Вы находитесь на странице: 1из 12

Soil Dynamics and Earthquake Engineering 46 (2013) 5263

Contents lists available at SciVerse ScienceDirect

Soil Dynamics and Earthquake Engineering


journal homepage: www.elsevier.com/locate/soildyn

Piles-induced ltering effect on the Foundation Input Motion


Raffaele Di Laora a,n, Luca de Sanctis b
a
b

Department of Civil Engineering, Second University of Napoli, Aversa (CE), Italy


Department of Technologies, University of Napoli Parthenope, Napoli, Italy

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 13 September 2012
Received in revised form
8 December 2012
Accepted 10 December 2012
Available online 22 January 2013

The inertial interaction analysis of a structure founded on piles is conventionally performed by


imposing that the Foundation Input Motion is merely that of the free eld, thus neglecting the
kinematic interaction between piles and soil generated by the passage of seismic waves. This would
lead to unnecessary overconservatism in the design, as there is evidence that the free-eld motion may
be thoroughly ltered out by piles (generally reduced), especially in the case of soft soils, where piles
are recurrently required to carry out the total load transmitted by the superstructure and/or to reduce
foundation settlements. Results provided from analytical and numerical tools elucidate the crucial
aspects controlling the mechanism of ltering effect. Reduced design spectra are also suggested to
account for the benecial effect coming from the piles when the inertial interaction analysis of the
superstructure is being performed.
& 2012 Published by Elsevier Ltd.

1. Introduction
The response of a structure to an earthquake is usually
predicted assuming that the support motion at the foundation
level is merely that of the free-eld. However, the superstructure
interacts with its foundation and the surrounding soil, creating
additional soil deformations, that add to those generated from the
passage of seismic waves, so as the motion in the vicinity of the
foundation can differ substantially from that of the free-eld.
Assuming a linear soil-foundation-superstructure response, the
analysis of the complete system can be performed according to
three consecutive steps: (i) predict the motion of the foundation
in the absence of the superstructure, i.e. the so-called Foundation
Input Motion (FIM); (ii) determine the dynamic impedance
functions associated to swaying, vertical, rocking and cross
swaying-rocking oscillation of the foundation; (iii) evaluate the
response of the superstructure supported on the springs and
dashpots and subjected to the motion of the foundation determined at the rst step. This procedure is commonly referred to in
literature as kinematic-inertial decomposition or substructure
method [13]. Once the response of the structure has been
computed, the pilesoil interaction effects can be readily calculated by superposition of kinematic and inertial effects. This
method has been extensively adopted to study the response of
structures and foundations under seismic excitation and is truly
attractive as alternative to fully 3D analyses involving the

Corresponding author. Tel.: 00390815010385.


E-mail address: raffaele.dilaora@unina2.it (R. Di Laora).

0267-7261/$ - see front matter & 2012 Published by Elsevier Ltd.


http://dx.doi.org/10.1016/j.soildyn.2012.12.007

complete pilesoil-superstructure interaction, that are very complex and rarely performed in engineering practice [4].
Dynamic impedance functions at the foundation level are a
fundamental ingredient for inertial interaction analysis of both the
superstructure and the piled foundations. Impedance functions of
soil foundation systems show frequency dependent characteristics. However, the analysis of structures is currently oriented
towards performance based criteria, for which non-linearity of
structural members is mandatory. In this case, the structural
analyses cannot be but performed in time domain, and the
frequency dependency of impedance functions makes problematic
the numerical computations. To overcome this difculty, a better
choice is represented by the so-called lumped-parameter models,
LPMs [5], capable of accounting for frequency dependency of
impedance functions. LPMs can be easily incorporated into a
non-linear analysis of the structure. An early application of this
type of model was presented by Ciampoli and Pinto [6] to analyze
the response of bridge piers. A similar work has been recently
published by Carbonari et al. [7]. They performed a non-linear
inertial analysis in time domain of a wall-frame structures with
concentrated plasticity by using LPMs to model pile group foundations embedded in a two-layer soil, focusing on the comparative
behaviour of compliant vs xed base models. It is argued from
these studies that compliance-base models behave quite differently from xed base models because SSI has a remarkable effect
on the dissipative behaviour of the structure. One out of the
limitation of LPMs is the inability to model impedance functions
whose components are neither quadratic nor linear with frequency. A second option that has received great attention is the
non-linear macro-element approach, where the foundation and

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

List of symbols
Latin symbols
a0
ap, as
c
Es, E1
Ep
f
f1
fm, fp
FIM
h1, h2
H
Ip
Iu
k
L
ML
q
T
T1

dimensionless frequency
pile top acceleration, free-eld surface soil
acceleration
Winkler dashpot coefcient
soil Youngs modulus, Youngs modulus in layer 1
pile Youngs modulus
excitation frequency
fundamental natural frequency of soil (site frequency)
mean and predominant frequency of earthquake
record
foundation input motion
thickness of the rst (surface) soil layer, thickness of
the second soil layer
thickness of the soil deposit
pile cross-sectional moment of inertia
translational kinematic response factor
Winkler spring modulus
pile length
local magnitude of the earthquake event
soil wavenumber
structural period
site period

the soil are represented by a single macro-element, NLME, which


is capable to predict non-linear permanent deformations of the
foundation in terms of settlement, sliding and rotation. As a
consequence of dissipative effects at the soil foundation level,
the seismic demand of the superstructure may also be reduced
signicantly. The early formulation of a macro-element was
published by Nova and Montrasio [8]; since then a number of
NLEMs have been proposed, like those suggested in Refs. [9,10],
capable to account for permanent deformation related dissipative
effects. The use of NLEMs for predictive analyses has been limited
to shallow foundations and homogenous soil conditions, as in this
case the availability of well documented 1 g and centrifuge tests
allowed for model validation.
Despite the modern trend towards complex SSI models, the
ltering effect on the support motion at the foundation level
generated from the piles has not been adequately addressed. The
most attractive application of the substructure method is to
assume that the support motion equals the free-eld seismic
motion. By contrast, the free-eld motion is ltered out by the
piles, especially in the case of soft soils, where piles are recurrently required to increase the bearing capacity of the foundation
and/or to reduce settlements [11,12], but this potential has not
been yet exploited in engineering practice. The goal of this paper
is threefold: (i) to outline the importance of the ltering effect for
analyzing efciently the dynamic response of the superstructure;
(ii) to offer insight into the ltering action exerted by the piles;
(iii) to propose a correction to design spectra that can be of
assistance for seismic risk reduction strategies.

2. Literature review
Pilessoil kinematic interaction has been addressed by many
researchers in the last decade. Emphasis has been placed on
kinematic bending effects, evaluated on the basis of analytical
studies [1316] and numerically-based parametric analysis
[1721]. As a result of this research effort, a number of ready-

53

Tmin, Tcrit characteristic structural periods


Vs, Vs1, Vs2 soil shear wave velocity, soil shear wave velocity in
layer 1 and 2
Greek symbols
soil hysteretic damping coefcient
dimensionless response coefcient
dimensionless transient response parameters
Winkler stiffness coefcient ( k/Es)
Winkler wavenumber
pile characteristic wavelength
Wavelength in the soil medium
n1, n2
soil Poissons ratio in layer 1 and 2
np
pile Poissons ratio
rs
soil mass density
x0, xmin, xcrit characteristic spectral ratios for transient motions
x0,av, xmin,av characteristic average spectral ratios
rp
pile mass density
o
cyclic excitation frequency (2pf)
os
characteristic cyclic excitation frequency of earthquake record
os, av
average characteristic cyclic excitation frequency of
earthquake records employed

bs
G
Gt1, Gt2
d
l
lp
ls

to-use methods are available to predict kinematic bending effects


at both the interfaces between layers and the pile head. The
problem of the ltering effect exerted by the piles has also
received some attention. The existence of ltering effect has
conrmed through published works referring to theoretical
[2229] and experimental evidence [2,3032], even if this observation has not been so far taken into consideration in engineering
practice.
Filtering effect was examined in the seminal work by FloresBerrones and Whitman [22], who expressed the ratio of pile and
soil accelerations as a function of excitation frequency in a closed
form solution, for the case of an innitely-long pile in homogeneous halfspace. As an outcome, piles were found to reduce
seismic motion with increasing excitation frequency and pile
diameter, whereas an increase in soil stiffness leads to a decrease
in the ltering effect.
In general the support motion at the foundation level is
different from that of the free-eld because of the scattered wave
eld generated from the difference between pile and soil rigidities. For motions that are rich in high frequency components,
even practically exible piles may not be able to follow the wavy
movements of the free-eld. On the other hand, if low frequency
components of the input motions are predominant, the scattered
eld is weak, and the support motion can be expected to be
approximately equal to that of the free-eld [23,2527].
Fan et al. [25] extended the investigation on ltering effect to
pile groups under steady-state conditions. They found that group
effects, although clearly depending on pile spacing, are not
relevant for lateral vibrations. By contrast, they may strongly
affect the rotational component of motion, which, however, lies
beyond the scope of this paper.
All theoretical studies mentioned above highlight the prominent role exerted by excitation frequency. Such a result is
conrmed by the experimental evidence about the ltering effect
available in the literature.
Kawamura et al. [30] have reported the case history of a 7-storey
residential building in Japan, as shown in Fig. 1. Acceleration

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

5,63

Plan view

4,72
1,40

13,59

1,84

54

8,36

8,36

8,36

5,63
55,79

8,36

8,36

8,36

Acceleration BUILDING LINE T = 0.33 s (RF-1F)


T = 0.20 s (Underground)
Acceleration SOIL LINE
Velocity
T=5s
Recorder

RF

Land Side
N-Value
0 10 20 30 40 50
Sand

Sea Side
1F

N-Value
0 10 20 30 40 50 60

GL

Sandy
Silt

S04

Silty
Sand

10

S12

12

10

Sand

Sand

Fine
Sand

15

15
4

Sand

20

20
S24

24

Clay
Sand

15

Fig. 1. Case history: plan view and elevation of the building (modied from Kawamura et al. [30]).

recordings were available since 1971 at two vertical alignments, the


building line and the soil line. By comparing the records of the two
alignments during 20 earthquakes, they found that the maximum
amplication at the ground surface was 1.5 times higher than the
one recorded at the base slab of the building. They also plotted the
Fourier spectral ratio between the building line and the soil line at
the level of the base slab, concluding that for structural periods
smaller than 0.3 s the ratio of the two accelerations is 0.5 as an
average. With increasing structural periods the Fourier spectral
ratios were found to be about unity. It is argued from this case
history that the high frequency components of the free-eld motion
are ltered out by the pilesoil-superstructure interaction.
Tajimi [31] described an example of (large diameter) bored
piles for which acceleration recordings were available at both the
pile-cap and at the ground surface on a far-distant axis representing free-eld conditions. By comparing the acceleration spectra of
the two motions he noticed that the spectrum of the far distant
motion included frequency contents higher than those composing
the pile cap motion, in agreement with the conclusion by
Kawamura et al. [30]. The same example has been later reported
by Otha et al. [32] and Gazetas [2]. This last work is particularly
signicant for the scope of this study. The example under
examination consists of an 11-storey building supported on
cast-in-place piles and founded on an alluvial deposit of alternating layers of sand and silt, as shown in Fig. 2. The foundations
system response was monitored during seven earthquakes
through 27 accelerometers that were placed on the pile building
axis and on a far-distant axis representing free-eld soil conditions. Fig. 2 rst published by Gazetas [2] shows the complete
eld recordings supplied by the 27 accelerometers in the form of
ratios between Fourier amplitude spectrum of the foundation
motion over that of the free-eld. It is worthy of note that
frequencies in the range between the fundamental frequency of
the soil deposit, f1, and the fundamental frequency of the superstructure, fst, are amplied due to pilesoil-structure interaction.

In addition, high frequency components of the seismic motion


(f 41.5f1) are ltered out by the pilesoil-structure interaction,
while low frequency components (f of1) are not affected by both
the piles and the structure. However, at frequencies larger than
1.5f1, there is no evidence however of the amount of the ltering
effect generated from the piles, as the recorded accelerations are
the result of the pilesoil-structure interaction (i.e., the sum of
kinematic and inertial interaction). Nevertheless, this case history
clearly demonstrates that piles have the potential of ltering out
the high frequency components of the input motion.
Makris et al. [33] have reported the case history of Painter
Street Bridge, in northern California, shaken on 25 April 1992 by
the Petrolia earthquake (with Magnitude ML 7.1 and 18 km from
the fault). Painter Street Bridge is a continuous, two span, 16 m
wide, cast in place prestressed bridge. One span is about 45 m
and the other 36 m (Fig. 3). It is about 16 m wide. The bent is
supported by two pile-groups, each consisting of 20 (4  5) driven
concrete piles. The bridge was instrumented in 1977. Motions
were recorded in all accelerographs, including the one in the freeeld, that on the footing of one pier, that above the pier at the
underside of the bridge girder and that on the deck near the west
abutment. The shear wave velocity has an average value of
Vs 225 m/s, and the pile diameter was only 0.36 m; the dimensionless frequency, a0 2pfd/Vs, is of the order of only 0.1. From
studies on vertical propagating shear waves in homogeneous soil
deposits [25], the authors concluded that the support motion is
practically equal to that of the free eld. Considering the free eld
input motion, they found a very satisfactory agreement between
the computed and the recorded response of the pile cap and the
bridge deck. It is worthy of note that the soil consists of
moderately stiff/dense soil layers, with the SPT blowcounts
ranging between 8 near the surface and 34 at 10 m depth and
the underlying layer being a very dense gravelly sand. This is the
case of a exible pile embedded in a moderately stiff/dense soil
deposit. In addition, the free-led motion has a low-frequency

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

content. Under this circumstances there is evidence that the


ltering effect is negligible.
The limited amount of experimental evidence demonstrates
that the frequency content of the input signal might exert a

55

remarkable inuence on the support motion. The parametric


study reported in this work conrmed the trend coming from
these eld observations.

3. The mechanism of ltering effect


Following the contributions by Flores-Berrones and Whitman
[22], Makris and Gazetas [26] and Nikolaou et al. [17], it may be
shown that in a homogenous halfspace the ratio between the
acceleration atop a xed-head innitely long pile, ap, and that at
the soil surface, as, is given by:

8.35 m Plan view


78 m
Accelerometer

Iu

Pressure gauge
Pore water
pressure gauge
30

depth [m]

Vs [m/s]
100

200

300

where G is a complex-valued interaction factor which may be


approximately expressed as [34]:

400

10

GC

20
30
100

f1 = fundamental frequency of the soil stratum


f1

pile-soil acceleration ratio

ap
G
as

fst

fst = fundamental frequency of the superstructure

4l

4l q4

in which q ( o/Vs) is the wavenumber of the harmonic SH wave


travelling in the soil, o the excitation circular frequency, and l is
the well-known Winkler parameter, which may be faithfully
approximated through the following expression, as the term
associated to inertia contribution of the pile mass is neglected:

1
k ioc 4
lC
3
4Ep Ip
Ep and Ip being pile Youngs modulus and cross-sectional moment
of inertia. The Winkler spring reaction modulus k may be taken as
proportional to soil stiffness Es by a coefcient d, which typically
assumes values close to unity [35], whereas the dashpot coefcient is expressed as [23]:

Recorded Fourier
amplitude ratio

c 6o4 rs V 4s d4 2bs k=o

3
frequency [Hz]

Fig. 2. Case history: (a) plan and section of the building; (b) recorded ratio of
Fourier amplitude spectra atop the pile and at the free-eld ground motion
(modied from Gazetas [2]).

Fig. 3. Cross section of the Painter Street Bridge (modied from Makris et al. [3]).

in which d is the pile diameter, rs and bs the density and the


hysteretic damping ratio of the soil. Eq. (2) is plotted in Fig. 4a
against excitation frequency for d 1.2 and different values of
stiffness ratio Ep/Es. Pile-to-soil acceleration ratio is always
smaller than unity (i.e. piles play always a benecial role in
reducing the seismic motion that excites the superstructure) and
decreases with frequency. This effect is physically due to the
resistance that the pile offers in adapting to the short wavelengths in the soil, as the dimensionless frequency a0 may be
interpreted as the ratio of pile diameter d and soil wavelength ls
( Vs/o). Moreover, at a given frequency pilesoil acceleration
ratio is progressively smaller as pilesoil stiffness ratio increases.
The physical interpretation of this phenomenon deserves special
discussion as reported below.
The ability of the pile to follow soil displacements is physically
connected to both pile diameter and pilesoil stiffness ratio. A

Fig. 4. Acceleration ratio as function of excitation frequency for the case of an innitely-long pile (Iu G).

56

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

characteristic pile wavelength may be therefore dened as:


 14
E
lp d p
Es

It is straightforward to verify that the real part of the


wavenumber l in Eq. (3) is proportional to the reciprocal of the
above dened wavelength lp. Therefore, assuming d 1 and
neglecting the term ioc in Eq. (3), and substituting Eq. (5) into
Eq. (2) the amplitude of G can be expressed as:
"

4 #1
1 olp
GC 1
6
20 V s
The new dimensionless parameter olp/Vs can be viewed as the
ratio of pile characteristic wavelength, lp, and soil wavelength, l.
Eq. (6) is plotted in Fig. 4b against olp/Vs. The curve expressed by
Eq. (2) for d 2 is also added for comparison. It can be seen that
the ltering effect, represented by the kinematic interaction
factor G, is governed by the ratio of the two wavelengths. The
larger is the characteristic wavelength of the pile compared to the
soil wavelength, the larger is the amount of the ltering effect.

4. Parametric analysis
To investigate quantitatively the ltering effect exerted on the
Foundation Input Motion by a piled foundation, a comprehensive
set of Finite Element analyses has been performed. Owing to the
second-order inuence of group effects for lateral vibration [25], a
xed-head single pile embedded in a two-layer soil has been
considered, as shown in Fig. 5a. Harmonic S-waves applied at the
bedrock level and propagating up and down (after reection)
constitute the seismic excitation.
According to the Buckingham theorem [36] pilesoil interaction is governed by 7 dimensionless ratios (H/d, L/d, Ep/E1, Vs2/Vs1,
h1/d, od/Vs1, bs). Despite the fact that only the complete set of
such parameters would sufce to control kinematic response,
some simplications are possible and were thereby employed in
this study. Specically: (a) considering long piles, whose length is
greater than the active one, as shown in [37], the rst two
dimensionless parameters do not affect pilesoil interaction;
(b) damping ratio does affect both pile and soil response, yet
not their ratio as proven in the ensuing, so that damping level
may be set at a constant value; (c) in light of the analytical
derivations reported above, dimensionless frequency od/Vs1 and
pilesoil stiffness ratio Ep/E1 may be combined to give the unique
parameter olp/Vs1. Kinematic harmonic response of long piles in

1 2

two-layer soils may be therefore conveniently expressed as


function of the 3 parameters h1/d, Vs2/Vs1 and olp/Vs1. Consequently, interface depth h1/d assumes values of 5, 10, 15 and 19
whereas layer impedance contrast is considered equal to 2 and 4.
The shear wave velocity in the rst layer is taken equal to 50 and
100 m/s, corresponding to pilesoil stiffness ratios of about 600
and 2500, respectively.
To investigate the response to real earthquakes a geometrical
quantity (e.g., total height of the soil deposit) and a material
stiffness (e.g. pile Youngs modulus) have to be dened. Accordingly, the total height H is set to 30 m, whereas the pile
has diameter d 1 m, length L 20 m and Youngs modulus
Ep 30 GPa. Soil and pile density are set to 1.75 and 2.5 Mg/m3,
respectively, whereas the value of soil Poissons ratio is 0.4.
The parametric analysis has been performed assuming a viscoelastic soil model, which is intrinsically unable to predict real
soil behaviour. Nevertheless, the results coming from this assumption have a general validity. Martinelli [4] has performed a study of
the kinematic response of piles embedded in a two layer soil
modelled through the advanced consititutive model by Dafalias
and Manzari [38]. Results of this study, undertaken by nite
element analysis, show that kinematic interaction effects can be
predicted within reasonable accuracy even assuming that the soil
behaves like a viscoelastic material, provided that the soil stiffness
prole is preliminary evaluated through an equivalent linear viscoelastic analysis, or ELE analysis. This conclusion, however, only
applies by taking a soil permeability larger than 10  4 m/s or
assuming drained soil conditions. Under circumstances where soil
permeability is lower than 10  4 m/s, it is necessary to perform a
coupled analysis to achieve a successful prediction of earthquakeinduced kinematic interaction effects. It is worthy of note, however, that even for medium to dense sand, the excess pore
pressures lead to a reduction of the soil stiffness pertaining to
the upper layer. In this respect, the ltered motion predicted by the
drained analysis represents an upper bound of that predicted by
the coupled analysis and can be safely employed in the evaluation
of the inertial response of the superstructure.

4.1. Numerical analyses


Although the problem is 3D, the geometry is axisymmetric
whereas the load is anti-symmetric. To simplify the analysis,
stresses and displacements are expanded into a Fourier series in
the circumferential direction, according to the technique introduced by Wilson [39]. For the example at hand, only the rstorder term of the series is needed. Owing to this procedure, the

Vs/ Vs1

h1

Vs1

h2

Vs2

Bedrock

SH

SH

Fig. 5. (a) Problem considered; (b) nite element mesh employed in the analyses.

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

original three-dimensional problem is conveniently reduced to a


2D one [40].
Numerical analyses were conducted using the commercial FE
code ANSYS [41]. Four-noded axisymmetric 2D elements are used
to mesh soil and pile. The assumption of a line of anti-symmetry
implicitly implies the use of periodic or tied boundaries. As a
consequence, the model under examination does not allow for the
transmission of outgoing waves generated at the pilesoil interface, which remains trapped inside the model. A common procedure to check the amount of inaccuracy coming from this choice
is the one originally suggested by Zienckiewicz et al. [42] for both
linear and non linear models. It consists of comparing the results
of the complete model with those predicted by the free-eld
conditions at the location of the boundary. For the problem under
examination, Di Laora et al. [43] have shown that results coming
from the assumption of tied boundaries are sufciently accurate,
provided that the width of the entire model is large enough.
To this aim, the lateral boundary of the model was set at 400 d
from the pile axis. Vertical displacements are restrained along the
lateral boundary of the mesh, while nodes at the base of the
model are restrained to both horizontal and vertical directions to
represent a rigid bedrock. The mesh employed in the analyses is
shown in Fig. 5b. The vertical size of the elements is set to
0.1 diameters for a width of 3 diameters close to layers interface
and 0.5 diamaters elsewhere. On the other hand, the horizontal

57

size is 1/8 pile diameters at pilesoil interface, thereby increasing


with radial distance up to 1.5 diameters at the free-eld. The
analyses performed in the parametric study were carried out in
the frequency domain, after extracting vibrational modes with
frequency up to 25 Hz, each having a pre-specied level of viscous
damping. In this way the drawbacks stemming from use of
common energy loss formulations such as Rayleigh damping
were avoided. An FFT algorithm was employed to transfer
responses from the frequency to the time domain and vice versa.

5. Harmonic response
Eq. (6) is compared with results from FE analysis in Fig. 6a.
Despite its simplicity, graph formula (6) provides accurate values
of pilesoil acceleration ratios. The advantage of the above
expression is twofold: (a) it offers an insight into the physical
interpretation of the interaction phenomenon; (b) it condensates
into a unique dimensionless parameter a number of physical
quantities such as pile diameter and Youngs modulus, soil
stiffness and density as well as excitation frequency.
The ability of such a parameter to describe the ltering effect
mechanism is conrmed in Fig. 6b, where results from Fan et al.
[25] are compared to pilesoil acceleration ratios obtained by FE
analyses for different types and levels of damping. It is noted that

Fig. 6. Acceleration ratio for different damping levels and types. Comparison between FE results from present work and those by Fan et al. (1991). In all cases, L/d 20, H/L1.5.

Fig. 7. Acceleration ratio in two-layer soil for different interface depth and soil stiffness. (a) h1/d5, (b) h1/d10, (c) h1/d15 and (d) h1/d19.

58

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

although both soil and pile responses are known to be dependent


of damping (primarily the level and secondarily the type), the
acceleration ratio is not.
The above considerations are valid under homogeneous soil
condition. From a conceptual point of view a hypothetical stiffer
layer would act as a partial xity for the embedded portion of pile.
It is hence expected that in two-layer soils pilesoil acceleration
ratio may deviate from Eq. (6), yet the entity of such perturbation
should vanish for smooth stiffness contrasts between soil layers
and deep layers interfaces. Fig. 7 depicts FE results in terms of
pilesoil acceleration ratio for a two layer soil, by varying interface depth, pilesoil stiffness ratio and layer stiffness contrast. It
is noted that for deep interfaces (h1/d Z10) Eq. (6) may be still
adopted for two-layer soils, at least for practical applications.

6. Transient response

spectral acceleration of the foundation over that of the free-eld


(Fig. 10c and d).
The spectral acceleration ratio at T0, x0, is a purely kinematic
interaction factor and is strongly dependent of the frequency
content of the input signal, as for xmin. For any particular earthquake event, it is convenient to dene an average circular
frequency os as:

os

2pf m 2pf p
2

The kinematic interaction factor, x0, is plotted in Fig. 11a


against the parameter:


o l 1
Gt1 1 0:15 s p
8
V s1
Fig. 11b shows the relationship between the minimum acceleration ratio, xmin, and the quantity

Gt2 2:5Gt1 1:5


Transient response is investigated by using real acceleration
time-histories, selected from an Italian database [44] and the
European database of Ambraseys et al. [45]. Time histories and
Fourier spectra of the signals are reported in Fig. 8. Also shown in
this graph are frequency content indices like the mean frequency
fm as dened by Rathje et al. [46] and the predominant frequency
fp, dened as the structural natural frequency corresponding to
the maximum spectral acceleration.
Some results of the parametric study are illustrated in Fig. 9,
where the spectral acceleration normalized by the peak rock
acceleration is plotted as a function of the structural period for
two different earthquake events, specically ULCINJ (Montenegro,
1979), characterized by low-frequency components, and NOCERA
UMBRA (Umbria-Marche, 1997), rich in high frequencies. The
comparison between the results coming from these two earthquakes allowed to investigate the role of the frequency content.
The ltering effect is remarkable for both these events when the
shear wave velocity of the upper layer, Vs1, is equal to 50 m/s. By
contrast, for Vs1 100 m/s, the ltering effect is negligible for
ULCINJ, and is less pronounced (but still signicant) for NOCERA
UMBRA. This is due to the fact that high-frequency components
are ltered out by the piles. The plots in Fig. 9 show that the
governing parameters of the ltering-effect are the soil stiffness
and the frequency content of the input signal. Note that this is in
agreement with the harmonic response shown in Figs. 4, 6 and 7.
An interesting result is that illustrated in Fig. 10, where the ratio x
of the spectral acceleration of the ltered motion, Sa,p, over that of
the free-eld Sa,s is plotted against the structural period. For a
given subsoil prole and any particular earthquake event it is
possible to recognize two critical points in (T:x) plane, T being the
structural period: (i) that corresponding to the minimum value of
the spectral acceleration ratio; (ii) that pertaining to the structural period after which the ltering effect becomes negligible.
The rst point has coordinates (Tmin, xmin), as reported in Fig. 10d,
while the second point has coordinates (Tcrit, xcrit) and can be
identied with the point of maximum curvature of the spectral
ratio function. The ordinate of the point at which the structural
period equals the critical value is nearly coincident with unity, so
as the coordinates of the second point can be conveniently
assumed to be (Tcrit, 1). The most interesting result is that for
any subsoil the structural periods Tcrit and Tmin are practically
unaffected by the earthquake event, as it will be discussed later.
On the other hand xmin strongly depends on the frequency
content of the input signal. The comparison between Fig. 10a
and b also shows that the amount of the ltering effect is
remarkably affected by the shear wave velocity of the upper
layer. By contrast, the layer impedance contrast, i.e. the heterogeneity of the soil prole, has only a little effect on the ratio of the

Note the similarity of the parameter Gt1 in Eq. (8) with the G
factor in Eq. (6). Gt1 may be therefore interpreted as a transient
kinematic interaction factor, whose coefcients naturally arise
from the choice in the denition of the frequency index os.
The results plotted in Fig. 11 indicate a clear trend for both x0
and xmin of increasing with the inverse of the frequency index os,
as expected. Based on linear regressions, the parameters x0 and
xmin can be calculated through the following expressions:

x0 1:71Gt1 0:64

10

xmin 0:91Gt2

11

Despite the amount of scattering visible in Fig. 11, the accuracy


of both the correlations is worthy of note, taking into account that
the frequency content of any earthquake event has been synthesized by a unique parameter, the circular frequency os dened by
Eq. (7).

7. Reduction of design spectra


In the previous section it has been shown that transient results
do not indicate a dependence of Tmin and Tcrit on the frequency
content of the input signal. This result could be actually anticipated by simple dimensional considerations. Indeed, considering
that such periods (which in dimensional terms represent a time)
clearly depend on soil shear wave velocity, which already contains the dimensions of time, excitation frequency cannot affect
Tmin and Tcrit. It may be therefore convenient to condensate the
results of the nite element analyses in terms of mean spectral
ratios, as in Fig. 12. For both Vs1 50 m/s and Vs1 100 m/s the
mean spectral ratios have an upper bound, corresponding to the
homogeneous subsoil (Vs2/Vs1 1). This can be physically motivated by the kinematical restraint exerted by the stiffer lower
layer. The larger is the degree of restraint exerted by the lower
layer, the greater is the resistance of the pile to follow the
deformations of the upper layer. On the other hand, the layer
stiffness contrast has only a weak effect on the structural periods
Tmin and Tcrit. Thus, Fig. 12 can be further synthesized, by referring
only to the upper bound represented by the homogeneous
subsoil. The results pertaining to the homogenous soil condition
are illustrated in Fig. 13, where the mean spectral ratios obtained
for Vs 75 m/s and 125 m/s have been also added for comparison.
The mean spectral curve of the homogeneous subsoil is wellrepresented by two parabolic functions, intersecting at TTcrit,
and a horizontal plateau. The vertex of the rst parabola has
abscissa T0, while the vertex of the second parabola has
coordinates (Tcrit, 1).

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

59

Fig. 8. Acceleration time-histories considered in the parametric analysis.

Interestingly, it is easy to verify that transient parameters Gt1


and Gt2 work very well even by employing a mean frequency
content. Specically, x0,av and xmin,av satisfy the following
equations:

x0,av Gt1
xmin,av Gt2

)
13

with

os os,av 10 rad=s

14

The characteristic periods Tmin and Tcrit may be taken as


equal to:
T min 12

d
Vs

15

60

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

Fig. 9. Spectral acceleration normalized by the peak rock acceleration. In all cases, h1/d 10 and Vs2/Vs1 4.

Fig. 10. Pile-head spectral acceleration over that of the free-eld. (a) h1/d10, Vs150 m/s, Vs2/Vs14, (b) h1/d10, Vs1100 m/s, Vs2/Vs14, (c) h1/d15, Vs150 m/s, Vs2/
Vs12 and (d) h1/d15, Vs150 m/s, Vs2/Vs14.

Fig. 11. Linear regressions for x0 and xmin.

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

61

Fig. 12. Mean spectral ratios for the subsoils considered in the parametric study. (a) Vs150 m/s and (b) Vs1100 m/s.

Fig. 13. Mean spectral ratios for homogeneous soil conditions. (a) Vs50 m/s, (b) Vs75 m/s, (c) Vs100 m/s and (d) Vs125 m/s.

T crit 3:5T min

16

All the ingredients needed to evaluate the reduced design


spectra are now available. It is suggested to adopt the reduction
factor for acceleration design spectra dened by the following
equations:
8
 2
T
>
>
xT x0  x0T2xmin T 2 x0 x0 xmin T min
; T rT min
>
>
min
<

2
T crit T
T min r T r T crit
>
> xT 11xmin T crit T min ;
>
>
:
xT 1
T ZT crit
17a; b; c

8. Conclusion
Inertial interaction analysis of the superstructure is usually
performed by imposing that the foundation input motion is
merely that of the free eld. By contrast, the free-eld signal is
ltered out by the piles, yet this potential in reducing seismic
demand is generally not exploited in engineering practice.
In the paper, emphasis has been rst placed on the ideal case
of an innitely-long pile embedded in a homogenous soil, to

elucidate the key parameters controlling the mechanism of


ltering effect. A comprehensive parametric study has been then
performed via Finite Element analyses for a subsoil consisting of a
two layer medium underlain by a rigid bedrock. Harmonic
response has been investigated emphasizing the role of interface
depth and layer impedance contrast. Transient response has been
analyzed with reference to 9 real acceleration time-histories,
selected in order to explore the effect of frequency content.
The results of this study are fully consistent with the experimental evidence about the ltering effect and can be summarized
as follows:
a. The mechanism of ltering effect is primarily governed by soil
stiffness layer, pile diameter and excitation frequency; a unique
parameter has been found to govern the phenomenon in harmonic oscillations, and is represented by the ratio of a characteristic
pile wavelength lp over the soil wavelength Vs/o;
b. Compared to the homogeneous case, in two layer soils ltering
effect is more pronounced due to the kinematic restraint
exerted by the stiffer layer. Nevertheless, should the interface
be not shallow (h1/d4 5), the deviation from the homogeneous
soil case is negligible even for large impedance contrasts;
c. With regard to transient response, the ratio of the spectral
accelerations has a square root shape and is characterized by

62

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

a critical point, at which the ltering effect becomes


negligible;
d. A reduction factor for acceleration design spectra has
been suggested to be adopted in the presence of piled
foundations.
The result of the present study can be accounted for in the
performance prediction analysis of new structures. In this case, it
is suggested to apply the substructure method, including the
calculation step nalized to the evaluation of the FIM. The results
presented herein can be of assistance to evaluate the opportunity
of carrying out the computational cost associated to the timedomain pilesoil kinematic interaction analysis. When the performance of the superstructure is being predicted by modal
response spectrum analysis, it is suggested to use the reduced
designed spectrum, according to Eqs. (17a,b,c). An equivalent
linear elastic analysis is strongly recommended to assess the soil
stiffness prole.
The reduced design spectrum equations could be also helpful
within strategies for seismic risk reduction at regional scale. In
this respect, it is suggested to incorporate the ltered effect
exerted by the piles within the seismic demand capacity, aPGA,
conventionally adopted for the so called rst-level evaluation of
the seismic vulnerability of the structure.
It is believed that further investigations are needed, for
subsoil conditions other than those considered in the parametric
analysis, to conrm the validity of the main conclusions of
this work.
Finally, it is fair to mention that the result of this study cannot
be applied to liqueable soils. In this case, indeed, the acceleration transmitted by the piles may be larger than that associated to
the free-eld conditions.

Acknowledgements
This research has been developed under the auspices of the
research Project ReLUIS 20092013, funded by the National
Emergency Management Agency.
References
[1] Roesset JM, Whitman RV, Dobry R. Modal analysis for structures with
foundation interaction. Journal of the Structural Division 1973;99(3):
399416.
[2] Gazetas G. Seismic response of end-bearing single piles. International Journal
of Soil Dynamics and Earthquake Engineering 1984;3(2):8293.
[3] Makris N, Gazetas G, Delis E. Dynamic soilpile-foundationstructure interaction: records and predictions. Geotechnique 1996;46(1):3350.
[4] Martinelli M, Comportamento dinamico di fondazioni su pali in sabbia. PhD
thesis, University of Roma La Sapienza, 2012.
[5] Wolf JP. Foundation vibration analysis using simple physical models. Englewood Cliffs, NJ: Prentice-Hall; 1994.
[6] Ciampoli M, Pinto PE. Effects of soilstructure interaction on inelastic seismic
response of bridge piers. Journal of Structural Engineering 2005;121(5):
80614.
[7] Carbonari S, Dezi F, Leoni G. Non linear seismic behaviour of wallframe dual
system accounting for soilstructure interaction. Earthquake Engineering and
Structural Dynamics 2012;41(12):165172.
[8] Nova R, Montrasio L. Settlements of shallow foundations on sands. Geotechnique 1991;41(2):24356.
[9] Di Prisco C, Nova R, Sibilia A. Shallow footing under cyclic loading: experimental behaviour and constitutive modeling. Geotechnical analysis of seismic vulnerability of historical monuments. In: Maugeri M, Nova R, editors.
Bologna: Pa tron; 2003.
[10] Figini R, Paolucci R, Chatzigogos CT. A macro-element model for non-linear
soilshallow foundationstructure interaction under seismic loads: theoretical development and experimental validation on large scale test. Earthquake Engineering and Structural Dynamics 2012;41(3):47593.
[11] de Sanctis L, Russo G. Analysis and performance of piled rafts designed
using innovative criteria. Journal of Geotechnical and Geoenvironmental
Engineering 2008;134(8):111828.

[12] Viggiani C, Mandolini A, Russo G. Piles and pile foundations. Spon Press,
Taylor and Francis Ltd, London and New York.
[13] Dobry R. O Rourke MJ. Discussion on Seismic response of end-bearing piles
by Flores-Berrones, R. and Whitman, R.V. Journal of the Geotechnical
Engineering Division (ASCE) 1983;109(5):77881.
[14] Mylonakis G, Nikolaou A, Gazetas G. Soilpilebridge seismic interaction:
kinematic and inertial effects part I: soft soil. Earthquake Engineering and
Structural Dynamics 1997;26(3):33759.
[15] Mylonakis G. Simplied model for seismic pile bending at soil layer interfaces. Soils and Foundations 2001;41(4):4758.
[16] Anoyatis G, Di Laora R, Mandolini A, Mylonakis G. Kinematic response of
single piles for different boundary conditions: analytical solutions and
normalization schemes. Soil Dynamics and Earthquake Engineering 2013;44:
18395.
[17] Nikolaou AS, Mylonakis G, Gazetas G, Tazoh T. Kinematic pile bending during
earthquakes analysis and eld measurements. Geotechnique 2001;51(5):
42540.
[18] Maiorano RMS, de Sanctis L, Aversa S, Mandolini A. Kinematic response
analysis of piled foundations under seismic excitations. Canadian Geotechnical Journal 2009;46(5):57184.
[19] de Sanctis L, Maiorano RMS, Aversa S. A method for assessing kinematic
bending moments at the pile head. Earthquake Engineering and Structural
Dynamics 2010;39(10):113354.
[20] Sica S, Mylonakis G, Simonelli AL. Transient kinematic pile bending in
two-layer soil. Soil Dynamics and Earthquake Engineering 2011;31(7):
891905.
[21] Di Laora R, Mylonakis G, Mandolini A. Pile-head kinematic bending in layered
soil. Earthquake Engineering and Structural Dynamics 2012. http://dx.doi.org/
10.1002/eqe.2201.
[22] Flores-Berrones R, Whitman RV. Seismic response of end bearing piles.
Journal of the Geotechnical Engineering Division 1982;108(4):55469.
[23] Mamoon SM, Banerjee PK. Response of piles and pile groups to travelling SH
waves. Earthquake Engineering and Structural Dynamics 1990;19(4):
597610.
[24] Kaynia A, Kausel E. Dynamics of piles and pile groups in layered soil media.
Soil Dynamics and Earthquake Engineering 1991;10(8):386401.
[25] Fan K, Gazetas G, Kaynia A, Kausel E, Ahmad S. Kinematic seismic response
analysis of single piles and pile groups. Journal of the Geotechnical Engineering Division 1991;117(12):186079.
[26] Makris N, Gazetas G. Dynamic pilesoilpile interaction part II: lateral
and seismic response. Earthquake Engineering and Structural Dynamics
1992;21(2):14562.
[27] Kaynia AM, Novak M. Response of pile foundations to Rayleigh waves and
obliquely incident body waves. Earthquake Engineering and Structural
Dynamics 1992;21(4):30318.
[28] Gazetas G, Fan K, Tazoh T, Shimizu K, Kavvadas M, Makris N Seismic response
of soilpile-foundationstructure systems: some recent developments.
Piles under Dynamic Loads, Geotech. Special Publication no. 34, ASCE, 1992,
5693.
[29] Padron LA, Aznarez JJ, Maeso O. Dynamic analysis of piled foundations in
stratied soils by a BEMFEM model. Soil Dynamics and Earthquake
Engineering 2008;28(5):33346.
[30] Kawamura S, Umemura H, Osawa Y Earthquake motion measurement of a
pile-supported building on reclaimed ground. Proceedings of the 6th World
Conference on Earthquake Engineering, India, 1977, 103108.
[31] Tajimi H. Seismic effects on piles. State-of-the-art-report special session,
Proceedings of the 9th international conference on soil mechanics, Tokyo,
1977, 112.
[32] Ohta T, Uchiyama S, Niwa M, Ueno K. Earthquake response characteristics of
structure with pile foundation on soft subsoil layer and its simulation
analysis. In: Proceedings of the 7th world conference on earthquake engineering. vol. 3. Istanbul, Turkey, 1980.
[33] Makris N, Gazetas G, Delis E. Dynamic pilesoil-foundationstructure interaction: records and predictions. Geotechnique 1996;46(1):3350.
[34] Mylonakis G, Nikolaou A Design methods for earthquake-induced pile
bending. International conference and exposition. Deep foundation institute,
Nice, 2002.
[35] Roesset JM. The use of simple models in soilstructure interaction. ASCE
specialty conference, Knoxville, TN, Civil engineering and nuclear power, vol.
2, 1980.
[36] Buckingham E. On physically similar systems; illustrations of the use of
dimensional equations. Physical Review 1914;4(4):34576.
[37] Randolph MF. The response of exible piles to lateral loading. Geotechnique
1981;31(2):24759.
[38] Dafalias YF, Manzari MT. Simple plasticity sand model accounting for fabric
change effects. Journal of Engineering Mechanics 2004;130:62234.
[39] Wilson EL. Structural analysis of axisymmetric solids. American Institute of
Aeronautics and Astronautics Journal AIAA 1965;3:226974.
[40] Potts DM, Zdravkovic L Finite element analysis in geotechnical engineering.
Application, Thomas Telford, London, 2001.
[41] ANSYS Inc. ANSYS Theory Reference 10.0. Canonsburg, Pennsylvania, US,
2005.
[42] Zienckiewicz OC, Bicanic N, & Shen FQ. Earthquake input denition and the
transmitting boundary conditions. Conference on advances in computational
non-linear mechanics: 109130. Editor St. Doltnis I, 1988.

R. Di Laora, L. de Sanctis / Soil Dynamics and Earthquake Engineering 46 (2013) 5263

[43] Di Laora R, Mandolini A, Mylonakis G. Insight on kinematic bending of


exible piles in layered soil. Soil Dynamics and Earthquake Engineering
2012;43:30922.
[44] Scasserra G, Lanzo G, Stewart JP, D Elia B. SISMA (Site of Italian Strong
Motion Accelerograms): a web-database of ground motion recordings for
engineering applications. Seismic Engineering Conference commemorating
the 1908 Messina and Reggio Calabria Earthquake, Reggio Calabria, Italy
2008;2:164956.

63

[45] Ambraseys N, Smit P, Sigbjornsson R, Suhadolc P, Margaris B. Internet-Site for


European Strong-Motion Data. European Commission; 2002 ResearchDirectorate General, Environment and Climate Programme.
[46] Rathje M, Abrahamson NA, Bray JD. Simplied frequency content estimates of
earthquake ground motions. Journal of Geotechnical and Geoenvironmental
Engineering 1998;124(2):1509.

Вам также может понравиться