Computers and Geotechnics 36 (2009) 718–724

Contents lists available at ScienceDirect

Computers and Geotechnics

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

Nonlinear analysis of laterally loaded rigid piles in cohesionless soil

Lianyang Zhang *
Department of Civil Engineering and Engineering Mechanics, University of Arizona, 1209 E 2nd Street, Room 200A, Tucson, Arizona 85721, USA

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, a method is developed for nonlinear analysis of laterally loaded rigid piles in cohesionless
Received 28 August 2008 soil. The method assumes that both the ultimate soil resistance and the modulus of horizontal subgrade
Received in revised form 11 November 2008 reaction increase linearly with depth. By considering the force and moment equilibrium, the system
Accepted 1 December 2008
equations are derived for a rigid pile under a lateral eccentric load. An iteration scheme containing
Available online 7 January 2009
three main steps is then proposed to solve the system equations to obtain the response of the pile.
To determine the ultimate soil resistance and the modulus of horizontal subgrade reaction required
Keywords:
in the analysis, related expressions are selected by reviewing and assessing the existing methods.
Rigid piles
Lateral loading
The degradation of the modulus of horizontal subgrade reaction with pile displacement at ground sur-
Cohesionless soil face is also considered. The developed method is validated by comparing its results with those of cen-
Ultimate soil resistance trifugal tests and three-dimensional finite element analysis. Applications of the developed method to
Modulus of horizontal subgrade reaction laboratory model and field test piles also show good agreement between the predictions and the exper-
Nonlinear analysis imental results.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction An alternative approach is the elastic analysis which makes use

of boundary element or finite element methods assuming that the
Rigid piles have been used frequently for supporting axial and soil is an elastic continuum [10–13]. This approach does not ac-
lateral loads for different structures, including high buildings, count for soil yielding and is, therefore, only valid for the analysis
transmission lines, power stations, and highway structures [1]. of laterally piles at very low load levels. However, such analyses
Lateral loads govern the design of piles in many cases. Design of have often been used at working load levels at which significant
piles under lateral loading is commonly based on two criteria. soil yielding may be expected. Satisfactory solutions can usually
The first criterion calculates the ultimate load and considers a gen- only be obtained in such cases if very low secant soil moduli are
eral factor of safety or specific resistance reduction factors and assumed. However, the selection of appropriate secant moduli is
load factors. The second criterion is based on an allowable lateral difficult, and such analyses cannot be used to predict the effect
displacement [2,3]. A design based on an allowable lateral dis- of overload in any rational manner [14].
placement provides a more rational approach because it can allow This paper presents a method for the nonlinear analysis of lat-
the design to incorporate the displacement tolerance of the struc- erally loaded rigid piles in cohesionless soil. For the developed
ture. Over the years, several methods have been developed to pre- method, the ultimate soil resistance and the modulus of horizontal
dict the load–displacement response of laterally loaded piles [1–5]. subgrade reaction are assumed to vary linearly with depth. Based
The most popular method of these in engineering practice is the on the review and assessment of the existing methods, the related
semiempirical method known as the p–y curve analysis (see, e.g., expressions are selected for determining the ultimate soil resis-
[6–9]). In the p–y curve method, a beam is used to represent the tance and the modulus of horizontal subgrade reaction. The degra-
pile, and the soil pressures acting on the pile are modeled by dis- dation of the modulus of horizontal subgrade reaction with pile
crete independent nonlinear springs. The major advantage of this displacement at ground surface is also considered. By considering
method lies in its ability to simulate the nonlinearity and nonho- force and moment equilibrium, the system equations for the pile
mogeneity of the soil surrounding the pile. The obvious disadvan- are derived. An iteration scheme is then proposed to solve the sys-
tage is that the continuous nature of the soil medium is ignored. tem equations to obtain the response of the pile under lateral load-
This method requires the input of a series of p–y curves, which ing. To validate the developed method, it is first compared to
are often difficult to select in the absence of data from instru- centrifugal model pile tests and three-dimensional finite element
mented lateral pile load tests. analysis results. Then it is applied to analyze laboratory model
and field test piles. The results show that the developed method
* Tel.: +1 520 6260532; fax: +1 520 6212550. can provide satisfactory prediction of the response of laterally
E-mail address: lyzhang@email.arizona.edu loaded rigid piles in cohesionless soil.

0266-352X/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2008.12.001
 L. Zhang / Computers and Geotechnics 36 (2009) 718–724 719

2. Method of analysis a y0 b
M M
Fig. 1a shows a pile of length L and diameter B, embedded in a H o y H o y
cohesionless soil and under a lateral eccentric load H. The basic ω
assumptions for the proposed method are as follows:
a a
1. The pile is rigid so that the displacement of the pile at ground
surface is a consequence of the rotation of the pile around a L L
pu
point located at some depth below the ground surface (see
Fig. 1a).
2. Both the ultimate lateral soil resistance (pu) and the modulus of p pu
horizontal subgrade reaction (kh) vary linearly with depth as
shown in Fig. 1b and c, i.e.,
z z mL
pu ¼ mz ðFL1 Þ ð1Þ
kh ¼ nz ðFL2 Þ ð2Þ c M d M
H o y H o y
3. The soil reaction (p) and pile displacement (y) relation is nonlin-
ear (see Fig. 2). The modulus of horizontal subgrade reaction b
(kh) will decrease with increasing pile displacement (y). As the b
a a
pile displacement increases to a certain level, the ultimate soil
resistance (pu) will be reached. L L c
pu
pu
With the above assumptions, the load–displacement behavior
p
of the pile can be analyzed as described in the following.
p
pu pu
2.1. Case 1
z mL -mL
Lateral force H and moment M are small and no soil reaction z
reaches the ultimate soil resistance (see Fig. 3b). Assume that the Fig. 3. (a) A laterally loaded rigid pile; (b) soil reaction distribution with no
pile rotates about a point at depth z = a and the displacement of yielding; (c) soil reaction distribution with yielding only in a region above the
rotation point and (d) soil reaction distribution with yielding in regions both above
and below the rotation point.

y0 the pile at the ground surface is y0 (see Fig. 3a), the displacement
H of the pile along the depth can then be expressed by
ω
e y pu kh
o o o ða  zÞy0
y¼ ð3Þ
a
Since there is no soil-yielding, the soil reaction along the depth
L
can be expressed, by using Eqs. (2) and (3), as
pu = mz kh = nz
ny0 ða  zÞz
p ¼ kh y ¼ ð0  z  LÞ ð4Þ
a
Horizontal force equilibrium yields
 
z z ny0 L2 a L
z H  ¼0 ð5Þ
a 2 3
Fig. 1. (a) A laterally loaded rigid pile; (b) variation of pu with depth and (c) Considering moment equilibrium about point 0, we have
variation of kh with depth.
 
ny0 L3 a L
Mþ  ¼0 ð6Þ
a 3 4

Thus we have two Eqs. (5) and (6) in two variables (y0 and a). By
solving these two equations, the pile displacement y0 correspond-
p ing to force H and moment M can be obtained. After y0 is obtained,
pu the displacement, bending moment and shear force of the pile
along the depth can also be determined.
(y, p)
2.2. Case 2
k h = p/ y
Lateral force H and moment M are in such magnitudes that only
part of the soil reaction in the region above the rotation point
0 y reaches the ultimate soil resistance (see Fig. 3c). Assume that the
pile rotates about a point at depth z = a and the displacement of
Fig. 2. Soil reaction and pile displacement relationship. the pile at the ground surface is y0 (see Fig. 3a), the displacement
720 L. Zhang / Computers and Geotechnics 36 (2009) 718–724

of the pile along the depth can then be expressed by Eq. (3). If the A. A simple computer program has been written to run this
depth of soil-yielding is z = b, the soil reaction along the depth can procedure.
be expressed, by using Eqs. (1)–(3), as

p ¼ pu ¼ mz ð0  z  bÞ ð7Þ 3. Ultimate soil resistance pu

ny ða  zÞz
p ¼ kh y ¼ 0 ðb  z  LÞ ð8Þ Several methods are available for determining the ultimate lat-
a
eral resistance pu to piles in cohesionless soils [2,8,15,16]. These
At depth z = b, we have pu = khy. So b can be related to a, by methods are either of a semiempirical nature or employ approxi-
using Eqs. (7) and (8), as mate analysis which often involves considerable simplifications
  [17]. Zhang et al. [18] did a comprehensive review of the existing
m
b¼a 1 ð9Þ methods for predicting the ultimate lateral resistance to piles in
ny0
cohesionless soils. For simplicity, the following simple expression
Consider, respectively, horizontal force equilibrium and bend- proposed by [16] is used to calculate pu:
ing moment equilibrium about point 0, we have
 pu ¼ K 2p cBz ð18Þ
1 ny a  2  1 
2 3
H  mb þ 0 b  L2  b  L3 ¼ 0 ð10Þ 
2 a 2 3 where K p ¼ tan2 ð45 þ /0 =2Þ is the passive earth pressure coeffi-
 
1 3 ny0 a  3 3
 1
4
cient, in which /’ is the effective internal friction angle of the soil;
M þ mb  b L  b  L4 ¼ 0 ð11Þ c is the effective unit weight of the soil; z is the depth from the
3 a 3 4
ground surface; and B is the diameter of the pile. Eq. (18) is consis-
Thus we have three Eqs. (9)–(11) in three variables (y0, a and b). tent with the experimental results of [19] and the recommendation
By solving these three equations, the pile displacement y0 corre- of [18].
sponding to force H and moment M can be obtained. With known Comparing Eqs. (1) and (18), the m in Eq. (1) can be simply
y0, the displacement, bending moment and shear force of the pile derive as
along the depth can also be obtained.
m ¼ K 2p cB ð19Þ
2.3. Case 3
4. Modulus of horizontal subgrade reaction kh
In this case, force H and moment M are great enough to yield the
soil not only in the region above but also below the rotation point, For cohesionless soils, it is reasonable to assume that the mod-
i.e., the soil reaction in both regions reaches the ultimate soil resis- ulus of horizontal subgrade reaction, kh, vary linearly with depth
tance (see Fig. 3d). The soil–pile interaction will be elastic only in [8,20]. According to Terzaghi [20] and Reese et al. [8], the variation
the middle region b 6 z 6 c. The displacement of the pile of kh with depth can be expressed by
along the depth can still be expressed by Eq. (3) (see Fig. 3a) and
thus the soil reaction along the depth can be expressed as kh ¼ nh z ð20Þ

p ¼ pu ¼ mz ð0  z  bÞ ð12Þ where nh is the constant of horizontal subgrade reaction [FL3].

Comparison of Eqs. (2) and (20) gives
ny0 ða  zÞz
p ¼ kh y ¼ ðb  z  cÞ ð13Þ n ¼ nh ð21Þ
a
p ¼ pu ¼ mz ðc  z  LÞ ð14Þ Terzaghi [20] recommended values of nh, respectively, for dry or
At depth z = c, we have pu = khy. So c can be related to a, by moist sands above groundwater level and submerged sands (see
using Eqs. (13) and (14), as Fig. 4). Research results show that the values of nh recommended
  by Terzaghi [20] are too low [8] and may only be representative
m of the ultimate (near failure) conditions [21]. Based on the results
c ¼a 1þ ð15Þ
ny0 of Meyer and Reese [22], Murchison and O’Neill [9] recommended
the correlations shown in Fig. 4 for determining nh based on the
Consider, respectively, horizontal force equilibrium and mo- relative density of sand.
ment equilibrium about point 0, we can have The modulus of horizontal subgrade reaction, kh, decreases with
 
1 2 ny a 2 1 3 1 increasing displacement, y (see Fig. 2). Many researchers have
H  mb þ 0 ðb  L2 Þ  ðb  L3 Þ þ mðL2  c2 Þ ¼ 0 ð16Þ studied the relation between kh and y, including Bhushan et al.
2 a 2 3 2
  [21], Mwindo [23], Kumar [24], Prakash and Kumar [25], Ikeda
1 3 ny0 a 3 1 4 1
M þ mb  ðb  L Þ  ðb  L Þ  mðL2  c3 Þ ¼ 0
3 4
ð17Þ and Matuzawa [26], and Kumar et al. [27]. Based on field pile test
3 a 3 4 3
data, Bhushan et al. [21] proposed nh (=kh/z) versus (y0/B) relations
So we have four Eqs. (9), (15), (16), and (17) in four variables (y0, for different sand relative density values (see Fig. 5), where y0 is
a, b and c). By solving these four equations, displacement y0 corre- the pile displacement at ground surface and B is the pile diameter.
sponding to force H and moment M can be obtained. After y0 is Mwindo [23], Kumar [24], Prakash and Kumar [25], and Kumar
determined, the displacement, bending moment and shear force et al. [27] studied extensively the variation of kh with y. Based on
of the pile along the depth can also be obtained. their study, the following exponential function can be used to de-
To determine the displacement y0 under force H and moment M, scribe the variation of nh (=kh/z) with y:
one first has to determine whether the pile is in Case 1, 2 or 3 and nh
then solves the corresponding equations. Since Eqs. (5), (6), (10), ¼ 0:052c0:48 ð22Þ
nh max
(11), (16), and (17) are high order nonlinear equations, an iterative
numerical procedure is proposed to solve them. The complete pro- where nhmax is the maximum (low strain) constant of horizontal
cedure for solving the nonlinear equations to determine the dis- subgrade reaction at a strain of c 6 0.002; and c is the shear strain
placement y0 at force H and moment M is described in Appendix which can be related to displacement y0 as follows
 L. Zhang / Computers and Geotechnics 36 (2009) 718–724 721

5. Rigidity of piles
Very Very
Loose Medium dense Dense
loose dense
80 The proposed method is applicable only to rigid piles in cohe-
sionless soil. A pile in cohesionless soil can be considered to be ri-
70 gid for practical purposes if the following condition is satisfied
Terzaghi (1955) nhmax
[28]:
Reese et al. (1974)
Constant of subgrade reaction, nh (MN/m3)

60 (a) for sand above the water table L

2 ð25Þ
(b) for sand below the water table
T
pffiffiffiffiffiffiffiffiffiffi
50 (a) in which T ¼ 5 EI=n; L is the length of the pile; and EI is the bending
stiffness of the pile.
40 nhmax
6. Comparison with existing solutions

30 (b)
In this section, the developed method is applied to analyze the
centrifugal model tests of piles and compare the results with those
20 (a) from three-dimensional finite element analysis by Laman et al. [1].
To model the behavior of a prototype pile with B = 1 m and L = 2 m,
10 Nazir [1] conducted model tests, respectively, at a centrifugal
(b)
acceleration of 50 g on a pile with a diameter B = 20 mm and an
embedment length of L = 40 mm (Test 1), at 40 g on a pile with
0
B = 25 mm and L = 50 mm (Test 2), and at 33.3 g on a pile with
0 10 20 30 40 50 60 70 80 90 100
B = 30 mm and L = 60 mm (Test 3), where g is the gravity accelera-
Relative density, Dr (%)
tion. The tests were carried out in dense sand with a lateral force
Fig. 4. Constant of horizontal subgrade reaction, nh, versus relative density, Dr, applied at a height of 120 mm above the soil surface. The effective
(modified after [9]). unit weight and the effective internal friction angle of the sand
were, respectively, 16.4 kN/m3 and 46.1°. The pile rotation angle
(x) was measured under various moments (M = He) during the
test, and it is plotted in Fig. 6 in the prototype scale.
Laman et al. [1] conducted three-dimensional finite element
analysis of these centrifugal model test piles by adopting a hyper-
bolic stress–strain model. The predicted moment M versus rotation
x curve is also plotted in Fig. 6. It compares well with the average
value of the three centrifugal models tests.
To analyze the prototype pile using the developed method,
the two parameters m and n need be determined. For
 
/0 ¼ 46:1 ; K p ¼ tan2 ð45 þ /0 =2Þ ¼ 6:16. Using Eq. (19), the m for
the prototype pile (B = 1.0 m) can be estimated as 621.7 kN/m2.

Fig. 5. Variation of constant of horizontal subgrade reaction, nh, with pile head
displacement, y0/B (modified after [21]).

y0
c¼ ð23Þ
1:667B
Combining Eqs. (22) and (23) gives

nh y 0:48
¼ 0:066 0 ð24Þ
nh max B

Since the nh values recommended by Murchison and O’Neill [9]

(see Fig. 4) are for determining the initial portion of the p–y curve,
they can be considered as nhmax. By selecting the values of nhmax
from Fig. 4, the variation of nh with (y0/B) based on Eq. (24) can
be obtained as shown in Fig. 5. It can be seen that the variation
of nh with (y0/B) based on Eq. (24) and the nhmax in Fig. 4 is in good
agreement with the nh versus (y0/B) relations proposed by Bhushan
et al. [21]. Therefore, in this paper, Eq. (24) and the nhmax in Fig. 4
will be used to determine nh (=kh/z), which is, as shown in Fig. 2, Fig. 6. Comparison of current predictions with centrifugal model test data and
the secant value. three-dimensional finite element analysis results [1].
722 L. Zhang / Computers and Geotechnics 36 (2009) 718–724

The sand is dense with relative density Dr = 85% and thus nhmax can Table 1
be estimated from Fig. 4 to be 75.9 MN/m3. With the obtained Parameters for analysis of model test piles of Prasad and Chari [19].

nhmax and assuming a Young’s modulus of 207 GPa for the pile, Dr (%) 25 50 75
the pile can be checked using Eq. (25) to be rigid. With known c (kN/m3) 16.5 17.3 18.3
nhmax, n (=nh) can be determined from Eq. (24). With the obtained /’ (°) 35 41 45.5
m (kN/m2) 22.9 40.9 66.6
m and n, the M versus x curve is obtained using the developed nhmax (MN/m3) 6.23 25.3 59.3
method and it is also shown in Fig. 6. It can be seen that the results nh (MN/m3) nh
nh max ¼ 0:066ðyB0 Þ0:48
of the developed method agree well with those of the centrifugal
model tests and the three-dimensional finite element analysis.
of 5.2 m and 5.5 m. The tests were conducted at three sites with
7. Applications medium dense to very dense sands. For piles 1–3 at Site A, the load
was applied at a point about 0.38 m below the ground surface. For
In this section, the developed method is applied to analyze piles 4–5 and 6–7, respectively, at Sites B and C, the point of the ap-
model test piles of Prasad and Chari [19] and field test piles of Bhu- plied horizontal load was at the ground surface. Since piles 1–3
shan et al. [21]. (with a diameter of 1.07 m) were constructed with a 1.52 m diam-
eter bell in the bottom 0.61 m, only piles 4–5 and 6–7 which are
7.1. Model test piles straight along the depth are analyzed.
Site B contains silty sand (depth 0–0.9 m) and silty sand with
Prasad and Chari [19] conducted 15 tests on smooth steel model gravely layers (depth 0.9–5.5 m), with relative density of 77% and
piles in well graded angular dry sand. Each pile was 1135 mm long, 88%, and internal friction angle of 36° and 42°, respectively. The
102 mm in outside diameter and 5.6 mm in wall thickness. The unit weights of the two layers are both 16.5 kN/m3. Piles 4–5 are
sand had three relative densities of 25%, 50% and 75%, effective unit both 5.5 m long but pile 4 has a diameter of 0.61 m and pile 5 of
weight of 16.5, 17.3 and 18.3 kN/m3, and effective internal friction 0.91 m. Using an internal friction angle of 40°, the m for piles
angle of 35°, 41° and 45.5°, respectively. Fig. 7 shows the measured 4–5 can be estimated as 212.7 kN/m2 and 319.1 kN/m2, respec-
load–displacement curves for the pile at three different relative tively. Since the soil is medium dense to very dense, nhmax can be
densities. The pile was buried to a depth (L) of 612 mm, and the lat- estimated from Fig. 4 as 60.9 MN/m3. With the obtained nhmax,
eral load was applied at an eccentricity (e) of 150 mm. the piles can be checked using Eq. (25) to be rigid. n (=nh) can be
To analyze the model test piles, the two parameters m and n are determined from Eq. (24) with known nhmax. The lateral load ver-
first estimated as shown in Table 1. With the obtained nhmax, the sus groundline displacement curves are then predicted as shown
piles can be checked using Eq. (25) to be rigid. The lateral load ver- in Fig. 8. The predictions are in good agreement with the field
sus ground line displacement curves are then predicted and shown measurements.
in Fig. 7. It can be seen that the predictions are in reasonable agree- Site C contains silty sand (depth 0–1.8 m) and silty sand with
ment with the test data. gravely layers (depth 1.8–5.5 m), with relative density of 38% and
92%, and internal friction angle of 36° and 42°, respectively. The
7.2. Field test piles unit weights of the two layers are also both 16.5 kN/m3. Piles
6–7 have the same length as piles 4–5, with pile 6 having a diam-
Bhushan et al. [21] tested seven full scale drilled and cast-in- eter of 0.91 m and pile 7 1.22 m. Since Site C has the top layer
place piles with diameters of 0.61–1.22 m and embedded lengths which is thicker and looser than that at Site B, an internal friction
angle of 38° is selected and the m for piles 6–7 can then be esti-
mated, respectively, as 266.6 kN/m2 and 355.5 kN/m2. Again, nhmax
can be estimated from Fig. 4 as 60.9 MN/m3 and n (=nh) can be
determined from Eq. (24). The piles can be checked using

Fig. 7. Comparison of current predictions with model test data [19]. Fig. 8. Comparison of current predictions with field test data for piles 4 and 5 [21].
 L. Zhang / Computers and Geotechnics 36 (2009) 718–724 723

element analysis lends confidence in the applicability of this meth-

od to practical problems.

Appendix A. Solution technique

Since there exist three possible cases for given force H and mo-
ment M, one has to determine which case the pile is in before solv-
ing the equations to determine the pile displacement y0. Since Eqs.
(5), (6), (10), (11), (16), and (17) are high order nonlinear equa-
tions, an iterative numerical procedure is proposed to solve them.
The complete procedure for solving equations to determine the
pile displacement y0 at force H and moment M are described
below:

A.1. Main step1

Assume that the pile is in Case 2 and solve Eqs. (9)–(11) using
the following iterative procedure:
(1a) Starting with a small value for the pile displacement, Dy0,
and a series of trial displacements can be constructed as follows:
Fig. 9. Comparison of current predictions with field test data for piles 6 and 7 [21]. ðiÞ
y0 ¼ iDy0 ði ¼ 1; 2; . . .Þ ð26Þ
(1b) Assume a small value for a, denoted by Da. For a
ðiÞ
y0 ði ¼ 1; 2; . . .Þ, a series of trial values for a can be constructed as
Eq. (25) to be rigid. The lateral load versus groundline displace- follows:
ment curves are then predicted as shown in Fig. 9. Again, the pre- ðiÞ
dictions are in good agreement with the field measurements. aj ¼ jDa ðj ¼ 1; 2; . . .Þ ð27Þ

and the corresponding values for b can be calculated from (9) as

7.3. Comments on current predictions !
ðiÞ ðiÞ m
 To obtain the nonlinear response of rigid piles under lateral bj ¼ aj 1 ðiÞ
ð28Þ
ny0
loading, the three stages shown in Fig. 3 need be considered.
ðiÞ
Although theoretically rather complicated, the process can be If bj is equal to or smaller than zero, then the pile is in Case 1
easily completed using the simple computer program and the unbalanced force can be calculated from Eq. (5) as
developed. " ðiÞ #
ðiÞ
 Figs. 4 and 5 are recommended for quick and easy determination ny0 L2 aj L
DHðiÞ
j ¼ H  ðiÞ  ð29Þ
of the modulus of horizontal subgrade reaction. It need be noted a 2 3
j
that the effect of free length of loading and the pile diameter is
not considered in the correlations shown in Figs. 4 and 5. This Otherwise the pile is in Case 2 and the unbalanced force can be
may explain why piles 4 and 5 in Fig. 8 and piles 6 and 7 in calculated from Eq. (10) as
Fig. 9, respectively, have the same predicted elastic response, " ðiÞ #
1  ðiÞ 2 ny0 aj  ðiÞ 2  1
ðiÞ 3

but the measured load–displacement responses are different DHðiÞ
j ¼ H  m bj þ ðiÞ 2
ðbj Þ  L  ðbj Þ  L 3
2 a 2 3
for different diameter piles at the same site. To consider the j
effect of pile diameter on the modulus of horizontal subgrade ð30Þ
reaction, the correlations of Guo and Lee [29] and Guo [30]
can be used. Now a series of values for the unbalanced force can be obtained
 After pile displacement y0 at ground surface is obtained, the pile from Eqs. (29) or (30). The process will be continued until j = k + 1
ðiÞ ðiÞ
displacement (y) and soil reaction (p) along the depth can be at which point DHk  DHkþ1 < 0: That means the solution for a
ðiÞ ðiÞ
easily determined, using the equations for the corresponding corresponding to y0 and satisfying Eqs. (5) or (10) is between ak
ðiÞ ðiÞ ðiÞ ðiÞ ðiÞ
cases shown in Fig. 3. With known soil reaction (p) along the and akþ1 . Denoting ak as ½aL 0 and akþ1 as ½aU 0 and correspond-
ðiÞ ðiÞ ðiÞ ðiÞ
depth, the bending moment along the depth can also be deter- ingly, DHk as ½DHL 0 and DHkþ1 as ½DHU 0 , then the solution of a
ðiÞ ðiÞ ðiÞ
mined. Since the paper focuses on the prediction of displace- corresponding to y0 is in the range h½aL 0 ; ½aU 0 i. By reducing the
ðiÞ
ments, the expressions of bending moment are not presented. solution range, the solution of a corresponding to y0 can be ob-
tained. Introducing ½aðiÞ 0 as
ðiÞ ðiÞ
½aL 0 þ ½aU 0
8. Summary and conclusions ½aðiÞ 0 ¼ ð31Þ
2
A method for nonlinear analysis of laterally loaded rigid piles in and using Eqs. (28) and (29) or (30), a ½DHðiÞ 0 corresponding to ½aðiÞ 0
ðiÞ ðiÞ ðiÞ
cohesionless soil has been presented. This method considers the can be obtained. If ½DHL 0  ½DHðiÞ 0 < 0, then let ½aL 1 ¼ ½aL 0 and
ðiÞ ðiÞ ðiÞ ðiÞ ðiÞ ðiÞ
ultimate soil resistance and the modulus of horizontal subgrade ½aU 1 ¼ ½a 0 , otherwise let ½aL 1 ¼ ½a 0 and ½aU 1 ¼ ½aU 0 . By so
ðiÞ ðiÞ
reaction linearly increasing with depth. The degradation of the doing, a reduced solution range is obtained as h½aL 1 ; ½aU 1 i. This
ðiÞ ðiÞ
modulus of horizontal subgrade reaction with pile displacement process will be continued until a solution range h½aL m ; ½aU m i is
(i)
at ground surface is also considered. The good agreement obtained achieved at which the absolute value of [DH ]m is smaller than a
between the results of this method and published data of labora- very small prescribed value. At this stage, ½aðiÞ m can be regarded
ðiÞ
tory model and field pile tests and three-dimensional finite as an approximate solution of a corresponding to y0 , denoted as a(i).
724 L. Zhang / Computers and Geotechnics 36 (2009) 718–724

ðiÞ
(1c) For each y0 and the corresponding a(i) obtained from step References
(1b), the unbalanced moment DM(i) can be obtained using Eqs.
(9) and (6) or (11) as [1] Laman M, King GJW, Dickin EA. Three dimensional finite element studies of the
moment-carrying capacity of short pier foundations in cohesionless soil.
If a(i) 6 0, the pile is in Case 1 and Comput Geotech 1999;25(3):141–55.
  [2] Poulos HG, Davis EH. Pile foundation analysis and design. New York, NY: John
ðiÞ
ny0 L3 aðiÞ L Wiley & Sons; 1980.
DMðiÞ ¼ M þ  ð32Þ [3] Zhang L. Drilled shafts in rock – analysis and design. London (UK): Balkema;
aðiÞ 3 4 2005.
[4] Zhang L, Ernst H, Einstein HH. A non-linear analysis of laterally loaded rock-
If a(i) > 0, the pile is in Case 2 and socketed shafts. J Geotech Geoenviron Eng ASCE 2000;126(11):955–68.
1 [5] Guo W. On limiting force profile, slip depth and response of lateral piles.
ðiÞ
DMðiÞ ¼ M þ mðb Þ3 Comput Geotech 2006;33:47–67.
3 [6] Matlock H. Correlations for design of laterally loaded piles in soft clay. In:
ðiÞ  
ny aðiÞ  ðiÞ 3  1
ðiÞ
Proceedings of 2nd annual offshore technology conference, Houston, Texas,
 ðiÞ0 ðb Þ  L3  ðb Þ4  L4 ð33Þ vol. 1; 1970: p. 577–94.
a 3 4 [7] Matlock H, Reese LC. Generalized solutions for laterally loaded piles. J Soil
Mech Found Div ASCE 1960;86(5):63–91.
ðiÞ [8] Reese LC, Cox WR, Koop FD. Analysis of laterally loaded piles in sand. In:
(1d) Using the values of y0 from Eq. (26), a series of values for
Proceedings of 6th offshore technology conference, Houston, TX, 1974; 2: pp.
the unbalanced moment can be obtained from Eqs. (32) or (33). 473–83.
The process will be continued until i = l + 1 at which point [9] Murchison JM, O’Neill MW. Evaluation of p–y relationships in cohesionless
DM(l)  DM(l+1) < 0. That means the solution of y0 satisfying force soils. In: Analysis and design of pile foundations: proceedings of a symposium
ðlÞ ðlþ1Þ ðlÞ sponsored by the asce geotechnical engineering division and a session
and moment equilibrium is between y0 and y0 . Denoting y0
ðlþ1Þ (i) L sponsored by the ASCE technical council on codes and standards in
L U
as ½y0 0 and y0 as ½y0 0 and correspondingly, DM as ½DM 0 conjunction with the ASCE national convention, San Francisco, California;
and DM(i+1) as ½DM U 0 , then the solution of y0 is in the range 1984: p. 174–91.
[10] Poulos HG. Behavior of laterally loaded piles: I. Single piles. J Soil Mech Found
h½yL0 0 ; ½yU0 0 i. By reducing the solution range, the solution of y0 can
Div ASCE 1971;97:711–31.
be obtained. Introducing [y0]0 as [11] Poulos HG. Difficulties in prediction of horizontal deformations in foundations.
J Soil Mech Found Div ASCE 1972;98(8):843–8.
½yL0 0 þ ½yU0 0 [12] Banerjee PK, Davies TC. The behavior of axially and laterally loaded piles
½y0 0 ¼ ð34Þ embedded in non-homogeneous soils. Geotechnique 1978;28:309–26.
2 [13] Randolph RF. The response of flexible piles to lateral loading. Geotechnique
1981;31(2):247–59.
and using Steps (1b) and (1c), a [DM]0 corresponding to [y0]0 can be [14] Budhu M, Davies TG. Nonlinear analysis of laterally loaded piles in
obtained. If ½DM L 0  ½DM0 < 0, then let ½yL0 1 ¼ ½yL0 0 and cohesionless soils. Can Geotech J 1987;24:289–96.
½yU0 1 ¼ ½y0 0 , otherwise let ½yL0 1 ¼ ½y0 0 and ½yU0 1 ¼ ½yU0 0 . By so doing, [15] Broms BB. Lateral resistance of piles in cohesive soils. J Soil Mech Found Div
ASCE 1964;90(2):27–63.
a reduced solution range is obtained as h½yL0 1 ; ½yU0 1 i. This process [16] Fleming WGK, Weltman AJ, Randolph MF, Elson WK. Piling
will be continued until a solution range h½yL0 m ; ½yU0 m i is achieved at engineering. Glasgow and London (UK): Surrey University Press; 1992.
which the absolute value of [DM]m is smaller than a very small pre- [17] Jamiolkowski M, Garassino A. Soil modulus for laterally loaded piles. In:
Proceedings of 9th international conference on soil mechanics and foundation
scribed value. At this stage, [y0]m can be regarded as an approximate
engineering, Tokyo, Japan; 1977: p. 87–92.
solution of y0. [18] Zhang L, Silva F, Grismala R. Ultimate lateral resistance to piles in cohesionless
soils. J Geotech Geoenviron Eng ASCE 2005;131(1):78–83.
[19] Prasad YVSN, Chari R. Lateral capacity of model rigid piles in cohesionless soils.
A.2. Main step 2 Soils Found, Tokyo, Japan 1999;39(2):21–9.
[20] Terzaghi K. Evaluation of coefficient of subgrade reaction. Geotechnique
Check if the pile is in Case 3. From the results of Main step 1, the 1955;5:297–326.
[21] Bhushan K, Lee LJ, Grime DB. Lateral load tests on drilled piers in sand. In:
displacement of the pile at the pile base can be obtained from Eq. Drilled Piers A, Caissons A, editors. Proceedings of a session sponsored by the
(3) as geotechnical engineering division and the ASCE national
convention. Missouri: St. Louis; 1981. p. 114–31.
ða  LÞy0 [22] Meyer BJ, Reese LC. Analysis of single piles under lateral loading, Research
yL ¼ ð35Þ report no. 244-1, Center for Transportation Research, The University of Texas,
a Austin, Texas; 1979.
[23] Mwindo JM. Strain dependent soil modulus of horizontal subgrade reaction.
The corresponding elastic soil reaction can be obtained from Eq. MS thesis, University of Missouri, Rolla, MO; 1992.
(2) as [24] Kumar S. Non-linear load deflection prediction of single piles in sand using a
subgrade reaction approach. MS thesis, University of Missouri, Rolla, MO;
nLðL  aÞy0 1993.
pL ¼ ð36Þ [25] Prakash S, Kumar S. Nonlinear lateral pile deflection prediction in sands. J
a
Geotech Eng ASCE 1996;122(2):130–8.
If pL from Eq. (36) is smaller than the ultimate soil resistance at [26] Ikeda Y, Matuzawa H. Evaluation of coefficient of lateral subgrade reaction
based on horizontal loading test for piles. In: Geotechnical site
the pile base (i.e., pu = mL), then the pile is not in Case 3 and the
characterization: proceedings of 1st international conference, Atlanta,
solution from Main step 1 is the correct answer. Otherwise, the pile Georgia; 1998: p. 825–30.
is in Case 3 and goes to Main step 3. [27] Kumar S, Lalvani L, Omar M. Nonlinear response of single piles in sand
subjected to lateral loads using khmax approach. Geotech Geol Eng
2006;24:163–81.
A.3. Main step 3 [28] Tomlinson MJ. Foundation design and construction. 5th ed. London
(England): Pitman Books Ltd.; 1986.
[29] Guo WD, Lee FH. Load transfer approach for laterally loaded piles. Int J Numer
The pile is in Case 3 and the pile displacement y0 should be pre- Anal Methods Geomech 2001;25:1101–29.
dicted by solving Eqs. (7), (15), (16), and (17). A similar procedure [30] Guo WD. Laterally loaded rigid piles in cohesionless soil. Can Geotech J
as used in Main step 1 can be used to obtain the answers. 2008;45:676–97.