See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245411107

Kinematic pile bending during earthquakes: Analysis and field measurements

Article  in  Géotechnique · January 2001

DOI: 10.1680/geot.51.5.425.39973

CITATIONS READS

174 854

4 authors, including:

Sissy Nikolaou George Mylonakis

WSP University of Bristol
43 PUBLICATIONS   599 CITATIONS    229 PUBLICATIONS   3,452 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PREPARE: Enhancing PREParedness for East African Countries through Seismic Resilience Engineering View project

Response Analysis and Behavior Interpretation of Inelastic Structures View project

All content following this page was uploaded by George Mylonakis on 01 December 2014.

The user has requested enhancement of the downloaded file.

Nikolaou, S., Mylonakis, G., Gazetas, G. & Tazoh T. (2001). GeÂotechnique 51, No. 5, 425±440

Kinematic pile bending during earthquakes: analysis and ®eld measurements

S . N I KO L AO U , G . M Y L O NA K I S { , G . G A Z E TA S { a n d T. TA Z O H }

The passage of seismic waves through the soil surrounding a Le passage d'ondes sismiques dans le sol qui entoure une
pile imposes lateral displacements and curvatures on the pile impose des deÂplacements lateÂraux et des courbures aÁ la
pile, thereby generating `kinematic' bending moments even pile, engendrant de la sorte des couples de ¯exion `cineÂma-
in the absence of a superstructure. These moments are tiques' meÃme en l'absence d'une superstructure. Ces couples
concentrated in the vicinity of interfaces of alternating soft se concentrent dans le voisinage des interfaces de couches
and stiff soil layers and, in the case of restrained-head piles, alterneÂes de sol tendre et de sol rigide et, dans le cas de
at the pile head. The scope of this paper is threefold: (a) to piles aÁ teÃte retenue, au sommet de la pile. Cet expose a donc
critically review some existing design methods for kinematic trois objectifs : (a) passer en revue et juger certaines des
pile loading; (b) to develop new analytical results for piles in meÂthodes de conception existantes relatives au chargement
homogeneous and layered soils; (c) to present a case study in de pile cineÂmatique ; (b) deÂvelopper de nouveaux reÂsultats
which theoretical predictions are tested against ®eld mea- analytiques pour des piles dans des sols homogeÁnes et des
surements. To this end, an approximate beam-on-dynamic- sols constitueÂs de couches ; (c) preÂsenter une eÂtude de cas
Winkler-foundation (BDWF) model is implemented, speci®- dans laquelle les preÂdictions theÂoriques sont compareÂes aux
cally developed for the seismic response of piles in layered mesures prises sur le terrain. Dans ce but, nous avons utiliseÂ
soil. Both ®xed- and free-head piles, and different boundary une maquette approximative de Winkler (BDWF), deÂvelop-
conditions at the pile toe, are considered. It is shown that peÂe speÂci®quement pour eÂtudier la reÂaction sismique des
the magnitude of kinematic moments depends mainly on the piles dans des sols constitueÂs de couches. Nous eÂtudions les
stiffness contrast between the soil layers, the pile±soil stiff- piles aÁ teÃte ®xe et les piles aÁ teÃte libre ainsi que les
ness contrast, the excitation frequency, and the number of diffeÂrentes conditions limites au pied de la pile. Nous mon-
excitation cycles. A unique case history involving the instru- trons que la magnitude des couples cineÂmatiques deÂpend
mented pile foundation of a multistorey building in Japan is principalement du contraste de rigidite entre les couches de
presented. Time histories of bending and axial strains re- sol, du contraste de rigidite entre la pile et le sol, de la
corded at six locations along two piles are successfully com- freÂquence d'excitation et le nombre de cycles d'excitation.
pared with results computed from simple formulae and Nous preÂsentons une histoire de cas unique sur les fonda-
methods presented in the paper. tions de piles instrumenteÂes d'un baÃtiment de plusieurs
eÂtages au Japon. Les histoires de temps pour les deÂforma-
tions ¯eÂchies et axiales enregistreÂes aÁ six emplacements le
long de deux piles montrent une bonne correÂlation avec les
KEYWORDS: case history; dynamics; earthquakes; numerical reÂsultats calculeÂs d'apreÁs les formules et les meÂthodes sim-
modelling and analysis; piles; soil structure interaction. ples preÂsenteÂes dans cet exposeÂ.

INTRODUCTION damage with the presence of strong discontinuities in strength

Pile damage due to seismic shaking has been observed in and, especially, stiffness in the soil pro®le. The most likely
numerous post-earthquake investigations around the world (Ross cause is the relatively large curvatures imposed on the piles by
et al., 1969; Margason, 1975; CNEL-ENEL, 1976; Okamoto, the surrounding soil, as it deforms while excited by the up- and
1983; Nishizawa et al., 1984; EEFIT, 1986). Mizuno (1987) down- (after re¯ection) propagating seismic waves. The reason
reported 28 cases involving seismic failures of piles in Japan. is that soil shear strain is discontinuous across interfaces
More recently, pile damage was observed in the Loma Prieta because of the different shear moduli between the layers, and
earthquake (1989) and particularly in the Kobe earthquake thereby the associated soil curvature (the derivative of strain) is
(1995). Identi®ed or suspected causes of failure in the above in®nite. Accordingly, this type of distress is called kinematic, to
cases include: distinguish it from the inertial distress (type (b)) due to head
loading arising from the inertia forces in the superstructure.
(a) large pile movements due to liquefaction and subsequent
Tajimi (1969) and Penzien (1970) were among the ®rst
lateral soil spreading
to study the problem, by using an analytical and numerical
(b) excessive bending and shear forces transmitted to the piles
approach respectively. Following these early efforts, the
from the superstructure
problem was analysed by Margason (1975), Blaney et al.
(c) bending due to vibratory deformations induced by the
(1976), Kagawa & Kraft (1980), Flores-Berrones & Whit-
passage of seismic waves through the soil.
man (1982), Kaynia & Kausel (1982), Dobry & O'Rourke
Support for the third scenario comes from the fact that damage (1983), Barghouthi (1984), Tazoh et al. (1987), Mineiro
has often been observed too deep to have been caused by (1989), Mamoon & Banerjee (1990), Kavvadas & Gazetas
loading coming from the pile top, in soils that could not (1993), Nikolaou et al. (1995), Kaynia & Mahzooni (1996),
possibly have suffered a severe loss of strength (e.g. liquefac- Guin & Banerjee (1998) Luo & Murono (2001) and others.
tion). Analytical and ®eld evidence (Dobry & O'Rourke, 1983; Most of these studies focus on the dynamic response of the
Mizuno, 1987; Tazoh et al., 1987) have associated this type of pile head; the associated curvatures and bending moments
along the pile have received less attention. Reviews on the
subject have been presented by Novak (1991), Pender
Manuscript received 26 July 2000; revised manuscript accepted 9 March (1993), and Gazetas & Mylonakis (1998).
2001.
Despite the above research efforts, and some documented
Discussion on this paper closes 2 November 2001, for further details
see the inside back cover.
cases of kinematically induced damage to piles, this mode of
 Mueser Rutledge Consulting Engineers, New York. response does not usually receive proper attention by engineers.
{ City University of New York. Instead, seismically loaded piles are traditionally designed to
{ National Technical University, Athens. withstand only the ¯exural stresses generated from the oscilla-
} Shimizu Corporation, Tokyo. tions of the superstructure. Nevertheless, the importance of

425
426 NIKOLAOU, MYLONAKIS, GAZETAS AND TAZOH
kinematic loading has started to be recognised in recent seismic elastic piles are ®nite. This gives rise to an interaction between
provisions. For example, Part 5 of the recently published the pile and the soil at the vicinity of the interface that cannot
European seismic code EC-8 (1996) states: be captured by equations (2) or (3). Third, results from these
methods may lead to erroneous design rules if carelessly inter-
`Piles shall be designed for the following two loading
preted. For instance, the peak bending strain, åp , in a cross-
conditions:
section of the pile is
(a) inertia forces from the superstructure . . . åp ˆ (1=R) r (4)
(b) soil deformations arising from the passage of seismic
where r is the distance from the neutral axis to the farthest
waves which impose curvatures and thereby lateral strain
®bre in the cross-section. Bending strain is useful in evaluating
on the piles along their whole length . . . Such kinematic
the seismic performance of a pile because
loading may be particularly large at interfaces of soil
layers with sharply different shear moduli. The design (a) it is dimensionless
must ensure that `no plastic hinge' develops at such (b) it is directly measurable experimentally
locations . . .' (c) it can be used to quantify damage
(d ) ultimate (`failure') bending strains do not vary signi®cantly
An analogous statement can be found in the Seismic guidelines
among common structural materials. (Typically, strains of
for ports (TCLEE, 1998), which in Part 6 deals with kinematic
the order of one thousandth are enough to in¯ict damage in
loads. The increasing awareness of practising engineers of the
conventionally designed concrete or steel beams)
importance of kinematic loading can be noticed from reports in
professional journals (e.g. European Foundations, Spring 1998; As mentioned earlier, Margason's method determines 1=R based
Pappin et al., 1998). only on the properties of the soil and the excitation (equations
This paper is aimed at improving current understanding of (2), (3)). Therefore, one would conclude from equation (4) that
the importance of kinematic loading on the seismic performance the peak bending strain increases proportionally with the pile
of piles. This is done in three parts: ®rst, two existing design radius r. This may lead to the conclusion (see Bertero et al.,
methods are critically reviewed; second, new analytical results 1974; Margason, 1975) that small-diameter piles are superior to
for piles in homogeneous and layered soil deposits are devel- large-diameter piles for kinematic loading. As will be shown
oped by implementing a pertinent bean-on-dynamic-Winkler later, this may or may not be true, depending on the circum-
foundation (BDWF) model; third, a case study is presented in stances.
which theoretical predicitons are tested against ®eld measure-
ments.
Dobry & O'Rourke (1983)
A simple model for determining kinematic pile bending
REVIEW OF AVAILABLE DESIGN METHODS moments at the interface of two soil layers has been proposed
Margason (1975) and NEHRP (1997) by Dobry & O'Rourke (1983). Their main assumptions are as
In one of the earliest methods for kinematic pile bending, follows:
Margason (1975) assumes that a long pile follows the motion of
(a) The pile is long, and the two soil layers are suf®ciently
the surrounding soil. Based on this assumption, the kinematic
thick for the response of the pile outside these layers (e.g. at
bending moments are determined by considering the peak cur-
the pile head or at the pile toe) not to in¯uence the response
vature developing in the free-®eld soil:
at the interface.
M ˆ Ep I p (1=R) (1) (b) The soil is subjected to a uniform shear stress, which
generates uniform strain within each layer.
where M ˆ peak pile bending moment; (1=R) ˆ peak `soil'
curvature; (Ep I p ) ˆ pile ¯exural stiffness. To compute (1=R), Based on these assumptions, and modelling the pile as a beam
Margason proposes the following relation: on Winkler foundation, Dobry & O'Rourke derived an explicit
solution for the pile bending at the interface. Expressed in terms
(1=R)  2 ÄUff =Äz 2 (2)
of bending strain, their solution can be cast as
in which ÄUff is the relative lateral displacement between two åp ˆ 2r ë1 ã1 F (5)
points in the soil separated by a vertical distance Äz. This
relation is based on approximating the de¯ected shape of the where ã1 is the soil shear at the interface, and ë1 is the well-
pile by a circular arc: that is, assuming that the pile is subjected known Winkler parameter (Scott, 1981),
to pure bending. Margason (1975) argues that peak soil curva-  1=4
tures during severe earthquakes are not likely to exceed about k1
ë1 ˆ (6)
0´02 mÿ1 (if liquefaction does not develop). 4Ep I p
An analogous approach is proposed in the NEHRP (1997)
seismic provisions: assuming that the pile follows the free-®eld referring to the properties of the ®rst layer. k 1 denotes the
soil motion, and considering only vertically propagating S modulus of the Winkler springs in the ®rst (top) layer, and it
waves, the curvature in the free-®eld soil is obtained from the was taken by Dobry & O'Rourke as three times the shear
one-dimensional wave equation (Newmark, 1968; NEHRP, modulus of the material: that is, k 1 ˆ 3G1 . F denotes the
1997): dimensionless function
(1=R) ˆ aff =Vs 2 (3) F ˆ cÿ3 (c ÿ 1)(c2 ÿ c ‡ 1) (7a)
where aff denotes the free-®eld soil acceleration and Vs the in which c expresses the ratio of the shear moduli of the two
propagation velocity of shear waves in the soil material. layers:
The accuracy of the above equations will be examined below.
In the interim, the following points are worthy of note. First, in c ˆ (G2 =G1 )1=4 (7b)
equations (2) and (3) the interaction between pile and soil is
neglected. Accordingly, several important parameters such as The Dobry±O'Rourke (1983) model provides a practical tool
the pile±soil relative stiffness, pile length to diameter ratio for determining kinematic interface moments. In contrast to the
(`slenderness ratio') and radiation damping are not incorporated. simplistic Margason/NEHRP method, the model takes into ac-
Second, the methods are inapplicable to interfaces between count the interaction between soil and pile and thus overcomes
different layers. As mentioned earlier, soil strains are discontin- the problem of singular soil curvature at the interface. Never-
uous across such interfaces, and thereby the corresponding soil theless, the dynamic nature of the excitation and the effect of
curvatures are theoretically in®nite. In contrast, curvatures in the ®nite thickness of the soil layers are not incorporated. In
 KINEMATIC PILE BENDING DURING EARTHQUAKES 427
addition, as will be shown later on, this solution by substantially & Ventura, 1984) can then be utilised to obtain the response in
overpredict the actual pile bending in certain cases. the time domain. The model can accommodate precisely the
frequency dependence of k and c, in contrast to methods that
require frequency-independent parameters to obtain the response
BEAM-ON-DYNAMIC-WINKLER-FOUNDATION MODEL directly in the time domain. Moreover, moderate levels of non-
The response to vertical S-wave excitation of a single pile in linearity in the soil surrounding the pile could also be handled
a layered soil can be obtained numerically using suitable ®nite- approximately with this model. Such non-linearity may arise
element (FE) formulation with `wave transmitting' boundaries from the large stresses induced in the immediate vicinity of the
(e.g. the one described by Blaney et al., 1976), or a boundary- pile, and can be modelled approximately with a linear analysis
element-type code (e.g. the one described by Mamoon & Bane- of a radially inhomogeneous soil (Sheta & Novak, 1982;
rjee, 1990 and Banerjee 1995). However, such rigorous tools, Veletsos & Dotson, 1986; Michaelides et al., 1998). The springs
even if available, have well-known limitations when used in and dashpots resulting from such an analysis would re¯ect the
seismic design. This is particularly true if seismic analysis using non-linearities due to pile±soil interaction, rather than the non-
actual or simulated ground motions is to be performed in the linearities due to vertically propagating shear waves in the free-
frequency domain. Under these conditions, pile response must ®eld soil.
be computed for a large number of frequencies covering the However, for the kinematic problem studied herein, such
frequency content of the seismic signal. Therefore a simpli®ed non-linearities are not in our prime interest for two reasons:
analytical model would be quite useful provided it had been (a) The maximum bending moments occur at layer interfaces
shown to be in accord with the rigorous results for a wide range located at a certain depth, where the soil is well con®ned
of pile types, soil pro®les, and excitation frequencies. and offers greater resistance (e.g. ultimate lateral load,
The simpli®ed model adopted in the present study is based pu  9 Su compared with pu  2 Su at the surface, where
on the beam-on-dynamic-Winkler-foundation (BDWF) approach Su ˆ undrained soil shear strength).
(Fig. 1), in which the role of the soil±pile interaction is (b) The motion of the pile into the soil is not as signi®cant as
simulated through a set of springs and dashpots continuously in the inertially (head) loaded pile.
distributed along the pile, the frequency-dependent parameters
of which [k ˆ k(ù); cd ˆ cd (ù)] are determined through theor- Recall that a kinematically loaded pile follows all but the
etical models (Novak et al., 1978; Gazetas & Dobry, 1984) or details of the free-®eld soil motion (Kagawa & Kraft, 1980;
using calibrations with rigorous numerical solutions (Roesset, Kavvadas & Gazetas, 1993). Nevertheless, when applying the
1980; Dobry et al., 1982; Kavvadas & Gazetas, 1993). This results of the paper to practical problems, a judicious choice of
approach has been used extensively to estimate the dynamic soil moduli should account for some unavoidable non-linearity
impedance of piles in relation to inertial interaction analyses: and/or disturbance due to pile installation. Only linear analyses
that is, for loads applied at the pile head (Novak, 1991). A few are performed in this study, based on the ®nite-element-derived
studies have also used Winkler-type models to determine kine- springs and dashpots of Gazetas & Dobry (1984) and Makris &
matic response of piles (Penzien, 1970; Flores-Berrones & Gazetas (1992).
Whitman, 1982; Barghouthi, 1984; Kavvadas & Gazetas, 1993; Note also that the dynamic Winkler model developed here
Nikolaou et al., 1995). In these formulations, the springs and differs from those in the foregoing studies in that it is applied
dashpots connect the pile to the free-®eld soil; the wave-induced to layered (rather than homogeneous) deposits, it focuses on
motion of the latter (computed with any available method, such pile bending (as opposed to the motion of the pile head), and it
as Schnabel et al., 1972; Roesset, 1977) serves as the support proposes rational closed-form expressions for kinematic curva-
excitation of the pile±soil system (Fig. 1). tures and bending strains developing along the pile. Results are
Frequency-domain solutions to the kinematic problem using presented in three parts: (a) homogeneous soil; (b) two-layer
BDWF models have been presented, among others, by Nogami soil; (c) relation between harmonic and transient response.
et al. (1991), Kavvadas & Gazetas (1993), and Mylonakis et al.
(1997). The discrete fourier transform (DFT) method (Veletsos
HOMOGENEOUS SOIL: ANALYTICAL SOLUTIONS
As a ®rst application, the bending of a pile in homogeneous
Seismic free-field motion Seismic pile motion, Y(z) soil will be examined using the BDWF method. The analytical
aspects of the method are well known (see aforementioned
ρ1, E1, ν1, β1 references); only results are presented below.
Uff (z)

In®nitely long piles

Based on analytical studies by Flores-Berrones & Whitman
ρi, Ei, νi, βi (1982), Barghouthi (1984), and Makris & Gazetas (1992), the
de¯ected shape Y (z) of a long ®xed-head pile in a homogeneous
halfspace, excited by vertically propagating harmonic S waves,
is given by
Y(z)
Y (z) ˆ ÃUff (z) (8)
where Uff (z) denotes the free-®eld soil motion, z the depth from
the soil surface, and à the dimensionless response factor:
k ‡ iùcd
k(z) È (9)
Ep I p (q 4 ‡ 4ë4 )

c(z) in which (k ‡ iùc) denotes the so-called dynamic impedance of

the Winkler bed, with ù being the cyclic vibrational frequency
and i the imaginary unity; q ˆ ù=Vs is the wavenumber of the
Vertical
harmonic SH waves in the soil; ë is given by equation (6), but
shear waves
with the spring constant k replaced by the complex dynamic
Fig. 1. The beam-on-dynamic-Winkler-foundation (BDWF) model impedance (k ‡ iùc).
for a pile in a multi-layer soil pro®le, used in this study. The pile Differentiating equation (8) twice with respect to depth, z,
is excited by vertically propagating S waves after Fan, 1992 the ratio of the peak pile and soil curvatures is obtained as
428 NIKOLAOU, MYLONAKIS, GAZETAS AND TAZOH
(1=R)p in the bedrock such that the bending moment at the toe is zero
ˆÃ (10)
(1=R)s (hinged-toe pile), the corresponding expression is
8 9
For a free-head pile, the corresponding solution gives < cos qh [cosh ëh cos ëh =
(Mylonakis, 1999) (1=R)p ‡2 sinh ëh (ë=q)2 [(1=Ã) ÿ 1]sin ëh]
ˆ Ã:1 ÿ 2 ;
(1=R)s cosh 2ëh ‡ cos 2ëh
(1=R)p (14)
ˆ à max[eÿëz (cos ëz ‡ sin ëz) ÿ cos qz] (11)
(1=R)s
Note that, with increasing pile length, the second term in the
The above expressions clearly indicate that pile curvature is not right-hand side of equations (13) and (14) converges to 1, and
equal to soil curvature (as assumed by Margason (1975) and both expressions duly reduce to equation (10).
NEHRP (1997)); its value depends on the characteristics of the Numerical values for the above equations are plotted in Fig.
soil and the pile, the excitation frequency, and the depth from 3. It can be seen that the curvature ratio may exceed 1 for a
the surface. wider range of frequencies than for the in®nitely long pile (Fig.
Equations (10) and (11) are plotted in Fig. 2 as functions of 2), and it is understandably higher with the hinged-toe than the
the dimensionless frequency factor a0 ˆ ùd=Vs, for a solid ¯oating-toe piles. Contrary to the trends incited by the Marga-
cylindrical pile of diameter d. It is seen that at low frequencies son/NEHRP equations (equations (2), (3)), no clear relation
pile and soil curvatures are approximately equal, which implies between curvature ratio and pile diameter can be drawn from
that the pile follows the free-®eld soil motion (à ˆ 1). With the graph. The jump observed at the fundamental natural period
increasing frequency, however, the curvature ratio decreases of the layer is associated with the sudden appearance of radia-
monotonically with frequency, and tends to zero as a0 tends to tion damping in the medium at that frequency (Blaney et al.,
in®nity. This can be interpreted as a progressively increasing 1976; Novak, 1991; Kavvadas & Gazetas, 1993; Guin & Bane-
destructive interference of the high-frequency (short-wavelength) rjee, 1998). The jump is clearly visible in the short layer
seismic waves exciting the pile. The trend is understandably (h=d ˆ 5), but less pronounced in the thicker ones.
stronger with large pile±soil stiffness contrasts. It is also ob- The response of the pile head can be determined analytically
served that the curvature ratio is always smaller than 1, and is with the BDWF model. The solution can be conveniently ex-
higher in free-head than in ®xed-head piles. An exception pressed through the so-called kinematic response factors I u and
occurs near a0 ˆ 0: at such low frequencies and at shallow I f (Blaney et al., 1976; Kaynia & Kausel, 1982; Fan et al.,
depths, the last term in equation (11), cos qz, is practically equal 1991), which are de®ned, respectively, as the maximum trans-
to 1. Accordingly, it is straightforward to show that the maxi- lation, Y (0), and rotation, È(0), of the pile head normalised by
mum pile curvature occurs when ëz ˆ ð: the corresponding maximum displacement at the surface of the
free-®eld soil (i.e. I u ˆ Y (0)=Uff (0); I ö ˆ È(0)d=Uff (0)). For a
(1=R)p long pile whose head is free to rotate (free-head conditions), the
ˆ 1 3 (eÿð ‡ 1) ' 1:04 (12) following closed-form solutions were obtained:
(1=R)s "  2 #
1 q
which in agreement with Fig. 2. Iu ˆ à 1 ‡ 2 (15)
ë
and
Piles of ®nite length q2 d
Iö ˆ à (16)
For a ®xed-head pile of ®nite length in a homogeneous layer ë
of thickness h over a rigid base, the curvature ratio is given by In the case of zero rotation at the pile head (®xed-head
the expression (Mylonakis, 1999) conditions), the kinematic ratio, I u , is given by the simpler
(1=R)p expression
(1=R)s Iu ˆ à (17)
2 3
 2 cos qh (sinh ëh cos ëh ÿ cosh ëh sin ëh) which elucidates the physical meaning of parameter Ã. Natu-
4 q ‡(q=ë) cosh ëh cos ëh sin ëh5 rally, with ®xed-head piles, I ö ˆ 0. Equation (17) was appar-
ˆÃ 1‡ (13)
ë sinh 2ëh ‡ sin 2ëh ently ®rst derived by Flores-Berrones & Whitman (1982) and
Barghouthi (1984).
The above expression corresponds to stress-free conditions at
the pile toe (¯oating-toe pile). For a pile with its toe socketed 1·5
Floating-toe pile (Equation (13))
1·5 Hinged-toe pile (Equation (14))
Fixed-head pile (Equation (10)) h/d = 20
Curvature ratio, (1/R)p/(1/R)s

Free-head pile (Equation (11)) h/d = 10

Margason (1975)
Curvature ratio, (1/R)p/(1/R)s

NEHRP (1997) 1·0

1·04
1·0
h/d = 5

0·5
Ep/Es = 1000
0·5
Fundamental frequency
10 000 of soil layer

100 000 0
0 0·1 0·2 0·3 0·4 0·5
0 a0 = ωd/Vs
0 0·1 0·2 0·3 0·4 0·5
a0 = ωd/Vs Fig. 3. Ratio of peak pile and soil curvatures in a ®xed-head pile in
homogeneous soil layer over rigid rock, for various layer thicknesses
Fig. 2. Ratio of peak pile and soil curvatures for an in®nitely long and boundary conditions at the pile toe; Ep =Es 1000, rs =r rp 0:7,
pile in homogeneous halfspace: rs =rrp 0:7, í 0:4, â 0 í 0:4, â 0:05
 KINEMATIC PILE BENDING DURING EARTHQUAKES 429
LAYERED PROFILES: PARAMETRIC ANALYSIS Note that three parameters remain invariable in all analyses:
With reference to layered soil pro®les, a comprehensive the distance from the pile tip to the bedrock (h3 =L ˆ 1), the
parameter study is presented in this section. It refers to harmo- Poisson's ratio of the soil (í1 ˆ í2 ˆ 0:40), and the material
nic state±state bending strains in a pile in a two-layer soil damping in the soil ( â1 ˆ â2 ˆ 10%). The damping values are
deposit subjected to vertically propagating seismic shear waves roughly consistent with strain levels caused by moderately
(Fig. 4). To conform with the analyses presented earlier, the pile severe shaking. The importance of the parameter (h3 =L), which
is modelled as a solid elastic cylinder of Young's modulus Ep , is related to the thickness of the soil pro®le, is examined later
moment of inertia I p , diameter d, and mass per unit length m. on.
Each soil layer j ( j ˆ 1, 2) is characterised by its shear wave Selected results are given in Figs 5±7. Fig. 5 presents the
velocity Vj, thickness hj , mass density rj, hysteric damping ratio envelopes of harmonic bending strain pro®les for cases A1 and
âj , and Poisson's ratio íj . The harmonic excitation is described A2, for both ®xed-head and free-head piles. The moments are
through the harmonic base rock acceleration ar exp(iùt). presented in terms of maximum dimensionless bending strain:
Peak harmonic steady-state bending strains along the pile are
determined for different groups of ®xed dimensionless para- M d
åp ˆ (18)
meters, with the ratio V1 =V2 being the main variable (Table 1). Ep I p 2
The analyses are divided into four groups (A±D). For each
group, four cases (1±4) are investigated, corresponding to dif- Several interesting points are worthy of note:
ferent stiffness contrasts between the two layers. Both free-head
(a) The kinematic bending strain in a free-head pile in homo-
and ®xed-head conditions are considered atop the pile, while
geneous soil (case A1) increases with depth, reaching its
the pile toe is assumed stress free. Thirty-two sets of parametric
maximum at approximately the mid-length of the pile. In
results were obtained in total.
contrast, for a ®xed-head pile, bending strain is a decreasing
function of depth, having its maximum at the pile head.
(b) For a two-layered pro®le (case A2), the bending strain
as eiωt diagram exhibits a pronounced peak close to (but not
exactly at) the interface between the two layers. This is in
agreement with the foregoing discussion about the con-
V1, ρ1 centration of kinematic moments in the vicinity of layer
β1, ν1 h1 interfaces.
Ep, lp, m (c) The bending strain diagrams of free-head and ®xed-head
piles converge with depth and become practically identical
beyond a certain distance from the surface. This depth
coincides with the so-called `active pile length', beyond
which a head-loaded pile behaves as in in®nitely long beam.
h2 Using the well-known approximate formula (Poulos & Davis,
1980; Randolph, 1981; Velez et al., 1983)
V2, ρ2  1=4
β2, ν2 Ep
La  1:5 d (19)
Es
d
h3 one obtains La =d ˆ 1:5 3 (1000)1=4 ˆ 8:4, which is agreement
with the results of Fig. 5. The values shown in the plots
ar eiωt represent response envelopes, which may entail different excita-
tion periods. However, in most cases examined the envelopes
were dominated by the fundamental period of the soil pro®le.
SH Evidence on this is presented immediately below.
waves Case A2 is examined further in Fig. 6. In this graph, transfer
functions relating bedrock acceleration to surface acceleration
Fig. 4. The system used in the parameter study: a single pile and bedrock acceleration to pile bending strain are portrayed as
embedded in a two-layer pro®le on rigid bedrock, excited by functions of frequency for two characteristic points: the pile top
harmonic SH seismic waves (or the soil surface when referring to the free-®eld motion), and

Table 1. Parameter cases for maximum harmonic steady-state kinematic bending strains. In
all cases í1 í2 0:4; â1 â2 10%; h3 =L 1
Group Case L=d Ep =E1 h1 =L V1 =V2 r1 =r2
A 1 20 1000 2/3 1 1
2 0´5 0´8
3 0´25 0´7
4 0´1 0´6
B 1 20 5000 1/2 1 1
2 0´5 0´8
3 0´25 0´7
4 0´1 0´6
C 1 40 5000 1/2 1 1
2 0´5 0´8
3 0´25 0´7
4 0´1 0´6
D 1 20 10 000 1/2 1 1
2 0´5 0´8
3 0´25 0´7
4 0´1 0´6
430 NIKOLAOU, MYLONAKIS, GAZETAS AND TAZOH
0 10
Case A1
Soil surface

Ratio of peak surface to bedrock acceleration

Interface
8
5

6
Depth, z/d

10

15
Fixed-head pile 2
Free-head pile

20 0
(a)
0 15
Case A2
Pile top
Interface, fixed-head pile
12
5 Interface, free-head pile

Pile bending strain, εp: 10–4

Depth, z/d

9
10 Equation (20)
Equations (5)–(7)
(Dobry & O'Rourke, 1983)
6

15
3
Equation (20)
Equations (5)–(7) (Dobry & O'Rourke, 1983)
20
0 5 10 15 20 0
0 1 2 3
Pile bending strain, εp: 10–4 Normalised excitation period, T/T1
(b)
Fig. 5. Steady-state kinematic bending strains along a ®xed-head
and a free-head pile in a homogeneous deposit (top) and a two-layer Fig. 6. Case A2: amplitudes of steady-state response for (a) free-®eld
deposit (bottom): L=d 20; Ep =E1 1000; V1 =V2 1=2. The response and (b) pile bending strain in a two-layer soil pro®le
dashed line indicates the depth of the interface

the point at some depth where a peak occurs in pile bending M=(as r1 h1 d 3 ), which accounts for the different free-®eld soil
strain. The similarity between the free-®eld transfer functions ampli®cation effects, the difference in the results is less than
(Fig. 6(a)) and the bending strain transfer functions (Fig. 6(b)) 10% even for high values of h3 =L.
is evident. The maximum bending strains occur at or near The most important conclusions that have emerged from
resonance (T =T1  1) for both the interface and the pile top. these analyses are as follows (see also Nikolaou et al., 1995):
(The small deviation from unity can be attributed to damping in
the system.) Also, although the free-®eld soil response at the (a) For a given excitation frequency, the kinematic bending
surface and the interface are quite different, the corresponding strain depends mainly on: the stiffness contrast (V1 =V2 )
transfer functions of bending strain are very similar and resem- between any two consecutive soil layers in the deposit; the
ble closely that of the free-®eld surface response. boundary conditions at the pile head or cap; the proximity
Cases B2, B3 and B4 are portrayed in Fig. 7. In case B2 it of the excitation period, T , to the fundamental (®rst) natural
can be seen that with a stiff pile (Ep =Es ˆ 5000) the active pile period, T1 , of the soil deposit, and to a lesser degree to the
length moves below the interface (from equation (19): second natural period, T2 , of the deposit; and the relative
La =d ˆ 1:5 3 (5000)1=4  12:6, in accord with the plotted re- depth, h1 =La , measured from the top of the pile down to the
sults). Apparently, this makes the kinematic distress at the pile interface of the layers with the sharpest stiffness contrast,
top (®xed-head conditions) more pronounced than that at the normalised with respect to the active length, La , of the pile.
interface. In cases B3 and B4 the shear wave velocity of the (b) The bending strains are largest either at the pile head or at
second soil layer, V2 , has been increased, respectively, to four the vicinity of the interface of soil layers with the sharpest
and ten times the V1 . Notice that the increase in stiffness stiffness contrasts. The strains at the interface for free-head
contrast leads to signi®cantly larger bending strain at the inter- and ®xed-head piles are almost identical, except when the
face, while bending at the top remains almost practically pile is `short' and/or `rigid' (i.e. when h1 , La ).
unaffected (see B2). With the sharpest impedance contrast (c) In most cases, the maximum harmonic bending strain
between the two layers in case B4, the bending strain at the occurs at the fundamental natural period of the soil deposit.
interface exceeds that at the pile head. The pile strain transfer functions display a very rapid
The effect of the dimensional parameter h3 =L is presented in reduction when moving away from resonance (Fig. 6). The
Table 2. Changing this ratio will dramatically affect the natural variation of kinematic bending strain with frequency
period of the pro®le and alter the free-®eld soil response. follows, more or less, the ampli®cation of the free-®eld
Nevertheless, it is evident from Table 2 that an increase or acceleration: as =ar (Figs 6(a), 6(b)). This shows the great
decrease in h3 =L does not have any major effect on the in¯uence of the ®rst mode of vibration on the magnitude of
maximum pile bending strain at the interface. In particular, bending strain, and contradicts some earlier statements in
when speaking in terms of the dimensional moment the literature that higher modes would produce larger
 KINEMATIC PILE BENDING DURING EARTHQUAKES 431
0 Table 2. Effect of relative height, h3 =L, on peak steady-state
Case B2 kinematic pile bending
max M
Case h3 =L åp (10ÿ4 ) as r1 h1 d 3
5
A3 1 16´8 27´9
0´5 14´6 27´3
4 21´2 32´4
Depth, z/d

C3 1 92´6 69´1
10
0´5 76´2 66´1
4 76´6 79´6
D3 1 62´6 90´8
0´5 50´0 85´7
15
4 54´3 97´2

20 proportional to the actual shear stress that is likely to

0 develop at the interface, as a function of the free-®eld
Case B3 acceleration at the soil surface, as :

ôc  as r1 h1 (20a)
5
The ®tted formula is written
 0:30  0:65  0:50
3 L Ep V2
Depth, z/d

:
M  0 042 ôc d (20b)
10 d E1 V1
Predictions obtained using the above formula are shown in Figs
5±7. Corresponding predictions using the Dobry & O'Rourke
15 (1983) solution are also shown for comparison. The good
Fixed-head pile performance of equation (20) is evident in all graphs. In
Free-head pile contrast, the Dobry±O'Rourke solution overestimates pile bend-
ing in certain cases (Figs 5, 7).
20
0
Case B4 TIME AGAINST FREQUENCY-DOMAIN RESPONSE
Harmonic steady-state results can only rarely be used directly
in design. This is because only a hypothetical harmonic excita-
5 tion with a very large number of cycles would produce a
response with amplitude equal to the steady-state value. A more
transient excitation, as in earthquake shaking, would tend to
produce smaller response.
Depth, z/d

10 To correlate steady-state and transient peak response, a

comprehensive numerical parameter study has been carried out,
using (a) three actual soil pro®les (I, II, and III, as depicted in
Fig. 8), (b) nine accelerograms (eight actual records and one
15 arti®cial motion; see Table 3 and Fig. 9).
The three deposits were chosen to represent cases where the
Equation 20 use of piles is the most likely engineering solution. The use of
Equations 5–7 (Dobry & O'Rourke, 1983) realistic soil pro®les was considered necessary in these analyses
20 to better represent the complicated transfer functions in actual
0 20 40 60 80 soil deposits. Sketched in Fig. 8, the soil pro®les are:
Pile bending strain, εp: 10–4
(a) a `soft clay' pro®le, a two-layer idealisation of an actual soil
Fig. 7. Steady-state kinematic bending strains along a ®xed-head that consists of a very soft clay underlain by stiff sand
and a free-head pile in a two-layer deposit; L=d 20, Ep =E1 having thickness that varies parametrically from 30 m to
5000. From top to bottom: V1 =V2 1=2, 1=4 and 1=10 94 m. The shallow version of the pro®le (total thickness
h ˆ 30 m) has a natural period of 0´52 s, while the deep
version ( h ˆ 94 m) has a period of 1´1 s.
(b) a `Boston' pro®le, an idealisation of an actual pro®le from
downtown Boston. The calculated natural period is
kinematic bending (Dobry & O'Rourke, 1983; Kavvadas &
approximately 0´5 s.
Gazetas, 1993). Indeed, while higher frequencies tend to
(c) a `Bay Area' pro®le, an idealised pro®le typical of the
generate `wavy' shapes of de¯ection (and thus have the
stiffer San Francisco Bay Area formations. The distribution
potential for inducing relatively large curvatures at the
of the shear wave velocity with depth does not show the
interface), the actual curvature is also affected by the overall
abrupt changes of the previous pro®les. The fundamental
drift between the top and the bottom of the pile. This drift
natural period of this pro®le is roughly 0´8 seconds.
usually becomes maximum at the ®rst natural mode, and
thereby produces the largest bending at the ®rst resonance. Additional information on the soil properties is given in
(d ) A closed-form expression has been developed for comput- Nikolaou et al. (1995).
ing approximately the maximum steady-state bending Since only linear analyses were performed, the wave veloci-
moment at the interface between the two layers (Nikolaou ties shown in Fig. 8 are considered as strain-compatible (`effec-
et al., 1995; Nikolaou & Gazetas, 1997). The expression is tive') quantities. In view of the strong seismic excitation that is
based on a `characteristic' shear stress ôc , which is presumed to be imposed, this means that the low-strain (`elas-
432 NIKOLAOU, MYLONAKIS, GAZETAS AND TAZOH
Vs: m/s Vs: m/s Vs: m/s
0 200 400 0 200 400 0 200 400
0 0 0

Soft Fill
5
clay
7
9
9·5

Bay
Blue mud
clay

Dense
20
sand
22

27
Glacial
30 till Dense
(or 94) clayey
sand
Vsr = { ∞1200 35

Vsr = 1400
40

Vsr = 550

Profile I Profile II Profile III

Idealised soft clay Idealised Boston Idealised Bay Area
profile profile profile

Fig. 8. The three actual soil pro®les used in this study

Table 3. The ground motions used in the time-domain analyses

Event Record Peak ground Approximate Range of
acceleration: number of predominant
g{ cycles, Nc periods, T : s
p

EC-8, S1 soil Arti®cial 0´13 .20 0´10±0´50

Northridge (1994) Pac_down, ch. 1 0´43 2±3 0´15±0´50
Pyrgos (1993) Pyrtran 0´46 1 0´12±0´45
La116th, ch. 1 0´39 4±5 0´10±0´25
Whittier (1987) Pacoima, ch. 1 0´16 3±4 0´10±0´30
Tarzana, ch. 3 0´4 10 0´30±0´40
Loma Prieta (1989) Anderson, downstr 0´25 6±7 0´15±0´30
Kobe (1995) Kobe JMA, NS 0´83 4 0´3±0´9
Mexico (1985) La Villita, 08 0´12 3 0´5±0´6
{ Before normalisation.
 Based on the 5%-damped acceleration spectrum (Fig. 9).

tic') S-wave velocities would be larger than the shown ones by corresponding maximum steady-state strains. It is evident from
a factor of the order of 1´5. A variation in damping ratio from this ®gure that the envelope of peak moments (in the `time
10% at the surface to 7% at the bottom was considered in all domain') has a distribution with depth that is of the same shape
pro®les. The piles used in the analyses are ®xed head with as the distribution of steady-state amplitudes (in the `frequency
d ˆ 1:3 m, Ep ˆ 25 MPa, and r ˆ 2:5 Mg=m3 . Their lengths domain'). But the values of the latter are about three to ®ve
are shown in Fig. 9. times larger than the former, depending on the excitation. It is
The nine selected accelerograms cover a broad range of worth mentioning that some of the peak strains are of the order
possible rock motions. They are assumed to consist solely of of one thousandth (0´1%), which is enough to damage the pile.
vertically propagating S waves, and are all normalised to peak Based on these analyses, it is proposed that a reduction factor
rock acceleration of 0´20 g. The characteristics of the records ç be applied to the maximum steady-state pile bending strain
are shown in Table 3. The duration of these motions, expressed (or moment) in the frequency domain to arrive at the corre-
as number of strong motion cycles, ranges between one and sponding peak value in the time domain:
seven cycles. Their dominant periods extend from about 0´1 to
0´9 s. The normalised acceleration response spectra of all mo- max åp (t)
ç (21)
tions are depicted in Fig. 9. max åp (ù)
The more than 80 parameter analyses performed showed that
the basic conclusions drawn from the harmonic analyses are still Data for ç obtained from the parameter analyses are presented
valid, except that the transient strains are of signi®cantly smaller in Fig. 11. The data are plotted as function of the effective
magnitude. Fig. 10 portrays the envelopes of pile bending number of cycles in the record, Nc , grouped in two categories:
strains calculated using the nine excitations, plotted against the (a) resonant conditions (denoted by the solid dots), which
 KINEMATIC PILE BENDING DURING EARTHQUAKES 433
6 itudinal (vertical) pile strains, arising from axial and bending
Tarzana, Whittier
loading, at six different depths. At each elevation four gauges
were installed in the periphery of the piles, from which axial
and bending strains in the x and y directions were recorded
5 (Fig. 12).
The soil pro®le is shown in Fig. 13. It consists of about 39 m
of soft saturated high-plasticity silty clay (PI  50) having a
LA116, Anderson, Loma Prieta
Whittier
measured shear wave velocity, Vs , ranging between 135 and
240 m=s, underlain by stiff mudstone with Vs  480 m=s. These
4 La Villita, Mexico
velocities are low-strain (Vs, max ) values determined in situ using
PS logging (plank hammering method). Soil properties are also
Spectral amplification

listed in Table 4. The earthquake event analysed here took place

under the Tokyo Bay area (358129N; 1398489E) on 2 February
3 1992 with a magnitude 5´9, focal depth 93 km, and epicentral
distance from the site 32 km. The largest acceleration recorded
on the ground surface was about 0´05g.
JMA Kobe
Acceleration time histories in the x and y directions just
Pacoima,
2 Northridge
under the ground surface (GL ÿ 2 m) and the base mudstone
(GL ÿ 40 m) are plotted in Fig. 14. Their response spectra are
shown in Fig. 15. Predictions using the equivalent-linear com-
puter code SHAKE are also shown for comparison. Computed
and recorded motions are in very good agreement despite the
1
fact that in the theory the soil is modelled as a series of
Pacoima, Whittier Pyrgos perfectly horizontal layers subjected to vertically propagating
shear waves. The prevailing period of approximately 1 s obser-
ved in the surface spectra is in agreement with the fundamental
0
0 0·25 0·50 0·75 1·00 1·25
natural period of the soil layer: T1  4 h=Vs  43
Structural natural period, T: s 39=150 ˆ 1 s. It is worth mentioning that because of the
relatively low accelerations in the soil (ar  0:025 g) and the
Fig. 9. Normalised 5%-damped spectra of the earthquake motions high plasticity index of the material (PI  50), the reduction in
considered soil shear modulus predicted by SHAKE using the Vucetic &
Dobry (1991) G±ã curves does not exceed a mere 10% (see
Fig. 13). Similarly, soil damping is estimated to be about 3±
correspond to the case when the fundamental natural period of 4%. However, a slightly higher material damping of about 5%
the deposit lies within the range of the predominant periods of was found necessary to match the amplitude of accelerations at
the excitation; (b) non-resonant conditions (denoted by the open the soil surface. To achieve this using the aforementioned G±ã
dots), when the fundamental period of the deposit lies outside curves, a plasticity index PI ˆ 30 had to be used (Fig. 15). This
the dominant periods of the earthquake motion. Least-square small discrepancy can be attributed to the material character-
®tted lines to the data are also presented. They are described by istics of the clay. The G(ã), â(ã) curves for PI ˆ 30 were used
ç ' 0:04Nc ‡ 0:23 (22a) thereafter.
The ratio of the recorded response spectra at the surface
and (ÿ2 m) and the base (ÿ40 m) of the pro®le is plotted in Fig.
ç ' 0:015Nc ‡ 0:17 ' 0:2 (22b) 16, for both x and y directions. The substantial ampli®cation
(about eight times) observed close to the fundamental natural
for the resonant and non-resonant conditions respectively. It is period of the pro®le is evident in the graph. The slightly higher
seen that for non-resonant conditions ç is quite small, and fundamental natural period and smaller ampli®cation in the x
practically independent of Nc . By contrast, for resonant condi- direction could be attributed to the stronger excitation imposed
tions, ç is understandably much more cycle-dependent, and may along the x axis. The similar ampli®cation patterns observed in
exceed the value of 0´5. the two directions lends further support to the assumption of
An analogy with the single-degree-of-freedom oscillator is one-dimensionality of the wave propagation, as adopted in our
worthy of note: while in such systems the steady-state spectral analyses.
ampli®cation at resonance (for 5% damping) is approximately The computed and measured time histories of pile bending
equal to 1=(2â) ˆ 1=(2 3 0:05) ˆ 10, the maximum spectral strain at the depth of 31 m (interface 1; see Fig. 13) are
ampli®cation in code acceleration spectra is usually taken as 2´5 compared in Fig. 17. Note that our method refers to a single
(NEHRP, 1997; Eurocode EC-8, 1996), which is equivalent to pile, and therefore the computed results for piles A and B are
an ç factor of 2:5=10 ˆ 0:25. This value is somewhat smaller identical (only one curve shown). Several interesting trends are
than the average observed in Fig. 11 for resonant conditions. worthy of note. First, the bending strains in the corner and
This `discrepancy' can be attributed to, among other reasons, centre piles are very similar. This indicates that group effects
the higher-mode resonances that can develop in a soil layer but are minor for kinematic response, which is in agreement with
not in a simple one-degree-of-freedom oscillator. the analytical studies of Kaynia & Kausel (1982), Fan et al.
(1991), Kaynia & Mahzooni (1996), and others. (By contrast,
pile-to-pile interaction due to dynamic loading at the pile cap
CASE STUDY: SEISMIC RESPONSE OF INSTRUMENTED `ERVIC' has been shown to have a profound effect on the response of
BUILDING PILE FOUNDATION' the group; see Kaynia & Kausel (1982), Dobry & Gazetas
The Ervic building is a 12 storey structure located in (1988).) Second, the prevailing period of the records is very
Yokohama, Japan. Its foundation consists of a one-¯oor base- close to 1 s, which elucidates the role of soil response (as
ment supported by 20 reinforced concrete piles, 35 m long and opposed to loading due to structural inertia forces) on the
1´7 m in diameter. Comprehensive instrumentation of this build- development of kinematic pile bending at depth. Third, the
ing was conducted by the Institute of Technology of Shimizu computed peak values tend to overestimate the measurements
Corporation. Accelerometers were installed on the building and by as much as 50%. While modelling errors are a `usual
in the free-®eld soil, as depicted in Fig. 12. Two piles, one suspect', this may also imply that the location of the strain
located at the corner (pile A) and the other near the centre (pile gauges might not be exactly at the interface between the two
B) of the foundation plan, were chosen for measuring long- layers, as was assumed in the analyses. Another possibility is
434 NIKOLAOU, MYLONAKIS, GAZETAS AND TAZOH
0 0
Soft clay Boston

5
10
Depth, z: m

Depth, z: m
15
10

20

15 25
0 10 20 30 0 5 10 15 20
0
Bay Area

5 Steady state
Pacoima, Whittier
Artificial, EC8
10
Tarzana, Whittier
Depth, z: m

Pacoima, Northridge
15 Anderson, Loma Prieta
JMA, Kobe
La villita
20
Pyrgos
LA116, Whittier
25

0 5 10 15
Pile bending strain, εp: 10–4

Fig. 10. Comparison of envelopes of pile bending strains for various earthquake motions, plotted against the
maximum steady-state strains, for four different soil pro®les (scaling to 0´2 g peak rock acceleration was
considered for all motions)

that the stiffness contrast between the two layers may be less
than the V2 =V1 ˆ 240=135  1:8 value assumed in our ana-
0·8 Data for T1 ≠ Tp (non-resonance) lyses, thereby leading to smaller bending strain. Nevertheless,
Data for T1 ≈ Tp (resonance) the agreement between the predicted and recorded time histories
Least-square fitted lines
is quite satisfactory from an engineering point of view.
The distributions along the pile of the peak values of
0·6 computed and measured bending strains are compared in Fig.
18. The computed values are solely the result of loading from
the surrounding soil (kinematic response): that is, the inertia of
the superstructure is ignored. However, the measured values are,
η 0·4 of course, the superposition of both kinematic (i.e. due to soil)
and inertial (i.e. due to structure) strains. The results of the
0·23
Nc + comparison show that:
0·04

0·015 Nc + 0·
17 (a) Theory and reality are in reasonable accord at the four
0·2
deeper elevations below the active pile length, estimated as
(equation (19))
La ˆ (1:5)(350)1=4 (1:7)  11 m (23)
0
0 2 4 6 The under-prediction by the analysis of the strains close to
Number of 'effective' excitation cycles, Nc
the pile head is understood, as such bending strains arise
mainly from structural loading.
Fig. 11. Proposed design curves for the frequency-to-time reduction (b) Kinematically induced bending strains are pronounced close
factor ç as a function of the number of `effective' excitation cycles to interface 1. This con®rms the signi®cance of kinematic
and resonance conditions. The dots correspond to the motions bending in deposits containing consecutive layers with
shown in the legend of Fig. 10 and in Fig. 9 sharply different stiffness.
 KINEMATIC PILE BENDING DURING EARTHQUAKES 435

24·8 m

m
·4
28

RF
28·4 m
46·6 m

7F

24·8 m
Pile B

Pile A

GL

B1F 0 10
y (+) Scale: m

x (+)

y
Soil Building accelerometer
Strain
transducer Ground accelerometer
z a
Dynamic strain transducer
x b
y d x

Pile B
Pile A c
GL –39 m
Supporting
layer

Fig. 12. Arrangement of accelerometers and strain gauges at Ervic Building

(c) Strains at the (even sharper) interface 2 are less pronounced L=d ˆ 34:9=1:7 ' 20:5 (26)
in our analysis because of the lack of adequate pile length V2 =V1 ˆ 240=135 ' 1:78 (27)
below the interface to mobilise bending moments. Unfortu-
nately, the authors could not ®nd a convincing explanation The resonant pile bending moment is obtained using equation
for the relatively high strain recorded at that depth in the y (20a):
direction. : : :
M  0:042 3 21 3 1:73 3 20:50 3 3 3500 65 3 1:780 5

 650 kNm (28)

Computation of kinematic pile bending strain using equations
(20)-(22) which corresponds to bending around the y axis.
As an example, the peak kinematic bending strain at inter- Assuming non-resonant conditions (a reasonable approxima-
face 1 (i.e. at z ˆ 31 m) will be computed using the approx- tion in view of the well-separated predominant periods of the
imate method proposed in equations (20)±(22). For simplicity, bedrock motion ( 0:1±0:3 s) and fundamental natural period of
only the strain in the x direction (along which the excitation is the soil layer ( 1 s)), and about three to four effective excita-
stronger) will be analysed. It will be shown that realistic tion cycles, the frequency-to-time conversion factor ç is ap-
estimates of the response can be obtained using this simpli®ed proximately (equation (22b), Fig. 11)
approach. ç  0:015 3 3 ‡ 0:17 ˆ 0:22 (29)
The characteristic stress at the interface is computed from
equation (20a): From equation (21), the peak bending moment at the interface
ôc ˆ (0:045 3 9:81) 3 1:55 3 31  21 kPa (24) is
max M(t)  0:21 3 650 ˆ 143 kN m (30)
The pile±soil stiffness contrast is approximately
Finally, the peak bending strain in the pile is obtained through
Ep 2:7 3 107 equation (18).
' ' 350 (25)
E1 135 3 1:5 3 2 3 (1 ‡ 0:4)
2
åp ˆ 143=(2:7 3 107 3 0:41) 3 1:7=2  1:1 3 10ÿ5 (31)
where 0´4 is an approximate estimate of the Poisson's ratio of
the clay. The pile slenderness ratio and stiffness contrast at which is in accordance with the recorded values in Fig. 18.
interface 1 are, respectively, As a ®nal remark, we should point out that the recorded peak
436 NIKOLAOU, MYLONAKIS, GAZETAS AND TAZOH

SPT blowcounts, N Shear wave velocity, Vs: m/s

0 10 20 30 40 50 100 200 300 400 500
0
Fill

Alluvial WT ρs = 1·6 Mg/m3

5 silty sand

A1
10
Low strain A2

15 A3

Strain compatible
20 Alluvial
Depth: m

silty clay

A4
ρs = 1·5 Mg/m3
25

30
Interface 1 A5

Sand with clay

35
ρs = 1·6 Mg/m3
Tuffaceous clay
Clay with sand Interface 2
Gravel A6
40
Mudstone
ρs = 1·6 Mg/m3
Fine sand
Mudstone
45

Fig. 13. Soil pro®le and properties; location of transducers in pile A and accelerographs in the building basement (open
circles) and the surrounding soil (open triangles)

Table 4. Summary of soil/foundation characteristics of Ervic Build- overall strains at the interface. More research is needed to
ing (mass density and shear-wave velocity values are shown in Fig. explore the interplay of these actions.
13.)
Depth Soil{ Pile
0±7 m Alluvial sand :
Ep ˆ 2 7 3 107 kPa CONCLUSIONS
rp ˆ 2:5 Mg=m3 A comprehensive analytical study was presented for the
Alluvial silty clay (CH-MH) L ˆ 35 m bending of single piles in homogeneous and layered soil depos-
wn ˆ 95% d ˆ 1 :7 m its under the passage of vertically propagating seismic SH
e ˆ 2 :6 waves. The method is based on a beam-on-dynamic-Winkler-
7±32 m W L ˆ 100% foundation (BDWF) model equipped with frequency-dependent
PI ˆ 55 spring and dashpot moduli. The main conclusions of the study
Su ˆ 45 kPa
are as follows:
Clay with sand (CH) (a) Pile curvatures are not equal in many instances to soil
wn ˆ 75% curvatures as assumed in some existing design methods, but
e ˆ 2 :0
32±40 m W L ˆ 30%
their values depend on the characteristics of the soil and the
PI ˆ 50 pile, the excitation frequency, and the depth from the soil
Su ˆ 75 kPa surface.
(b) Contrary to trends indicated by the Margason and NEHRP
.40 m Mudstone methods, no clear relation between imposed curvature and
{ Hysteretic damping 5% and Poisson's ratio 0´4 were considered for all
pile diameter can be inferred. It therefore follows that the
layers. seismic behaviour of slender piles is not always superior to
that of large-diameter piles.
(c) Kinematic bending strains are largest at the head of a
capped pile, and at the location of relatively deep interfaces
between layers with very different stiffnesses. The maxi-
mum harmonic steady-state moment at the interface occurs
strains are not large enough to in¯ict damage to the pile. In at the fundamental natural period of the soil deposit. Its
fact, even with a tenfold increase in the intensity of the value depends on a number of soil and pile parameters.
excitation (0´25 g on rock), the peak pile bending strain would Equation (20) was developed for estimating approximately
be of the order of 10ÿ4 (assuming linear response), which is this maximum moment in practical applications.
still below yielding. Sharper stiffness contrasts between the soil (d ) Peak pile bending due to transient excitation is only a
layers at the interface (V2 =V1 . 1:8) could, however, induce fraction, ç, of the maximum value (with respect to
bending distress. Also, other factors such as dead loads, nega- frequency) under harmonic steady-state excitation. ç can
tive skin friction and axial pile vibrations may increase the be estimated using equations (22) and the chart of Fig. 11.
 KINEMATIC PILE BENDING DURING EARTHQUAKES 437
x direction y direction
0·05
–2 m –2 m
computed computed

–0·05

0·05
Ground acceleration, aff: g

–2 m –2 m
recorded recorded

–0·05

0·05
–40 m –40 m
recorded recorded

–0·05
5 10 15 20 25 30 5 10 15 20 25 30
Time, t: s Time, t: s

Fig. 14. Predicted and recorded acceleration time histories of the free-®eld soil response

x direction y direction
0·15
–2 m –2 m
computed computed
Elastic Elastic
0·10 β = 5% β = 5%

50 50
0·05
SHAKE, PI = 30
SHAKE, PI = 30
0

0·15
Spectral acceleration, SA: g

–2 m –2 m
recorded recorded
0·10

0·05

0·15
–40 m –40 m
recorded recorded
0·10

0·05

0
0 0·5 1·0 1·5 0 0·5 1·0 1·5
Period, T: s Period, T: s

Fig. 15. Predicted and recorded 5% damped acceleration response spectra in the free-®eld soil
438 NIKOLAOU, MYLONAKIS, GAZETAS AND TAZOH
10 Peak pile bending strain: 10–5
9 0 1 2 3 4 0 1 2 3 4
Average 0
8 y direction, computed x direction y direction
recorded
7 5
SAsurface/SAbase

A1
6
10 Active A2
5 x direction, pile
recorded length
4 15 Strain A3
compatible
3 PI= 30
20 Strain

Depth: m
2 Elastic compatible
β = 5% PI= 30 A4
1 25 Elastic
β = 5%
0
0 0·5 1·0 1·5 30
Period, T: s A5

Fig. 16. Ratio of response spectra (RSS) at the surface (ÿ2 m) and 35
at rock (ÿ40 m). Note the similar ampli®cation patterns in the x
and y directions 40 A6

45

Computed Measured Measured

(e) Time histories of pile bending strain recorded at six corner pile centre pile
locations along two piles in an instrumented building in
Fig. 18. Measured and predicted peak pile bending strains
Japan compare favourably with results obtained from the
methods developed in the paper. Examination of the
recordings indicates that the prevailing period of the time ACKNOWLEDGEMENT
histories of pile bending strain at depth is close to the The data on the Ervic Building were kindly provided by the
fundamental natural period of the soil deposit. This Institute of Technology of Shimizu Corporation.
elucidates the role of soil response (as opposed to building
response) in the development of kinematic pile bending.
Group effects are of only minor importance for kinematic NOTATION
response. Comparison of recorded and computed strains a0 dimensionless frequency (ˆ ù d=V s )
shows that theory and reality match reasonably well at aff , as , ar free-®eld soil acceleration, surface soil acceleration, rock
elevations deeper than the active pile length. acceleration

x direction y direction
20
–31 m –31 m
computed computed

–20

20
Pile bending strain, εp: 10–6

–31 m, pile B –31 m, pile B

recorded recorded

–20

20
–31 m, pile A –31 m, pile A
recorded recorded

–20
5 10 15 20 25 30 5 10 15 20 25 30
Time, t: s Time, t: s

Fig. 17. Computed and recorded time histories of pile bending strain 31 m below the surface
 KINEMATIC PILE BENDING DURING EARTHQUAKES 439
c (G2 =G1 )1=4 Gazetas, G. & Mylonakis, G. (1998). Seismic soil±structure interaction:
cd Winkler dashpot modulus new evidence and emerging issues. Geotech Earthq. Engng & Soil
d pile diameter Dynamics III, vol. 2, pp. 1119±1174. ASCE.
Es , Ei Soil Young's modulus; Young's modulus in layer i Guin, J. & Banerjee, P. K. (1998). Coupled soil±pile±structure inter-
Ep pile Young's modulus action analysis under seismic excitation. J. Struct. Engng, ASCE
F dimensionless function related to pile bending moment at 124, No. 4, 434±444.
the interface Kagawa, T. & Kraft, L. M. (1980). Lateral load±de¯ection relationships
G soil shear modulus of piles subjected to dynamic loads. Soils Found. 20, No. 4, 19±36.
h, hi soil thickness; thickness of soil layer i Kavvadas, M. & Gazetas, G. (1993). Kinematic seismic response and
Ip pile cross-sectional moment of inertia
p bending of free-head piles in layered soil. GeÂotechnique 43, No. 2,
i ÿ1 207±222.
I u, I ö kinematic response factors Kaynia, A. M. & Kausel, E. (1982). Dynamic stiffness and seismic
k Winkler spring modulus response of pile groups, Research Report R82±03. Massachusetts
L, La pile length, active pile length Institute of Technology.
M pile bending moment Kaynia, A. M. & Mahzooni, S. (1996). Forces in pile foundations under
m mass per unit pile length seismic loading. J. Engng Mech., ASCE 122, No. 1, 46±53.
Nc number of effective excitation cycles Luo, X. & Murono, Y. (2001). Seismic analysis of pile foundations
q SH wavenumber (ˆ ù=V s ) damaged in the January 17, 1995 South Hyogo earthquake by using
r distance from neutral axis to farthest ®ber in pile cross the seismic deformation method. Proc. 4th Int. Conf. Recent Ad-
section vances in Geotechnical Earthquake Engineering and Soil Dynamics,
PI soil plasticity index San Diego.
R radius of curvature Makris, N. & Gazetas, G. (1992). Dynamic pile±soil±pile interaction.
SA spectral acceleration Part II: Lateral and seismic response. Earthquake Engng & Struct.
t time Dynamics 21, No. 2.
T, Ti period; ith natural soil period Mamoon, S. M. & Banerjee, P. K. (1990). Response of piles and pile
Tp predominant period of earthquake record groups to traveling SH-waves. Earthquake Engng & Struct. Dy-
U ff horizontal soil displacement namics 19, No. 4, 597±610.
Y horizontal pile displacement Margason, E. (1975). Pile bending during earthquakes. Lecture, 6 March
V s, V 1, V 2 soil shear wave velocities, soil shear wave velocity in 1975, ASCE-UC/Berkeley Seminar on Design Construction and Per-
layers 1 and 2 formance of Deep Foundations (unpublished).
z depth from soil surface Michaelides, O., Gazetas, G., Bouckovalas, G. & Chrysikou, E. (1998).
Approximate nonlinear analysis of piles. GeÂotechnique, 48, No. 1,
â hysteretic damping coef®cient 33±54.
à dimensionless kinematic response parameter Mineiro, A. J. C. (1989). Simpli®ed procedure for evaluating earthquake
ã soil shear strain loading on piles. In De Mello Volume (eds J. E. Morewira and L.
åp pile bending strain DeÂcourt), CõÂrculo do Livro SA, p. 567.
ç frequency-to-time conversion factor Mizuno, H. (1987). Pile damage during earthquakes in Japan. In
ë Winkler wavenumber parameter Dynamic response of pile foundations (ed. T. Nogami), ASCE
í soil Poisson's ratio Special Publication, pp. 53±78.
r mass density Mylonakis, G., (1999). Analytical solutions for seismic pile bending.
ôc characteristic soil shear stress at layer interface Unpublished research report, City University of New York.
ù cyclic oscillation frequency Mylonakis, G., Nikolaou, A. S. & Gazetas, G. (1997). Soil±pile±bridge
seismic interaction: kinematic and inertial effects. Part I: Soft soil.
Earthquake Engng & Struct. Dynamics 26, 337±359.
NEHRP (1997). Recommended provisions for seismic regulations for
REFERENCES new buildings and other structures. Washington, DC: Building
Banerjee, P. K. (1995). The boundary element methods in engineering. Seismic Safety Council.
McGraw-Hill, New York. Newmark, N. M. (1968). Problems in wave propagation in soil and
Barghouthi, A. F. (1984). Pile response to seismic waves. PhD disserta- rock. Proceedings of the international symposium on wave propaga-
tion, University of Wisconsin, Madison. tion and dynamic properties of earth materials, University of New
Bertero, V., Lin, T. Y., Seed, H. B., Gerwick, B. C., Brauner, H. A. and Mexico Press.
Fotinos, G. C. (1974). Aseismic design of prestressed concrete Nikolaou, A. S. & Gazetas, G. (1997). Seismic design procedure for
piling. Proc. 7th Cong. Federation Internationale de la Precon- kinematically loaded piles. Proc. 14th Int. Conf. Soil Mech. Found.
trainte, New York. Engng, Hamburg, Special Volume, ISSMFE TC4 Earthquake geo-
Blaney, G. W., Kausel, E. & Roesset, J. M. (1976). Dynamic stiffness of technical engineering, 253±260.
piles. Proc. 2nd Int. Conf. Num. Methods Geomech., Blacksburg VA Nikolaou, A. S., Mylonakis, G. & Gazetas, G. (1995). Kinematic
2, 1010±1012. bending moments in seismically stressed piles. Report NCEER-95-
CNEL-ENEL (1976). Contribution to the study of the Fruili earthquake 0022, National Center for Earthquake Engineering Research. Buffa-
of May 1976, Rome. lo: State University of New York.
Dobry, R. & Gazetas, G. (1988). Simple method for dynamic stiffness and Nishizawa, T., Tajiri, S. & Kawamura, S. (1984). Excavation and
damping of ¯oating pile groups. GeÂotechnique 38, No. 4, 557±574. response analysis of a damaged rc pile by liquefaction. Proc. 8th
Dobry, R. & O'Rourke, M. J. (1983). Discussion on `Seismic response World Conf. Earthquake Engng 3, 593±600.
of end-bearing piles' by Flores-Berrones, R. & Whitman, R. V. J. Nogami, T., Jones, H. W. and Mosher, R. L. (1991). Response analysis
Geotech. Engng Div., ASCE p. 109. of pile-supported structure assessment of commonly-used approxima-
Dobry, R., Vincente, E., O'Rourke, M. J. & Roesset, J. M. (1982). tions, Third International Conference on recent advances in Geotech-
Horizontal stiffness and damping of single piles. J. Geotech. Engng nical Engineering and soil dynamics, University of Missouri-Rolla,
Div., ASCE 108, No. 3, 439±459. Missouri, 931±940.
EEFIT (1986). The Mexican earthquake of 19 September 1985. London: Novak, M. (1991). Piles under dynamic loads: state of the art. Proc.
Institution of Civil Engineers. 2nd Int. Conf. Recent Adv. Geotech. Earthquake Engng Soil Dy-
Eurocode EC-8 (1996). Structures in seismic regions, Part 5: Founda- namics, St Louis 3, 2433±2456.
tions, retaining structures, & geotechnical aspects. Commission of Novak, M., Nogami, T. & Aboul-Ella, F. (1978). Dynamic soil reaction for
the European Communities. plane strain case. J. Engng Mech. Div., ASCE 104, No. 4, 953±959.
Fan, K., Gazertas, G., Kaynia, A., Kausel, E. and Ahmad, S. (1991). Okamoto, S. (1983). An introduction to earthquake engineering, 2nd
Kinematic seismic response of single piles and pile groups. J. edn. University of Tokyo Press.
Geotech. Engng Div., ASCE 117, No. 12, 1860±1879. Pappin, J., Ramsey, J., Booth, E. & Lubkowski, Z. (1998). Seismic
Fan, K. (1992). Seismic response of pile foundations evaluated through response of piles: some recent design studies. Proc. Inst. Civ. Engrs,
case histories. PhD thesis, State University of New York at Buffalo. Geotech. Engng 131, 23±33.
Flores-Berrones, R. & Whitman, R. V. (1982). Seismic response of end- Pender, M. (1993). Aseismic pile foundation design analysis. Bull. NZ
bearing piles. J. Geotech. Engng Div., ASCE 108, No. 4, 554±569. Nat. Soc. Earthquake Engng 26, No. 1, 49±160.
Gazetas, G. & Dobry, R. (1984). Simple radiation damping model for piles Penzien, J. (1970). Soil±pile foundation interaction. In Earthquake
and footings. J. Geotech. Engng Div., ASCE 110, No. 6, 937±956. engineering (ed. R. L. Wiegel), Ch. 14. New York: Prentice Hall.
 440 NIKOLAOU, MYLONAKIS, GAZETAS AND TAZOH
Poulos, H. G. & Davis, E. H. (1980). Pile foundation analysis and Tajimi, H. (1969). Dynamic analysis of structure embedded in elastic
design. John Wiley & Sons. stratum. Proc. 4th World Conf. Earthquake Engng, Santiago, 53±69.
Randolph, M. F. (1981). The response of ¯exible piles to lateral loading. Tazoh, T., Shimizu, K. & Wakahara, T. (1987). Seismic observations
GeÂotechnique 31, No. 2, 247±259. and analysis of grouped piles. In Dynamic response of pile founda-
Roesset, J. M. (1977). (eds C. S. Desar and J. T. Christian). Soil tions: experiment, analysis and observation, Geotechnical Special
ampli®cation of earthquakes. Numerical Methods in Geotechnical Publication No. 11. ASCE.
Engineering, McGraw-Hill, 639±682. Technical Council on Lifeline Earthquake Engineering (TCLEE) (1998).
Roesset, J. M. (1980). Stiffness and damping coef®cients of foundations. Seismic guidelines for ports (ed. S. D. Werner). ASCE.
Proc ASCE Geotech. Engng Div. Nat. Conv. 3, 1±30. Veletsos, A. S. & Dotson, K. W. (1986). Impedances of soil layer with
Ross, G. A., Seed, H. B. & Migliaccio, R. (1969). Bridge foundations in disturbed boundary zone. J. Geotechn. Engng, ASCE 112, No. 3.
the Alaska earthquake. J. Soil Mech. Found. Engng, ASCE, 95. Veletsos, A. S. & Ventura, C. E. (1984). Ef®cient analysis of dynamic
Schnabel, P. B., Lysmer, J. & Seed, H. B., (1972). SHAKE: A computer response of linear systems. Earthquake Engng & Struct. Dynamics
program for earthquake response analysis of horizontally layered 12, 521±536.
sites, EERC 72-12. Berkeley: University of California. Velez, A., Gazetas, G. & Khrishnan, R. (1983). Lateral dynamic
Scott, R. F. (1981). Foundation Analysis, Prentice Hall, New York. response of constrained-head piles. J. Geotech. Engng Div., ASCE
Sheta, M. and Novak, M. (1982). Vertical vibration of pile groups, 109, No. 8, 1063±1081.
Journal of the Geotechnical Engineering Division, ASCE, 108, GT4, Vucetic, M. and Dobry, R. (1991). Effect of soil plasticity on cyclic
570±590. response. J. Geotech. Engng Div., ASCE 117, No. 1, 89±107.

View publication stats