the test is shown in the previous
report of KajimaJCUREe project.').
4.1 Numerical model of a pile foundation in nonlinear soil
In this nonlinear correlation analysis, the employed numerical procedure are
described below,
The 2-dimensional Finite Element Method(FEM) for spatial discretization.
Step by step direct time integration by Newmark P.method.
The modified R-O model and plastic theory is applied to the soil element.
Finite element model
The finite element model investigated in this analysis is shown in Fig.4. 1. The soil
is divided to three parts, the inner soil is the part in which the pile-foundation-soil
interaction effects may be induced strongly. On the contrary, the outer and side soil are the
region where can be treated as free field soil deposit. And the both of vertical and
horizontal direction of inner and outer soil is connected by the dashpot element expressed
as C = pVA, in which p = density, V = shear wave velocity, A = governed area of nodal
points. This model is developed to be able to represent 3-dimensional effect by 2-
dimensional FEM solution.
The piles are modeled by flexural beam element, and the soil is modeled by four-
nodes 2-dimensional plane -strain element. Physical constants for the dry sand are
indicated in Table 4.1. The initial shear modulus G0 listed in this table are estimated from
equation (2.2).
-91-
Side Soil r Outer Soil - Inner Soil
Pile2 / Pile,,,,/" -i
CD 00
I Pilel Pile4 00
475 828 475
1778 (Unit ; cm)
Dashpot
Outer S
tl
Je Soil
Fig.4. 1 Analysis Model
-92-
I
Table 4.1 Physical Constants
Level (cm)
Unit Weight (ton/cm3)
Shear Velocity (cm/s)
Poisson's Ratio
1045.0 8.205E+03 985.0 1.161E+04 867.0 1.378E+04 749.0 1.520E+04
Inner 637.0 1.616E+04 Soil 548.0 1.470E-06 1.691E-i-04 0.33
448.0 1 .766E-i-04 342.0 1.834E+04 236.0 1.899E+04 118.0 1 .957E+04
Thickness: t= 119.0 (cm ; Inner) 771.0 (Outer) 890.0 (Side)
-93-
a Step by step time integration
The Newmark 3 method (a = 1/2, = 1/4) and Cholesky factorization by skyline
matrices formulation are employed to step by step time integration. This conservative
method became to be enough efficient for nonlinear dynamic response analysis, because of
recent progress of computer performance.
Nonlinear model
The Drucker-Prager's yield criterion and modified Prager's hardening rule is
applied to the judgment for yield and unloading. The yield function is expressed as
f=aI1 +i2 (4.1) a in which I,, is n-th stress invariant, and the plastic stress-strain relation matrix is
represented by partial differentiation off with respect to stress cr, elastic stiffness matrix,
and hardening coefficient Ct The modified Rumberg-Osgood model is employed to
determine the hardening coefficient. According to the associated flow rule, the equivalent
stress 0e is expressed as the same form with yield functionf (4.1), and the hardening
coefficient (tangent stiffness in this case) Cf 15 decided as the function of equivalent stress
Cf=bI(YeO+b) (4.2)
in which the constant a,b is determined by the expression below,
a=2ith/2(1-y)-h (4.3.a) I
b = z (yG0'y) / (1 - y)(a + 1) (4.3.b)
where, h, yc, ,yc is reference value of damping factor, stiffness degrading ratio, and
shear strain respectively. The shear modulus and damping factor versus shear strain
relation assumed in this correlation analysis is shown in Fig.4.2.
—94—
The correlation analyses for three case of dry sand test are conducted. Table 4.2
summarizes the maximum acceleration measured in each tests. The case of Dry06 is
excited by the smallest acceleration level. The measured acceleration observed at the base
level is subject to FEM model described above, and investigate the correlation of the test
results and analyses, according to the bending moment of piles and acceleration response
spectra at the points of L15, L9, and Pile cap.
—95—
1.0
0.0 1.0x10 6 1.0x10 5 1.0x10 4 1.0x1Q 3 0
1.0x10 2
PIC
(%)
40
On
Refference value
h= 0.12 = 1.0x1O
Fig.4.2 Nonlinear Model
-96-
L15
L9
Table 4.2 Measured Acceleration (gafl
DRY04 DRY05 DRY06
L15 255.4 342.1 189.4
L9 190.9 270.0 137.3
Pile cap 331.9 416.1 304.4
Base 203.1 339.4 85.2
-97-
4.2 Result of correlation analysis
The acceleration response spectra are shown in Fig.4.3 - 5. At the points of side
soil(L9 and L15), the results of test and analyses agree very well but the analysis shows
slightly smaller spectrum value in the case of Dry06. The damping effect in lower strain
range may be overestimated for the nonlinear characteristic assumed in this analyses.
Otherwise, at the point of pile cap, the predominant vibration mode of soil-pile-foundation
system is found at about 0.4 sec. The analysis of Dry06 shows smaller value at also this
point, and the other case show about 1.5 times lager than the test results. The additional
damping effects to material nonlinearity, such as slide and separation between the pile and
soil, are expected especially higher exciting case.
The transition of hardening coefficient are shown in Fig.4.6 - 8. It is found that the
change of yielding stage happens more frequently at the inner soil region.
The bending moment of pile is shown in Fig.4.9. The analysis of Dry06 shows a
little smaller than test results, but they represent good agreement for the response value as
well as the mode shape which the node point is produced at near surface.
As the 2-dimensional F.E.modeling technique for soil-pile-foundation system, the
method which the soil domain is divided to multi-region according to the degree of soil-
pile interaction effects is proposed. And the proposed method is verified according to the
correlation analysis for the centrifuge test of pile model in dry sand.
—98—
-600.0 1 0 0 20 30 (SEC)
L15 OF BOX
-600.0 L._._._,,_._.._. -coo.o[ to zo 30 scc I 0 10 20 30 tOEd 0
PILE CAP L9 OF BOX
Fig.4.3 Response Spectra and Time Histories (DRY04)
1 ouuu
- - 2000
- - 1000
0 0.10 0.20 0.50 1.00 1.40 2.00 0.
I GIlL) 600.0 1 600.0
0 0.20 0.50 1.00 1.40 2.00
[GIlL)
CAP
TEZT - PNflLYSIS 3000
2000
1000
0 0.
600.0
0.20 0.50 LOU 1.
3000
2000
1000
0
L15 OF BOX L9OFBOX
TEST -------------------- INELYSIS TEST --------------------ANIlLYSIS
3000
2000
CAP
PNcLYSIS 3000
2000
L15 OF BOX L9OFBOX NIN
TEST ---- .--------------- flNnLYSIS tsi .................... flNflLYSIS
2000
- 1000 C C
0 0.
600.0
10 0.20 0.50 1.00 1.40 2.1
(CclL)
0 10 0
600.0
10 0.20 0.50 1.00 1.40 2.1
(COL]
(000
0 0 0.
600.0
10 0.20 0.50 1.00 1.40 2.00
(Ct1L)
-600.0 1 0
-600.0 -600.0 .,. 0 20 30 (500) .
0 0 20 30 (SOd 1 0 10 20 30 (SECJ
PILE CAP L9 OF BOX L15 OF BOX
Fig.4.4 Response Spectra and Time Histories (DRYO5)
L15 OF BOX
ANYSIS
L9 OF BOX
TEST --------------------FNLYSIS TEST
3000
2000
1000
3000
2000
- 1000 Q
0 0.
600.0
CAP
N$LYSI6
3000
2000
1000
II 0 0.20 0.50 1.00 1.40 2.00 0.
600.0
0.20 0.50 1.00 1.40 2.00 0.10 0.20 0.50 1.00 1.40 2.00 (CPU
600.0 .
-600.0 0 ;o 30 (SEC) 0 ID 20 30 (SECI 0 0 20 30 (5CC)
PILE CAP L9 OF BOX
L15 OF BOX
Fig.4.5 Response Spectra and Time Histories (DRYO6)
0.0 os 1.0 0.0 0.5 1.0 0.0 0.5 1.0 II ______________________ I
lOsec
20s ec 3 Os ec
Fig.4.6 Hardening Coefficient ( DRYO4)
0 ___________
S
0.0 0.5 1.0 0.0 0.5 1.0 0.0 0.5 1.0
I 1 I I I I
lOsec
2Osec
3Osec
Fig.4.7 Hardening Coefficient ( DRY05)
0.0 0.5 1.0 0.0 0.5 1.0 0.0 0.5
mmmRw=k I _____________
lOsec
2Osec
3 Os ec
Fig.4.8 Hardening Coefficient ( DRYO6
(cm) 0 TEST
1395 LEVEL
1045
0
/ /
/ o
/ \ \ \
o \
o \
ID I,
1000.0 2000.0 3000.0 4000.0 (ton cm)
ANA. (cm) 0 TEST
M. LEVEL
1045
1i I.
\
0
/ \ \
O\ \.
\\
/0
/
/ /
1000.0 2000.0 3000.0 4000,0 (ton 'cm)
Nfl. RNR.
/ /
0/
0
\ 0
I C
/ 0
I
\
1000.0 2000.0 3000.0 4000.0 (ton 'cm)
DRYO4 DRYO6
Fig.4.9 Pile Moment
(cm) 0 TEST
1395 LEVEL
1045
5. Conclusions
The results in this study can be summarized as follows:
The proposed numerical model, which consists of beam elements and nonlinear lateral
Winkler springs, taking into account the changing effective stress, can effectively predict
the centrifuge test of pile foundation model in dry or saturated sand. It can be confirmed
that the proposed model represents the response characteristics of a pile foundation in
nonlinear liquefied soil.
Earthquake responses of the building supporting on composite foundation are strongly
affected by nonlinear liquefied soil-foundation interaction during strong earthquake.
Response properties of superstructure change depending on the magnitude of input motion
and soil nonlinearity of surrounding soil. Particularly, Composite foundation responses
significantly change by large deformation and degraded resistance of soil due to
liquefaction.
2-dimensional finite element model incorporating soil nonlinearity, which the soil
domain is divided into multi-regions according to the degree of soil-pile interaction effects
is proposed. And the proposed method is verified by the correlation analysis for the 41 centrifuge test of pile foundation in dry sand.
—106—
References
Kajima-CUREe Research Project Report: Dynamic Interaction between Pile Group and Non-linear Soil, 1991. 2
Finn, W.D.L., Lee, K.W. and Martin, G.R. :An Effective Stress Model for Liquefaction, Journal of the Geotechnical Eng. Division, ASCE, Vol.103, pp.517-533, 1977
Richart, F.E., Hall, J.R. and Woods, R.D. :Vibrations of Solids and Foundations, Prentice-Hall Inc., 1970
Kausel, E. and Peek, R. :Dynamic Loads in the Interior of a Layered Stratum-An Explicit Solution-, Bulletin of the Seismological Society of America, Vol.72, pp.1459-1481, 1982
Broms, B.B. :Design of Laterally Loaded Piles, Journal of the Soil Mechanics and Foundations Division, ASCE, Vol.91, pp.79-99, 1965
Masuda, K., Sasaki, F., Urao, K., Ueno, K. and Miyamoto, Y. :Simulation Analysis of Forced Vibration Test for Actual Pile Foundation by Thin Layer Method, Proc. of Reliability and Robustness of Eng. Software Conf., 1987
Takemura, M. and Ikeura, T. :A Semi-Empirical Method using a Hybrid of Stochastic and Deterministic Fault Model-Simulation of Strong Ground Motions during Large Earthquakes-, J. Phys. Earth, 36, 1986
Masuda, K., Kitamura, E., Miura, K. and Miyamoto, Y. :Dynamic Soil-Structure Interaction Analysis Considering Diaphragm Wall and Pile Group, Tnt. Conf. on Computational Eng. Science, 1991
—107—
CUREe-KAJIMA RESEARCH PROJECT
A METHODOLOGY FOR NONLINEAR SOIL-STRUCTURE INTERACTION
EFFECTS USING TIME-DOMAIN ANALYSIS TECHNIQUES
Ronaldo I. Borja and H. Allison Smith
Dept. of Civil Engineering
Stanford University
February 15, 1991 - May 14, 1992
A METHODOLOGY FOR NONLINEAR SOIL-STRUCTURE INTERACTION EFFECTS USING TIME-DOMAIN ANALYSIS
TECHNIQUES
Ronaldo I. Borja and H. Allison Smith Department of Civil Engineering, Stanford University
June 15, 1992
PROJECT SUMMARY
The dynamic response of rigid foundations on an elasto-viscoplastic half-space is inves-
tigated in the context of nonlinear finite element (FE) analysis. A deviatoric viscoplastic
theory with a linear combination of isotropic and kinematic hardening is used to model the
soil constitutive response. Large-scale nonlinear FE computations are made feasible by the
use of a composite Newton-PCG iteration technique, which requires the factorization of the
consistent tangent operator no more than once during the solution process. Time-domain
analyses are used to investigate the nonlinear responses of vertically oscillating circular,
square, and rectangular foundations to harmonic loads, using.two- and three-dimensional
FE modeling. It is shown that for low frequency excitations, resonance is created which
amplifies the motion of the foundation at amplitudes well above those obtained at the
zero-frequency level.
In addition, horizontal, rocking, and torsional vibration modes of strip and square
foundations are considered using the same methodology developed for vertically oscillating
foundations. The foundation responses for horizontal, rocking, and torsional modes are
characterized by increased vibrational amplitudes due to material stiffness degradation.
Furthermore, one or more resonance frequencies are created which resemble those observed
for vertically oscillating finite-size foundations. Nonlinear soil effects are shown to be
dominant over a wide range of excitation frequencies for foundations vibrating in torsional
and horizontal modes. In contrast, nonlinear soil effects are shown to be dominant over a
much narrower range of excitation frequencies for the vertical and rocking modes.
1
Contents
1 Introduction 3
2 Model Description 5
2.1 Soil-Structure Model ............................... 5
2.2 Solution Algorithm ............................... 6
2.3 Constitutive Model ................................. 7
2.4 Forcing Function and Radiation Boundary Condition ............ 10
3 Foundations Subjected to Vertical Vibrations 11
3.1 Vertical Vibration of Circular Foundations ..................11
3.2 Vertical Vibration of Square Foundations ...................15
3.3 Discussion of Results ..............................16
3.4 Vertical Vibration of Rectangular and Strip Foundations ..........18
3.5 Summary and Conclusions ...........................21
4 Foundations Subjected to Horizontal, Rocking, and Torsional Vi-
brations 23
4.1 Symmetry/Anti-Symmetry and Forcing Functions ..............24
4.2 Horizontal and Rocking Vibrations of Strip Foundations ..........25
4.3 "Critical Frequency" and "Optimal" FE Grid ................28
4.4 Horizontal, Rocking, and Torsional Vibrations of Square Foundations . 30
4.5 Summary and Conclusions ...........................32
5 Boundary Element Model for Plane Strain 34
5.1 Time-Domain BE Formulation .........................34
5.2 Boundary Integrals for Plane Strain .......................36
6 List of Publications 38
7 Acknowledgements 39
P 1 Introduction
Accurate representation of soil-structure interaction effects is a crucial part of earthquake
engineering analyses. A large number of soil-structure models exist of varying degrees of
complexity. The choice of the most appropriate model for a particular analysis is based on
the building-foundation system of interest, the desired level of accuracy, and experienced
"engineering judgment." In many cases, multiple models must be analyzed and experi-
mentally verified before one can select the most effective model for a particular system.
It is generally recognized that the dynamic response of soils is nonlinear, hysteretic, and
irreversible. At sufficiently high strain levels, soil deformations also are known to exhibit
pronounced rate-dependent effects. While nonlinear effects and other unusual behavioral
features dominate soil responses in all but a few exceptional cases, very little work has
been done to incorporate them into soil-structure interaction (SSI) models (Lysmer et al.
1974; Kausel et al. 1976; Jakub and Roesset 1977).
Part of the difficulty associated with nonlinear SSI analyses stems from the lack of ef-
ficient mathematical tools with which one can transform a hyperbolic structural dynamics
problem into a more tractable elliptic problem as in linear analyses. Nonlinearities usually
force the analyst to abandon frequency-domain analyses in favor of more computationally
intensive time-domain analyses. Without numerical enhancements, "conventional" nonlin-
ear time-domain codes usually break down from the rigors of heavy computation associated
with time-stepping and iterations. Thus, analysts typically are forced to neglect nonlin-
ear effects due to the lack of effective SSI models which can accurately model the soil
complexities in a computationally feasible manner.
This report presents a finite element (FE) methodology for modeling nonlinear SSI
effects and applies the model to the problem of machine foundations. While the FE formu-
lation presented here is generalized to be applicable to any nonlinear structural dynamics
problem, the cornerstone of the present model is a state-of-the-art algorithm that empha-
sizes computational efficiency. Thus, the model avoids repeated triangular factorization of
3
the effective mass matrix in the nonlinear analysis while preserving the optimal order of
convergence of an iterative method (Ortega and Rheinboldt 1970).
Hysteresis, rate-dependence, and irreversible deformation are dominant features of
cyclic soil behavior (Perzyna 1971; Prevost 1977; Mróz et al. 1979; Dafalias and Herrmann
1982). This suggests that elasto-plastic and elasto-viscoplastic models may be appropriate
for modeling the soil constitutive behavior. In particular, the presence of hysteretic damp-
ing in machine foundation problems demands that the soil constitutive model be of the
"kinematically-hardening" type (Prager 1956; Mróz 1967). While it can always be argued
that one constitutive model may be more appropriate than another, it is not the objective
of this paper to present such an argument. It is the aim of this paper to demonstrate,
with the simplest possible constitutive model, the effect of soil stiffness degradation, rate
of loading, and hysteresis on the dynamic response of machine foundations. For this pur-
pose, the elasto-viscoplastic theory of Duvaut and Lions (1976) is used to model the soil
behavior.
The effect of repeated loading on machine foundations is to create a zone of intense
yielding and inelastic deformation beneath the foundation. This results in "dynamic non-
homogeneity" with respect to soil elasticities, which perturbs the overall response of an
otherwise homogeneous soil-structure system. This report shows that plastic deformation
(and soil stiffness degradation) for vibrating rigid foundations results in an overall increase
in vibrational displacement amplitudes and the creation of resonance frequencies where 41 motion is amplified above those at zero-frequency level. The resOnance phenomenon is
explained in the context of a single-degree-of-freedom (SDOF) oscillator analog which has
been used successfully by previous investigators to model prototype continuum SSI prob-
lems (Hsieh 1962; Lysmer 1965; Richart and Whitman 1967; Richart et al. 1970; Veletsos
and Wei 1971; Kausel 1974; Luco 1974; Gazetas 1983).
4
2 Model Description
This section contains a description of the soil-structure interaction model investigated in
this report. The equations of motion are formulated for a nonlinear soil-structure sys-
tem, and the constitutive model representing the nonlinear soil properties is presented.
In addition, the composite Newton-PCG iterative solution algorithm is formulated and
discussed.
2.1 Soil-Structure Model
Consider a massless, rigid foundation of arbitrary shape shown in Fig. 2.1, and let the
forcing function be denoted by the nodal vector FEXT(t). The forcing function FEXT
may arise from an externally applied dynamic load (as in machine foundation) or from
propagating incident waves (such as seismic excitations). The FE dynamic equation of
motion then reads
Mx + FINT = FEXT (1)
where M = mass matrix; x = nodal acceleration vector; and FINT = internal nodal force vector, which takes the form
FINT = if BTcrdfI
(2)
where BT = s train- displacement transformation matrix; o = Cauchy stress vector; and S Q = problem domain. The stress vector o, may depend in general on the nodal velocity
vector * and on the nodal displacement vector x either in a linear manner or through a
more complex relationship.
Assuming that FEXT is given over a certain time domain of interest, then Eq. (2) may
be time-integrated in the form
Ma+j + (1 + c)(FINT)fl+l - a(FINT) = FEXT(tn+l+a) (3)
5
where a+1 is an approximation to (+1) and a is a parameter which takes on a parallel
meaning to the parameter introduced by Hilber et al. (1977) for linear structural dynamics.
Defining the predictor values for displacements and velocities as
2(1 - 23) = d + LtV + 2
a
Vn+i=Vn+Lt(17)an (5)
where /3 and are integration parameters, then the finite difference (FD) equations for
displacements and velocities take the form
d 1 = + Lt2/9a+1 (6)
n+l = Vn+l + Ltya+i (7)
Note that when a = 0, Eq. (3) degenerates to Newmark's method. The purpose of in-
troducing the parameter a in Eq. (3) is to achieve an optimal balance between effective
numerical dissipation and loss of accuracy compared with the trapezoidal rule (Hughes
1987).
2.2 Solution Algorithm
Eq. (3) can be written in the following residual form:
r(a 1) = FEXT(tn+1+a) - Ma1 - (1 + a)(FINT ) +l + a(FINT ) 0 (8) 41 The vector a+1 can then be determined iteratively for the nonlinear case by Newton's
method. The iteration takes the form
= r(a +1); a+1 .' a +1 + ak k+1k
__ (9)
where k is an iteration counter. For a fully implicit solution, —r'(a +1) is the consistent
tangent operator derived from
-r'(a 1) M + (1 + a)LiyCT + (1 + a)1t2/3KT (10)
-- 6
where CT = ôFINT/ôv' +i = algorithm material damping matrix, and KT =
= algorithmic material stiffness matrix. Iterations commence by allowing d+1 and v+i
to initially take on their respective predictor values.
A fully implicit solution to Eq. (8) requires simultaneous equation solving for the lin-
earized problem at each iteration. This is done by a sequence of computations involving
triangular factorization of the effective mass matrix (which also serves as the tangent
operator) followed by a backsubstitution. In general, triangular - factorization is a very
costly operation for large systems and should be avoided whenever possible. To cut down
on computing costs associated with triangular factorization, Eq. (9) is herein solved itera-
tively using a linear equation solving technique based on preconditioned conjugate gradients
(PCG), in which the preconditioner is the elastic component of the effective mass matrix.
Thus, for the nonlinear case, equation solving is done by the use of so-called composite
Newton-PCG iteration (Borja 1991a). The use of a composite iteration guarantees that
with a constant time step, no additional subiterations will be required for linear analysis
since the preconditioner is identical to the tangent operator itself. Furthermore, since the
tangent operator is factored only once throughout the entire analysis, the savings engen-
dered by the Newton-PCG iterations are significant particularly for nonlinear analysis of
large systems (Borja 1991a).
2.3 Constitutjve Model
The constitutive model used in the present study is the deviatoric elasto-viscoplastic model
of Duvaut and Lions (1976). The inviscid counterpart of this model is the Von-Mises elasto-
plastic model with a linear combination of isotropic and kinematic hardening (Borja 1991b).
This model can replicate the following features considered typical in soils subjected to
dynamic loads: path-dependence, plasticity, hardening/softening, the Bauschinger effect,
and viscous damping. Though this constitutive model may seem simplistic compared to the
complexities of real soil behavior, its well-behaved nature facilitates a clear interpretation
of the significance of each of the model features. Furthermore, this simple model provides a
backdrop against which the performance of more robust constitutive models (e.g., Prevost
7
1977; Mróz et al. 1979; Dafalias and Herrmann 1982) may be assessed for calibration and
response predictions.
For the inviscid case the evolution equations for the proposed constitutive theory are
summarized in Hughes (1984). Specifically, the radius R of the yield surface is assumed to
vary linearly with the effective plastic strain, while the backstress a is assumed to evolve
according to Prager's translation rule (Prager 1956). An accurate and stable integration
algorithm for this model is embodied in the generalized radial return concept of Krieg
and Key (1976), which is also summarized in Hughes (1984). This algorithm produces the
time-discrete evolution equations for the inviscid solutions ã,+1, R, 4, and
For the viscoplastic case the general rate-dependent constitutive model takes the form
= tr —g()( —a); g(ij) = 1
where & = inviscid solution of the rate-constitutive equation, btl' is the usual trial rate-of-stress, and ri = viscosity coefficient. Other forms for g(i) also may be considered, such
as the exponential and rational laws (Simo et al. 1988), but the simpler form given by
Eq. (11) is chosen here. Employing the one-step backward difference scheme on Eq. (11)
yields
tr — o. + ( t/)a1 12 — 1 + ( t/)
where zt/77 step size-to-damping ratio. Note that the elastic solution is recovered from • Eq. (12) as —* 0, while the inviscid solution is recovered as --* x.
We also postulate the following time-discrete evolution equation for radius:
R1 — (13) — 1 + (t/77)
and time-discrete evolution equation for backstress:
= a + ( t/)a +1 1 + (i.t/77) (14)
It then follows that the algorithmic material and damping stiffness matrices specialize to
CT = O (15)
KT = 1 K + 1 + ( 1 + (t/) J BTa+l(Efl+l)Bd (16)
where K = elastic stiffness matrix and consistent algorithmic moduli of
Simo and Taylor (1985). Note that for the present constitutive model, viscous damping
is embedded in the KTterm and not in the CTterm. The use of the above viscoplastic
model allows modeling of material stiffness degradation via a plastic-modulus parameter H', modeling of the Bauschinger effect via a kinematic hardening parameter 0, and modeling
of nonproportional damping via the viscosity coefficient i.
' The significances of the model parameters are shown in Fig. 2.2. Note in Fig. 2.2
that there is no upper bound on the value of the plastic hardening parameter H'; as H' —poo, the inviscid elasto-plastic solution approaches the elastic solution. Also note
that the model does not preclude softening plasticity, i.e., H' < 0; the only restriction is that H' > -3i, where y = elastic shear modulus. The kinematic hardening parameter
6 allows the yield surface to translate in the deviatoric stress space: 0 = 0 implies pure
kinematic translation, 0 = 1 implies pure isotropic hardening/softening, and 0 < 0 < 1
results in a linear combination of isotropic and kinematic hardening features. However, this
constitutive model does not have the capability to replicate the shear-volume relationship
that could be important in predicting the onset of liquefaction in saturated sands.
' Not all nonlinear constitutive models may be used for steady-state forced-vibration
analysis of rigid foundations. It is important to note that, in order to be useful for
time-domain vibration analysis, a constitutive model must first exhibit a steady-state
(closed-loop) hysteretic stress-strain response in the limit as the number of load cycles
approaches infinity. This requirement rules out any elasto-plastic model possessing some
degree of isotropic hardening since these models have an elastic stress-strain limiting re-
sponse. Therefore, even though the model can replicate the isotropic hardening effect,
but this feature is not as meaningful as the kinematic hardening effect in the analysis of
steady-state response, and is thus suppressed altogether. Furthermore, from a numerical
-- 9
analysis standpoint, a constitutive model must also exhibit a finite elastic region—bounding
surface-type models (e.g., Dafalias and Herrmann 1982) and constitutive theories which
use the notion of a vanishing elastic region (e.g., Mroz et al. 1979) are unacceptable—in
order that the elastic infinite-domain solutions may be exploited in the far-field region.
The latter requirement thus guarantees that in the limiting condition the extent of the
yield zone formed in the supporting soil medium is finite. The above requirements are
satisfied by the elasto-viscoplastic constitutive model used in this work.
2.4 Forcing Function and Radiation Boundary Condition
In this report, the nonlinear responses of vibrating rigid foundations to harmonic forcing
functions are investigated. Vibrations in vertical, horizontal, rocking and torsional modes
are considered. Harmonic forcing functions are used so that the nonlinear responses can be
interpreted against the corresponding linear responses where closed-form or semi-analytical
solutions are available for most foundation shapes (Cazetas 1991).
Long-duration analyses require the use of an accurate radiation boundary operator
to allow an effective transmission of the outgoing waves (Lysmer and Kuhlemeyer 1969;
Manolis 1983; Givoli 1988; Pinsky and Abboud 1991). Radiation boundary conditions are a
separate computational issue which can introduce additional complexities and uncertainties
into the solution. Fortunately, with a sufficiently large FE -mesh, steady-state responses
can be established from an initially quiescent condition before the waves reflecting from
the mesh boundaries reach the model foundation. Thus, with a judicious choice of mesh S size, time step, and duration of analysis, the model responses can be made completely
free from the inaccuracies brought about by the use of the various artificial boundary
operators. These computational issues will be elaborated further in Chapter 4. For long-
duration analyses, it may be necessary to introduce artificial boundary operators to make
computations economically feasible.
10
3 Foundations Subjected to Vertical Vibrations
Here we compare the responses of vertically oscillating rigid foundations on a visco-plastic
half-space to the corresponding linear elastic responses. The FE code used in the present
study is an enhanced nonlinear version of DLEARN (Hughes 1987) and runs in 64-bit-
per-word single precision operation on a Cray Y-MP supercomputer at San Diego Super-
computer Center. This code has a dynamic storage allocation feature that automatically
calls the composite Newton-PCG routine when the problem size exceeds 1K degrees of
freedom; otherwise, for smaller problems the solution is carried out by the conventional
triangular factorization technique. Two error tolerances are prescribed in the code for
execution: a global Newton iteration control which determines the overall accuracy of the
solution, TOLl = iO, and a local iteration control which determines the accuracy of
the PCG linear equation solver, TOL2 = 10-2 (Borja 1991a). In all of the cases studied
the elastic constants are assumed to be as follows: Young's modulus E3 = 240,000 kPa, Poisson's ratio v = 1/3 and shear velocity v = 200 rn/sec for the soil; E f = 109E3 for the
"rigid" foundation. Material responses are sampled at the Gauss points via a four-point
integration for two-dimensional quadrilateral elements and an eight-point integration for
three-dimensional brick elements, both considered "standard" for these elements. Finally,
the time-integration parameters of Hilber et al. (1977) are set to /3 = 0.3025, y 0.60, and o = —0.10, and the time step is taken to be /t = T/16 (where T = period of the exciting force).
3.1 Vertical Vibration of Circular Foundations
Consider the well-studied problem of a vertically oscillating circular foundation bonded to
the surface of an elastic half-space (Lysmer 1965; Luco and Westman 1971; Veletsos and
Wei 1971; Veletsos and Verbic 1973). Under a harmonic vertical forcing function 1(t), the response 5(t) of the system is also harmonic and is related to f(t) by the so-called dynamic impedance function
Kd = = K5(ã + iao ) (17)
11
where K3 = static vertical impedance of the foundation, a and /3 are frequency- depende nt
coefficients, i = unit complex number, and
wr a0 = -
V3 (18)
is a dimensionless frequency parameter which accounts for the combined effects of the
excitation frequency w, the foundation radius r, and the shear wave velocity v3. The ratio between the dynamic displacement amplitude i and the static displacement L is called the amplitude ratio, and is given by
LK 3 1 -JIkJ - a2 + aj2 (19)
In practice, the dynamic responses of rigid foundations to harmonic forcing functions are
usually described in terms of the variation of the amplitude ratio /L with the dimen- sionless frequency a0.
The FE mesh investigated is composed of 936 nodes and 875 bilinear axisymmetrjc
elements shown in Fig. 3.1, which gives rise to a total of 1,751 unknown degrees of freedom.
The rigid foundation is represented by five ring elements subjected to a sinusoidal vertical
forcing function f(t) is applied at the center of the foundation.
Fig. 3.2 shows how the mesh of Fig. 3.1 was studied to determine if the spatial dis-
cretization pattern adequately represents the dynamic behavior of the SSI model in the
linear regime. For each value of a0, the foundation was subjected to a harmonic forcing I function f(t) starting from an initially quiescent condition. The steady-state displacement amplitude for each a0 was then normalized with respect to the corresponding displacement
amplitude at zero frequency and plotted in Fig. 3.2a to construct an amplitude-frequency plot for this mesh. The good agreement between the time-domain FE solution and the
frequencydomajn amplitude-frequency curve, determined from Eq. (19) together with the values of a and reported by Veletsos and Tang (1987), suggests that the mesh of Fig. 3.1
is adequate for this study.
12
It is crucial that the steady-state condition be reached in the solution prior to the arrival
of the reflected waves since no artificial boundary operators are used in the present analysis.
In general, the time required for the arrival of the reflected waves can be determined from
the mesh dimensions and the P-wave velocity for the soil material, and should be long
enough to obtain a meaningful physical interpretation of the results. Fig. 3.2b further
affirms the adequacy of the mesh of Fig. 3.1 as the harmonic foundation response lags the
forcing function by a phase distance 0 and continues to manifest after only the first cycle
of loading from an initially quiescent condition. The fact that the steady-state condition
can be established so soon for the half-space problem is hardly surprising considering the
significant influence of radiation damping on the system response, a phenomenon which
will be discussed further in the next section.
Having established the accuracy of the mesh of Fig. 3.1, the half-space is now assumed
to be elasto-viscoplastic in which the yield surface is assumed to translate (kinematic
hardening), but not expand (isotropic hardening), in the stress space. It is important to
note that for harmonically vibrating foundations, isotropic hardening does not influence
the foundation response at steady state; expansion of the yield surface contributes only
to permanent deformation during the transient period of building up the dynamic motion
from an initially quiescent condition. Thus, in all of the nonlinear test problems, a value
of 0 = 0 was assumed.
Results of nonlinear time-domain analyses are shown in Figs. 3.3 and 3.4. In Fig. 3.3, S the soil was assumed to be a strain-hardening material with H' = 0.20E3 and uniaxial
yield stress cr, = 0.001E3. An implication of these values of model parameters is a 50%
increase in static displacement due to plastic deformation.
Amplitude-frequency plots and time-history responses (normalized with respect to the
half-space elastic zero-frequency displacement are portrayed in Figs. 3.3a and 3.3b, re-
spectively, showing the influence of material damping coefficent 77 on the system response.
The following observations can be made from Fig. 3.3: (1) plasticity results in increased
vibrational displacement amplitudes, and hence, in smaller effective dynamic impedance
13
compared to the elastic impedance; (2) resonance frequencies are created at sufficiently low
values of 77 and a0; (3) increasing the material damping coefficient 77 decreases the vibra-
tional displacement amplitudes (however, note that the material damping in the present
model is nonproportional); (4) resonant peaks disappear at sufficiently large values of
i; (5) the elasto-viscoplastic zero-frequency displacements for different values of material
damping coefficient are the same; and (6) the elasto-viscoplastic response approaches the
elastic response as a0 becomes large (i.e., a0 > 2).
The last two observations directly result from the deviatoric elasto-viscoplastic model
proposed in this study. At the zero-frequency level (ao = 0), the increment of time step At
approaches infinity and consequently, Lt/77 - oo for any finite value of i. As indicated
previously, this corresponds to the inviscid solution. On the other hand, as a0 becomes large, At decreases to very small values causing Lt/71 -+ 0 for any finite, nonzero value of
i, implying an elastic solution. Furthermore, it is well known that the amplification factor
(amplitude ratio) is independent of damping at high frequencies. Hence, the solution for
= 0 also approaches the elastic curve as a0 becomes large.
In Fig. 3.4, the soil was assumed to be elasto-plastic (i 0) with the same uniaxial
yield stress as that used in Fig. 3.3. Amplitude-frequency plots then are constructed for
different values of H' using the same FE methodology to show the influence of material
stiffness degradation on the system response. The following additional observations can
be made from Fig. 3.4: (1) as H' approaches zero, the resonant peak increases and, al-ternately, as H' becomes large the resonant peak disappears; and (2) H' <0 results in
an unstable time-history response, i.e., no steady-state solution appears to exist for rigid
foundations oscillating on a strain-softening elasto-plastic soil. The second observation is
understandable considering that strain-softening causes the yield surface to collapse with
harmonic vibration.
A more illustrative picture of the nonlinear SSI phenomenon is shown in Fig. 3.5 where
snapshots of deformed meshes for the circular foundation problem are portrayed. This
figure also shows the zones of plastification at the indicated time instants, where "full
14
plastification" implies that all of the four Gauss points in an element yielded while "par-
tial plastification" implies yielding of at least one, but not all four, Gauss point. Note
that the effect of repeated loading on machine foundations resting on an elasto-viscoplastic
half-space is to create a zone of intense yielding and inelastic deformation beneath the
foundation, which in turn results in "dynamic nonhomogeneity" with respect to soil elas-
ticities. The yield zone thus created is a finite zone where the soil moduli have degraded to
their elasto-plastic or elasto-viscoplastic values; however, it is not a fixed region in space.
Instead, at steady-state the yield zone oscillates in such a way that it defines the same
region in space it occupied one period of time earlier. The effect of radiation damping, in
which outgoing waves quickly die out, is also portrayed in Fig. 3.5 to illustrate the ease
with which the steady-state waves are established in a half-space.
3.2 Vertical Vibration of Square Foundations
The three-dimensional FE mesh for this problem is shown in Fig. 3.6. The mesh is com-
posed of 3,856 nodes and 3,159 trilinear brick elements including the foundation elements,
giving rise to a total of 9,896 unknown degrees of freedom. In this problem, the forcing
function f(t) and temporal discretization are the same as those used in Example 1.
A mesh convergence study similar to that discussed in Example 1 was performed to
assess the adequacy of the spatial and temporal discretizations. Results of this study are
summarized in Fig. 3.7, which shows that the time-domain FE elastic solution compares
10
very well with the frequency-domain amplitude-frequency curve for a square foundation on
an elastic half-space presented by Wu (1991). It must be noted that for square foundations,
the expression for a0 takes the form
a0 = vs (20)
where b = half-width of the foundation.
Fig. 3.7 also shows the nonlinear responses derived from running three-dimensional FE
analyses where the half-space is represented by an elasto-plastic (ij = 0), strain-hardening
15
(H' = 0.20E3) material. In this study, the free variable is the uniaxial yield stress
which controls the size of the yield zone. The following observations can be made from
Fig. 3.7: (1) as o, decreases the dynamic displacement amplitude increases; (2) for the
range of values of o, shown in Fig. 3.7, resonance is created in all three nonlinear cases;
(3) as a, decreases, the resonant peak increases; and (4) the displacement amplitudes for
the elasto-plastic case approach those for the elastic case as a0 becomes large (i.e., a0 > 2).
Finally, the deformed meshes and yield zones at time instants I = 1.5T (corresponding to "deepest push") and t = 2.OT (corresponding to "highest pull") are shown in Figs. 3.8
and 3.9, respectively. As seen previously, he effect of radiation damping for the three-
dimensional case is to enhance the formation of steady-state waves over a short period of
time.
3.3 Discussion of Results
The response of a foundation on a soil stratum containing a local yield zone is similar
to the response of a foundation on a nonhomogeneous half-space. It is known that the
response of a foundation on a nonhomogeneous half-space can be substantially different
from the response of an identical foundation resting on a uniform half-space (Kausel 1974;
Kausel and Ushijima 1979). For example, the presence of a bedrock at a relatively shallow
depth causes the amplitude-frequency curves for vertically oscillating circular foundations
to exhibit resonant peaks (Gazetas 1983). Since the effect of local yielding is to create a
finite zone that is generally softer than the surrounding medium, resonant peaks can also S be expected to occur under this condition.
The creation of resonant peaks can be explained better with the aid of a simple
single-degree-of-freedom (SDOF) oscillator analog. The idea here is to replace the proto-
type foundation-soil system with an SDOF mass-spring-dashpot oscillator with frequency-
dependent stiffness and damping coefficients (Hsieh 1962; Lysmer 1965). Under a harmonic
exciting force the dynamic impedance function for an oscillator with a natural frequency
w is given by Eq. (19), with a = 1 and 9 = cv3/r, where c = dashpot coefficient.
16
The dashpot serves to represent the effect of radiation damping which arises when stress
waves propagate outward from the contact surface as the foundation moves against the
soil (Reissner 1936). Thus, the amplitude ratio, i = / (a2 + a2)_h/2, can be plotted
against a0 for each of the equivalent oscillators.
Fig. 3.10 shows the amplitude-frequency plots for vertically oscillating circular foun-
dations on (1) a homogeneous half-space, (2) a stratum of thickness H = 2r underlain
by a firm bedrock, and (3) a cylindrical medium of radius R = 2r and thickness H = 2r
surrounded by a firm bedrock (a physical meaning for the third case is that of a soil in a
valley). The rationale for presenting these curves is that as the ratio between the soil-to-
rock moduli approaches unity, the problem approaches that of a uniform half-space. On
the other hand, as the moduli ratio approaches zero, the problem degenerates to any of the
remaining two specific cases mentioned above. For values of the moduli ratio between zero
and unity, the problem reduces to a composite soil system in which the foundation soil is
either underlain by or is embedded in a stiffer medium, the latter resembling the case of a
half-space containing a local yield zone.
For the half-space problem, radiation damping is so significant that resonance is not
possible. The amplitude-frequency plot for the stratum-over-bedrock problem was con-
structed from known variations of the oscillator parameters a and 3 with a0, as developed
by Kausel (1974), Kausel and Ushijima (1979), and Tassoulas (1981). Here, the dimen-
sionless parameter 9 for the equivalent oscillator is an irregular, nonnegative function of
a0, which has a value nearly equal to zero when a0 < 1. Since the damping coefficient
fi is small for low-frequency excitations, the foundation response exhibits resonant peaks.
Finally, the foundation response for the "valley" problem was constructed directly by the
authors by running extended time-domain analyses to achieve steady-state conditions. The
response curve portrayed in Fig. 3.10 shows more pronounced resonant peaks and multiple
resonance frequencies. For convenience in presentation, the curves plotted in Fig. 3.10 all
have been normalized with respect to their corresponding elasto-static displacements so
that they all converge to A*/A = 1 at a0 = 0.
17
It is evident from the examples presented in this section that local yielding in an other-
wise homogeneous elastic half-space tends to reduce the effective dynamic impedance of the
half-space and create resonance frequencies where the foundation motions are amplified to
values well above those at zero-frequency levels. There was no attempt on the part of the
writers to develop nonlinear "design curves" similar to those available for linear systems
because the system response is generally a function of the specific constitutive model cho-
sen to describe the soil behavior. Furthermore, the response curves are nonlinear functions
of the constitutive model parameters. However, with a well-validated and calibrated con-
stitutive model, a FE methodology for generating nonlinear response curves was presented
herein so that more complex soil behavioral features can be incorporated into the solution
for routine engineering applications.
3.4 Vertical Vibration of Rectangular and Strip Foundations
In this section the responses of vertically oscillating rectangular and strip foundations on
elasto-plastic half-space are compared to the corresponding linear elastic responses. The
FE mesh investigated, shown in Fig. 3.11, is composed of about 4K nodes and over 3K
trilinear brick elements, resulting in a total of about 10K unknown degrees of freedom.
Note that this mesh represents one-fourth of the total 3-D mesh, and can thus be used for
analysis of vibrational response in the vertical direction only.
The foundation dimensions are represented by the fixed half-width b and the varying half-length 1, with i/b > 1.0 representing the various foundation aspect ratios. In the limit I as i/b -* oc which corresponds to a strip foundation, a plane strain condition results, and
the mesh of Fig. 3.12 will degenerate to an equivalent two-dimensional mesh. No artificial
absorbing boundary was used in the numerical model.
Harmonic vibrations were established by applying a sinusoidal vertical excitation force,
f (t), in the time domain. For convenience, and for consistency with results for founda-
tions having different aspect ratios, the harmonic force was computed for each rectangular
foundation in such a way that the mean pressure, p = f(t)/bl, is unchanged (in the plane
IV
strain limit, 1(t) was computed on the basis of a unit i). More specifically, the ampli-
tude of f(t) is proportional to the foundation area. It is important to note that since no
artificial absorbing boundaries were used in the analyses, the only meaningful results are
the computed foundation responses prior to the arrival of the reflected waves. The longest
possible duration of analysis can be estimated as the total time it takes for a travelling
P-wave to radiate from the short edge of the foundation (in the i-direction), reflect on the
boundary, and arrive back to the same point as a reflected wave, see Fig. 3.11.
Fig. 3.12 shows the results of linear elastic FE analyses in the form of amplitude-
frequency curves. For ease in presentation, the steady-state amplitudes were normalized
with respect to the zero-frequency elastic amplitude corresponding to a square foundation.
The excitation frequencies were nondimensionalized through the use of the parameter a0 =
wb/v8, where w is the excitation frequency and v is the shear wave velocity.
The following observations can be made from Fig. 3.12: (1) the higher the aspect ratio,
the larger the steady-state amplitude; (2) as i/b -p oo, the zero-frequency amplitude ap-
proaches infinity; (3) at sufficiently low frequencies (e.g., a0 < 1.0) steady-state amplitudes
are a strong function of the foundation aspect ratio; (4) at the zero-frequency level, the
static amplitudes are nearly proportional to the square root of the foundation areas; and
(5) at sufficiently high frequencies (e.g., a'0 > 2.0), the amplitudes are nearly indepen-
dent of the foundation aspect ratio. In general, these behaviors are consistent with those
presented in Wu (1991).
Some explanations of the foregoing observations are given in the following. It has
been proven that the vertical static stiffness of a typical rectangular foundation can be
approximated with good accuracy by the corresponding value of a circular foundation
having the same area (Gazetas 1983). Since the static stiffness of a circular foundation
has an analytical expression and is proportional to its radius (or the square root of its
area), observations (3) and (4) mentioned above are expected. Furthermore, since a strip
foundation does not have a characteristic length scale, the static foundation impedance
degenerates to zero in the limit as i/b -i 00, i.e., the amplitude becomes unbounded. On
19
the other hand, at high-frequency excitations radiation damping dominates the system
response. Since the radiation energy flux travels predominantly in the vertical direction as
the excitation frequency increases (Gazetas 1987), for the same force per unit foundation
area the steady-state amplitude remains nearly constant regardless of the foundation aspect
ratio. Conclusively, there is an important point to make from the above observatons. As
a0 -* 0, the magnitudes of vertical impedance functions for rectangular foundations are
approximately proportional to the square root of their areas, whereas the corresponding
magnitudes are proportional to their areas at high frequencies.
Fig. 3.13 illustrates results from the nonlinear elasto-plastic analyses as well as ob-
servations noted from the elastic case. Additional observations from Fig. 3.13 can be
summarized as follows: (1) local soil yielding does amplify the motion of the foundation;
(2) resonance is created when the aspect ratio becomes reasonably close to unity; and (3)
at sufficiently high frequencies (e.g. a0 > 1.0), the steady-state amplitudes are nearly the
same regardless of the foundation aspect ratio. This last point also is true for the elastic
case, but at higher values of a0.
Of particular interest in this example is the creation of resonance frequencies when the
foundation aspect ratio becomes reasonably close to unity, i.e., when foundation becomes
nearly square. Recall that the effect of local soil yielding is to create a finite yield zone
below the foundation where the soil moduli have degraded to their elasto-plastic values.
Furthermore, the effect of having a soft zone embedded in a stiff medium is to reduce the
effective radiation damping of the system. Consequently, resonance is created for square
and nearly square foundations. Alternately, it was noted previously that the magnitude of
the impendance function approaches zero in the static limit as the aspect ratio approaches
infinity; consequently, the radiation damping is large enough to suppress the resonant
peak in the limit as i/b - oo. As illustrated in Fig. 3.13, the radiation damping per unit
foundation area increases with the increasing aspect ratio such that the resonant peak does
not occur in slender foundations (e.g., i/b = 3.2). Observation (3) mentioned above can
be explained by the degraded soil moduli which cause the "high frequency" level (relative
- 20
to the shear velocity of the soil) to be reached under lower excitation frequencies.
Figs. 3.14 through 3.17 illustrate the extent of oscillating yield zones for the rectangular
(b/i = 2) and strip foundation problems under an excitation frequency of a0 = 2.0 (consid-
ered high-frequency excitation, see Fig. 3.13). The deformations for the strip foundation are
shown in perspective view so that they may be compared directly with those for rectangular
foundation, but the former problem was actually analyzed under a two-dimensional plane
strain condition. Although the yield zones for the two foundations clearly differ in extent
and geometry, note that they produced nearly the same steady-state amplitudes, a stark
contrast to low frequency behavior where amplitudes increse noticeably with increasing
aspect ratio. This points to the previously mentioned fact that at high frequency excita-
tions, the extent and geometry of yield zones balance in such a way that the magnitude of
impedance function per unit foundation area is the same regardless of the foundation as-
pect ratio. Furthermore, note that surface waves dissipate rapidly with increasing distance
from the foundation. This implies that "steady-state" waves can be established reasonably
rapidly in time before the reflected waves reach the foundation. In general, the required
time to reach a "steady-state" condition is longer for elasto-plastic problems than it is for
the elastic half-space problem since the presence of a local yield zone embedded in a stiffer
medium does tend to lessen the radiation damping effects. Time-domain models designed
to handle material nonhomogeneity generally require artificial absorbing boundaries so that
extended time-stepping computations may be made feasible.
S 3.5 Summary and Conclusions
The dynamic response of vertically-excited rigid foundations on an elasto-viscoplastic half-
space has been investigated in the context of nonlinear FE analysis. The effect of repeated
loading on machine foundations is to create a zone of intense yielding and inelastic defor-
mation beneath the foundation, resulting in "dynamic nonhomogeneity" with respect to
soil elasticities. In general, radiation damping is smaller for a half-space problem contain-
ing a local yield zone than it is for the same half-space problem without a yield zone. Local
yielding thus creates resonant peaks on the amplitude-frequency curves which increase as
21
the ratio between the elastic to elasto-plastic (or elasto-viscoplastic) moduli increases. This
phenomenon is consistent with known results for the problem of finite-size foundations on a
uniform layer over bedrock and for the problem of finite-size foundations on a finite region
embedded in a half-space.
Furthermore, local yielding influences the foundation response at low frequency excita-
tions, but its influence in high frequency regime is generally insignificant. Another factor
which is of special interest in this report is the effect of the foundation aspect ratio on the
amplitude-frequency response of vertically vibrating rectangular foundations. Although
low frequency amplitudes depend significantly on the foundation length 1 for a given fixed foundation width b, it has been shown that high-frequency amplitudes do not.
A notable contribution of this report is the advance in nonlinear soil-structure inter-
action analysis. A nonlinear model such as the one presented in this report provides a
means for understanding the influence of other factors that could dominate the SSI re-
sponse, such as the hysteretic soil behavior and soil stiffness degradation. When validated
and calibrated, nonlinear models allow the extrapolation of numerical results well beyond
the limits or capabilities of those that employ traditional linear models. With the advent
of supercomputers and with more efficient numerical tools, it is shown that large-scale
nonlinear SSI analyses can be carried out in a computationally feasible manner despite the
costly time-stepping and iterations. Specifically, three-dimensional nonlinear SSI analyses
of the order of 10K degrees of freedom and requiring a total of about 100 iterations can
be run on the Cray supercomputer in about two hours of CPU time. With efficient par-
allel computation techniques coming of age, the feasibility of performing these complex,
computationally intensive nonlinear analyses will be enhanced.
22
4 Foundations Subjected to Horizontal, Rocking, and Torsional Vibrations
It is shown in Chapter 3 that the effect of local yielding on the dynamic response of
vertically excited circular and rectangular foundations (including strip foundations) does
induce the creation of resonant peaks on the amplitude-frequency curves. Furthermore,
parametric studies indicate that as the excitation frequency increases, the nonlinear effect
decreases (i.e., the elastic behavior dominates the foundation response in the high-frequency
range). The next phase of this project involves extending this investigation to include
horizontal, rocking, and torsional modes of vibration.
While the results presented for vertically oscillating foundations are helpful in demon-
strating the significance of plastic deformation on the dynamic response of vibrating foun-
dations, they are not directly applicable to problems involving seismically induced excita-
tions where typically the foundation motion is dominated by lateral and rocking modes.
In this chapter we employ the same soil constitutive model and methodology for nonlin-
ear analysis of vertically oscillating foundations to study the nonlinear responses of finite
size foundations in lateral, rocking, and torsional modes. The methodology is generalized
further by considering both two- and three-dimensional problems with implicit integration
in the time domain. Whenever possible, symmetry and anti-symmetry are exploited in
the analyses so that the size of the problem is reduced. For example, symmetry cuts the
size of the problem by one-half as does anti-symmetry; thus, a three-dimensional analysis
of rectangular foundations for either lateral, rocking or torsional vibration modes can be
performed successfully with the use of a quarter-size mesh.
The enhanced nonlinear version of DLEARN is again used and run on a CRAY Y-
MP 8/864 supercomputer to construct time-history responses and steady-state amplitude-
frequency curves for various modes of oscillation. The amplitude-frequency curves are used
to assess the impact of soil nonlinearities on the foundation behavior as well as determine
the range of excitation frequencies over which nonlinear effects may be dominant. Two
23
error tolerances, TOLl = iO and TOL2 = 10-2, were used in the code for control of global
Newton and local PCG iterations, respectively (Borja 1991). The elastic parameters for the
foundation-soil model are: Young's modulus E3 = 213 MPa, Poisson's ratio ii = 1/3, and elastic shear wave velocity v3 = 200 rn/sec for the soil; and Young's modulus E f = 109E3 for the "rigid" foundation. Material responses were sampled at the Gauss points using
a full four-point Gaussian integration for 2D analysis and an 8-point integration for 3D
analysis. Time-stepping was done via the a-method of Hilber et al. (1977) using time-
integration parameters of /9 = 0.3025, -y = 0.60, a = —0.10, and a time step of At = T/16, where T = period of excitation.
4.1 Symmetry/Anti...Symmetry and Forcing Functions
Except for the axisymmetric case of vertically vibrating circular foundations, all problems
associated with oscillating finite-size foundations are three-dimensional which induces im-
mense increases in the bandwidths of the storage matrices as compared to two-dimensional
analysis. Therefore, whenever possible, nonlinear time-domain analyses of the 3D foun-
dation vibration problems requires utilization of symmetry and anti-symmetry conditions
where each reduces the size of the problem by one-half to make the computation feasible.
Fig. 4.1 illustrates how these conditions can be exploited to reduce the overall size of the
matrix problem.
Consider a rigid square foundation resting on a half-space. Under vertical excitation this
foundation will oscillate in such a way as to form two vertical planes of symmetry passing S through the foundation axes. Vertical planes of symmetry can be prescribed numerically
using horizontal roller supports in the manner shown in Fig. 4.1. Thus, this problem can
be analyzed by modeling only one-quarter of the half-space, as we did in Chapter 3.
Consider next a square foundation subjected to torsional excitation. In this case, only
horizontal shear waves propagate; therefore, the two vertical planes cutting the foundation
axes represent planes of anti-symmetry. Planes of anti-symmetry can be prescribed numer-
ically by releasing the constraints normal to these planes in a manner shown in Fig. 4.1.
24
Thus, the solution also requires only one-quarter of the half-space to be modeled.
Finally, consider a square foundation subjected to either lateral or rocking excitation.
In this case, one vertical plane passing through a foundation axis represents a plane of
symmetry while the other represents a plane of anti-symmetry. Thus, the size of the
matrix problem can be reduced analogously. The appropriate boundary conditions for FE
vibration analysis of rectangular foundations are shown schematically in Fig. 4.1.
Forcing functions for forced vibration analysis are sinusoidal nodal forces acting on the
foundation. These forcing functions used in this study are shown in Fig. 4.1 for various
modes of vibration. Note that the amplitudes of these forces (or moments) are only 25%
of their total values since only one-fourth of the total half-space is modeled.
Vertical and lateral forcing functions are concentrated forces applied on the foundation
in a manner shown in Fig. 4.1. For a rocking foundation the forcing function is a couple
represented by a pair of eccentric and oppositely directed vertical nodal forces. However,
due to the fact that the center of the foundation lies on a plane of anti-symmetry where
vertical motion is prohibited, only one of the two forces need be prescribed. Finally,
torsional forcing functions are represented by a pair of orthogonal oscillating forces acting
in phase on the foundation sides. Because of the small strain formulation adopted in the
numerical model and since the finite elements have no rotational degrees of freedom, it
is not possible to simulate torsional excitations with a single force applied only on one
foundation side. For output interpretation purposes, the "responses" of the foundation
to lateral, rocking, and torsional excitations are the horizontal displacement L, angular rotation e, and angular twist , respectively.
4.2 Horizontal and Rocking Vibrations of Strip Foundations
As discussed earlier, nonlinear analysis of 3D meshes are prohibitively expensive to perform
unless special techniques are utilized to reduce the model size. An exception is the strip
foundation analysis which can be reduced to plane strain conditions with uniform loading
along the longitudinal direction. Hence, a two-dimensional FE mesh is sufficient in this
- 25
case, and the choice of strip foundations makes it feasible to investigate the horizontal and
rocking vibrations of foundations by the FE methodology without intensive computation.
The two-dimensional FE mesh used here is similar to the one shown in Fig. 3.1 where
the vertical vibration is considered. To make this mesh also suitable for investigating
horizontal and rocking vibrations, utilization of anti-symmetry conditions mentioned earlier
is required. Harmonic vibrations are established by applying a sinusoidal horizontal force
or rocking moment along the longitudinal direction of the foundation.
Classical theory of elasticity predicts that the static vertical and horizontal impedances
of strip foundations are. zero, which implies that the zero-frequency foundation displace-
ments are infinite. To generalize the presentation of steady-state amplitude-frequency I response curves for horizontally vibrating strip foundations, the horizontal displacement
amplitudes L are herein normalized with respect to their elastic value, at a nondi- mensionalized frequency of a0 = wblv, = 1, where w = excitation frequency and b =
foundation half-width. No such singularity arises for rocking strip foundations, however,
since their rocking impedances are finite even at the zero-frequency level. Thus, for rocking
foundations the rotational amplitudes E can simply be normalized with respect to their
elastic value at the zero-frequency level, W. The normalized nonlinear response curves are
then plotted versus a0 in Figs. 4.2 and 4.3 for strip foundations vibrating in the lateral and rocking modes.
Fig. 4.2 shows the influence of plastic modulus H' and viscosity coefficient 77 on the I steady-state amplitude-frequency responses of horizontally vibrating strip foundations. For
the first case, an elasto-plastic soil (77 = 0) with an initial uniaxial yield stress of o, = 0.0013E3 was assumed, and H' was allowed to take on different values. For the second case, H' was set to 0.40E3, o, to 0.0013E3, and i, was allowed to vary. It is clear from
the response curves of Fig. 4.2 that (a) local soil yielding results in increased vibrational
amplitudes; (b) increasing H' and/or 77 decreases the vibrational amplitudes; (c) minor
resonant peaks are created at sufficiently low values of H' and i; and (d) nonlinear soil effects are dominant over a wide range of excitation frequencies (ao = 0 .- 5). The first two
RIi
observations are similar to those observed in the vertical vibration of rigid foundations and
can be explained by the stiffness degradation of the soil. The third observation is a peculiar
feature and may have resulted from the participation of higher natural frequencies of the
system. The last observation differs remarkably from the results in Chapter 3, which show
that for vertically oscillating strip foundations the nonlinear amplitude-frequency response
curves approach the elastic solution at a much lower excitation frequency (e.g., a0 < 2).
Thus, this example shows that while for a vertically vibrating foundation nonlinear soil
effects may be relatively unimportant in the high-frequency range, they could be significant
over a wide range of excitation frequencies for a horizontally vibrating foundation.
A similar parametric study was conducted for rocking strip foundations and the results
are shown in Fig. 4.3. Note that for rocking strip foundations nonlinear soil effects are
dominant only in the low-frequency range (e.g., a0 <3). However, Fig. 4.3 shows that at
or near the resonance frequencies the angular displacements can be amplified dramatically
by nonlinear soil effects. One can explain the similarity in responses between a vertically
vibrating foundation and a rocking foundation from the fact that both systems generate
predominantly P-waves. On the other hand, a horizontally vibrating foundation generates
mostly S-waves, which radiate at a slower speed than P-waves. Thus, the effects from
radiational damping are reduced, resulting in the nonlinear model for horizontal modes
converging to the elastic model at larger frequency values than those observed for vertical
and rocking modes.
Finally, Figs. 4.4 and 4.5 show snapshots of the deformed meshes and yield zones at
a0 = 2.0 for strip foundations vibrating in the lateral and rocking modes, respectively.
For presentation purposes, a full mesh was employed to show the implications of the anti-
symmetric feature exploited in the analyses. Fig. 4.4 shows the predominantly S-waves
generated by the horizontally vibrating strip foundation, while Fig. 4.5 suggests that a
rocking strip foundation generates mostly P-waves. The similarity of the yield zones even
at the early pairs of time instants (e.g., t = 1.75T and t = 2.25T) suggests that "steady-
state conditions" can be achieved over a reasonably short period of time, thus affirming
- 27
the potential of a time-domain solution methodology for nonlinear FE analysis of vibrating
foundations.
4.3 "Critical Frequency" and "Optimal" FE Grid
In dynamic analysis the accuracy of a FE grid is a function of the excitation frequency.
Because of spatial discretization error, accurate representation of higher-order vibration
modes requires increasingly finer meshes. However, the cost of a three-dimensional FE
analysis grows at an enormous rate as the resolution of the mesh is increased.
Fig. 4.6 shows two FE grids consisting of eight-node trilinear brick elements, which
are considered "optimal" for three-dimensional forced vibration analysis in the low- and
high-frequency ranges. These grids have the same total number of degrees of freedom
(approximately 10K) but have different mesh dimensions. The coarse mesh (Fig. 4.6(a))
is optimal for low-frequency excitations where high-frequency modes are not significant.
The long dimensions of this mesh allow a sufficient number of long-period load cycles to
be applied at the foundation prior to the arrival of the reflected waves. Alternately, the
mesh of Fig. 4.6(b) is appropriate for high-frequency excitations where the capturing of
higher-order modes requires finer resolution. Although the dimensions of this mesh were
reduced for reasons of economy, this representation still provides an optimal modeling since
it allows a sufficient number of short-period load cycles to be applied at the foundation
prior to the arrival of the reflected waves. Since the total number of degrees of freedom is
the same for both grids, the cost of running a full three-dimensional FE dynamic analysis
is also approximately the same for either mesh.
Fig. 4.6 also shows two FE half-meshes (Figs. 4.6(c) and 4.6(d)) for two-dimensional
analysis of forced vibration of strip foundations in both lateral and rocking modes (note
that the centerline for each mesh is a plane of anti-symmetry). Since the increase in
computing costs with mesh size is not as dramatic for 2D as it is for 3D analyses, the high-
resolution mesh of Fig. 4.6(d) may be used both for low- and high-frequency excitations,
as we did in the previous section on investigating the horizontal and rocking vibrations of
NEZ
strip foundations. The low-resolution mesh of Fig. 4.6(c) was used in this study only to
verify the adequacy of the fine mesh.
In interpreting the FE results, one should be careful in evaluating the adequacy of a
FE mesh since its accuracy is generally a function of the mode order of the system and
the excitation frequency. Mesh convergence studies are thus an important part of the nu-
merical analysis since the predicted dynamic responses could vary dramatically depending
upon the quality of the FE mesh. There is no a priori rule for cho6sing an "optimal" mesh
particularly for nonlinear applications, but in general, for low-frequency excitations, the
coarser the mesh, the stiffer the predicted response. For a mesh dominated by nonhomo-
geneous material response arising from local plastification, an opposite trend is possible
depending upon the magnitude of the excitation frequency.
Figs. 4.7-4.10 show the predicted time-history responses of strip foundations vibrating
in the horizontal and rocking modes using the two plane-strain FE grids of Fig. 4.6(c) and
an extended version of Fig. 4.6(d). For a horizontally vibrating foundation, Fig. 4.7 shows
that the resolution of the coarse mesh in the elastic regime is adequate for low-frequency
excitation (ao = 1.0) but not for high-frequency excitation (ao = 5.0). For this mode of
vibration, the "critical frequency" below which the coarse mesh is acceptable but above
which it is unacceptable is approximately a0 = 3.0. On the other hand, for a rocking
foundation Figs. 4.9 and 4.10 show that the coarse mesh is adequate for both low- and
high-frequency excitations (ao = 1.0 and a0 = 5.0, respectively). The adequacy of the fine
mesh was established from a separate mesh convergence study employing a yet finer mesh.
In general, the accuracy of the mesh degrades with increasing values of the ratio h/k, where h = element dimension and ,\ = wavelength, which equals v/w, where v = prop-agating wave velocity and w = excitation frequency. For an elastic soil with a Poisson's ratio of v = 1/3, the P-wave velocity v, is two times greater than the S-wave velocity v
(or, equivalently, the wavelength ratio in a homogeneous soil is ),/) = 2 under the same
excitation frequency), which explains why the coarse mesh may not be sufficient at high
frequencies for the S-wave dominated horizontal motion even though it is adequate for the
- 29
P-wave dominated rocking motion.
Much can be learned from the time-history responses of Figs. 4.7 and 4.8 for a hor-
izontally vibrating strip foundation. For an excitation frequency of a0 = 5.0, Fig. 4.7
indicates that, for the coarse mesh, reflected waves arrive at the foundation much sooner
than expected (the theoretical arrival time of reflected S-waves is t = 9.9T). The early
arrival time might be due to the fact that while the elements near the foundation are
fine enough to effectively transmit outgoing S-waves, the exterior elements could be too
coarse to transmit the same waves on to the reflecting mesh boundaries. Thus, the coarse
elements might have acted as wave reflectors, reducing the effectiveness of the mesh.
Contrasting observations are indicated in Fig. 4.8 which shows that nonlinear soil effects
tend to 'improve' the performance of the non-uniform coarse mesh of Fig. 4.6(c). With local
yielding the velocity of propagation inside the plastic zone decreases to v1 < v3; however, it is possible that h/v is still below the critical value inside the yield zone allowing the waves
to continue propagating effectively, beneath the foundation. As the shear waves reach
the iinyielded zone, they regain their initial elastic propagation speeds. Thus, it takes
longer for the reflecting waves in the nonlinear case to reach the foundation, resulting in
the performance of the coarse mesh being improved by the introduction of the nonlinear effects.
Since local plastification near the foundation could retard the propagation of body
waves, we define an "optimal" FE grid as one where the elements inside the probable yield I zone are small enough to effectively transmit waves traveling at reduced speeds, and with
exterior elements that are effective enough to transmit elastic waves. An "optimal" mesh
thus provides a balance between solution accuracy and computational cost.
4.4 Horizontal, Rocking, and Torsional VibraCions of Square Foundations
Classical theory of elasticity predicts that the static displacements of finite-size founda-
tions under vertical, lateral, rocking, and torsional loads are finite, and can thus be used
30
to normalize the corresponding amplitude-frequency response curves. For the horizontal,
rocking, and torsional modes, the normalized responses shown in this paper are the horizon-
tal displacements */e, angular rotation e*/ee, and angular twist */e, respectively.
Figs. 4.11-4.13 show the predicted amplitude-frequency plots for a vibrating square
foundation subjected to lateral, rocking, and torsional excitations. Each plot demonstrates
the respective influence of the plastic hardening modulus H', viscosity coefficient i, and
initial yield stress o, on the foundation's dynamic response. All the response curves of the
elastic cases are compared with the frequency domain solutions in Wu (1991) to show the
accuracy of the time domain analyses. It is clear from the response curves of Figs. 4.11-4.13
that (a) local soil yielding increases the predicted vibrational amplitudes of the foundation
motion particularly at decreasing values of H', 77, or o; (b) primary resonance peaks
are created at sufficiently low values of a0 in all modes; (c) compared to the rocking
mode, nonlinear soil effects are more dominant over a wider range of excitation frequencies
(ao = 0 5) for the lateral and torsional modes. The latter observation is consistent with
those observed for a vibrating strip foundation and can be explained from the type of body
waves that each mode of vibration generates.
In generating the above response curves, the results from the mesh convergence studies
of the previous section were utilized to determine which FE grid is appropriate for each
analysis. Thus, for the lateral and torsional vibration modes which produce mostly S-
waves, the coarse mesh of Fig. 4.6(a) was used for low-frequency excitations (ao < 3.0), and the fine mesh of Fig. 4.6(b) for high-frequency excitations (ao > 3.0). On the other
hand, for the rocking mode which generates mostly P-waves, the coarse mesh of Fig. 4.6(a)
was used for all values of excitation frequencies. Results of an additional mesh convergence
study for the S-wave dominated torsional vibration are shown in Fig. 4.14. Nonlinear soil
and high-frequency excitations are chosen to show the inadequate performance of the coarse
mesh in this case. This phenomenon is similar to that of the S-wave dominated horizontal
vibration of strip foundations and reveals that the methodology used for the 2D strip
foundation problem also applies to the 3D square foundation problem.
31
Finally, Figs. 4.15-4.17 show snapshots of the deformed meshes and yield zones at
a0 = 2.0 for square foundations vibrating in the lateral, rocking, and torsional modes.
A low-frequency excitation was chosen to plot the relevant yield zones since it is at low
frequencies where nonlinear soil effects tend to be more dominant. Again, the similarity of
the yield zones at the early pairs of time instants suggests that "steady-state" conditions
had been achieved over a short period of time, and affirms the potential of a time-domain
solution methodology for nonlinear FE vibration analysis.
4.5 Summary and Conclusions
The dynamic response of harmonically excited rigid foundations has been investigated
in the context of two- and three-dimensional nonlinear FE analysis. Lateral, rocking, and
torsional vibration modes were considered. Localized soil yielding beneath the foundation is
shown to result in increased vibrational amplitudes of the system response and the creation
of resonance particularly at low-frequency excitations. At high-frequency excitations (ao '
5.0), amplification of the harmonic response due to nonlinear soil effects is also possible
particularly when the motion is dominated by shearing modes. The latter result is contrary
to common understanding that nonlinear soil effects are generally unimportant in the
high-frequency range. Results presented in this paper apply only to soil conditions where
the deviatoric elasto-viscoplastic constitutive model is appropriate (e.g., soft soils and
undrained conditions).
Some computational issues which impact the quality of the numerical results have been I addressed. Mesh convergence studies are extremely important in assessing the adequacy of
a FE grid. The time-domain solution methodology is shown to be effective in the analysis
of nonlinear machine foundation response in all possible modes of vibration. Furthermore,
with a supercomputer and an enhanced FE code, it is shown that implicit time-stepping
schemes can be performed in a feasible manner even for large-scale nonlinear 3D systems.
Studies are now underway to investigate the potential of the methodology for nonlinear
analysis of seismically excited structures and for developing nonlinear transfer functions
that are critical in nonlinear soil-structure interaction analysis.
32
5 Boundary Element Model for Plane Strain
The objective of this chapter is to describe an on-going research at Stanford University on
coupled finite element (FE)-boundary element (BE) modeling for nonlinear soil-structure
interaction analysis. New implementational ideas that pertain to the BE formulation for
plane strain are also described. We are now testing segments of the BE code for eventual
integration into the nonlinear FE code. Note that all analyses are done in the time domain.
5.1 Time-Domain BE Formulation
We consider an exterior domain çE bounded by a regular surface 49ç1E in the sense of
Kellogg (1929). Let u and v be two single-valued, twice continuously differentiable vector
fields. We can look at u as the real displacement field and v as the virtual field. The
boundary integral equation for linear elastodynamics with quiescent initial condition and
zero gravity forces then reads (Kobayashi 1987)
c.u()=j(u*v_u*v)dS (21)
where c is a 'jump' tensor, V = n cc : V is a gradient mapping operator which maps a
given displacement field onto a traction field n is the unit normal to 811E, is
the source point, Ce is the elasticity tensor, V is the gradient operator, and the symbol *
denotes a time convolution.
An important element of a boundary integral formulation is the fundamental solutibn
Here, we employ Green's function for a point pulse t at source point for our fundamental
solution to determine the response of point x in an infinite domain. The fundamental solu-
tion for the response v(x, ; 8; t) has the explicit form (Eringen and Suhubi 1975; Manolis
and Beskos 1988)
1 1r1r3 ,' r v2(x,;63;t) = - - -) + -( - r
4 [ c1 c2 r r3 -
3rr 6 c' ' '
IV
8(t - r) d] (22) r
33
while for Vv(x, ; ; t) the fundamental solution is
=
--- { [( - )7jfl - c (i2 '1rirmnm - 2rjn1 + r1n + Sii'mflm)]ö(t
- -f-) 4ir r3 c 12 r5 r3
+(l2 i'm flm - 3rn3 + 6iirmnm) -
_64(5r - - r,rz1 + ran, + öljrmnm - fc;~
ir3 b(t
+1 2 i'i'ji'mm 1 -
2 rjrjrmnm 1 rn, + Siirmnm)( - (23) C2 r4 c2 r2 c2
I where r2 = - x2 J, r = - xli, c1 is the P-wave velocity, and c2 is the S-wave velocity.
Substituting Eqs. 22 and 23 in Eq. 21 results in the exact boundary integral equation
cu() 4ir p (24)
where
1, = I [ -i'd-
u( r 1 - ri x,t - -) + t - - -) c r3 Cl c2 r r3 c2
3rr - r3 r IC u(t - r) d] dS (25) 1
and
[I
JE { [(24_l)rL nj
C (i'n - 2rn + rn + Sjjrmnm x, t -
r5 r3 )]u1(
Cl -) — 2 rjn2 - 3r1n + 6ijrmnm
)u1(x,t - .1r5
r3 r3)
rn2 + r1n, +
r5 r3
{ )rni
IC-1 u
-
(x, t - \r)\ dA l
+ 2 C rrjrmnm
+ - r)c
(4 2 ,c 4 r 1 c Cl 2 rjrjrmnm - 1 rn + Siji'mflm r 2
)üi(x,i--)}dS (26)
34
5.2 Boundary Integrals for Plane Strain
Consider a plane strain problem on the x, y-plane. The appropriate boundary integral
equation for this type of problem can be obtained by viewing &E as the surface of an
infinite cylinder parallel to the z-axis. Let us take only the region z > 0 (and then multiply
the results- by two) so that 81E E (x, y) x [0, oo). Equation (4) then reads
c .u(e) = 2lrfL.L 0 dzdL
(27)
where L is the projection of 91E on the x, y-plane.
We next discretize L into n,1 boundary elements and integrate over the elements Lq
numerically to obtain
1 el g c . u() =
(Jo dz)dLq F9 W9 (28) q=1 g=1
where n9 is the number of Gauss integration points, W9 are the Gauss weights, and
F9 = fo D dz (29)
Because of the time-translation property of Vu and u, Eq. 29 should be convoluted with
z (or, equivalently, with t).
Performing a convolution in the time domain requires a temporal discretization. Due
S to the presence of a velocity term that multiplies the tranction kernel, the temporal dis-
cretization needs to be at least C0 -continuous. We consider the following piecewise-linear
temporal discretization for particle displacement (Manolis and Beskos 1988):
UX,i_ (
t_tn _r/c) nlI ' r'
C n At C _)
[(t_tn_i_rIc) (30)
and the following piecewise constant temporal interpolation for particle velocity:
u(x,t__)=>(u Ltn_un-i
) On (-_r) CI (31)
Also, consider the following piecewise linear temporal discretization for the gradient field:
35
___________ t—i,, —r/c Vux,t - = [(_ — —_r)u - ( )'u] (t - ) (32) C n
where c represents the relevant wave speed, n is a time step counter, At is the time increment, and
q(t) = H(t - (n - 1)zt) - H(t - nLt) (33)
in which H is the Heaviside function. Furthermore, the causality condition allows Eq. 29 to be written in the form
F91fR
Dg dz (34)
where a2 = x 2 + y, which is fixed at a given integration point, while R represents the
relevant distance traveled by the singular-point disturbance. Also, the integrands D 9 have typical forms (a2 + 2)_m/2, where m = 1,... , 4, and are therefore analytically integrable.
Note in this formulation that the convolution is done at the Gauss integrations points,
unlike in most conventional formulations where the convolution is done at the nodes. Con-
sequently, piecewise continuous space-time interpolations need to be assumed. To illustrate
this point, assume that the spatial and temporal interpolations are piecewise linear; then
the space-time shape functions must be bilinear. Studies are underway to investigate the
performance of the BE code for eventual integration into the nonlinear FE code.
36
6 List of Publications
Borja, R. I., Wu, W.H., and Smith H. A., "Nonlinear response of vertically oscillating
rigid foundations," submitted to ASCE Journal of Geotechnical Engineering (in review).
Borja., R. I., Wu, W.H., and Smith H. A., "Nonlinear vertical vibration of rectangular
foundations," to appear in the Proceedings of the Tenth World Conference in Earthquake
Engineering Madrid, Spain, July, 1992.
Borja, R. I., Wu, W.H., Amies, A. P., and Smith H. A., "Nonlinear lateral, rocking,
and torsional vibration of rigid foundations," submitted to ASCE Journal of Geotechnical
Engineering (in review).
"Coupled finite element-boundary element formulation for nonlinear dynamics problem
in plane strain" (in preparation).
"Numerical modeling of kinematic interaction for nonlinear soil-structure interaction
analysis" (in preparation).
S
37
7 Acknowledgements
Financial support for this research was provided by the California Universities for Research
in Earthquake Engineering (CUREe), and by the Kajima Corporation, Japan which is
gratefully acknowledged. Computer time was provided by a grant from the San Diego
Supercomputer Center. The authors also wish to acknowledge the help of Mr. Wen-
Hwa Wu and Mr. Alexander Amies, both Ph.D. students studying under the direction of
Professors Smith and Borja.
References
Borja, R. I. (1991a). "Composite Newton-PCG and quasi-Newton iterations for non-
linear consolidation." Comput. Methods Appi. Mech. Engrg., 86(1), 27-60.
Borja, R. I. (1991b). "Modeling the monotonic and cyclic viscoplastic soil behavior."
Proc., Second hit. Conf. Recent Advances Geotech. Earthquake Engrg. Soil Dyn., S.
Prakash, ed., St. Louis, MO, 37-40.
Dafalias, Y. F., and Hefrmann, L. R. (1982). "Bounding surface formulation of soil
plasticity." Ch. 10 in: Soil Mechanics—Transient and Cyclic Loads, G. N. Pande and
0. C. Zienkiewicz, eds., John Wiley and Sons, 253-282.
Duvaut, C. and Lions, J. L. (1976). Inequalities in mechanics and physics. Springer-Verlag, New York, N.Y.
Eringen, A. C. and Suhubi, E. S. (1975). Elastodynamics, Vol. 2: Linear Theory, New York, Academic Press.
Gazetas, G. (1983). "Analysis of machine foundation vibrations: state of the art." Soil Dyn. Earthquake Engrg., 2(1), 2-42.
Gazetas, G. (1987). "Simple physical methods for foundation impedances." Ch. 2 in:
Dynamic Behaviour of Foundations and Buried Structures, P. K. Banerjee and R.
Butterfield, eds., Elsevier Applied Sciences, London: 45-93.
Gazetas, G. (1991). "Formulas and charts for impedances of surface and embedded
foundations." J. Geotech. Engrg., ASCE, 117(9), 1363-1381.
Givoli, D. (1988). "A finite element method for large-domain and problems," thesis
presented to Stanford University, Stanford, Calif., in partial fulfillment of the require-
ments for the degree of Doctor of Philosophy.
Hilber, H. M., Hughes, T. J. R., and Taylor, R. L. (1977). "Improved numerical
dissipation for the time integration algorithms in structural dynamics." Earthquake Engrg. Struct. Dyn., 5(3), 283-292.
Hsieh, T. K. (1962). "Foundation vibrations." Proc. Inst. Civil Engrs., 22(June), 211-226.
Hughes, T. J. R. (1984). "Numerical implementation of constitutive models: rate-
independent deviatoric plasticity." Chapter 2: in Theoretical Foundations for Large-
Scale Computations for Nonlinear Material Behavior, S. Nemat-Nasser, R. J. Asaro,
and G. A. Hegemier, eds., Martinus Nijhoff Publishers, Boston, MA, 26-57.
Hughes, T. J. R. (1987). The finite element method. Prentice-Hall, Englewood-Cliffs,
N.J.
Jakub, M., and Roesset, J. M. (1977). "Dynamic stiffness of foundations: 2-1) vs 3-D
solutions." Research Rep. R77-36, MIT, Cambridge, MA.
Kausel, E. (1974). "Forced vibrations of circular foundations on layered media." Re-search Rep. R74-11, MIT, Cambridge, MA.
Kausel, E., Roesset, J. M., and Christian, J. T. (1976). "Nonlinear behavior in soil-
structure interaction." J. Geotech. Engrg. Division, ASCE, 102(11), 1159-1170.
Kausel, E. and Ushijima, R. (1979). "Vertical and torsional stiffness of cylindrical
footings." Research Rep. R76-6, MIT, Cambridge, MA.
Kellogg (1929). Foundations of Potential Theory. Springer-Verlag, NY.
Kobayashi, S. (1988). "Elastodynan-ijcs." Chapter 4 in: Boundary Element Methods in Mechanics, North-Holland, 191-255.
Krieg, R. D. and Key, S. W. (1976). "Implementation of a time-in dep en dent plasticity
theory into structural computer programs." in Constitutive Equations in Viscoplastic-
40
ity: Computational and Engineering Aspects, J. A. Stricklin and K. J. Saczalski, eds.,
AMD-20, ASME, New York, N.Y.
Luco, J. E., and Westman, R. A. (1971). "Dynamic response of circular footings." J.
Engrg. Mech. Div., ASCE, 97(EM5), 1381-1395.
Luco, J. E. (1974). "Impedance functions for a rigid foundation on a layered medium."
Nuci. Engrg. Des., 31(2), 204-217.
Lysmer, J. (1965). "Vertical motions of rigid footings," thesis presented to the Uni-
versity of Michigan, Ann Arbor, Michigan, in partial fulfillment of the requirements
for the degree of Doctor of Philosophy.
Lysmer, J. and Kuhlemeyer, R. L. (1969). "Finite dynamic model for infinite media."
J. Engrg. Mech. Div., ASCE, 95(EM4), 859-877.
Lysmer, J. et al. (1974). "LUSH—A computer program for complex response analysis
of soil-structure systems." Rep. No. EERC Univ. of California, Berkeley, CA.
Manolis, G. D. (1983). "A comparative study on three boundary element method
approaches to problems in elastodynamics." mt. J. Numer. Methods Engrg., 19(1), 73-91.
Manolis, G. D. and Beskos, D. E. (1988). Boundary Element Methods in Elastody-
namics, Unwin Hyman Ltd., MA.
Mróz, Z. (1967). "On the description of anisotropic work-hardening." J. Mech. Phys.
Solids, 15(3), 163-175.
Mrdz, Z., Norris, V. A., and Zienkiewicz, 0. C. (1979). "Application of an anisotropic
hardening model in the analysis of elastoplastic deformation of soils." Ge'otechnique, 29(1), 1-34.
Ortega, J. M., and Rheinboldt, W. C. (1970). Iterative solution of nonlinear equations
in several variables, Academic Press, Inc., San Diego, Calif.
41
Perzyna, P. (1971). "Thermodynamic theory of viscoplasticity." in Advances in Applied Mechanics, Vol. 11, Academic Press, New York, N.Y.
Pinsky, P. M., and Abboud, N. N. (1991). "Finite element solution of the transient
exterior structural acoustics problem based on the use of radially asymptotic boundary
operators." Comput. Methods Appi. Mech. Engrg., 85(3), 311-348.
Prager, W. (1956). "A new method of analyzing stresses and strains in work-hardening
plastic solids." J. Appi. Mech., Trans. ASME, 78(December), 493-496.
Prevost, J. H. (1977). "Mathematical modelling of monotonic and cyclic undrained
clay behaviour." mt. J. Num. Analyt. Methods Geomech., 1(2), 195-216.
Reissner, E. (1936). "Stationare, axialsymmetrjsche, durch eine schut-telnde Masse
erregte Schwingungen eines homogenen elastischen Haibraumes." Ingenieur Archiv, 7(6), 381-396 (in German).
Richart, F. E., and Whitman, R. V. (1967). "Comparison of footing vibration tests
with theory." J. Soil Mech. Fdn. Engrg. Div., ASCE, 93(SM6), 143-168.
Richart, F. E., Woods, R. D., and Hall, J. R. (1970). Vibrations of soils and founda-tions. Prentice-Hall, Englewood-Cliffs, N.J.
Roesset, J. M. (1980). "Stiffness and damping coefficients of foundations." in: Dy- namic Response of Pile Foundations, ASCE, O'Neill and R. Dobry, eds., 1-30. S Simo, J. C., and Taylor, R. L. (1985). "Consistent tangent operators for rate-
independent elasto-plasticity." Comput. Methods Appi. Mech. Engrg., 48(1), 101-118.
Simo, J. C., Keinedy, J. G., and Govindjee, S. (1988). "Non-smooth multisurface plas-
ticity and viscoplasticity. Loading/unloading conditions and numerical algorithms."
mt. J. Num. Methods Engrg., 26(10), 2161-2185.
Tassoulas, J. L. (1981). "Elements for the numerical analysis of wave motion in layered
media." Research Rep. R81-2, MIT, Cambridge, MA.
42
Veletsos, A. S., and Wei, Y. T. (1971). "Lateral and rocking vibrations of footings."
J. Soil Mech. Fdn. Div., ASCE, 97(SM9), 1227-1248.
Veletsos, A. S., and Verbic, B. (1973). "Vibration of viscoelastic foundation." mt. J.
Earthquake Engrg. Struct. Dyn., 2(1), 87-102.
Veletsos, A. S. (1977). "Dynamics of Structure-Foundation Systems." in: Structural
and Geotechnical Mechanics, W. J. Hall ed., Prentice-Hall, Inc., Englewood Cliffs,
New Jersey, 333-361.
Veletsos, A. S., and Tang, Y. (1987). "Vertical vibration of ring foundations." mt. J.
Earthquake Engrg. Struci. Dyn., 15(1), 1-21.
Wu, W. H. (1991). "Soil-structure interaction effects of simple structures supported
on rectangular foundations," thesis presented to Rice University, Houston, Texas, in
partial fufflilment of the requirements for the degree of Master of Science.
n
43
MASSLESS RIGID EXT0)
FOUNDATION
a
Figure 2.1. Soil-Structure Model.
Ca
I
71=00
H'=O / tj>0
H'<O
PimEa (a) (b)
Ca
S
Cl
0=
0<0<
0=1
ta
L STRESS
>0
10
:0< 1
(d) a3
(C)
t
Figure 2.2. Geometric Interpretation of ConstitutiveModel Parameters: (a) Hardening, Perfect Plasticity, and Softening; (b) Viscous and Inviscid Plasticity; (c) Isotropic, Kinematic,
and Combined Hardening; (d) Viscous Translation and Expansion of Yield Surface on Pi-Plane.
MESH SCALE: lOm
Figure 3.1. Finite Element Mesh for Analysis of Vertically Oscillating Circular Foundation.
S
1.2
p1.0
60.8 - 0 FREQUENCY-DOMAIN SOLUTION 0 TIME-DOMAIN (FE) SOLUTION
0.2
0.0 L
0
2 4 6
8 FREQUENCY PARAMETER, a0 = ar/v3
(a)
TIME, tIT
Figure 3.2. Mesh Convergence Study. Vertically Oscillating Circular Foundation on Elastic Half-Space: (a) Amplitude Versus
Frequency Plot; (b) Time-History Plot.
I 2 4 6 8
FREQUENCY PARAMETER, a0 = cor/v,
(a)
I
z' u..
0 0.4
I— z 0.0 - TIME, tIT w
\t
() -04
=
0 -j 1 !! -. -. -. 0.0002 a
- 0 . 8 17 0.0005 ESTIC SOLUTION
(b)
Figure 3.3. Vertically Oscillating Circular Foundation on Kinematically Hardening Viscoplastic Half-Space: (a) Amplitude
Versus Frequency Plot; (b) Time-History Plot.
2.5
2 4 6 8 FREQUENCY PARAMETER, a0 =
0.0 0
(a)
'.1 ' I'. ! 0 8
S / •-'% S O-r* , \.
20.4 1 A it
I \\
0.0 TIME, tIT ii i\ '.3
-0.4 /1 Ilk '. / f \\ I _o_H'=._o.oSEs I ------ H'=O.1E5 - 0.8 I l \j - - - H' = 0.2Es
1 - ELASTIC SOLUTION
(b)
Figure 3.4. Vertically Oscillating Circular Foundation on Elasto-Plastic Half-Space. Influence of H' on (a) Amplitude Versus
Frequency Plot; (b) Time-History Plot.
E
C
E
E
C
C
-U----
Iliflunaus IIIIIiiiiiu•
FULL PLASUFICAI1ON PARTiAL PLAS11FICA11ON
Figure 3.5. Establishing Steady-State Waves and Yield Zones for Circular Foundation Problem: Arrows Denote Instantaneous Direction
of Force; E = Extensional Yielding; C = Compressional Yielding.
Figure 3.6. Finite Element Mesh for Analysis of Vertically Oscillating Square Foundation; Mesh Dimensions are 60x60x60 m.
0
2.5
.0 FREQUENCY-DOMAIN SOLUTION
TIME-DOMAIN (FE) SOLUTION:
D a=O.00O9E3
L. c7=O.0011E
a=O.0016E
o ELAS11C SOLUTION
0 1 2 3 4 5
FREQUENCY PARAMETER, a0 = t-ob/v
Figure 3.7. Amplitude-Frequency Plot for Vertically Oscillating Square Foundation on Elasto-Plastic Half-Space.
.5
[s]
I
S
Figure 3.8. Square Foundation Problem: Deformed Mesh and Steady-State Yield Zone at 'Deepest Push".
TS:
Figure 3.9. Square Foundation Problem: Deformed Mesh and Steady-State Yield Zone at "Highest Pull".
14
12
10 0
O T
Uj
0
J1JIH
VALLEY
2 2r
1H NUllIllIll IIIIIIIIII
STRATUM OVER BEDROCK
IJ 4 =O.O5
I
1 2 3 4 5 6
7 8
FREQUENCY PARAMETER, a0 = o.,r/v 3
Figure 3.10. Influence of Boundary Conditions on Amplitude-Frequency Plots and Resonant Peaks for Vertically
Oscillating Circular Foundation (Hir = 2): as Soft Region Becomes More Finite, Resonant Peak Increases.
Figure 3.11. Finite Element Mesh for Analysis of Vertically Oscillating Rectangular Foundations; Mesh
Dimensions are 60x60x60 m.
ASPECT RATIO:
1/b =1.0 G i/b = 2.0 L i/b=3.2
i/b=oo
2 3 4 5 6
FREQUENCY PARAMETER, a0 = tob/v,
Figure 3.12. Amplitude-Frequency Plot, Elastic Case.
.v
.5
.0
.5
5
ELASTIC
® i/b = 1.0 ELASTO-PLASTIC
i/b=1.0 ' i/b =2.0. t i/b =3.2
l/b=oo
0.01 I I I I I 0 1 2 3 4 5
6
FREQUENCY PARAMETER, a0 =
Figure 3.13. Amplitude-Frequency Plot, Elasto-Plastic Case. r
S PARTIAL PLASTIFICATION
FULL PLASTIFICATION
Figure 3.14. Deformed Mesh at t=1.5T for a Rectangular Foundation with 1/b = 2.0.
PARTIAL PLASTIFICATION FULL PLASTIFICATION
Figure 3.15. Deformed Mesh at t=2.OT for a Rectangular Foundation with I/b = 2.0.
a
PARTIAL PLASTIFICATION
FULL PLASTIFICATION
Figure 3.16. Deformed Mesh at t=1.5T for a Strip Foundation.
PARTIAL PLASTIFICATION FULL PLASTIFICATION
Figure 3.17. Deformed Mesh at t=2.OT for a Strip Foundation.
VERTICAL
ROCKING
LATERAL
TORSIONAL
Figure 4.1 Symmetry and Anti-Symmetry for Vertical, Lateral, Rocking, and Torsional Vibration Modes. Arrows Denote Free
DOFs, Circles Denote Fixed DOF's.
3.0
2.5
< 2.0 0
Li H'=O.3E
ô
1
0
2 3 4 5 6
3.0
2.5
< 2.0 0
Li
Ai=O.002
<> 11= 0.005 0 ELAS11C SOLUTION
1. 2 3 4 5 6
FREQUENCY PARAMETER, a0 = o.th/v
Figure 4.2 Amplitude-Frequency Plots for Horizontally Vibrating Strip Foundations Showing Effects of Plastic Modulus and Viscosity Parameter.
3.5
3. go
Ii
H'=O.4E
A H'=O.7E
<C> H'=Es ELASTIC SOLUTION
0.0 0
1 2 3 4 5 6
2.5
0
A i=O.002
a ELASTIC SOLUTION
1 2 3 4 5
FREQUENCY PARAMETER, a0 = ob/v M.
Figure 4.3 Amplitude-Frequency Plots for Strip Foundations Subjected to Rocking Vibration Showing Effects of Plastic Modulus and Viscosity Parameter.
(b)
(a)
MESH (C) (d) SCALE:
10 m
Figure 4.6 Three-Dimensional FE Grids for (a) Low-Frequency Excitation and (b) High-Frequency Excitation. Plane Strain FE Grids: (c) Coarse Mesh and (d) Fine Mesh. Note: Foundation
Dimensions are the Same for all Grids.
a0 = 1.0
1
2
c 08
1 0.4 I- z LU
0.0 LU C.) 4 -J CIL Cl) a0 = 3.0
-u.a 0
4
6
1.0
< 0.5
I- z LU
0.0 LU C-) 4
Cl)
_1 fl
a0 =5.0
0 2 4 6 8 10
TIME,t/T
Figure 4.7 Normalized Displacement-Time History Responses of Horizontally Vibrating Strip Foundation on Elastic Soil; Solid Lines
Denote Fine Mesh, Dashed Lines Denote Coarse Mesh.
1.0
2.0
I— z w 0.0 w C.)
-J -2.0
Cl) 0
a0= 1.0
0 1 2
0 2 4 6
1.0
< 0.5
I- z LU
0.0 LU C.)
-0.5
Cl) 0
_1 fl
a0 =5.0
2 4 6 8 10 TIME,t/T
Figure 4.8 Normalized Displacement-Time History Responses of Horizontally Vibrating Strip Foundation on Elasto-Plastic Soil (H' = O.4E , i = 0.00I3Es); Solid Lines Denote Fine Mesh, Dashed
Lines Denote Coarse Mesh.
S
a0 = 1.0
1.0
O 1 2
a0 3.0
-1.0' 0 2
0.6 1 a0=5.0 C)
0.3
00 0•
-0.3
-0.6 0 2 4 6 8 10
TIME,t/T
Figure 4.9 Normalized Rotation-Time History Responses of Rocking Strip Foundation on Elastic Soil; Solid Lines Denote Fine Mesh, Dashed
Lines Denote Coarse Mesh.
4
6
a0= 1.0
0 1 2
10
z 00.0
0 -1.0
a0= 3.0
0 2 4 6
0.8 a0 =5.0
-0.8 0 2 4 6 8 10
TIME,t/T
Figure 4.10 Normalized Rotation-Time History Responses of Rocking Strip Foundation on Elasto-Plastic Soil (H' = 0.4E, 1 = O,y= 0.0013E);
Solid Lines Denote Fine Mesh, Dashed Lines Denote Coarse Mesh.
2.0
--@- FREQUENCY-DOMAIN SOLUTION
1.5 TIME-DOMAIN (FE) SOLUTION:
El i=O.O i=O.002
1.0
0 ELAS11C SOLUTION
0.5
1 2 3 4 5
FREQUENCY PARAMETER, a0 = wb/v
Figure 4.11 Amplitude-Frequency Plot for Horizontally Vibrating Square Foundation Showing Effects of Viscosity Parameter.
3.0
t) 2.5 1-6
-- FREQUENCY-DOMAIN SOLUTION
2.0 0
TIME-DOMAIN (FE) SOLUTION: H' =O.4E
AH'=E ELASTIC SOLUTION
me 0 1 2 3 4 5
7.
FREQUENCY PARAMETER, a0 = wb/v
Figure 4.12 Amplitude-Frequency Plot for a Horizontally Rocking Square Foundation Showing Effects of Soil Plastic Modulus.
--®-- FREQUENCY-DOMAIN SOLUTION
TIME-DOMAIN (FE) SOLUTION: a=O.00l4E
t a=O.0028E ELASTiC SOLUTION
0
I I I I
0 1 2 3 4 5
FREQUENCY PARAMETER, a0 =
Figure 4.13 Amplitude-Frequency Plot for a Torsionally Vibrating Square Foundation Showing Effects of Initial Yield Stress.
4
KI
S
a0=4.0
-o i 2 3 4 5
4
e2
z 0O I-
0 -2 I
I
V
a0 = 5.0 1
2 4 6 8
TIME,t/T
Figure 4.14 Normalized Torsional Rotation-Time History Responses of Square Foundation on Elasto-Plastic Soil (H' = O.4E5 , 0, = 0.00I3Es); Solid Lines Denote Fine Mesh, Dashed Lines Denote Coarse
Mesh.
t= 1.50T t = 1.75T
t = 2.00T t = 2.25T
— FULL PLASTIFICATION EM PARTIAL PLASTIFICATION
Figure 4.15 Establishing Steady-State Waves and Yield Zones for Horizontally Vibrating Square Foundation: Arrows Denote
Instantaneous Direction of Force.
t= 1.50T t=1.75T
S
t = 2.00T t = 2.25T
FULL PLASTIFICATION
E.x.x.:. PARTIAL PLASTIFICATION
Figure 4.16 Establishing Steady-State Waves and Yield Zones for Rocking Square Foundation: Arrows Denote Instantaneous
Direction of Moment.
t=1.50T
t=1.75T
6M
()
t = 2.00T
t = 2.25T
FULL PLASTIFICATION ::: PARTIAL PLASTIFICATION
Figure 4.17 Establishing Steady-State Waves and Yield Zones for Torsionally Vibrating Square Foundation: Arrows Denote
Instantaneous Direction of Torque.
UNIVERSITY OF SOUTHERN CALIFORNIA
Department of Civil Engineering
Research Center for Computational Geomechanics
A STUDY OF SOIL-PILE-STRUCTURE INTERACTION IN LIQUEFIABLE SOILS
DURING STRONG EARTHQUAKES
by
J.P. Bardet, S.W. Chi, Q. Huang, and G.R. Martin
Kajima-CUREe project report
August 1992
Table of contents
INTRODUCTION................................................................................................................ 1
SIMULATION OF CENTRIFUGE TESTS ON PILE GROUPS ................................................... 1 Summary of experimental results ........................................................................ 1 Definition of finite element model......................................................................2
Geometry................................................................................................. 2 Calibration of material parameters..........................................................2
Equivalent 2D structural properties.............................................2 Linear soil properties................................................................... 3 Nonlinear soil properties............................................................. 3
Soil permeability .......................................................................... 5
Eigenvalueanalysis............................................................................................. 5 Linearanalysis ...................................................................................................... 5
Method.................................................................................................... 5 Results............................................ ......................................................... 6
Nonlinearanalysis............................................................................................... 6 Freefield response .................................................................................. 6
Response of pile group............................................................................ 7 Conclusion........................................................................................................... 8
PREDICTION OF SEISMIC RESPONSE OF K-BUILDING FOUNDATION .............................. 8 Definition of finite element model......................................................................9
Geometry................................................................................................. 9 Calibration of soil parameters................................................................. 9
Linear soil properties................................................................... 9 Nonlinear soil properties............................................................. 9
Eigenvalueanalysis.............................................................................................10 Linearanalysis..................................................................................................... 10 Nonlinearanalysis............................................................................................... 11
Freefield response .................................................................................. 11 Moderate earthquake loading (100 gal) ..................................................11 Strong earthquake loading (407 gal).......................................................12
Conclusion........................................................................................................... 12
REFERENCES................................................................................................................... 13
ACKNOWLEDGMENT....................................................................................................... 13
LIST OF FIGURE CAPTIONS .............................................................................................. 15
LIST OF TABLES............................................................................................................... 86
INTRODUCTION
The interaction between building, foundation and soil during earthquakes is a
challenging problem for researchers which has practical and financial implications in the
design of buildings on liquefiable soils in Japan and the United States of America. Until
now, mostly simplified methods have been employed to describe the nonlinear effects of
soil-pile response in interaction problems. These methods use nonlinear, hysteretic and
degrading springs to represent the soil response in the pile vicinity. However, there is still
very limited experience in using continuum techniques based on nonlinear finite elements
and nonlinear constitutive models. This report summarizes the findings of the University of Southern California
(USC) during the research on building-foundation-soil interaction conducted under the
) sponsorship of Kajima-CUREe. The purpose of this research is (1) to verify a finite
element procedure developed at USC for analyzing liquefaction problems by comparison
of numerical and centrifuge results in the case of a pile group, and (2) to apply the
procedure to predict the response of an existing building founded in liquefiable soils.
The report is organized into two distinct parts. The first part is related to the
verification and calibration of our procedure by comparison to centrifuge results. The
second part is concerned with the application of the finite element procedure to a
particular building (the K-building) in Tokyo.
SIMULATION OF CENTRIFUGE TESTS ON PILE GROUPS
Summary of experimental results
Scott and Hushmand (1991) carried out a series of seven centrifuge tests at the
) California Institute of Technology in the framework of the Kajima-CUREe research
project on building-foundation-soil interaction effects. The objective of their
experimental program was to better understand the interaction between pile-foundations
and liquefiable soil during earthquakes. The prototype foundation was made of a group of
four piles, about 1 im in length and covered with a rigid cap. The pile group model was
placed in a laminar box filled with dry or saturated Nevada sand of various initial
densities. The laminar box was laterally shaken at its base by earthquake input motions of
various magnitudes. The response of the structure was monitored by recording the
bending moment in the piles, the acceleration in soil and pile cap, the horizontal
displacement, and the pore pressure in the case of saturated tests. The test geometry and
position of sensors are shown in Fig.1. Further detail on the description of the test
procedure and results may be found in Scott and Hushmand (1991).
- page 1 -
The objective of the present analysis is to simulate the results of the centrifuge
tests by using nonlinear finite element analysis. As for analyses by Miura et al (1991,
1992), four centrifuge tests, out of seven tests, have been selected for analysis. These tests
are referred to as tests 4,6, 8 and 9 by Scott and Hushmand (1991) and D2, Dl, S2 and
Si by Miura et a! (1992), respectively.
Definition of finite element model
Geometry
Fig.! shows the finite element mesh used in the analysis. The geometry is similar
to that of Miura et al (1992), except that the mesh has only a single layer of elements,
instead of several layers as described by Miura et al (1992). The soil layers, piles and pile
cap are represented by using four noded isoparametric elements.
In the nonlinear and saturated analyses, the pore-pressure is allowed to dissipate
and diffuse, owing to soil permeability. There are three degrees of freedom per node.
Two degrees are for the horizontal and vertical solid displacements, and one for the pore
pressure. This finite element formation is referred to as U-p formulation (Zienkiewicz and
Shiomi, 1984) Nodes 'on the left and right.vertical boundaries at the same e1evation•have
identical horizontal displacement, which prescribes the lateral boundary conditions of the
laminar box. However, the vertical component of the boundary nodes is left free. The
earthquake input motion is applied by specifying the horizontal acceleration at all the
lowest nodes. The earthquake input motion was provided under digitized format by R.F.
Scott and B. Hushmand. The acceleration corresponds to record No.13 of Scott and
Hushmand (1991) data.
Calibration of material parameters
The structural properties are assumed to be linear; they are scaled to account for
three-dimensional effects. The soil which surrounds the pile group is considered in both
linear and nonlinear cases.
Equivalent 2D structural properties
One of the problems in the two-dimensional (2D) modeling of three-dimensional
(3D) soil-structures is to preserve the dynamic characteristics of 3D soil-structures. The
inclusion of nonlinear soil behavior requires that both initial stresses (applied by the static
weight of structures) and dynamic stresses (applied by the dynamic motion of structures)
be calculated with a reasonable degree of accuracy. The dynamic motion of structures
generates transient stresses on the soils, in addition to those directly resulting from the
earthquake waves. Owing to their highly nonlinear nature, soil properties cannot be
simply scaled without altering their characteristic responses. It is therefore important that
- page 2 -
the 2D scaling of 3D structures preserve the fundamental modes and periods in the range
of earthquake ground motion. This scaling requires not only selecting appropriate unit mass for soil and
structure, but also the careful choice of structural stiffnesses. We propose to verify the
assumption on the 2D scaling of structural masses and stiffnesses by performing an
eigenvalue analysis using elastic properties for both soil layers and structures. We will
also verify the assumption on structural unit masses by examining the initial soil stresses
at the end of the gravity loading, before the application of earthquake input motion. The scaled 2D and 3D properties for the pile and pile cap are summarized in table
1 and the corresponding calculation steps are reported in tables 2 and 3. The plane strain
analyses are performed on a 1-rn thick layer of soil and structure. The scaling of the structural properties assumes that the soil interacting with the pile group has the same
thickness as the pile spacing (1.67rn).
Linear soil properties The linear soil properties selected in the analysis are similar to those of Miura et
al (1991, 1992) that were derived from laboratory test results on Nevada sand compiled
by Scott and Hushmand (1990). The following relation is used to calculate the low-strain
shear modulus G0 in kPa:
G0 -- A (1+e) m
(2. 17-e)2 3 0.5
(1)
where e is the void ratio and (Nm the mean effective stress in kPa. The coefficient A is set
equal to 5840. (Miura et al, 1992). The variation of G and Young's modulus E with depth
is reported in tables 5 and 6 in the case of tests 4,6, 8 and 9. The Poisson ratio is set
equal to 0.3.
Nonlinear soil properties
)
The nonlinear soil behavior of Nevada sand is simulated by using the BDP model.
This model which derives from previous work (Bardet, 1990) is currently in publication
and is detailed in Proubet (1991). This section only provides a summary necessary to
understand the calibration of BDP model parameters. The BDP model has 11 material constants that are calibrated, in our analysis, from
experimental tests on the Nevada sand.. Table 7 lists the constant values used in the
analysis. The following procedure has been adopted to define these constants.
The elastic properties in the BDP model depends on mean pressure p. The
Young's modulus E is:
E =E0 po (2)
- page 3 -
where the elastic parameters E0, no, and V are calculated from the initial slope of stress-
strain response during triaxial or simple shear tests. When using the results of low-strain-
amplitude resonant column tests, it is recommended to multiply E0 obtained from Eq. 1
by a factor of 0.6 to account for larger but reversible strain levels.
The failure parameters a and c are defined from the friction angle (1) and the
apparent cohesion of Nevada sand. In the simple shear test, a and 4) are related through:
a=sin4) (3)
The parameter Pa that is the pressure associated with undrained shear strength 'tmax is
calculated by using the following relation:
tmax - C Pa=
(4) a
The parameter 0 controlling the transition between contracting and dilating behavior is
defined by selecting a positive attractive pressure Pb close to zero: I
R 'max ' Pa - Pb
The pressure Pb attracts the effective stress paths during undrained loadings. The dioser
Pb is to zero, the softer is G, the mean-pressure-dependent shear modulus.
The material parameters H and n, that control the variation of plastic deformation
when the stress point approaches the failure surface, are calibrated by using a one-
element program called SOIL (Bardet, 1989). SOIL is based on linearized integration
techniques (Bardet and Choucair, 1991). H and n are adjusted by trial and error so that the
simulated response fits the experimental data points. The tests selected for fitting the
material parameters is the simple shear test at 8 psi (55 kPa) initial pressure reported by
Scott and Hushmand (1990).
The tenth and eleventh parameters of the BDP model, Hu and nu, control only the
evolution of plastic deformation during cyclic loading. Independent of other constants
calculated from monotonic tests, Hu and flu are calibrated from liquefaction stress curves.
The liquefaction is assumed to take place when the pore pressure becomes minimum
during a cyclic loading. The BDP model assumes that the pore pressure remains positive
so that the shear modulus remains defined. Figs.2 and 3 show the simulated evolution of
mean pressure and shear strain during cyclic simple shear tests at prescribed shear stress
amplitude. All the tests are carried out from 55 kPa initial effective stress.
The liquefaction strength curve of Fig.4 is constructed from the simulated results
of Figs.2 and 3. This simulated liquefaction curve is in good agreement with the
experimental liquefaction curve (Scott and Hushmand, 1990).
(5)
- page 4 -
Soil permeability The permeability coefficients used in the analysis were selected as recommended
by Miura et al (1992). The values used in the analysis are reported in table 4. The
selected values are higher than the experimental values (Scott and Hushmand, 1990) to
account for the centrifuge scaling of diffusion coefficient.
Eigenvalue analysis
The eigenvalue analyses were performed to verify not only the distribution of
stiffness and mass in the finite element models but also boundaiy conditions. All the
eigenvalue analyses are carned in the linear case, by using linear properties for soil and
structure.
The analysis results are shown in Figs. 5 to 6 in the case of dry and saturated
Nevada sands. The first four fundamental frequencies are 2.49, 5.05,7.44, and 8.21 Hz in
the dry case and 2.28, 4.57, 6.37, and 7.02 in the saturated case. In the dry case, effective
stresses are calculated by using dry unit mass, whereas, in the saturated case, effective
stresses are calculated by using buoyant unit mass. The natural modes of the model are
similar in the dry and saturated cases. The first mode is a simple shear mode. In the
second mode, the pile group and soil undergo shear modes of opposite direction. In the
third and fourth modes, the soil mass moves vertically while the pile barely moves.
Linear analysis
Before performing nonlinear analyses, it is useful to perform a few linear
analyses. Linear analyses are generally faster to perform than nonlinear analyses, and
allow the assessment of nonlinear effects by comparing the results of linear and nonlinear
analyses.
Method
The elastic properties used in the linear models are assumed constant, i.e.
independent of mean pressure and shear strain amplitude, but they vary with depth as
specified in tables 5 and 6. A Rayleigh damping factor 8 , that creates viscous damping
proportional to the stiffness matrix, is included to simulate the effect of soil structural
damping caused by hysteretic soil behavior and wave radiation. The following
proportionality constants 8 were used in the linear analyses:
test number 4 6 8 9
test condition dry dry saturated saturated
maximum acceleration (g) 0.203 0.085 0.255 0.112
damping factor 8 1 0.2 0.005 0.15 0.03
- page 5 -
The values of 8 were found by trial and error in order to simulate the peak acceleration.
As expected, the damping increases with the level of earthquake ground motion owing to
more extended irreversible soil behavior. The linear analyses in the saturated cases are
carried out without considering pore pressure. There are only two degrees of freedom per
node in the linear analysis instead of three in the saturated nonlinear analysis. The shear
modulus G, which depends on the initial effective stress, is calculated by using buoyant
unit mass in the saturated cases ( 8 and 9) and dry unit mass in the dry cases (4 and 6).
Results
Figs. 7 to 22 show the results of linear analyses for tests 4, 6, 8 and 9. The input
acceleration in test 4, 6, 8 and 9 and corresponding acceleration response spectra are
shown in Figs. 7, 11, 15, and 19. The computed and experimental accelerations at
location L15 and L9 and in the pile cap are also shown.
Linear analyses show that the dry response of the pile group can be realistically
simulated by using a Rayleigh type damping. The dominant period of the pile group can
be simulated by using elastic soil properties. The maximum acceleration amplitude can be
simulated by adjusting the amount of viscousdamping; However, in the saturated cases,
the measured response spectra shift toward lower frequency, owing to the softening of the
elastic properties of liquefied sands. This spectral shift, not taken into account by linear
analyses, requires the use of nonlinear analyses.
Nonlinear analysis
Before analyzing the pile group, a free field analysis is performed to characterize
the soil response, assuming that the soil layers, free of the pile group, are of infinite
extent.
Free field response I
The soil colunm of the free field analyses Consists of a single vertical column Cut
into the fmite element mesh of Fig. 1. The pore pressure, and horizontal and vertical
displacements of the nodes on the leftand right vertical bOundaries are set to coincide.
The stresses and pore pressure are initialized by applying a constant gravity loading. This
gravity initialization ensures that the model is in an equilibrium state before beginning the
earthquake loading.
The results of the free field responses are shown in Figs. 23 to 25 for test 8. Figs.
23 shows the computed accelerations and acceleration response spectra at L15 and L9.
The results of free-field simulations may be compared to the experimental results on the
pile group, keeping in mind that the pile group may influence the dynamic response of
- page 6 -
the soil layers. The computed acceleration is similar to the experimental acceleration in
Fig.30. The input acceleration is slightly attenuated at the ground surface, indicating that
the soil layers do not amplify the input ground motion. Fig.24 shows that the computed
excess pore-pressure build-up at PP2 and PP4 is larger than the experimental values (with
the pile group) in Fig.33. Fig. 25 shows that the relative displacement is the same order
of magnitude as the experimental value in Fig. 32.
The results of the free field responses are shown in Figs. 26 to 28 for test 9. The
conclusions drawn for test 8 may be extended to test 9. The peak acceleration at ground
surface is 0.3 g and 0.2 g for test 8 and 9, respectively, which shows that the
amplification of acceleration is larger in test 9 than in test 8. Both spectral peaks of
computed acceleration (1.5 g for test 8 and 0.9 g for test 9) are higher than the
experimental ones (1 .g for test 8 and 0.6 g for test 9). The computed and measured
displacements have the same order of magnitude. The computed pore-pressures in tests
8 and 9 are larger that the measured pore-pressures.
Response of pile group
Figs.29 to 33 show the results of the nonlinear finite element analysis for test 8.
They compare the computed and experininta1 time histories of acceleration at L9 and
L15, and at pile cap, acceleration response spectra, displacement at L3, L9 and L15, and
finally pore pressure at PP1, PP2, PP3 and PP4.
As shown in Figs. 29 and 30, the computed and experimental accelerations are in
agreement. However, as shown in Fig.31, the computed acceleration of the pile cap has a
dominant frequency larger than the experimental one, which indicates that the numerical
model is stiffer than the centrifuge model. As shown in Fig. 32, the analysis predicts
permanent displacement slightly larger than the experiment. As shown in Fig. 33, the
computed pore pressures are larger than the measured ones. The computed and measured
pore pressures increase at the same rate but the computed pore-pressure diffuses less
rapidly than the measured one. This difference is expected since two-dimensional
drainage is more constrained than three-dimensional drainage.
Figs.34 to 38 show the results of the nonlinear finite element analysis for test 9.
As shown in Fig.34 to 36, the agreement between computed and measured accelerations
is better in test 9 than in test 8. The acceleration level in test 9 is lower than in test 8, and
generates less nonlinear effects in the soil layers. As shown in Fig. 37, the computed
displacements are slightly larger than the measured ones. The square erratic lines in
Fig.37 result from a round-off error during the post-processing phase, but this does not
invalidate the computed results. As shown in Fig.38, the computed pore pressure is much
larger than the measured one. This discrepancy may come from faster three-dimensional
drainage. The same problem was noticed in the analysis of Miura et al (1992).
- page 7 -
The permeability coefficient is an important factor that controls the build-up and
dissipation of pore-pressure. 2D models do not easily explain 3D pore-pressure
dissipation mechanisms. Close to the piles, the 3D pore pressure builds up more slowly
and dissipates more rapidly than in 2D cases. This 3D effect could be artificially
simulated in 2D analyses by decreasing the model capacity to generate pore-pressure and
by locally increasing the permeability coefficient. The 2D scaling of 3D permeability and
drainage mechanism needs to be further investigated.
Conclusion
The response of pile groups in dry sands may be simulated by using linear
analysis, provided that an adequate value of Rayleigh damping be selected. However, the
simulation of the response of pile groups in saturated sands requires the use of nonlinear
models having the capability of building up and diffusing pore pressure to account for
the structural softening observed during pore pressure build-up.
in the context of the nonlinear finite element analyses with pore-pressure
diffusion, the BDP model is a useful model to describe the cyclic response of saturated
sands. It is capable of predicting the pore-pressure build-up as well as the softening of the
material properties resulting from pore-pressure build-up. The BDP model is a useful tool
to analyze the experimental results of soil-structure-foundation interaction in saturated
sands.
PREDICTION OF SEISMIC RESPONSE OF K-BUILDING FOUNDATION
The finite element technique that was previously calibrated by comparing
numerical and centrifuge results is now applied to predict the seismic response of a
particular building, referred to as K-building. The foundation and superstructure of the K-
building is defined by Miura et al (1992). Our objective is to illustrate the application of
our procedure in a particular case of practical interest to designers. It is not our intent to
perform a parametric study, as Miura et a! (1992) have reported in order to understand the
influence of various factors on the dynamic response of the K-building.
Following the finite element model formulation of the K-building and its
foundation, an eigenvalue analysis and a linear transient analysis are carried out. These
linear analyses are followed by a free-field response for an earthquake input motion equal
to 100 gal. The nonlinear dynamic response of K-building is then predicted in the case of
moderate (100 gal) and strong (407 gal) earthquake input motion.
- page 8 -
Definition of finite element model
Geometiy
Fig.39 shows the finite element model used in the analysis of the 34-story K-
building. The superstructure model is identical to the one defined by Miura et a! (1992).
However, since our plane-strain analysis applies to a im thick layer, all the properties of
the three-dimensional structure are divided by a depth thit is assumed equal to the width
of the K-building, i.e. 30.m. The superstructure is represented by beam elements, the
properties of which are listed in table 9. The beam elements are assumed weightless. The
superstructure inertial mass is lumped at the floors by using concentrated mass elements,
the mass of which is listed in table 9. The static weight or, the superstructure is applied by
using vertical nodal loads distributed on the first floor. The piles underlying the K-
building are represented by truss elements that are capaNe of transmitting only axial
forces. The properties of the truss elements are listed in table 8. The basemat and
diaphragm wall are represented by using four noded isopzrametric linearly elastic
elements. The soil layers are also represented by four noded isoparametric elements that
have the capability of describing nonlinear soil behavior and pore-pressure diffusion. The
earthquake input motion (see Fig.46) is specified by presribing the horizontal
acceleration at depth 53.5m. This input acceleration was provided by Miüra et al (1992).
Calibration of soil parameters
Linear soil properties The linear soil properties selected in the analysis (see table 10) are similar to those
defined by Miura et a! (1992). The shear modulus G varies with the initial effective
stress, whereas Poisson ratio v is constant. v is assumed ro be equal to 0.3 for sands and
0.48 forclays (Miura et al, 1992).
Nonlinear soil properties
The clay layers, referred to as clay-i, -2, and-3, are modeled by using the elastic-
perfectly plastic von Mises model. The elastic properties.. i.e., Young's modulus and
Poisson ratio, are identical to those used in the linear analyses. The maximum size of the
von Mises failure surface is directly related to clay cohesion. Their values are reported in
table 10. The sand layers, referred to as sand-i, -2, -3, and-4, are modeled by using the
BDP model. The model constants have been calibrated zs outlined in the previous
chapter. The elastic constants E0 were selected to be 60 of the ones calculated from low
amplitude resonant tests. The slope a of the failure surfae was calculated from the
friction angle 0 by using Eq.3. The apparent cohesion was arbitrarily taken equal to zero.
- page 9 -
The slope of the characteristic state was selected to have Pb close to zero. The attractive
pressure Pa , which is ielated to the undrained shear strength through Eq.4, was assumed
to be at least twice the value of the initial effective pressure. This assumption is useful for
controlling the position of the elbow in the effective stress path during simple shear tests.
This assumption was necessary since no laboratory data was available for monotonic
simple shear tests on sand-i, -2, -3 or -4. The parameter H0 and n were adjusted by trial
and error so that the effective stress path has an elbow at the undrained shear strength
given by Miura et al (1992) (see table 9). The parameters Hu and nu were also adjusted to
fit the liquefaction strength curves provided by Miura et a! (1992). The liquefaction
strength curve of FigAO is constructed by simulating the response of sand- 1, -2, -3 and-4
during cyclic simple shear tests at constant shear stress amplitude. In these simulations,
the initial effective stress is equal to the in situ geostatic stress. The soil permeability
selected in our analysis (see table 10) comes form Miura et a! (1992).
Eigenvalue analysis
Fig.41 shows the first six eigenmodes of K-building. In the first mode (0.65 Hz)
deformation is concentrated in the building. The first mode has a higher frequency than
the one calculated (0.48 Hz) by Miura et al (1992). This discrepancy results from a larger
rotational stiffness for the foundation, which Miura et al (1992) softened by adding a
rotational spring. The five other modes have frequencies in the range determined by
Miura et al (1992), who found 1.60 and 2.79 Hz for the natural frequency of second and
third mode, respectively, in the case of a simplified spring model. The soil layers deform
simply in shear at 1.58 Hz, and in more complicated patterns when the frequency
increases. The elastic soil layers have a fundamental frequency higher than the building,
which implies that soil layers will probably interact, not with the fundamental mode of
the building, but with its second and thinl modes. This remark pertains to elastic soil
behavior, it does not rule out that the first structural mode may interact with soil layers if
the soil elastic modulisoften due to pore-pressure build-up.
Linear analysis
The linear analysis was carried out to assess the contribution of nonlinear soil
behavior to the response of the K-building using the 100 gal acceleration input shown in
Fig.42. No Rayleigh damping was introduced in the linear analysis. Fig.43 shows the
time history and acceleration response spectra of the calculated acceleration at the first
floor and top of K-building. The first structural mode (0.65 Hz) is clearly seen in the
response spectra. As shown in Fig.44, the maximum displacement at roof and first floor is
8.5 cm and 2 cm, respectively. Figs. 45. and 46 show the variation of maximum values of
horizontal acceleration, lateral displacement, bending moment and shear force in the
- page 10 -
superstructure. The results of Figs.45 and 46 are in agreement with those of Miura et al
(1992).
Nonlinear analysis
Free field response
Figs.47 to 51 show the free-field response of K-building in the case of nonlinear
soil properties and a 100 gal maximum input acceleration. Fig. 47 shows the acceleration
and acceleration response spectra at various depths. The acceleration is larger at 10.8m
than at the ground surface due to the truncation of acceleration that takes place in the
sand-2 layer. Fig.48 shows that the permanent displacement localizes in the sand-2 layer,
and that layers above and under sand-2 layer behave elastically. Fig.49 shows that pore-
pressure builds up mainly in sand-2. Fig.50 shows the effective stress path and the stress-
strain response in the middle of sand-2 layer. The vertical effective stress is initially about
40 kPa. It decreases toward zero after. 10 sec as shear stress cycles are applied by the
earthquake input motion. The stress-strain response shows that the shear strain reaches
8% in sand-2 layer. Fig. 51 shows that pore-pressure builds up also in sand-3, but to a
lésserextent, generating less-softening of the stress-strain response. '-
Moderate earthquake loading (100 gal)
Figs. 52 to 59 show the response of K-building for a 100 gal input acceleration
earthquake. Fig. 52 shows the time histoiy and acceleration response spectra on the roof
and first floor. This nonlinear response is similar to the linear response of Fig. 43, which
indicates that nonlinear effects do not dominate the building response. This remark is
corroborated by the results of Fig.53, 54 and 55 that show the time history of relative
displacement on the roof and first floor, and the maximum values of horizontal
acceleration, lateral displacement, bending moment, and shear force in the
superstructure. In spite of similarity between linear and nonlinear analyses, soil layers
undergo plastic deformation as shown in Figs. 56, 57, 58 and 59. Fig.56 shows that the
acceleration is truncated in sand-2 layer as in the free-field response. Fig.57 shows that
permanent displacements are observed in the sand-2 layer, in spite of the K-building
foundation. Fig.58 shows that pore-pressure builds up in all sand layers, but -
predominantly in sand-2 which liquefies. Finally, Fig. 59 shows that the vertical
effective stress is reduced in all sand layers, which should result in reducing the
fundamental frequency of soil layers.
The similarity of structural response for linear and nonlinear analyses indicates
that the effects of nonlinear soil behavior are reduced by the elastic diaphragm wall and
- page 11 -
piles, that rest directly on elastic soil layers. The stresses in the diaphragm wall and piles
should however remain tolerable for sustaining this conclusion.
Strong earthquake loading (407 gal)
Figs. 60 to 67 shows the response of K-building for a 407 gal input acceleration
earthquake, which corresponds to a maximum acceleration four times larger than in the
previous analysis. Fig.60 shows that the time histories of the horizontal acceleration on
the roof and first floor are different from the ones obtained in linear analysis. The
acceleration response spectra are wider, which indicates that there are more nonlinear
effects for 407 gal than for 100 gal input acceleration. Fig. 61 shows that the
displacement reaches 20 cm at the top of K-building. Figs. 62 and 63 shows that the
maximum values of horizontal acceleration, lateral displacement, shear force, and
bending moment has a similar distribution, but a larger magnitude, in the 407 gal case
compared to the 100 gal case. The computed values are in agreement with those
calculated by Miura et al (1992). Figs. 64 to 67 show that the soil layers undergo severe
nonlinear behavior. Pore-pressure build-up is more extended than in the previous case.
S and- i and -2 undergo liquefaction for the complete duration of the earthquake while
sand-3 and -4 undergo liquefaction for a shorter time.
As noted in the 100 gal analysis, the elastic diaphragm wall and elastic piles,both
resting on elastic soil, limits the nonlinear effects arising from soil liquefaction.
Additional analysis would be appropriate to examine the assumption of elastic support
and behavior for the piles and diaphragm wall.
Conclusion
The fmite element procedure utilizing the BDP model was calibrated by using
centrifuge test data. Applications of the model were then demonstrated to predict the
dynamic response of the K-building in Tokyo. Although the nonlinear soil constants were
calibrated in an approximate way owing to a lack of laboratory data, the overall
procedure produces results in general agreement with results obtained by Miura et al
(1992) with different methods. The procedure is capable of describing structural
response during earthquake loadings while soil experiences pore pressure increases up to
the point of liquefaction. Our analysis indicates that the structural response of the K-
building is strongly controlled by the assumption that the diaphragm wall and piles are
elastic and rest on elastic soil layers.
-page 12-
REFERENCES
Bardet., J.P., 1989, "SOIL: an interactive program to simulate soil behavior" Civil
Engineering Department, University of Southern California.
Bardet., J.P., 1989, "LINOS: a finite element for geomechanics" Civil
Engineering Department, University of Southern California.
Bardet., J.P., 1990, "An hypoplastic model for sand't ASCE, Engineering
Mechanics, Vol.116, No.9, PP.1973-1994.
Bardet., J.P., and W. Choucair, 1991, "Linearized integration techniques for
incremental constitutive equations," mt. J. Numerical and Analytical Methods in
Geomechanics, Vol.15, pp.1-19.
Miura, K, K. Suzuki, Y. Miyamoto, K. Masuda, and S. Uchiyama, 1991,
"Earthquake response of a structure on pile group in liquefiable sand deposit,"
Kajima-CUREe report, Kajima Corporation, Japan, p.123.
Miura, K, Y. Miyamoto, A. Yamada, M. Nagano, Y. Sago, Y. Hyodo, K.
Masuda, T. Maeda,E.Kitamura, K, Suzuki,-Y. Suzuki, A. Fukuoka, 1991,
"Building-Foundation-Soil Interaction Effects," Kajima-CUREe report, Kajima
Corporation, Japan, p.107.
Proubet, J., 1991, "Application of numerical methods to geomechanics problems,
PhD thesis, University of Southern California,.
Scott, R.F., and B. Hushmand, 1990, "Compilation of laboratory test results on
Nevada Sand," Private communication to K. Miura (Kajima Corporation), p.83.
Scott, R.F., and B. Hushmand, 1991, "Soil-pile response in the Santa Cruz (Loma
Prieta) earthquake, October 17, 1989." Kajima-CUREe report, California Institute
of Technology, Pasadena, California, University of California, Davis, California.
Zienkiewicz, O.0 and T. Shiomi, 1984, "Dynamic behavior of saturated porous
media: the generalized Biot formulation and its numerical solution." mt. J.for
Numerical and Analytical Methods in Geomechanics, Vol.8, pp.71-96.
ACKNOWLEDGMENT
The authors acknowledged the financial support of Kajima-CUREe and are
thankful for the interest, enthusiasm and patience of Kajima Corporation. The authors
thank B. Hushmand and R.F. Scott for communicating their centrifuge data. The finite
element analyses were performed and supervised by J.P. Bardet assisted by S.W. Chi and
- page 13 -
LIST OF FIGURE CAPTIONS
Geometry of finite element model for pile-group.
Shear-stress versus mean-pressure and shear-stress versus shear-strain simulated by
BDP model during cyclic simple shear test on Nevada Sand (At/a0= 0.232 and
0.193, (= 55 kPa)
Shear-stress versus mean-pressure and shear-stress versus shear-strain simulated by
BDP model during cyclic simple shear test on Nevada Sand (ttt/sY0= 0.154 and
0.128, = 55 kPa)
Experimental and simulated liquefaction strength curves for Nevada sand during
cyclic simple shear tests at c= 55 kPa.
First four natural modes of pile group for test No.6.
First four natural modes of pile group for test No.8.
Time history and acceleration response spectrum of input horizontal acceleration for
test No.6.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L9 for test No.6.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L15 for test No.6.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at pile cap for test No.6.
S11. Time history and response spectrum of input horizontal acceleration for test No.4.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L9 for test No.4.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L15 for test No.4.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at pile cap for test No.4.
Time history and acceleration response spectrum of input horizontal acceleration for
test No.8.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L9 for test No.8.
-page 15-
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L15 for test No.8.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at pile cap for test No.8.
Time history and response spectrum of input horizontal acceleration for test No.9.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L9 for test No.9.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L15 for test No.9.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at pile cap for test No.9.
Computed time histories and acceleration response spectra of horizontal acceleration
at L9 and Li 5 for free field of test No.8
Computed time histories of excess pore pressure at PP2 and PP4 for test No.8
Computed time histories of relative displacement at W. L9, and L15 for test No.8.
Computed time histories and acceleration response spectra of horizontal acceleration
at L9 and Li 5 for free field of test No.9
Computed time histories of excess pore pressure at PP2 and PP4 for test No.9
Computed time histories of relative displacement at L3, L9, and L15 for test No.9.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L9 for test No.8.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L15 for test No.8. S Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at pile cap for test No.8.
Computed and experimental time histories of relative displacement at L3, L9, and
L15 for test No.8.
Computed and experimental time histories of excess pore-pressure at PP1, PP2, PP3
and PP4 for test No.8.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L9 for test No.9.
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at L15 for test No.9.
- page 16 -
Computed and experimental time histories, and acceleration response spectra of
horizontal acceleration at pile cap for test No.9.
Computed and experimental time histories of relative displacement at L3, L9, and
L15 for test No.9.
Computed and experimental time histories of excess pore-pressure at PP1, PP2, PP3
and PP4 for test No.9.
Geometry of finite element model for K-building.
Experimental and simulated liquefaction strength curves for sands 1, 2, 3 and 4
during cyclic simple shear tests.
Fundamental modes of K-building
Time history and acceleration response spectrum of input horizontal acceleration for
K-building.
Time histories, and acceleration response spectra of horizontal acceleration on roof
and first floor of K-building computed by linear analysis.
Time histories of relative displacement on roof and first floor of K-building
computed by linear analysis
Distribution of maximum acceleration and relative displacement versus elevation
computed in linear analysis of K-building.
Distribution of maximum bending moment and shear force versus elevation
computed in linear analysis of K-building.
Time histories, and acceleration response spectra of horizontal acceleration at depth
0, 10.8, and 53.5 m for the free field of K-building.
48. Time histories of relative displacement at depths 0,6.9, 7.6, and 10.8m for free field
of K-building.
Time histories of excess pore-pressure computed at the middle of sand layers 1, 2, 3
and 4 for free field of K-building.
Shear-stress versus mean-pressure and shear-stress versus shear-strain simulated by
BDP model at the middle of sand-2 layer during free field response of K-building.
Shear-stress versus mean-pressure and shear-stress versus shear-strain simulated by
BDP model at the middle of sand-3 layer during free field response of K-building.
Time histories, and acceleration response spectra of horizontal acceleration on roof
and first floor of K-building computed for 100 gal input acceleration.
- page 17 -
Time histories of relative displacement on roof and first floor of K-building
computed for 100 gal input acceleration.
Computed distribution of maximum acceleration and relative displacement versus
elevation of K-building (100 gal).
Computed distribution of maximum bending moment and shear force versus
elevation of K-building (100 gal).
Time histories, and acceleration response spectra of horizontal acceleration at depth
6.9 and 7.6 m for K-building (100 gal).
Time histories of relative displacement at depths 0, 6.9, 7.6, and 10.8m for K-
building (100 gal).
Time histories of excess pore-pressure computed at the middle of sand layers 1, 2, 3
and 4 for K-building (100 gal).
Time histories of vertical effective stress computed at the middle of sand layers 1, 2,
3 and 4 for K-building (100 gal).
Time histories, and acceleration response spectra of horizontal acceleration on roof
and first floor of K-building computed for 407 gal input acceleration.
Time histories of relative displacement on roof and first floor of K-building
computed for 407 gal input acceleration.
Computed distribution of maximum acceleration and relative displacement versus
elevation of K-building (407 gal).
Computed distribution of maximum bending moment and shear force versus
elevation of K-building (407 gal).
Time histories, and acceleration response spectra of horizontal acceleration at depth
6.9 and 7.6 m for K-building (407 gal). I
Time histories of relative displacement at depths 6.9 and 7.6 m for K-building (407
gal).
Time histories of excess pore-pressure computed at the middle of sand layers 1, 2, 3
and 4 for K-building (407 gal).
Time histories of vertical effective stress computed at the middle of sand layers 1, 2,
3 and 4 for K-building (407 gal).
- page 18 -
rr,z##,zs rt,z,,ztI
L3
Figure 1. Geometry of finite element model for pile-group.
- page 19 -
Cf C\1 -
0
0 c:z
mean pressure p (kPa) mean pressure p (kPa)
Figure 2. Shear-stress versus mean-pressure and shear-stress versus shear-strain
simulated by BDP model during cyclic simple shear test on Nevada Sand (Lxt/ 0= 0.232 and 0.193, = 55 kPa).
- page 20-
C)
20 40
mean pressure p (kPa)
strain €12
! I
oressure p (kPa) C)
Sr
V tr
C,
Tr
—iO —5xiO 4 0 5x10 4 10
strain €12
Figure 3. Shear-stress versus mean-pressure and shear-stress versus shear-strain
simulated by BDP model during cyclic simple shear test on Nevada Sand
(At/ 0= 0.154 and 0.128, y= 55 kPa).
-page2l -
0
0 0 4-
Cl) 0 (I)
- 4-- Cl)
10 Ca
Cl)
0
Figure 4. Experimental and simulated liquefaction strength curves for Nevada sand
during cyclic simple shear tests at 55 kPa
- page 22 -
Figure 5. First four natural modes of pile group for test No.6.
t'J Centrifuge Test No.08 uode No.4 Fraq ?.025 Oçarmode dot. I Ornox -- L222C-01
x,xx,otIox x.00OC,O ...J 2.IO
Figure 6. First four natural modes of pile group for test No.8.
0 0
0 0
.=' N a
C 0
U
0 N
10 20
time (sec)
0 0 Ln
0 0 0 N
0 o 0 '- 0 C if) 0 -
S U 0
0 0 In
due
frequency (Hz)
Figure 7. Time history and acceleration response spectrum of input horizontal
acceleration for test No.6 (1.g = 1000.gal).
- page 25 -
C 0
- ANALYSIS
C 0 - (N
01
C 0
C) C) U U o 0
0 (N
0
0 0
TEST
0 0
10
0 0 In (N
0 0 0 (N
0 0
0 C In 0 - 0 1.
() 0
0 0 In
onalysis test
10: 20
time (sec)
frequency (Hz)
Figure 8. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L9 for test No.6 (l.g = 1000.gal).
- page 26 -
0
I V
OM
C
(N - analysis
C -- test 0 0 (N
0 0' o
0 C tfl 0'-
0 '- 0
0 0
C C
0 L
0.1
ANALYSIS
0 0
TEST
h, ii
?
4 k~v~
10 20
time (sec)
frequency (Hz)
Figure 9. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L15 for test No.6 (l.g = 1000gaI).
- page 27 -
0 0
0 0 LU N
0 0 0
_ N 0
0 c LU 0- 0 - 0
U 0
0 0 LU
0 0
ANALYSIS 0 0
- N C 01
0
G) a) U
0 N
0 0
TEST
J~ i i1ji!
10 20
time (sec)
onolysis -- test
I'
'I
0- 0.1
frequency (Hz)
Figure 10. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at pile cap for test No.6 (l.g = 1000.gal).
- page 28 -
I
0 0
0 0 N 0
cDn
0
CD C-)
0 N
0 0
i0
10 20
time (sec)
41
0 0 0 N
0 C 0
0 C Lfl 0 - 0 '- 0
So 0
0 0LO
frequency (Hz)
Figure 11. Time history and response spectrum of input horizontal acceleration for test
No.4 (1.g = 1000.gal).
- page 29 -
C C
ANALYSIS
0 0
TEST 0 0 (N
:.
C U U C 0
0 IN
fo 20
time (sec)
- analysis I 1
-- test 1
C 0 0
0 c if)
o
0 0 if)
WE F
frequency .z)
Figure 12. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L9 for test NoA (1.g = 1000.gal).
- page 30 -
0 0
10
C C to (N
0 0 0 (N
ANALYSIS
C C
0 0
- (N 0
C 0
0 0
TEST
20
time (sec) 0 0 C!)
C')
0 © 0 (N
0 o ©
0 C LU 0 -
0
U 0
0 0 C!)
- analysis -- test
OR
frequency (Hz)
Figure 13. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L15 for test No.4 (1.g = 1000gal).
-.page3l -
0 0
0 0
TEST
H'
11wtmj
10• 20
0 0 (N
0 0 0 (N
0 0
O In 0W-
0 0
— 0
C-, 0
0 0 tiq
time (sec)
0 '-- 0.1 1 10
frequency (Hz)
Figure 14. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at pile cap for test No.4 (1.g = 1000.gal).
- page 32 -
0 0
C 0
0 .10 20
(sec)
C 0 (N
0 0
- U•)
01
00
C o 0 0) ci) 0
So
C tO
0.1
1 10
frequency (Hz)
Figure 15. Time history and acceleration response spectrum of input horizontal
acceleration for test No.8 (Lg = l000.gal).
- page 33 -
ANALYSIS
LI 0
TEST
0 10 20
time (sec)
0 0 (N
- onaysis test
0 0
-' In 0
C 00
0 00 ci) ci) 0 0 0 0
0 (0
ra
0.1 1 10
frequency (Hz)
Figure 16. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L9 for test No.8 (l.g = 1000.gal).
- page 34 -
0 0 0 (N
0 0 (0
o - 0'
C 0 0 0
a) a) U 0
0 0 (C)
0 0
-' 0 0
C 0
Q) 0 o (N
0 0
ANALYSIS
0 0
0
0' (N
C 0
a) o (_) 0 o
0 0
TEST
0 10 20
time (sec)
- onolysis test
frequency (Hz)
Figure 17. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L15 for test No.8 (l.g = 1000.gal).
-page 35-
A
0
-' 0 0 (N
C 0
a) a)
0 (N
0 0
ANALYSIS
TEST
0 10 20
0 time (sec)
0 0 (N
- - anolysis - -- test
1 10
frequenc: (Hz)
Figure 18. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at pile cap for test No.8 (1.g = 1000.gal).
page 36 -
0 0
- In o
C 00
0 00 a) — a) 0 0
0 In
me
0 0
0 _ 0 - N 0
C 0
0 0
0 N
0 0 tO
10 20
time (sec)
0 0 0 N
0 0
0
C o 0
*ZI 0
O 00 C- 0 0 0 0 0 0
0 If)
no -
1 10
frequency (Hz)
Figure 19. Time history and response spectrum of input horizontal acceleration for test
No.9 (l.g = 1000.ga1)
- page 37 -
A
0
0 0
0 0
- (N 0 0'
C 0
C,
C) U 0
0 0 (N
0 0
TEST
'U 10 20
0 time (sec)
0 0 (N
- Oflolysis test
0 0
- U) 0
C o 0
0
C)
C) 0 C)
0 0 U)
I
10
frequency (Hz)
Figure 20. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L9 for test No.9 (1.g = 1000.gal).
- page 38 -
C L)
ANALYSIS
0 0
- C') 0 0'
C 0
0) 0) 0
C C')
0 0
I0
10 20
0 0 0 N
- onoysis -- test
time (sec)
S 0 0
- U) 0• 0'
C 0 0 0
4) 0) 0 0 0
0 LI)
TEST
0k-. 0.1
frequency (Hz)
Figure 21. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L15 for test No.9 (Lg = 1000.gal).
- page 39 -
C C It
0 0 - (N 0
C 0
a) 0
0 (N
ANALYSIS
0 0
0 0 ::'
0
C 0
a) a) 0 - 0
0 0 (N
TEST
0 0
0
10 20
.00 0 0 LO
0 o
C 00
0 2° a) —a) 0 0
0 0 LI•)
time (sec)
- analysis test
/
frequency (Hz)
Figure 22. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at pile cap for test No9 (Lg = 1000gal).
-page 40-
d
(N
C 0
d
(N
0
0 10 20
time (s)
O 0
0
frequency (Hz)
Figure 23. Computed time histories and acceleration response spectra of horizontal
acceleration at L9 and L15 for free field of test No.8 (1.g = IOO(igal).
- page 41 -
L
p iti:1;\'tc I ,
-
-
~-j
tL 1 -
bottom -- PP2
PP4 .. ground surface
-I .. . .. . . . 0 on
10
20
time (s)
I
Figure 24. Computed time histories of excess pore pressure at PP2 and PP4 for test No.8
- page 42-
L3
L15 -
C a) E 04 0
a) 0 0 U,
D 0 a) > o (N
-i3O '- 0
11
10 20 30
time (s)
Figure 25. Computed time histories of relative displacement at L3, L9, and L15 for test
No.8.
— page 43 —
L9
d
-.' (N
0
2° a) C) U
Q
0
115
E! TTT 5 10 15 20 25
time (s)
Ui
d 1
- input . L9 L15
- s
I
0.1 1 10
frequency (Hz)
Figure 26. Computed time histories and acceleration response spectra of horizontal
acceleration at L9 and L15 for free field of test No.9 (l.g = 1000.gal).
-page 44-
I
_I
boto-..-.. -- FF2 -. -. FF4
ground surface
...-r•... -I.-..--(-•, ....
5 10 15 20 25
S time (s)
Figure 27. Computed time histories of excess pore pressure at PP2 and PP4 for test No.9
-page45 -
L9
0
E (N
E C)
CN
:;
L15
S
5 10 15 20 25
:ime (s)
Figure 28. Computed time histories of relative displacement at L3, L9, and L15 for test
No.9.
- page 46 -
ANALYSIS
0 0
0 =. C o c'J 0'
C 0
C) C)
00 (N
C 0
TEST
5 10 15 LU
time (sec) 0 Ii) (N
0 0 0 (N
0 0 0 '- 0 C 0- 0 '- 0
(•) 0
0 0 In
- dnciyss - - - test
0 - 0.1 El
frequency (Hz)
Figure 29. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L9 for test No.8 (1.g = 1000.gal).
- page 47 -
C ' 0
(N
0 0
0
0 (N
0 0
0 0 If, (N
0 0 0 (N
0 0.. 0
0 C LI) 0-
0
2o 0. 0 0
0 0 a-,
'IT
ANALYSIS
TEST
0 o (N
0
C 0
Q)
C-, 0 0 (N
0 0
0 0 Nf-
5 10 15 20 25
- analysis test
I
frequency (Hz)
Figure 30. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L15 for test No.8 (1.g = 1000.gal).
-page 48-
0 0
0 :' o
(N o
C 0
a) a)
0 (N
0 0
0 0
C .= 0
(N o
C 0
a) a)
0 0 (N
0 0
ANALYSIS
TEST
5 10 15 20 25
:tme (sec)
- - onaIyss - - -. test
1 10
frequency (Hz)
Figure 31. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at pile cap for test No.8 (1.g = 1000.gal).
- page 49 -
0 0 Ln (N
0 0 0 (N
o 0
C to 0 -0 L0
0 0
0 0 to
10 15
time (sec)
Figure 32. Computed and experimental time histories of relative displacement at L3, 1-9,
and L15 for test No.8.
- page 50-
PP PP2 PP3 Pp4
NAL.YSIS
jA4
5 10 15 20 25
PP2 PP3 • TEST
5 10 15 20 25
time (sec'
Figure 33. Computed and experimental time histories of excess pore-pressure at PPI,
PP2, PP3 and PP4 for test No8.
-page5l -
0 0
0 171
0
C)
0 0 N
0 0 LU N
0 0 0 N
0 o0
0 C LU 0 - 0 '- 0
0 0
0 0 LU
ANALYSIS 0
-0 0
C 0
0 0 N
TEST
Hi 10 20
time (s'
- onolysis test
0 0.1
LIM
frequency (Hz)
Figure 34. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at L9 for test No.9 (1.g = 1000.gal).
- page 52 -
0 0 LN
ANALYSIS
0 0
TEST
10 20
time (s)
analyss
I
-- test
0 0.1 IN
frequency (Hz)
Figure 35. Computed and experimental time histories, and acceleration response spectra
- of horizontal acceleration at L15 for test No.9 (l.g = 1000.gal).
-page 53-
C C CN
0 (C)
0 0 0 04
0' 0
S.-
ctfl 0- 0 '-0
0 0
0 0 LU
0 0 N
0
C 0
C) C) 0 0 0
0 0 N
ANALYSIS
TEST
I
10 20
time (s) 0 0 LO N
0 0 0 N
0 0 '- 0 c ti) 0- 0 '- 0
0 0
0 0
- analysis test
0.1 1 10
frequency (Hz)
Figure 36. Computed and experimental time histories, and acceleration response spectra
of horizontal acceleration at pile cap for test No.9 (l.g = 1000.gal).
- page 54
time (s)
Figure 37. Computed and experimental time histories of relative displacement at 0, L9, and L15 for test No.9.
- page 55 -
S.'
F ppi
NALYSII if
a)
a.
0 'U 20
time (s)
_PP1 -- PP2
P?3 TEST PP4
S.'
a)
a. 0
10 20
time (s)
Figure 38. Computed and experimental time histories of excess pore-pressure at PPI,
PP2, PP3 and PP4 for test No.9..
- page 56 -
levation (m)
) -3.9 -6.9 -7.6 10.8 -9.5 15.95
23.3
29.85
-53-5
sand-i sand-2 sand-3 sand4
earthquake input motion
Figure 39. Geometry of finite element model for K-building.
- page 57 -
0.5
0.4
0.1
0
1 10 100 1000
Number of cycles
—a----- Sand-i (test)
0 Sand-i (analysis)
Sand-2 (Kajima)
0 Sand-2 (analysis)
Sartd-3 (test)
o Sand-3 (analysis)
-A- Sand-4 (test)
£ Sand-4 (analysis) I
S
Figure 40. Experimental and simulated liquefaction strength curves for sands 1, 2, 3 and
4 during cyclic simple shear tests.
- page 58 -
K-BuIdinq elQenoolue on&ysis
K-9uiIdng egenvoIue anolvss 4o4n No.6
- 2.619 62çowoodo 404.4
- 2.665C-02
K-Buildirg elgenvolue CnOIys iOCe NO.4 Feq - 2.t9 U.ço.000dn 404. 0— - 3.4724
Figure 41. Fundamental modes of K-building
- -page59-
0 0
0 10 20 .30 60
TIME ('.cc)
I
0.1 1
RE0U€NC (Hz)
Figure 42. Time history and acceleration response spectrum of input horizontal
acceleration for K-building (l.g = 1000.gal).
- page 60 -
0 0
0 0 c'4
C 0
C) C) 0 oO
0 ('4
0 0
40 0 20 40
time (sec)
0 20
time (sec)
roof first floor
0 0 0
0 0
—0
C 00
0 00
c'J C) U 0 00
0 0
S '.1 1
frequency (Hz)
- 0— IC 0.1
frequency (Hz)
IsJ
Figure 43. Time histories, and acceleration response spectra of horizontal acceleration
on roof and first floor of K-building computed by linear analysis (1.g
1000. gal)
- page 61 -
roof
20 40
time (sec)
first floor
20 40
time (sec)
Figure 44. Time histories of relative displacement on roof and first floor of K-building
computed by linear analysis.
- page 62 -
I 0(
8C
0
2 4 6 8
Max. displa cement (clu)
-40
0
EPZoX!I
100.0
80.0
60.0
20.0
0.0
-20.0
/ .400
0.00 100.00 200.00 300.00 400.00 500.00
Max. acceleraUon (ga!)
E
40.0
0
Figure 45. Distribution of maximum acceleration and relative displacement versus
- elevation computed in linear analysis of K-building (1.g = 1000.gal).
- page 63 -
120 120
100.0 100
80.0 80
60.0 0 C,
E
Z 60 0
Ron
20.0 20
0.0 01
0 0.0 2000.0 4000.0 6000.0
Max. shear force ((on)
5000000 10000000 15000000
Max. bendtng moment (ton an)
Figure 46. Distribution of maximum bending moment and shear force versus elevation
computed in linear analysis of K-building.
- page 64 -
6
0 20 40
time. (s)
20 . 40 60
time (s)
6 GL=-10.8 (m) -
0
- GL=-53.5m) - -- GL-15.95rn) -
GL=-10.8m) I ' - CL=0.Om) j \
i I' ......... i % t
o 20 40
L
time (s) 0.1
1 10
frequercy (Hz)
6' U
Figure 47. Time histories, and acceleration response spectra of horizontal acceleration at
depth 0, 10.8, and 53.5 m for the free field of K-building (1 g = 1000.gal).
- page 65 -
GLOm -- GL —6.9m
CL —7.6m CL —10.8m
0 20 40 60
time (s)
Figure 48. Time histories of relative displacement at depths 0, 6.9, 7.6, and 10.8m for
free field of K-building.
- page 66 -
0 10
0 20 40 Awl
time (s)
Figure 49. Time histories of excess pore-pressure computed at the middle of sand layers
1, 2, 3 and 4 for free field of K-building.
- page 67 -
99
'0
10 20 30 40 50
men-pressure (iPo) -.
0 5 10
shear strom
Figure 50. Shear-stress versus mean-pressure and shear-stress versus shear-strain
simulated by BDP model at the middle of sand-2 layer during free field
response of K-building.
- page 68 -
0 20 40 60
meon pressure (kPa)
shear strain y
Figure 51. Shear-stress versus mean-pressure and shear-stress versus shear-strain
simulated by BDP model at the middle of sand-3 layer during free field
response of K-building.
- page 69 -
0 0
-0 0 C4
C 0
a) a)
0 0 CN
0 0
0 20 40
time (sec) 0 0 - 01
0 0
-S. 0 -at')
0 0
a) 0 0
0 0 0
ii
C C
roof
C C
0
0 0
0 0
60
0 0 0
0 0 C r')
0 0 0 r'J
0 0 0
Ere I
frequency (Hz)
frequency (Hz)
Figure 52. Time histories, and acceleration response spectra of horizontal acceleration
on roof and first floor of K-building computed for 100 gal input acceleration
(1.g = 1000.gal).
- page 70 -
roof
E -0
0 20 40
time (sec)
E n first floor
0 20 40
time (sec)
Figure 53. Time histories of relative displacement on roof and first floor of K-building
computed for 100 gal input acceleration.
- page7l -
120.0
100.0
80.0
60.0
E
Z 40.0
0
20.0
0.0
-20.0
0.00 100.00 200.00 300.00 400.00 500.00
Max. acceleration (gal)
itt,.
100.
R0.(
60.0
E
Z 40.0
0
20.0
0.0
-20.0
40.0
0 1 2 3 4 5 6
Max. displacement (Cm)
Figure 54. Computed distribution of maximum acceleration and relative displacement
versus elevation of K-building (100 gal).
- page 72 -
120.0
100.0
80.0
E
60.0
0
40.0
20.0
0.0
120.0
100.0
80.0
40.0
20.0
0.0 0.OE+0 2.0E+6 4.OE+6 6.OE+6 8.0E+6 1.OE+7 1.2E+7
Max. bending moment (ton cm)
0.0 :c0.0 20010 3000.0 4000.0 5000.0
Max. s:%ear force (ton)
E
Z 60.0
0
Figure 55. Computed distribution of maximum bending moment and shear force versus
elevation of K-building (100 gal).
- page 73 -
0 0 (N
0 0 (N
0 0 (N
20 40
time (sec' 0 0 CO
0 0 aD
C 0
0 0
a) 0 0
0 0 N
- depth=6.9 m - .. depth=7.6
0. Ii
frequency (Hz)
Figure 56. Time histories, and acceleration response spectra of horizontal acceleration at
depth 6.9 and 7.6 m for K-building (100 gal) (1.g = 1000.gal).
- page 74 -
F II lL't It
tI' • r
' 1111
IIII
liii It U. jr
%l •'
I It
lf
- depth=6.9 m depth=7.6 m
20
time (sec)
Figure 57. Time histories of relative displacement at depths 6.9 and 7.6 m for K-
building (100 gal).
-page 75-
0 0
1,
V
I
- - -1.-
sand-4 - -- sand-3
sand—i
0)
0 0 0
LI)
20
time (sec)
Figure 58. Time histories of excess pore-pressure computed at the middle of sand layers
1, 2, 3 and 4 for K-building (100 gal).
-page 76-
I X~Vi* P~
I
- sand-4 -- sand-3
sand-2 sand—i
20
time (sec)
Figure 59. Time histories of vertical effective stress computed at the middle of sand
layers 1, 2, 3 and 4 for K-building (100 gal).
- page 77 -
C C C
0 0 0
0 0'
C 0 0 - 0 a,
0
0 0.1
frequency (Hz)
roof
0 0 LU
0 0 LU
0 0 0
20 40
time (sec)
0 0 0
0 0 0
0 . 0.1 1 Ic
frequency (Hz)
Figure 60. Time histories, and acceleration response spectra of horizontal acceleration
on roof and first floor of K-building computed for 407 gal input acceleration
(l.g = 1000.gal).
-page 78-
20 40
(N 0
time (sec)
first floor
0
20 40
time (sec)
Figure 61. Time histories of relative displacement on roof and first floor of K-building
computed for 407 gal input acceleration.
- page 79 -
60.0
E
100.0
80.0
Z 40.0
C,
20.0
120.0
-40.0
0.00 200.00 400.00 600.00 800.00 1000.00
Max. acceleration (gal)
120.
100.0
80.0
60.0
E
E 40.0
C,
20.0
0.0
-20.0
en a
0 5 10 15
Max. displacement (Cm)
Figure 62. Computed distribution of maximum acceleration and relative displacement
versus elevation of K-building (407 gal) (l.g = 1000gal).
- page 80 -
80.0
Z 60.0
C
40.0
20.0
5.OE+6 1.OE+7 1.5E+7 2.OE#7 2.5E+7
Max. bending moment (ton-cm)
120.0
100.0
80.0
E
a
60.0
40.0
20.0
0.0
0.OE+0 0.0 2000.0 40000 6000.0 8000.0 10000.0 12000.0
Max. shear force (ton)
1200
100.0
Figure 63. Computed distribution of maximum bending moment and shear force versus
elevation of K-building (407 gal).
- page 81 -
0 9
0 0 0 N
20 40
- depth=6.9 rn - - depth=7.6 m
0 0
0 0
- N 0 01
GL=-6.9 (m)
0 0
GL-7.6 (m)
0 - 0.1
frequency (Hz)
Figure 64. Time histories, and acceleration response spectra of horizontal acceleration at
depth 6.9 and 7.6 rn for K-building (407 gal) (Lg = 1000gal).
- page 82 -
tI -1 -4
(1•'
jj4t1.1
I / ' 1
depth=6.9 m depth=7.6 m
20
time (sec)
Figure 65. Time histories of relative displacement at depths 6.9 and 7.6 m for K-
building (407 gal).
- page 83 -
- sand—i .; - - sand-2 -- SQflcj-3 * ...' ..
I/; i Ih jflhl H l 4 L ' ' k — TI
t I
114
t('IA1/ft':;
20
NO
time (s)
Figure 66. Time histories of excess pore-pressure computed at the middle of sand layers 1, 2, 3 and 4 for K-building (407 gal).
- page 84 -
20
0 0
C') CO a) L0
Ln (I)
a) > I . 00
0 4- a)
0 0 0
Ln
a) >
0 0 (N
::z
Zz
sand—i sand-2 sand-3 sand-4
. 1
we
0 LC
time (s)
Figure 67. Time histories of vertical effective stress computed at the middle of sand
layers 1, 2, 3 and 4 for K-building (407 gal).
- page 85 -
LIST OF TABLES
Summary of properties of three- and two-dimensional pile group.
Three- and two-dimensional properties of pile for analysis of centrifuge tests.
Three-dimensional and two-dimensional properties of pile cap.
Summary of material properties for Nevada sand used in analysis.
Variation versus depth of horizontal and lateral effective stress and elastic
properties in tests No.6 and 4 on dry Nevada sand.
Variation versus depth of horizontal and lateral effective stress and elastic
properties in tests No.9 and 8 on saturated Nevada sand.
BDP model constants for Nevada sand used in analysis.
Two- and three-dimensional properties of piles of K-building.
Two- and three-dimensional properties of basemat and superstructure of K-building.
Material properties for soil layers of K-building.
BDP model constants for soil layers of K-building.
I
- page 86 -
Table 1. Summary of properties of three- and two-dimensional pile grouP.
3D PROPERTIES 2D PROPERTIES(input data)
PART E 6 unit mass Ref E Unit mass Poisso
kPa kPa ton/m3 kPa ton/m3 ratio
pile 2.O0E08 8.30E07 1.91E00 I 1.43E07 1.91E00 0.2
pile cap part A (steel) 2.O0E08 8.30E+07 7.67E00 1 2.00E08 5.90E+00 0.2(
part B (Al) 7.00E+07 2.56E07 2.74E00 I
part C (Al) 7.00E07 2.56E07 2.74E00 1
Aluminium block 700E07 2.56E07 2.74E00 1
Reference: E.P. Popov, Mechanics of Materials, Prentice-Hall. 1978, p.570.
Table 2. Three- and two-dimensional properties of pile for analysis of centrifuge tests.
pile in 3D - - pile in 2D
external diameter (m) 0.48 width (m) 0.48
thickness(m) 0.0127
pile length(m) 10.73 length(m) 10.73
pile cross-section area(m2) 1.86E-02 cross-section area(m2/m) 0A8
inertial (m4) 5.52E-04 inertial (m4/m) 922E-03
Young's modulus E (kPa) 2.00E08 YOuflgS modulus E (kPa) 1-13E07
pile EA(kN) 3.73E+06
pile EI(kN-m2) L10E05
steel unit mass (t/rn3) 7.67E00 unit mass (t/rn3/m) 1.91E00
pile mass (ton) 1.53E00
soil unit mass (t/rn3) 1.97
total depth (m) 1.67 total depth (rn) 1.00E-00
average mass per unit depth (ton/rn) 9.85E00 mass per unit depth (ton/rn) 9.85E00
average El per unit depth((kN-m2/rn) 1.32E05 El per unit deptli((kN-m2/rn) 1.32E05
average EA per unit depth (kN/m) 4.47E06 EA per unit depth (kN/m) 688E06
Note: El and mass per unit length are preserved, but EA is not preserved
total depth 20
- page 87 -
Table 3. Three-dimensional and two-dimensional properties of pile cap.
I 3D pile cap I 2D pile cap I
part A (circular) diameter Cm) 3.840
height Cm) 0.690 unit mass (ton/m3) 7.670
mass (ton) 61.291 part B (circular) diameter(m) 2.100
height.(m) 0.370 unit mass (ton/m3) 1.570
mass (ton) 2.012
part C (square) width (m) 3.270
height Cm) 0.520
unit mass (ton/m3) 2.735
mass (ton) 15.207 Alum Thium block (square) width Cm) 1.260
height Cm) 1.260
unit mass (ton/m3) 2.735
mass (ton) 5.471
total depth Cm) 1.67
vertical center of gravity Cm) 1.109
horizontal center of gravity (m) 0.061 average.mass per unit length (t/m) 50.29
iidth (m) 3.8W-00 eight (rn) 2:22E00 nit mass Ct/m3/m) 5.90E+00 iass per unit length (t/rn) 5.03E+01 ccentricity of center Qf qravity (m)6.06E-02
Alu1inium block 3D pile cap
A
C
3D eccentricity
2D pile cap
2D
- page XX -
Table 4. Summary of material properties for Nevada sand used in analysis.
- test number 8 9 6 4
void ratio 0.697 0.697 0.791 0.831
porosity 0.411 0.4 11 0.442 0.454
relative density (%) 5l.l 77 51.177 22.366 10.345
Poisson ratio 0.330 0.330 0.330 0.330
saturated unit mass (tim3) 1.978 1.978 1.927 1.907
dry unit mass (tim3) 1.568 1.568 1.485 1.453
buoyant unit mass (t1m3) 0.978 0.978 0.927 0.907
permeability (m/s) 1.50E-03 1.50E-03 1.50E-03 1.50E-03
Yours modulus CoflStaflt 1.35E04 135E+04 1-20E+04 1.14E04
CoefficientofearthpreSsUre 0.493 0.493 0.493 0.493
water unit mass (t/m3) soil specific gravity
maximum dry unit weight minimum dry unit weight
minimum void ratio maximum void ratio
MWIIIIII 2.660
16.970 kN/m3 13.980 kN/m3 0.5 36
9:865 .
reference: R.F. Scott, letter to Kajima. Caltech, 30 October 1990 pp.1-5
RE. Scott & B. Huslimand, letter to Kajima. Caltech, 16 November 1990.
Note: Permeability coefficieti is multiplied by 50 to account for centrifuge scaling
reference: Kajima report. February 1991. p.48&69
- page 89 -
Table 5. Variation versus depth of horizontal and lateral effective stress and elastic
properties in tests No.6 and 4 on dry Nevada sand.
tp't f
depth (m)
vertIcal elf stress (kPa)
lateral elf stress (kPa)
Youngs modulus E (kPa)
Shear modulus G (kPa)
0.00 437 2.15 2.50E01 9.40E03
0.60 17.34 8.54 4.98E0'1 1.87E-04
1.78 34.53 17.01 7.03E,04 2.64E04
2.96 51.28 25.26 8.56E+04 3.22E+04
408 65.92 32.47 9.71E04 3.65E+04
4.97 79.69 39.25 1.07E05 4.01E+04
5.97 94.69 46.64 1 . 1 6E05 4.37E+04
7.03 110.14 54.25 1.25E+05 4.72E04
8.09 126.45 62.28 1 .34E05 5.05E04
9.27 143.64 70.75 1.43E+05 5.39E04
10.45 152.24 . . 74.98
i-eference:Kajima report February 1991. pp.51-52
note: stresses are evaluated at middle of layer, except for the last row that is evaluated at bottom
depth(m)
vertical elf stress(kPa)
lateral elf stress(kPa)
Youngs modulus E(kPa)
Shear modulus 6(kPa)
0.00 4.28 2.11 2.35E-04 8.83E03
0.60 16.96 8.35 '1.68E04 1.76E04
1.78 33.78 16.64 6.61E04 2.48E04
2.96 50.17 24.71 8.05E-04 3.03E04
4.08 6450 31.77 9.13E~04 3.43E04
4.97 77.97 38.40 1 .00E+05 3.77E+04
5.97 92.65 45.63 I .09E05 4.11 E+04
7.03 107.76 53.08 1.1 8E05 4.'13E+04
8.09 123.73 60.94 1.26E05 '1.75E04
9.27 140.55 69.22 1.35E05 5.06E04
10.45 148.96 73.37
reference: Kaj ima report, February 199 I, pp.51-52
note: stresses are evaluated at middle of layer, except for the lait row that is cvilj;itcd at bottom
- page 90 -
Table 6. Variation versus depth oi horizontal and lateral elfective stress and elastic
properties in tests No.9 :nd 8 on saturated Nevada sand.
vertical elf lateral ef f Young's modulus Shear modulus
depth (m) stress (kPa) stress (kPa) E (kPa) G (kPa)
0.00 2.88 1.42 2.29E04 8.61E03
0.60 11.42 5.63 4.56E04 1.71E04
1.78 22.75 11.20 6.43E04 2.'12E04
2.96 33.79 16.64 7.84E+04 2.95E04
4.08 43.43 21.39 8.89E+04 3.34E+0-1
4.97 52.51 25.86 9.78E04 3.68E04
5.97 6239 30.73 1 .07E05 4.01 E04
7.03 72.57 35.74 1.1 5E05 4.32E04
8.09 83.32 41.04 1.23E05 4.63E04
9.27 94.64 46.62 1.31E05 4.93E+04
10.45 100.31 49.41
reference: Kajima report. February igi, pp.51-52 note: StressèS are evaluated at middle of layer, except for the last row thai is evaluated at bottom
vertical elf lateral eff Young's modulus Shear modulus
depth (m) stress (kPa) stress (kPa) - E (kPa) 6 (kPa)
0.00 2.88 1.42 2.29E,04 8.61E03
0.60 11A2 5.63 4.56E04 1.71 E04
1.78 22.75 11.20 6.43E,04 2.42E04
2.96 33.79 16.64 7.84E04 2.95E04
4.08 43.43 21.39 8.89E04 3.34E-04
4.97 52.51 25.86 9.78E04 3.68E+04
5.97 6239 30.73 1 .07E05 4.01 E04
7.03 72.57 35.74 1.15E05 4.32E04
8.09 83.32 41.04 1.23E+05 4.63E04
9.27 9464 46.62 1.31E05 4.93E04
10.451 100.31 49.41
reference: Kajima report, February 19l, pp.51-52 note: streseS are evaluated at middle c layer, except for the last row that is evaluated at bottom
- page9l -
Table 7. BDP model constants for Nevada sand used in analysis.
definition notation values unit
constant for elastic Youngs modulus E0 8000. kPa
exponent for elastic Youngs modulus no 0.5
constant Poisson ratio v 0-3
slope of failure surface a 0-6
apparent cohesion c 1. kPa
slope of characteristic line 0 0.63
attractive pressure Pa 70. kPa
constant for mpnotonic plastic modulus .. H0 0.1
exponent for nionotonic plastic modulus n 1.6
constant for cyclic plastic modulus Hu 0.8
exponent for cyclic plastic modulus flu 4.
I
I
I
- page 92 -
Table 8. Two- and three-dimensiorzal properties of piles of K-building.
pile row In 3D truss In 2D
external diameter (m) 1.8 length(m) 21.3
pile length(m) 243 cross-section area(m2) 0.121
pile cross-section area(m2) 2.51E-00 mass per unit length (tim) 0.251
inertial (m4) 5.15E-01 Young's modulus E (kPa) 2.52E07
Youngs modulus E (kPa) 2.52E-07 EA (kN) 1 .O7EO7
pile EA(kN) 6.41E07 total mass(t) 6.18E-00
pile El(kN-rn2) I .3OE07 total mass on foundation (t/m) I .26E,03
concrete unit mass (t/m3) 2.300
pile mass (ton) 1.12E02
average soil unit mass (t/m3) 1.700
total depth (m) 30
numberof piles . 5
total mass on foundtion (tim) I .26EO3
average EA per unit depth (kN/m) i.O7EO7 . .
Notes: EA and mass per unit length are preserved, but bending is neglected Piles are modelled by using truss elements; they transmit only axial forces.
0 0 0 0 0
top view 2DQ
- pace 93
Table 9. Two- and three-dimensional properties of basernat and supersulicture of K-
building.
m2t
structure mass per unit depth (tim) 1319.03
number of point for mass distribution 11.00
distributed nodal load (kNim) 1201.87
width(m) 30
depth(m) . 6
total mas(t). . 10657
mass per unit depth (t/m) 355.23
unit mass (t/m3) l.92J
references: Kajima report, 1-eDruary I'z, rlg....
Notes: The mass of the superstructure is applied by using 11 applied nodal forces The inertia mass elements do not apply any gravity loads.
averaging depth (m) 30
Young's modulus (kPa) 2.52E07
total building mass per unit depth (t/m) 1349.03 number of point for mass distribution 11.00
distributed nodal load (kN/m) 1201.87
references: Kajima report, February 1992, pA5
(Kajima report, Feb 1992, Fig.3.2)
-. page 94 -
9 continued) 20 data
M ass El (t.cm2) shear El corrected average average
(t) correction for shear inertial mass
GA (t) (actor Phi (t.cm2) (m4/m) (t/m)
1.078E+06 1470 4509E+13 2.32E+03 1.942E+10 2.518E-02 49.00
1.585E+06 1096 5.908E+13 5.32E+03 1.11 IE+10 1.440E-02 36.53
1.713E+06 998 6.758E+13 5.63E03 1.200E10 1.557E-02 33.27
1.754E+06 998 7.131E13 5.80E+03 1.229E10 1.594E-02 33.27
1.769E+06 998 7.383E+13 5.96E+03 1.240E+10 1.608E-02 33.27
1.798E+06 998 7.569E+13 6.01E+03 1.260E+10 1.634E-02 33.27
1.848E+06 1056 7.822E+13 6.04E+03 1.295E10 1.680E-02 35.20
1.862E+06 1002 8.086E+13 6.20E+03 1.305E10 1.692E-02 33.40
1.867E+06 1002 8.147E+13 6.23E+03 1.308E+10 1.697E-02 33.40
1.877E+06 1002 8.193E+13 6.23E+03 1.315E10 1.706E-02 33.40
1.894E+06 1002 8.251E+13 6.22E03 1.327E10 1.721E-02 33.40
1.948E+06 1056 8.137E+13 5.96E+03 1.365E10 1.770E-02 35.20
2.087E+06 1002 8.651E+13 5.91E+03 1.462E10 1.897E-02 33.40
2.361E+06 1048 9.581E13 5.60E+03 1.712E10 2.220E-02 34.93
2.398E+06 1061 9.589E+13 5.51E03 1.739E10 2.255E-02 35.37
2.415E06 1061 9.603E+13 5.48E+03 1.751E10 2.271E-02 35.37
2.435E06 1116 9.L184E+13 5.37E03 1.766E10 2.290E-02 37.20
2.464E06 1061 9.662E+13 5.41E+03 1.787E10 2.31.7E-P2 35.37
2.511E+06 1061 1.002E14 5.50E+03 1.821E+10 2.361E-02 35.37
2.521E+06 1061 1.007E14 5.51E+03 1.828E+10 2.371E-02 35.37
2.525E06 1061 1.007E+14 5.50E+03 1.8311+10 2.3751-02 35.37
2.533E06 1116 9.920E+13 5.40E+03 1.837E10 2.382E-02 37.20
2.554E+06 1061 1.044E14 5.64E+03 1.852E+10 2.402E-02 35.37
2.568E+06 1061 1.04-1E14 5.61E+03 1.862E10 2.415E-02 35.37
2.61 1E-06 1061 1.044E+14 5.51E+03 1.893E+10 2.455E-02 35.37
2.694E+06 1061 1.091E+14 5.58E+03 1.953E10 2.534E-02 35.37
2.747E+06 1133 1.093E+14 5.49E+03 1.992E10 2.583E-02 3777
2.753E+06 1054 1.102E+14 5.52E+03 1.996E+10 2.5891-02 36.13
2.760E06 1084 1.10'IE+14 5.52E+03 2.00IE+10 2.596E-02 36.13
2.787E+06 1054 1.102E14 5.45E+03 2.021E+10 2.621E-02 36.13
2.9331+06 1084 1.093E+14 5.14E03 2.1271+10 2.7581-02 36.13
3.762E+06 1149 1.077E+14 3.95E+03 2.728E+10 3.5381-02 38.30
5.659E+06 2486 1.199E+14 1.13E-03 1.0631+11 1.3791-01 82.07
3.082E+07 3797 1.482E+l1 3.191+02 4625E+1 I 998E-01 126.57
1.000E10 5967 i.000E+17 3.33E+02 2.9911+14 3.8791+02 198.90
43301-06 5520 5.5601+14 1.90E+05 2.923E+09 3.7911-03 184.00
43301+06 374 5.560E14 3.14E+05 1.768E+09 2.293E-03 12.47
43301+06 386 5.560E+11 1.71E-05 3.257E+09 422-1E-03 12.87
4.330E+06 -1.14 5.560E14 1.71E~05 3.2571+09 422-11-03 1480
4.330E+06 526 5.560E+11 9 12E04 6.098E+09 7.9091-03 17.53
4330E06 906 5.560E+I-1 2.33E+04 2.383E-10 3.0911-02 30.20
43301+06 1204 5.560E+14 2311+04 2.402E+10 3.1151-02 40.13
43301+06 1175 5.560E+14 257E+04 2.166E10 2.8091-02 39.17
t1.330E06 I I4 5.560E14 2.57E+04 2.166E10 2.8091-02 38.20
4.330E-06 I I- 5.560114 2571+04 2.166E10 2.8091-02 38.20
3D data (1
height (cm)
I .045E-O
9.985E+0
9.695E+0
9.-105E+O
9.1 l5E+0
8.825E+0:3
8.5 35E + 03
8.245E+03
7. 955E +03
7.665E+03
7.3 75E +03
7.085E03
6.795E03
6.505E03
6.210E+03
5.9 1 5E+03
5.620E03
5.32 5E + 0.3
5.030E+03
4.7 35E +03
441OE+03
4.1 45E+03
3.850E+03
3.555E+03
3.260E+03
2.965E+03
2.670E+03
2.375E+03
2.080E +03
I .785E+03
I .190E03
1.195E+03
9.000E02
4250E+02
0.000E+00
-6.000E+02
-6.900E +02
-7.600E 02
-8.550E +02
-9.500E +02
- I .Ot3OE+03
- I .337E+03
-i.595E+03
-1 .840E03
-2.085E +03
-2.330E 03
- pagc 95 -
Table 10. Material properties for soil layers of K-building.
total unit permeability Poisson fficient
- mass (m/s) ratio earth 1pre
material depth (m) (t/m3) _________ ssure
9 sand-I 0 1.7 6.90E-06 0.3 0.43
8 clay-i -3.9 1.6 1.00E-08 0.48 0.92
7 sand-2 -6.9 1.7 690E-06 0.3 0.43
6 clay-2 -7.6 1.6 1.00E-08 0.48 0.92
5 sand-3 -9.5 1.8 6.90E-06 0.3 0.43
4 sand_Li -10.8 1.9 2.60E-05 03 0.43
3 clay-3 -15.95 1.7 LOOE-08 0.48 0.92
2 eiastic-1 -23.3 2.0 0.3
1 -29.85 1.9 0.3 ela
!~j
-53.5
material sand-1 clay-i sand-2 clay-2 sand-3 sand-4 clay-3 elastic-I
eiastic-2
Specific gravity
2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66 2.66
void ratio 1.371 1.767 1371 1.767 1.075 0.844 1.371 0.660 0.844
porosity 0.578 0.639 0.578 0.639 0.518 0.458 0.578 0.398 0.4581
mises c (kPa)
15.277
19.237
75.252
shear modulus
(t/m2) 4170 1045 1910 2760 5625
20480 9180
41330
41330
Youngs modulus (kPa)
1.06E05 3.03E04 4.87E04 8.01E04 1.43E+05 5.22E+05 2.66E+05 1.05E06 1.05E06
layer 9 8 7 6 5 4 3 2 1
Notes: All layers are assumed saturated
layer material
vertical eff stress
(kPa)
hcizontal efI stress (Pa)
friction
angie (deg)
shear strength
t/m2)
9 sand-I 13. 5.7 282 0.040.64
8 clay-i 35.6 32.8 2.7
7 sand-2 46.8 20.1 21.6 0.540.58
6 clay-2 54.8 50.6 3.4
5 sand-3 65.5 28.1 30.3 1 .82 1.97
4 sand-4 93.3 40.0 1,1.3 3.585. 18
3 clay-3 141.2 130.3 13.3
2 elastiC 1
I elastic-2
17,
- p3gc 96 -
Table 11. BDP model constants for soil layers of K-building.
page 97 -