Vous êtes sur la page 1sur 19

ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission.

Copyright (c) 2012. American Society of Civil Engineers


SIMPLIFIED M O D E L FOR ANALYSIS
O F O N E OR T W O P I L E S
By Aristonous M . Trochanis, 1 Associate M e m b e r , A S C E ,
Jacobo Bielak, 2 a n d Paul Christiano, 3 M e m b e r s , A S C E

ABSTRACT: The main objective of this work is to use the experience gained from
a three-dimensional parametric study presented in a companion paper to develop
a simplified model that will reproduce the main nonlinear features of the behavior
of single piles, as well as the interaction between a pair of piles. Hence, guided
by the results of that parametric study, a simplified model consisting of coupled,
inelastically supported one-dimensional piles is developed and presented herein.
This model takes into consideration slippage and separation between the piles and
the soil, as well as the overall inelastic soil behavior including degradation. The
validity of the proposed model is assessed by an extensive comparison with nu-
merical results from the more accurate three-dimensional study, as well as with
results from experimental field tests. Close agreement is observed between these
results and the predictions from the simplified model. This represents an essential
step towards a simplified, yet realistic, method of analysis for general pile groups.

INTRODUCTION

The new simplified model proposed herein attempts to capture the main
characteristics of the nonlinear behavior of pile foundations, determined from
a three-dimensional parametric study presented in a companion paper (Tro-
chanis et al. 1990). That study revealed the important role of several factors
that have frequently been underestimated as affecting the response of pile
foundations, such as yielding of the soil surrounding the foundation, relative
pile-soil slippage, and separation between soil and pile. An additional im-
portant factor that was not covered by the three-dimensional analysis is the
degradation of the soil and the pile-soil interface properties.
Since three-dimensional analyses and detailed field tests of prototypes are
not feasible in most practical situations, several simplified models have been
developed in the past. Recently reported simplified models include those by
Nogami and Konagai (1987) and Nogami et al. (1988), who represented the
soil resistance at several points along the pile with multiple sets of springs
and dashpots in series, and by Chow (1987), who used theoretical curves of
soil resistance versus pile displacements and full coupling between piles in
groups by means of Mindlin's equations. However, many of the simplified
models described in the technical literature have been developed on the basis
of a limited number of field tests, and consequently, their range of appli-
cation may be limited to only the type of piles or soil corresponding to the
specific test site. Furthermore, most such models have been limited by sev-
eral simplifying assumptions.
'Engr., Odomikaniki, Athens, Greece; formerly, Grad. Res. Asst., Dept. of Civ.
Engrg., Carnegie Mellon Univ., Pittsburgh, PA 15213.
2
Prof., Dept. of Civ. Engrg., Carnegie Mellon Univ., Pittsburgh, PA.
3
Dean, Carnegie Inst, of Tech., Carnegie Mellon Univ., Pittsburgh, PA.
Note. Discussion open until August 1, 1991. Separate discussions should be sub-
mitted for the individual papers in this symposium. To extend the closing date one
month, a written request must be filed with the ASCE Manager of Journals. The
manuscript for this paper was submitted for review and possible publication on July
17, 1989. This paper is part of the Journal of Geotechnical Engineering, Vol. 117,
No. 3, March, 1991. ©ASCE, ISSN 0733-9410/91/0003-0448/$1.00 + $.15 per
page. Paper No. 25601.

448

J. Geotech. Engrg. 1991.117:448-466.


scelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.

We propose a new simplified model that incorporates, in a rational way,


the effects of several of the aforementioned parameters. Both single piles
and, to a lesser extent, pairs of piles are considered as a first step toward
the development of a full model for pile groups. The model is based on the
general behavior patterns that emerged from a careful analysis of the results
from our three-dimensional finite element model, as well as from results of
previous analytical and experimental studies. In particular, the proposed method
makes use of a nonlinear model for the soil resistance along the sides and
at the tip of the pile; furthermore, it accounts for pile-soil slippage and sep-
aration, as well as degradation of the soil properties. Although the model
has been formulated for dynamic loading by considering pile inertia effects
and radiation and material damping in the soil (soil inertia is incorporated
through these parameters) (Trochanis 1988) this paper is concerned only with
quasi-static problems. The parameters of the proposed model are readily as-
sociated with the physical parameters governing the response of actual pile
foundations, and widely accepted code recommendations or empirical for-
mulas are used for their estimation. The development of the model involves
standard finite element and finite difference theory procedures, which render
its implementation on commercial finite element packages highly feasible.

MODEL FOR SINGLE PILES

General Description
The model presented in this section deals with the analysis of single piles
subjected to axial and lateral loads. The pile is modeled as an elastic flexural
beam, while the surrounding soil is represented by a Winkler-type founda-
tion, i.e., the soil resistance at any depth along the pile due to the pile
displacement is coupled only to the resistance at that depth. The model is
essentially one-dimensional, since the solution is sought as a function of only
the longitudinal coordinate of the pile. The form of the solution within the
soil itself is not investigated here, even though it enters the problem im-
plicitly and crucially through the parameter selection for the pile and the
soil. The solution for the pile response is obtained by using finite element
spatial discretization along the longitudinal coordinate.
Schematic diagrams of the proposed model are shown in Figs. 1(a) and
1(b) for the cases of axial and lateral loading, respectively. The pile is com-
pletely defined by its Young's modulus Ep, cross-sectional area Ap, moment
of inertia Ip, and length L. Only piles of constant cross-section were studied
in this work, but the extension of the model to nonprismatic piles is straight-
forward. The soil resistance is simulated by distributed nonlinear springs
with stiffness acting vertically (ksu) or horizontally (ksw), depending on the
type of loading. Because of the possibility of different behavior on the two
sides of the pile, e.g., when separation from the soil takes place, distinct
stiffness values are used for the left and right sides of the pile for both
vertical and horizontal loading. These quantities, which are denoted by the
superscripts / and r for left and right sides, respectively (Fig. 1), initially
are equal on both sides of the pile, combining to the total corresponding
magnitude for the vertical case, whereas each one is taken to be the entire
soil resistance for the lateral case. The stiffness of the side elements can be
prescribed as functions of distance from the pile head in order to represent
the variation of the relevant soil properties with depth.

449

J. Geotech. Engrg. 1991.117:448-466.


ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers
»H,M

-p |-vwn-®i^w^ p-*- w
i
•Ep.Ap,L,pp

f •

IB JvWr—#i-WV^|
IK,
(a) (b)

FIG. 1. Schematic Diagram of Simplified Model for Single Piles: (a) Axial Load-
ing; (b) Lateral Loading

In the discretized form of the model, these distributed quantities are re-
placed by a set of concentrated springs and dashpots acting on the nodes of
the discretized pile. It should be noted that an additional vertical spring (Ksl)
is placed at the bottom of the pile to model the concentrated vertical reaction
of the soil underneath the pile tip. In the case of lateral loading, rotational
springs may be attached at the head and the tip of the pile (Krh,Kr,) in order
to model constraints imposed by a pile cap or by an enlarged pile base or
socket. The load is applied at the pile head, as concentrated vertical (V),
horizontal (//), or moment (M) loads. Any distribution of initial shear or
normal stresses along the pile can be accommodated as a precompression of
the appropriate springs.

Constitutive Law
In this work nonlinear response due to soil yielding and pile-soil slippage
or separation are modeled through the springs representing the soil. A de-
grading hysteretic model developed by Wen (1976) and by Baber and Wen
(1981) for the analysis of single- and multidegree oscillators is adapted here
as the constitutive model for the vertical and horizontal springs of the pro-
posed simplified model. This constitutive model possesses versatility and a
simple analytical form, two attractive characteristics for practical use. Ac-
cording to this model the distributed soil reaction on the pile, q, in the lateral
or axial direction, with corresponding displacement w (or u), can be ex-
pressed as:
q = akw + (1 - a)kz (1)
in which k = a reference stiffness, a = a weighting factor, and z = a trans-
formed displacement variable that controls the hysteresis and is governed by
the following nonlinear differential equation:

(2)
aw w

450
J. Geotech. Engrg. 1991.117:448-466.
scelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.

q 0.75,

0.50-

0.25

-0.5
-0.25-

-0.50 •

-0.75.

(a) (b) CO

FIG. 2. Wen's Model: (a) Effect of Parameter n; (b) Effect of A Degradation; (c)
Behavior with Pile-Soil Separation

A, P, 7 and n are parameters of the model that control the shape of the
hysteresis loop, and an overdot denotes differentiation with respect to time.
The solution of Eq. 2 for z as a function of w describes the hysteretic nature
as well as the versatility of the model. It is noted that sudden changes in
properties, as when the velocity changes sign, are built-in automatically, and
it becomes unnecessary to trace these changes explicitly. The model can
simulate both strain hardening and strain softening materials based on the
combination of values for (3 and 7 (Wen 1976), while the parameter n con-
trols the sharpness of the transition from the linear to the nonlinear range.
The effect of the parameter n is illustrated in Fig. 2(a), in which the hys-
teresis loops for a softening material are shown for different values of n;
elastic-perfectly plastic behavior is attained in the limiting case for which
n becomes infinity.
The model can also reproduce degrading behavior with either degrading
stiffness or strength, or both. Simultaneous stiffness and strength degradation
is introduced by making the parameter A a decreasing function of the history-
dependent total amount of dissipated hysteretic energy e(t), so that:
A(t) =A0- 8Ae(r) (3)

where A0 = the initial value of A and 8A = the degradation rate corresponding


to A. The type of degradation introduced by this parameter is reflected by
the hysteresis loops in Fig. 2(b). Pure stiffness or strength degradation can
be accomplished by the use of two additional history-dependent parameters
in Eq. 2. Also, if desired, a criterion other than hysteretic energy may be
used to control the amount of degradation.
The basic parameters of Wen's model can be associated readily with the
physical parameters of the problem of pile behavior. It can be shown that
for strain softening materials the yield level fy (strength) is given by:
1/FI

/, = ( ! - a)*| (4)
P+Y
On the other hand, the initial stiffness kt and the final stiffness kf of the
model are:
kt = ak + (1 — a)kA, kf=ak (5)
451

J. Geotech. Engrg. 1991.117:448-466.


ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers
The ratio of the final, post-yield stiffness to initial stiffness reduces to the
value of a when A = 1, giving a a direct physical interpretation. In order
to make the relationship between Wen's parameters and the physical system
parameters more direct, the following simplifications are introduced: (1) The
initial value of A is set to unity, so that the actual stiffness of the soil along
the pile can be used as the reference stiffness; (2) n is given the value 2 for
the lateral springs and the tip spring, and the value 3 for the vertical springs;
and (3) the parameters p and y are always taken to be equal; their value is
determined by solving Eq. 4 using the actual soil strength on the left-hand
side and the previous values for A and n on the right-hand side. It will be
shown later that this particular choice of parameters leads to good agreement
between the model predictions and the results from experiments and from
three-dimensional analyses. In addition, separate calculations have shown
that the pile response is not very sensitive to small changes in the values of
n or different combinations of p and -y (Trochanis 1988).
In order to model possible separation of the pile and the soil in the case
of lateral loading, the original constitutive law is modified so that no tensile
force will be sustained by the springs. The modified hysteretic behavior for
the nonlinear part of the restoring force in the springs is presented in Fig.
2(c) for five cycles with constant unit displacement amplitude, with the com-
pressive restoring force shown as a positive quantity. At some time during
the unloading branch of the first cycle the restoring force will tend to become
negative (tensile); the modified model assumes that a gap opens at that par-
ticular displacement, and thereafter the stiffness and restoring force in the
spring remain zero until this displacement is exceeded during a subsequent
cycle (closing of the gap). (This assumption is strictly valid only for ma-
terials with some cohesion; cohesionless soil will tend to flow into the gap,
resulting in some reduced, but nonvanishing, stiffness.) The initial stiffness
immediately after the closure of the gap is determined by the initial stiffness
of the model and the cumulative stiffness degradation up to the previous
cycle. The opening of gaps on both sides of the pile during lateral loading
is handled by using a different spring for each side of the pile (Fig. 1).

Mathematical Formulation
For the case of a concentrated horizontal load Hit) and moment M(i) ap-
plied to the top of the pile, the lateral motion of the pile is resisted by a
horizontal soil pressure q(x,t) distributed along its length. The equation of
equilibrium in terms of the deflection w can be written as:
EpIpwn(x,t) - q(x,t) = 0 (6)
in which x = the spatial coordinate along the pile. After substituting for the
stiffness (Wen's model) and damping contributions in q from the preceding
section, Eq. 6 can be rewritten as:
EpIpw'v + a X W w + (1 - <XsJrtK(x)z'sw(x, t) + asJ^sw(x)w
+ (1 - a X « 4 ( ^ , 0 = 0 (7)
The boundary conditions associated with Eq. 7, which incorporate the ro-
tational springs krh and k„ [Fig. 1(b)] at the pile head and tip, respectively,
are:
EpIpw"(0,t) + M(t) - krhw'(0,t) = 0 (8a)
452
J. Geotech. Engrg. 1991.117:448-466.
scelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.

EpIpw"'(0,t) - H(t) = 0 (8*)


EpIpw"{L,t) + krtw'(L,t) = 0 (8c)
EpIpw'"(L,t) = 0 (8rf)
Eqs. 7 and 8, together with Wen's basic constitutive relationships (Eqs. 1
and 2) written for ksw, zsw, and w, define the problem of the nonlinear re-
sponse of a single pile under lateral quasi-static load. A similar set of equa-
tions can be written for the axial response of the pile; those will involve the
second derivative of the axial displacement u with respect to x and the ver-
tical soil stiffness km. Details of the solution procedure of the discretized
equations for the axial and lateral displacements, and the computer imple-
mentation, can be found in Trochanis (1988).

Parameter Selection
The selection of appropriate parameters for the initial elastic stiffness and
strength of the soil along the pile is crucial for predicting satisfactory pile
response with any simplified model. In our calculations we have used the
well-known expressions given by Scott (1981) for estimating the linearly
elastic axial and lateral soil stiffnesses; for variable stiffness one can use
Randolph and Wroth (1978). We also adopted the rules given in the "Rec-
ommended Practice for Planning, Designing and Constructing Fixed Off-
shore Platforms" (1986) by the American Petroleum Institute (API) for es-
timating vertical (shear) and horizontal strength of the soil along a pile or
underneath its tip. These sets of recommendations were selected based on
their ease of implementation, their widespread acceptance, and their straight-
forward association with the input parameters needed by our simplified model;
furthermore, as will be seen later, these rules lead to accurate predictions of
the behavior observed during various static field tests, both qualitatively and
quantitatively. The basic rules used in this work from both references (Scott
1981; "Recommended" 1986) for estimating the static stiffness and strength
of the soil resistance are summarized in Trochanis (1988). These recom-
mendations are the result of the experience of numerous researchers and
practicing engineers but, naturally, they cannot cover every possible situa-
tion. Thus, if a better estimate for the stiffness and strength parameters of
this model is available, it should be used. Such was the case here, when the
one-dimensional model was compared with our three-dimensional model. As
an example, a better estimate could be made of the vertical soil stiffness in
these cases, since the properties of the pile-soil interface were accurately
known. The three-dimensional model employed, as part of the input, an
elastic stiffness for the frictional interface elements. The total stiffness along
the pile can then be modeled as two stiffnesses in series, the aforementioned
stiffness of the interface elements and the stiffness of the adjacent soil as if
it were bonded to the pile; the latter can be calculated based on some rational
assumption for the deformation of the soil around the pile; e.g., if we assume
that the deformation initially varies linearly with the distance from the pile,
and that it is already negligible at distances greater than two pile widths (as
noticed in the three-dimensional results), then this stiffness can be shown to
be equal to two times the shear modulus. This combination of interface and
soil stiffness in series worked very well for different pile diameters, lending
additional confidence to the aforementioned arguments. The shear strength

453

J. Geotech. Engrg. 1991.117:448-466.


ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. A
along the pile in the case of the three-dimensional model can also be easily
calculated from the known initial normal stresses and the friction coefficient
of the interface elements; thus, this value was used in the simplified model.

NUMERICAL RESULTS FOR SINGLE PILES

Model Validation
The proposed model was first evaluated by comparing its results with those
corresponding to our own three-dimensional analysis. A comparison between
the simplified and three-dimensional models for the case of an axially loaded
pile, 10-m long and 0.5-m square is shown in Fig. 3. Two cases are con-
sidered: elastic soil bonded to the pile, and inelastic soil with allowance for
relative pile-soil slippage. The pile and soil properties used in the three-
dimensional model are presented in the companion paper (Trochanis et al.
1990). The use of the aformentioned rules for selecting the parameters of
the simplified model led to a uniform vertical stiffness of 6,800 kPa along
the pile and a tip stiffness of 14,000 kPa; the side strength varied linearly
with depth from 1.4 kN/m at the soil surface to 84 kN/m at the pile tip
level, and the tip strength was 100 kN. The load-settlement curves as pre-
dicted by the two models are shown in Fig. 3(a), while Fig. 3(b) depicts
the distribution of the axial force along the pile for the peak load value (522
kN). Note that in the nonlinear case more force is transferred from pile to
soil at greater depths (indicated by the greater curvature of the nonlinear
axial force curve in Fig. 3(b) as opposed to the transfer at smaller depths,
due to the early pile-soil slippage near the surface. In the bonded case only
10% of the head load is transferred to the pile tip, while this percentage is
doubled for the final load in the nonlinear case. Similarly, good agreement
between the 1-D and 3-D models was found in the case of axially loaded
end-bearing piles (Trochanis 1988).
The two models also were compared in the case of lateral loading. Three
different combinations of soil and pile-soil interface models were used: elas-

(a) (b)

FIG. 3. Comparison of Simplified and Three-Dimensional Model for Axially Loaded


Piles: (a) Head Load-Settlement Curves; (b) Axial Force Distribution with Depth

454

J. Geotech. Engrg. 1991.117:448-466.


scelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. Al

0.75 1.50
Deflection (cm)

0 3D model
— - ID model, elastic soil 0.1
- — ID model, plastic soil

'-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2


Deflection (cm)

(a) (b)

FIG. 4. Comparison of Simplified and Three-Dimensional Model for Laterally


Loaded Piles: (a) Head Load-Deflection Curves; (b) Deflected Shapes

f. 0.0

o 3D model
- ID model

"100 0 100 200 300 400 500 100 0 100 200 300 400 500
Lateral noimal stress (kPa) Lateral normal stress (kPa)

(a) (b)

FIG. 5. Comparison of Normal Pressure Distributions along Pile as Computed


by Simplified and by Three-Dimensional Models: (a) Elastic Soil; (b) Inelastic Soil

tic soil bonded to the pile, and either elastic or inelastic soil allowed to
separate from the pile. Predictions of the two models for the last two cases
are depicted in Figs. 4(a) and 4(b), cast in the form of head load-deflection
curves and deflected shapes at the peak load, respectively. The lateral stiff-
ness for the simplified model was 20,000 kPa (equal to the elastic modulus
of the soil), and the lateral strength for the inelastic soil possessed a bilinear
distribution with depth, having values of 70 kN/m, 350 kN/m, and 450
kN/m at the surface, at depth of 1.5 m and at the pile tip level (10 m depth),
respectively. It was found that the simplified model captures the main char-
acteristics of the behavior of the three-dimensional model, e.g., no energy
dissipation from pile-soil separation alone, and concentration of pile deflec-
tions on the upper portion of the pile. Figs. 5(a) and 5(b) depict the normal
pressure distribution along the pile at the maximum load, as predicted by

455
J. Geotech. Engrg. 1991.117:448-466.
ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. the two models for the cases of elastic and inelastic soil, respectively; the
pressures at both the leading and trailing face of the pile are shown, plotted
on both sides of the vertical reference axis, which can be thought of as
representing the initial position of the pile. Areas of pile soil separation can
be recognized by the zero-normal pressure at the trailing face; also, the pres-
sure at the leading face in the case of inelastic soil reaches a maximum value
at a point below the surface, which moves downward with progressive soil
plasticity, whereas the peak pressure for elastic soil occurs always at the soil
surface. The simplified model reproduces accurately the pressure distribu-
tions and the depth of pile soil-separation for both soils and for both sides
of the pile. The same good agreement was found in the elastic bonded case
not shown here. It should be noted that the strength distribution was adjusted
in this particular case because no specific recommendation of the API ("Rec-
ommended" 1986) applies for a soil with both cohesion and friction angle
different than zero (c = 34 kPa, cj> = 16.7°), as was the case for the three-
dimensional model; however, if the strength values for a clay with c = 34
kPa and a sand with § = 16.7° are superimposed, the values of the strength
used herein are justified. Moreover, the same strength distribution was also
tested for two other piles with different lengths and widths resulting in the
same excellent agreement between the two models (Trochanis 1988).
The simplified model was assessed further by an extensive comparison
with actual field data on the axial and lateral response of single piles in both
clays and sands found in Mexico City. The data correspond to the static
vertical loading of a single concrete pile 15-m long and 0.3-m square with
a vertical load at its head; the pile and soil properties and the setup of the
specific test can be found in the companion paper. Use of Scott's formula
for the axial stiffness of the soil along a flexible pile leads to a value of
10,700 kPa; however, the effective stiffness calculated in the manner de-
scribed earlier, using the soil and pile-soil interface stiffnesses in series, was
only 5,100 kPa. The latter value was used in the simplified model along
with a tip stiffness of 7,400 kN/m. The shear strength varied linearly from
zero at the pile head to 75 kN/m at the pile tip, and the tip strength was
65 kN. The corresponding load-settlement curve is shown in Fig. 6(a), along
with the experimentally obtained curve and that obtained by the three-di-
mensional model; both models result in curves close to the experimental one.
In addition to the previous case of an axially loaded pile in clay, the sim-
plified model was applied to a case of an axially loaded pile in sand. The
pile, designated as No. 2 in the Arkansas River Project (Mansur and Hunter
1970), was a steel pipe with length 16.1 m, outer diameter 0.41 m and wall
thickness 12.7 mm, resulting in an axial stiffness EA = 3.08 X 106 kN. It
was loaded and unloaded in increments, for a total of five full loading and
unloading cycles, with peak values of 445; 890; 1,335; 1,735; and 2,225
kN. The settlement was recorded at the end of each loading branch, together
with the rebound after each unloading. The shear modulus of the sand at
the site increased linearly with depth, with a measured value of 15,860 kPa
at a depth of 7.62 m, and its friction angle of 32°. The detailed derivation
of the parameters for the simplified model based on Scott (1981) and the
API ("Recommended" 1986) is presented by Trochanis et al. (1987). The
load-settlement curve predicted by the simplified model for the five loading
cycles is shown in Fig. 6(b) along with the experimental peak settlements
and rebounds. The simplified model reproduces the behavior observed in the

456
J. Geotech. Engrg. 1991.117:448-466.
ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.

g 450

1.0 1.5 2.0 2.5 2 3 4 5


Head settlement (cm) Head settlement (cm)

(a) (b)

FIG. 6. Comparison of Head Load-Settlement Curves Obtained by Simplified Model


with Field Data for Axialiy Loaded Piles: (a) Pile in Clay (Mexico City Test); (b)
Pile in Sand (Mansur and Hunter 1970)

2 3 4 5 30 60 90 120
Head deflection (cm) Bending moment (kN-m)

(a) (b)

FIG. 7. Comparison of Simplified Model with Field Tests on Laterally Loaded Pile
(Brown et al. 1988): (a) Head Load-Deflection Curve; (b) Bending Moment Distri-
bution with Depth

field with good accuracy for all loading cycles. Similarly good agreement
was found for a test on pile No. 10 from the same project (Trochanis 1988).
The simplified model was used next to predict the behavior of laterally
loaded piles in stiff clay (Reese et al. 1975) and in sand (Long and Reese
1984; Brown et al. 1988). Although the agreement between the model's
predictions and field measurements was equally good in all three cases (Tro-
chanis 1988), only the comparison with the most recent test will be presented
here, for static problems. Brown et al. (1988) conducted lateral tests on a
single pile and on a 3 X 3 square pile group in medium dense sand. The
single pile was a steel pipe with outer diameter 27.3 cm, wall thickness 9.27
mm, and embedment depth 4.9 m; the load was applied 30 cm above the

457

J. Geotech. Engrg. 1991.117:448-466.


ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. ground surface. For the dense sand at the site (4> = 38.5°) the API rec-
ommendations led to a lateral stiffness proportional to the depth with a peak
value of 120,000 kPa at the pile tip, and to a strength distribution that varied
nearly from zero at the top to 1,620 kN/m at the bottom of the pile. The
model predictions for the head deflection versus the load, as well as the
bending moment distribution with depth at an intermediate value of the load
(86.8 kN) are plotted against the corresponding measured data in Figs. 7(a)
and 7(b), respectively. The agreement between the analytical model and the
experimental results is excellent. It should be noted that Brown et al. (1988)
suggested an adjustment of the friction angle to 50°, when using earlier API
criteria for p-y curve, in order to explain the experimental results; no such
adjustment was necessary with the present model.

Parametric Study
A parametric study was conducted with the simplified model in order to
assess the role of several pile and soil parameters on the monotonic response
of piles. A second objective was to study the importance of degradation of
the soil resistance during cyclic loading. The basis pile and soil properties
used in this parametric study are those pertaining to Mexico City. The API
recommendations were used to relate the soil properties to the simplified
model parameters.
Results for monotonic loading are shown in Fig. 8. These provide a suc-
cinct summary of the effects of various parameters, such as the elastic mod-
ulus of the soil, as measured by the shear modulus Gs, the undrained shear
strength of clays cu, and their distribution with depth (taken to be linear;
subscripts h and t denote corresponding values at head and tip, respectively);
the friction angle <>j of sands; and the pile slenderness ratio L/D (for D =
0.5 m), on the axial and lateral response of the pile head.
The effect of degradation of the soil properties on the cyclic response of
single piles is studied in Fig. 9. Recall that Wen's model can be used to
represent either stiffness or strength degradation or simultaneous degradation
of both; the latter degradation type is examined here. Results obtained for
a single pile loaded axially without load reversals showed no significant en-
ergy dissipation and, consequently, in accordance with Wen's model, no
significant degradation. The effects of degradation, though, in the axial case
are more pronounced when load reversals take place. To illustrate this point,
the head settlement response of a pile to five full loading cycles between
+300 kN (downward) and —300 kN (upward) is shown in Fig. 9(a) for a
degradation rate 8A = 0.1; this load represents 60% of the failure load for
this pile. The response of the same pile for a cyclic load with amplitude 400
kN (80% of the failure load) is shown in Fig. 9(b) for the same degradation
rate. For the smaller cyclic load the response is still stable after five cycles
although some deterioriation is evident; for the larger load, however, due to
the large amount of dissipated energy, which in turn causes substantial de-
terioration of the soil resistance, the pile fails in extraction from the soil
after just two cycles. This failure is demonstrated by the final portion of the
load-settlement curve in Fig. 9(b), which is parallel to the negative settle-
ment axis. This behavior would be expected to occur also for the smaller
load at higher rates of degradation.
Shown in Fig. 9(c) is the effect of the same type of degradation on the
lateral response of a pile that is allowed to separate from the inelastic soil.
458

J. Geotech. Engrg. 1991.117:448-466.


ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.

Gsh = Gst = 6800 kPa


- Gsh = Ost = 13,600 kPa
Gsh = 0,Gst= 13,600 kPa
Esh = Est = 20.000 kPa
Esh = Est = 40,000 kPa
Esh = 0, Est = 40,000 kPa

3 4 S 4 6 8 10
Head settlement (cm) Head deflection (cm)
(a) (b)

1000
800
600-
400
- Cuh = Cut = 20 kPa Cuh = Cut = 20 kPa
200 Cuh = Cut = 40 kPa Cuh = Cut = 40 kPa
•• Cuh = 0, Cut = 40 kPa Cuh = 0, Cut = 40 kPa
0
0 2 3 4 5 6 8 10
Head settlement (cm) Head deflection (cm)
(0 (d)

..••

4 = 25*
$=15*

500-

0- C—,—
3 4 5 4 6 8 10
Head settlement (cm) Head deflection (cm)
(e)

I 240n
L/D = 40 J ,80
L / D = 20 Z
L/D=10 a
S 120-
L / D = 40
60- L / D = 20
L/D=15
L/D=10
6 8 10
Head settlement (cm) Head deflection (cm)

(g) (h)

FIG. 8. Effect of Soil and Pile Parameters on Pile Head Response: (a) Effect of
Shear Modulus of Soil on Axial Pile Response; (b) Young's Modulus of Soil on
Lateral Response; (c) Undralned Strength of Clay on Axial Response; (d) Un-
drained Strength of Clay on Lateral Response; (e) Friction Angle of Sands on Axial
Response; (f) Friction Angle of Sands on Lateral Response; (g) Pile Slenderness
Ratio on Axial Response; (/?) Pile Slenderness Ratio on Lateral Response

The shape of the initial hysteresis loop has also been observed in our three-
dimensional analysis, while the shape of the subsequent loops is due to the
assumption in the proposed simplified model that the soil does not rebound
after it separates from the pile. Both the behavior with degradation (8A =
0.25) and without (8A = 0) are shown in Fig. 9(c); the degradation led to
the final amplitude (after five cycles) of the head deflection being approx-
imately 25% greater than the amplitude in the case without degradation. The

459

J. Geotech. Engrg. 1991.117:448-466.


ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.

(a) (b) (c)

FIG. 9. Effect of Soil Degradation on Cyclic Pile Response: (a) Vertical Load with
Amplitude 60% of Failure Load; (ft) Vertical Load with Amplitude 80% of Failure
Load; (c) Horiiontal Load

other two types of degradation included in Wen's model exhibited similar


quantitative effects (Trochanis 1988).

EXTENSION OF SIMPLIFIED MODEL TO PAIRS OF PILES

Description of Extension
The experience gained by the parametric study conducted on a pair of piles
using the three-dimensional model demonstrated the significant role that pile-
soil slippage and separation, as well as soil plasticity, play in reducing the
interaction between the two piles for axial or later loading (Trochanis et al.
1990). A first attempt at incorporating these effects into the analysis of the
two piles is described by Trochanis (1988). Some of the basic ideas, along
with select numerical results, are presented in the following paragraphs.
Recall that for the single pile model the soil resistance is represented with
a set of vertical or horizontal nonlinear springs distributed along the pile and
at the pile tip. The interaction between neighboring piles also is modeled
herein by a set of spatially distributed nonlinear vertical or horizontal springs
connecting the two piles. At the same time the stiffness of the fixed ground
springs (i.e., those connecting each pile to the surrounding soil) must be
reduced appropriately, in order to take into consideration the added soil re-
sistance represented by the interaction stiffness, and to avoid overestimating
the total stiffness of the system. The magnitudes of the interaction stiffness
and of the reduction in fixed ground stiffness can be calibrated using any
acceptable elastic solution for pile group response. The effects of pile-soil
slippage and separation, as well as soil plasticity, on the vertical and hori-
zontal interaction are incorporated in the present analysis through the con-
stitutive laws governing the behavior of the interaction springs.
A schematic representation of the proposed model is shown in Figs. 10(a)
and 10(b) for the cases of axial and lateral loading of the piles, respectively.
The symbols Kiu and Kiw are. used for the distributed interaction stiffness in
the two directions, and Ru, Rw are the vertical and horizontal reduction fac-
tors applied to the fixed ground springs on the sides of the piles facing one
another; the latter acquires different values in the axial and lateral cases. In
the discretized form of the model, only nodes of the two piles at the same

460
J. Geotech. Engrg. 1991.117:448-466.
scelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.

H
H-,,M-, H2, M2

n a ' 1 -JVW-
—w*-\ Jvwfc— '• — W M - |

•Ep,lp,L,pp-.

-VW—!
JTWA— —*W-| §~VWV—

n
^sw w*sw

(a) (b)

FIG. 10. Schematic Diagrams of Simplified Model for Pair of Piles: (a) Axial
Loading; (h) Lateral Loading

depth are connected through the nonlinear springs. The assumption of ver-
tically uncoupled interaction springs leads to a banded stiffness matrix for
the pile-soil-pile system, as opposed to the full matrix of other methods. The
effect of full coupling is taken implicitly into account by the method used
for estimating the magnitude of the interaction stiffness and the reduction
factor, i.e., by calibrating the model using well-known elastic results. It is
clear that both the interaction stiffness and the reduction factor are functions
of the pile spacing. In particular, the interaction stiffness must decrease rap-
idly with increasing spacing, becoming negligible for sufficiently large spac-
ings; on the other hand, the reduction factor should have a value less than
unity for close spacings and tend to unity for large spacings, when the effect
of the second pile on the first is negligible.
In the general case, any reliable elastic results on interaction of two piles
from the relevant literature, e.g., Poulos and Davis (1980), can be used for
determining the appropriate interaction stiffness and reduction factor for use
with the simplified model. The present work uses our results of the three-
dimensional model, which agreed with those from Poulos and Davis (1980).
Values obtained for the case of two elastic piles Ep = 2 x 107 kPa 10-m
long and 0.5 m (Es = 2 X 104 kPa), are shown in Fig. 11 for illustration;
only the case of one active pile is necessary for estimating the model pa-
rameters, and only one combination of uniform interaction stiffness and re-
duction factor can match the response of both the active and the passive
pile. Fig. 11(a) depicts the elastic axial interaction stiffness Kiu> normalized
with respect to the shear modulus of the soil, as well as the variation of the
reduction factor Ru as a function of pile spacing for the axial case. Fig. 11 (b)
shows the corresponding variations in the lateral case (the stiffness in this
case is normalized with respect to the Young's modulus of the soil rather
than the shear modulus). Evidently, interaction is practically negligible at
distances greater than 20 pile diameters; consequently, the reduction factor
is practically unity at this distance. The large values of normalized inter-
action stiffness for close spacings (greater than 2 in the axial case and 3 in
the lateral case), and its hyperbolic reduction with increased spacing are also
noteworthy.
461
J. Geotech. Engrg. 1991.117:448-466.
ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.

Ru •— Rw
Khi/Gs Kiw /Es

1 -
""
^ " • " ^ ^

o.
16 24 16 24
Pile spacing (s/D) Pile spacing (s/D)

(a) (b)

FIG. 11. Variation of Elastic Interaction Stiffness and Reduction Factor: (a) Axial
Loading; (ft) Lateral Loading

The reduction in interaction due to pile-soil slippage and separation and


to soil plasticity is incorporated in the present analysis by assuming that the
reduction in interaction stiffness between the two piles is similar to the re-
duction in the fixed ground stiffness caused by the same factors. The amount
of reduction in the fixed springs is expressed by the tangent stiffness of each
spring divided by its initial stiffness, i.e., the previous assumption means
that the tangent interaction stiffness can be calculated from the tangent stiff-
ness of the ground springs. The rules used for this calculation are different
for the cases of axial and lateral loading of the two piles. Details will be
found in Trochanis (1988).

Validation of Proposed Extension


Results from the simplified model for two piles using the rules just de-
scribed were first compared with results of the three-dimensional analysis
for several nonlinear cases with spacings of two diameters between two ax-
ially loaded piles. The interaction parameters used in this part of the analysis
are those presented in Fig. 11(a). Shown in Fig. 12(a) are load-settlement
curves obtained by the simplified model for two piles with the same prop-
erties as before, and allowed to slip relative to the surrounding soil. Either
one or both piles were considered active, i.e., subjected to loading. Cor-
responding curves associated with a three-dimensional analysis are presented
in the companion paper; the agreement between the two sets of results is
very encouraging.
A similar comparison of the two models was conducted in the case of
lateral loading. A representative set of load-deflection curves obtained by
the simplified model for two piles in inelastic soil (with pile-soil separation)
is shown in Fig. 12(b); the corresponding curves from the three-dimensional
model can be found in the companion paper. As in the case of axial loading,
the predictions by the two models are in close agreement, substantiating the
rules used in the development of the simplified model. Note that when only
one of the piles is active, the simplified model reproduces accurately the
stiffening effect due to the presence of the passive pile on the response of
the active one; by contrast, a softer behavior is predicted for the trailing pile

462
J. Geotech. Engrg. 1991.117:448-466.
scelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.

Head settlement (cm) Head deflection (cm)

(a) (b)

FIG. 12. Head Load-Displacement Curves for Two Neighboring Piles Obtained
by Simplified Model: (a) Axial Loading; (b) Lateral Loading

in the case of two active piles, i.e., for the same load the trailing pile under-
goes larger deflections. Conversely, for the same amount of deflection the
trailing pile would sustain a smaller load. This behavior agrees qualitatively
with recent experimental results on the individual response of piles in a lat-
erally loaded 3 X 3 group reported by Brown et al. (1988). Extension of
the model, as presently formulated, to multiple piles would require an elastic
analysis of the complete pile group in order to determine the reduction coef-
ficients and elastic stiffnesses of the coupling springs using available elastic
attenuation coefficients such as those by Poulos and Davis (1980). The in-
elastic analysis then would follow using the rules previously delineated.

SUMMARY AND CONCLUSIONS

Guided by the results of a three-dimensional model presented in the com-


panion paper, a simplified procedure was developed for incorporating the
effects of soil plasticity, and pile-soil slippage and separation into the anal-
ysis of one and two piles. The proposed model is based on a nonlinear Wink-
ler-type representation of the soil resistance to a single pile. Allowance for
slippage and separation on two sides of the pile, as well as inelastic soil
behavior including degradation in the immediate vicinity of the pile was
made by means of a versatile nonlinear constitutive law. Interaction between
piles was modeled by nonlinear springs connecting the piles, acting verti-
cally or horizontally depending on the type of loading. The validity of the
model was assessed through an extensive comparison with numerical results
of the more accurate three-dimensional study, as well as with results from
experimental field tests. Close agreement was observed between results per-
taining to the two analyses. This represents an important step toward the
development of simplified, yet realistic, models for the analysis and design
of more general pile groups.
Thanks to the greatly reduced computational cost of the simplified model
it was possible to conduct an additional numerical study in order to quantify
more systematically the effect of several additional parameters such as soil
resistance and, in particular, soil degradation on pile response, which were
463
J. Geotech. Engrg. 1991.117:448-466.
ascelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. not examined by the three-dimensional analysis. It was observed that a pile
that is subjected to a cyclic axial load equal to 60% of its ultimate capacity
can experience many load cycles without apparent damage. However, when
the load was increased to 80% of the ultimate load, failure occurred after
only two cycles. This may have been the mechanism that caused the ex-
cessive settlement and even failure of a number of buildings supported on
pile foundations in Mexico City during the great 1985 Michoacan earth-
quake. This conclusion must be regarded as tentative, as little information
is currently available on the degradation properties of soil in real situations.
Additional experiments are needed to better characterize the behavior of ac-
tual pile foundations under cyclic loads, thereby permitting the further de-
velopment of relatively simple predictive models.

ACKNOWLEDGMENTS

The research reported herein was supported by a grant (No. ECE-86/1060)


from the National Science Foundation. The authors are grateful to Dr. Clif-
ford Astill, the cognizant program officer at NSF. Thanks are also extended
to the Pittsburgh Supercomputing Center for the use of its computing facil-
ities. We would also like to thank Drs. Albert Jaime and Daniel Resendiz
for their helpful discussions and for making available results of the field tests
in Mexico City. We are grateful to the referees for their critical reviews and
helpful suggestions, which contributed to sharpen the focus of this paper.

APPENDIX I. REFERENCES

Baber, T. T., and Wen, Y-K. (1981). "Random vibration of hysteretic, degrading
systems." J. Engrg. Mech. Div., ASCE, 107(6), 1069-1087.
Brown, D. A., Morrison, C , and Reese, L. C. (1988). "Lateral load behavior of
pile group in sand." J. Geotech. Engrg., ASCE, 114(11), 1261-1276.
Chow, Y. K. (1987). "Axial and lateral response of pile groups embedded in non-
homogeneous soils." Int. J. Numer. Anal. Methods Geomech., 11(3), 621-638.
Long, J. H., and Reese, L. C. (1984). Testing and analysis of two offshore drilled
shafts subjected to lateral loads. ASTM, Philadelphia, Pa., 215-228.
Mansur, C. I., and Hunter, A. H. (1970). "Pile tests—Arkansas River project." J.
Soil Mech. and Found. Div., ASCE, 96(5), 1545-1582.
Nogami, T., and Konagai, K. (1987). "Dynamic response of vertically loaded non-
linear pile foundations." J. Geotech. Engrg., ASCE, 113(2), 147-160.
Nogami, T., Konagai, K., and Otani, J. (1988). "Nonlinear pile foundation model
for time-domain dynamic response analysis." Proc. 9th World Conf. on Earth-
quake Engrg., Japan Assoc. Earthquake Disaster Prevention, 3, 593-598.
Poulos, H. G., and Davis, E. H. (1980). Pile foundation analysis and design. John
Wiley and Sons, Inc., New York, N.Y.
Randolph, M. F., and Wroth, C. P. (1978). "Analysis of deformation of vertically
loaded piles." J. Geotech. Engrg., ASCE, 104(12), 1465-1488.
"Recommended practice for planning, designing and constructing fixed offshore plat-
forms," API Recommended Practice 2A (RP 2A), Sixteenth edition, pp. 47-53,
1986.
Reese, L. C , Cox, W. R., and Koop, F. D. (1975). "Field testing and analysis of
laterally loaded piles in stiff clay." Proc. 7th Offshore Tech. Conf., ASCE et al.,
671-690.
Scott, R. F. (1981). Foundation analysis, Prentice-Hall, Inc., Englewood Cliffs,
N.J.
Trochanis, A. M., Bielak, J., and Christiano, P. (1987). "On hysteretic dissipation
of piles under cyclic axial loading." J. Geotech. Engrg., ASCE, 113(4), 335-350.

464

J. Geotech. Engrg. 1991.117:448-466.


scelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.

Trochanis, A. M. (1988). "A three-dimensional nonlinear study of piles leading to


the development of a simplified model," thesis presented to Carnegie Mellon Uni-
versity, at Pittsburgh, Pa., in partial fulfillment of the requirements for the degree
of Doctor of Philosophy.
Trochanis, A. M., Bielak, J., and Christiano, P. (1990). "A three-dimensional non-
linear study of piles." J. Geotech. Engrg., ASCE, 116(3), 429-447.
Wen, Y-K. (1976). "Method for random vibration of hysteretic systems." J. Engrg.
Mech. Div., ASCE, 102(2), 249-264.

APPENDIX II. NOTATION

The following symbols are used in this paper:

A = Wen's parameter controlling initial stiffness;


A0 = initial value of A;
Ap = cross-sectional area of pile;
Cuh, C„, = undrained clay strength at pile head and tip level, respectively;
c = undrained strength for clay;
D = equivalent pile diameter;
EP,ES = Young's moduli of pile and soil, respectively;
Esh,Esl = Young's modulus of soil at pile head and tip level, respec-
tively;
fy = soil yield strength;
Gs = shear modulus of soil;
Gsh,Gst = shear modulus of soil at pile head and tip level, respectively;
H = concentrated horizontal load at pile head;
Hi,Hi = concentrated horizontal loads acting on two piles in pair;
= cross-sectional moment of inertia of pile;
Ktu,Km = vertical and horizontal interaction stiffness, respectively;
Krh,K„ = rotational soil spring constant at pile head and tip, respectively;
Ks, = vertical soil spring constant at pile tip;
k = reference stiffness of Wen's model;
khkf = initial and final stiffness of Wen's model;
ks = tangent stiffness of spring following Wen's model;
k k = distributed vertical and horizontal soil stiffness along pile sides,
respectively;
L = length of pile;
M = concentrated bending moment at pile head;
Ml,M2 = concentrated bending moments acting on two piles in a pair;
n = Wen's parameter controlling transition from linear to nonlinear
range;
q = distributed lateral soil resistance;
RU,RW = vertical and horizontal fixed ground stiffness reduction factors,
respectively;
t = time variable;
u = vertical pile displacement;
V = concentrated vertical load at pile head;
Vi,V2 = concentrated vertical loads acting on two piles in pair;
w = horizontal pile displacement;
x = spatial coordinate along pile;

465

J. Geotech. Engrg. 1991.117:448-466.


scelibrary.org by UNIVERSIDADE FEDERAL DO RIO DE on 10/01/12. For personal use only. No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers.
z = nonlinear transformed displacement variable in Wen's model;
=
Zsu,zSw nonlinear part of vertical and horizontal soil resistance, re-
spectively;
a = Wen's nonlinear stiffness weighting factor;
P,7 = Wen's parameters controlling shape of hysteresis loop;
8,4 = degradation rate corresponding to W e n ' s A parameter;
e = amount of dissipated hysteretic energy; and
4> = friction angle of soil.

J. Geotech. Engrg. 1991.117:448-466.

Vous aimerez peut-être aussi