Xu 2013
Xu 2013
Xu 2013
a r t i c l e i n f o a b s t r a c t
Article history: This paper proposes a modified strain wedge model for the nonlinear analysis of laterally loaded single
Received 27 October 2012 piles in sandy soils by using the Duncan–Chang model as well as the Mohr–Coulomb model to describe
Received in revised form 13 December 2012 the stress–strain behavior of soils in the strain wedge. The input soil property for sandy soils only needs a
Accepted 11 January 2013
relative density which can be easily estimated from in situ tests. The strain wedge depth is calculated by
Available online 19 March 2013
an iterative process and the subgrade reaction modulus below the strain wedge is assumed to increase
linearly with depth, though it does not change with the lateral load applied to the pile. Seven case histo-
Keywords:
ries are used to verify the applicability of the proposed method. The results show the following: (1) good
Piles
Lateral load
agreements are found between the predicted and the measured results of full scale tested piles; (2) the
Strain wedge model predicted deflections and moments using the Duncan–Chang model are almost the same as those using
p–y Method the Mohr–Coulomb model; and (3) the size effect of the pile diameter or width on the subgrade reaction
Nonlinear stress–strain relationship modulus should be considered.
Size effect Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction [8–15], and some computer programs (i.e., COM624P [16], LPILE
PLUS [17]) are available for routine use. This method is popular be-
The majority of the piles encountered in practice may be re- cause of the fact that the p–y curves employed are back calculated
garded as ‘flexible’ under lateral loading [1]. Over the years, a num- from a few full-scale field tests, which, however, can make the p–y
ber of different approaches have been proposed to determine the curves site-specific [2]. The main problem with the p–y method is
behavior of laterally loaded piles. These methods can be generally that it neglects the soil continuity and the soil shearing resistance.
classified as the elastic beam on an elastic foundation method, the In continuum analysis, the soil surrounding the pile is treated as
p–y method, and the continuum analysis [2]. a three-dimensional continuum with assumptions made on its
The first method is initially referred as to the Winkler founda- stress–strain behavior or its constitutive relationships. This work
tion and assumes that the beam is supported by an array of dis- was pioneered by Poulos [18], who treated the soil mass as an elas-
crete springs representing the soil. This method is later extended tic continuum and the pile as a strip. However, this method is less
to deal with the problem of laterally loaded flexible piles, which popular than the p–y method because of the complex mathematics
is treated as a beam-on-foundation problem rotated by 90° [3–6]. involved in the analysis [2]. The 3D finite element method, which
However, this method is only satisfactory for very small strain lev- takes into account the three-dimensional soil–pile interaction
els because the soil behaves nonlinearly, especially at the top of the and both the elastic and the nonlinear behavior of soil, is fre-
pile. quently used to calculate the response of piles under a lateral load
The p–y method (p = the soil pressure (resistance) per unit pile [19–22]. However, the long computation time and the enormous
length and y = pile deflection) was first developed by Matlock [7] resources required for such an analysis prevent engineers from
and Reese et al. [8] for a single pile system. This method is concep- using 3D finite elements in routine designs.
tually similar to the beam on an elastic foundation method except Based on the deficiencies in the different methods discussed
that the employed soil springs are inelastic and is most commonly above, the p–y method with a three-dimensional passive wedge
used in the current design of laterally loaded piles. Over the years, in front of the pile, which considers the soil continuity and the
different p–y curves have been developed for different soil types soil–pile interaction, seems more practical to engineers. This meth-
od was originally developed by Reese et al. [8,9] to calculate the
ultimate soil resistance of the p–y curve and was later modified
⇑ Corresponding author. Tel./fax: +81 277 30 1621.
by Kim et al. [14] to calculate the ultimate soil resistance of the
E-mail addresses: [email protected] (L.-Y. Xu), [email protected]
hyperbolic p–y curve for clay. Norris [23] extended this method
(F. Cai), [email protected] (G.-X. Wang), [email protected] (K. Ugai).
0266-352X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compgeo.2013.01.003
L.-Y. Xu et al. / Computers and Geotechnics 51 (2013) 60–71 61
where D symbolizes the diameter or the width of the pile cross-sec- (3) The calculation of the strain wedge depth: the depth of the
tion and x is the depth below the top of the strain wedge. strain wedge is determined by an iterative procedure that
assumes that the pile deflection at the ground level in the
2.2. The strain wedge model in layered soil strain wedge equals that calculated from the BNEF analysis
and that the amount of volumetric compression in the strain
A multisublayer technique dealing with the strain wedge model wedge equals that, from the pile at the ground level to the
in layered soil was first developed by Ashour et al. [24]. The deflec- first zero-deflection depth, calculated from the BNEF analysis
tion pattern of the embedded pile in layered soil is assumed con- (see Section 3.2).
tinuous, as shown in Fig. 3a. Each sublayer is assumed to behave (4) To determine the subgrade reaction modulus below the
like uniform soil and has its own properties, depending on the sub- strain wedge, the subgrade reaction modulus below the
layer location. All variations at the middle of each sublayer, such as strain wedge is assumed to increase linearly with depth,
the mobilized fan angle, um, and the horizontal stress change, Drh, though it does not change with the lateral load applied to
are treated as the values for the entire sublayer (see Fig. 3a). The the pile (see Section 3.3).
governing equations of the strain wedge configuration at any in- (5) The finite element grid in the strain wedge is the same as
stant can be rewritten as follows: that in the pile; however, the thickness of each sublayer in
um;i Ref. [24] must be between 0.3 and 0.6 m.
bm;i ¼ 45 þ ð5Þ
2
3. Proposed method
um;i
Hm;i ¼ 45 ð6Þ
2 3.1. Determining the subgrade reaction modulus inside the strain
wedge
Bi C i ¼ D þ 2ðh xi Þ tan bm;i tan um;i ð7Þ
where i donates the number of sublayer, D is the pile diameter or Determining the subgrade reaction modulus, Es, inside the
width, and x is the depth from the pile at the ground level to the strain wedge can be summarized as follows [24]: (1) finding a rela-
middle of each sublayer under consideration (see Fig. 3a). tion between the subgrade reaction, p, and the horizontal stress
An array of discrete springs representing the soil is used to ac- change, Drh, in the strain wedge; (2) finding a relation between
count for the soil–pile interaction (the nonlinear springs in the the deflection pattern, d, and the horizontal strain, e, in the strain
strain wedge and the liner springs below the strain wedge in this wedge; (3) based on steps (1) and (2), the relation between the
study), as seen in Fig. 3b. subgrade reaction modulus (Es = p/y) and the secant soil modulus
(E = Drh/e) can be established.
2.3. Modifications in the proposed method
3.1.1. The relation between the subgrade reaction and the horizontal
The strain wedge model in this study is modified in the follow- stress change
ing aspects: For a horizontal slice of the wedge at depth x (see Fig. 1b), the
subgrade reaction, p (force per unit length of the pile) is expressed
(1) The stress–strain relationship in the strain wedge model: the by the following equation based on the principle of force
Duncan–Chang and the Mohr–Coulomb models are equilibrium:
employed (see Section 3.1.1).
(2) The size effect of the pile diameter or width on the subgrade pi ¼ Drh;i Bi C i S1 þ 2si DS2 ð8Þ
reaction modulus is considered according to Carter [29] and
where S1 and S2 equal 0.75 and 0.5, respectively, for a circular pile
Ling [30] (see Section 3.1.3).
section, and both values equal 1.0 for a square pile [31]. s is the
mobilized shear stress along the pile sides (see Fig. 1b), and is ex-
pressed by the following equations according to Ashour et al. [24]:
In this paper, a hyperbolic curve is used to describe the Drh–e u ¼ 28 þ 12:5Dr =100 ðdeg :Þ ð19Þ
relationship, which is referred to as the Duncan–Chang model
The relative density, Dr, can be estimated from the SPT results.
[25], as shown in Fig. 4. Such a relationship is incorporated in a
In this study, a modified version of Meyerhof’s correlation is used
more advanced model–the Hardening Soil model in PLAXIS [32].
to estimate the relative densities of the sand [27]:
As a comparison, the Mohr–Coulomb model (bilinear stress–strain qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
relationship) is also adopted (see Fig. 4). The basic equation of Dun- Dr ¼ 14 ðN1 Þ60 ð20Þ
can–Chang model is summarized as follows:
e where N1 is the SPT blow counts (N1)60 is the normalized SPT blow
r1 r3 ¼ ð11Þ counts corrected for the overburden effective stress (100 kPa refer-
a þ be
ence effective stress). According to Skempton [34] (N1)60 is ex-
where e is the axial strain in the triaxial test corresponding to the pressed as:
horizontal strain in the strain wedge and r1 and r3 are the major
and minor principal stresses, respectively. r1–r3 corresponds to 170N 1
ðN1 Þ60 ¼ ð21Þ
the horizontal stress change, Drh, in the strain wedge. Thus, Eq. r0v 0 þ 70
(11) can be rewritten as:
where r0v 0 is given in kPa.
e According to the Mohr–Coulomb failure criteria, the horizontal
Dr h ¼ ð12Þ
a þ be stress change at failure, Drhf, for sandy soils is given by the follow-
where the constants a and b are expressed as follows: ing equation:
1 2 sin u 0
a¼ ð13Þ Drhf ¼ ðr1 r3 Þf ¼ r ð22Þ
E0 1 sin u 3
To calculate BC in Eq. (8), the mobilized fan angle, um, should be
1 Rf calculated in advance (see Eq. (7)) from the following equation:
b¼ ¼ ð14Þ
ðr1 r3 Þult ðr1 r3 Þf
sin um =ð1 sin um Þ
SL ¼ ð23Þ
where E0 is the initial elastic modulus of the soil (r1–r3)ult is the sin u=ð1 sin uÞ
asymptotic value of the deviatoric stress (r1–r3)f is the deviatoric
stress at failure, and Rf is the failure ratio in the range of 0.75–1 where SL is the stress level and is defined as:
and is taken as 0.9 in this study, as recommended by PLAXIS [32]. Drh
SL ¼ ð24Þ
The initial elastic modulus, E0, is expressed as: Drhf
2E50 Note that the definition of SL in the Duncan–Chang model in
E0 ¼ ð15Þ
2 Rf this paper is different from the original definition by Duncan and
where E50 is the secant modulus of elasticity at 50% of (r1–r3)f, Chang [25], which is the ratio of Drh to (r1–r3)ult. The maximum
which is given by the following equation for sandy soils: value of the stress level is limited to one, and the maximum value
m of the mobilized fan angle, um, is limited to the internal friction an-
r03 gle, u, of the soil.
E50 ¼ Eref
50 ð16Þ
pref
3.1.2. The relation between the horizontal strain and the pile deflection
where Eref
50 is a reference stiffness modulus corresponding to the ref-
The shear strain, c/2, can be derived from Mohr’s circle for soil
erence confining pressure pref (taken as 100 kPa), r03 is the confining
strain, as shown in Fig. 5, and is defined as:
pressure in a triaxial test corresponding to the effective vertical
stress, r0v 0 , in the strain wedge, and m is a material parameter. Note c 1 1
¼ ðe ev Þ sin 2Hm ¼ ð1 þ tÞe sin 2Hm ð25Þ
that E50 is used as the elastic modulus for the Mohr–Coulomb mod- 2 2 2
el, as shown in Fig. 4. According to Norris [23], the deflection pattern d equals c/2.
There exist some empirical relations for sandy soils to calculate
Eref
50 , m and u from the relative density, Dr (as a percentage) and are
1 e
d¼ ð1 þ tÞe sin 2Hm ¼ ð26Þ
given by Eqs. (17)–(19) according to Brinkgreve et al. [33]. 2 Ws
Eref 2
50 ¼ 60; 000Dr =100ðkN=m Þ ð17Þ
Fig. 4. Duncan–Chang model and Mohr–Coulomb model. Fig. 5. Mohr’s circle for soil strain (modified after Norris [23]).
64 L.-Y. Xu et al. / Computers and Geotechnics 51 (2013) 60–71
where D
Es mod;i ¼ Es;i ð35Þ
Dref
2
Ws ¼ ð27Þ
ð1 þ tÞ sin 2Hm where Dref is 1 m.
The parameter Ws varies with Poisson’s ratio of the soil and the
mobilized fan angle. Poisson’s ratio, t, for sandy soils is in the range 3.2. Determining the strain wedge depth
of 0.1–0.5, and Hm can vary between 45° (for um = 0° at e = 0) and
25° (for um = 40° at failure) [24]. For this range for t and Hm, the The strain wedge depth, h, is calculated based on an iterative
value of the parameter Ws is between 1.74 and 1.82 for sandy soils process. First, the initial strain wedge depth, h0, and the initial hor-
[24], and in this paper, it is taken as a constant with an average va- izontal strain e0 are specified at first trial to calculate the subgrade
lue of 1.78. reaction modulus, Es, for the FEM analysis of the pile with a lateral
The pile deflection at sublayer i in the strain wedge can be writ- load, P0, and a moment, M0, applied to the pile. Second, assuming
ten as: that the pile deflection at the ground level in the strain wedge
equals that calculated from the BNEF analysis and that the amount
e of volumetric compression in the strain wedge equals that, from
yi ¼ ðh xi Þd ¼ ðh xi Þ ð28Þ
Ws the pile at the ground level to the first zero-deflection depth, calcu-
where h is the depth of the strain wedge. The pile deflection at the lated from the BNEF analysis (X0, see Fig. 3a), the new strain wedge
ground level ySW0 (x = 0) is given as: depth, h1, and the new horizontal strain, e1, are then obtained from
e the following equations:
ySW0 ¼ hd ¼ h ð29Þ 8 R X0
Ws ( >
ySW0 ¼ h1 We1s ¼ yFEM0 < 2
h1 ¼ 0y
yFEM dx
R X
) FEM0 ð36Þ
1
h h d ¼ 0 0 yFEM dx >
: e1 ¼ yFEM0 Ws
3.1.3. The relation between the subgrade reaction modulus and the soil 2 1 1 1 h1
modulus
The subgrade reaction modulus, Es, at sublayer i can be written where yFEM is the pile deflection within the depth X0, and yFEM0 is
as: the pile deflection at the ground level. Finally, the iterative process
pi stops when the adjacent absolute error of the pile deflection at the
Es;i ¼ ð30Þ ground level is less than a tolerant error of 105.
yi
pi refers to Eq. (8), which can be alternatively written as: 3.3. Determining the subgrade reaction modulus below the strain
pi =D Bi C i S1 2si S2 wedge
Ai ¼ ¼ þ ð31Þ
Drh;i D Drh;i
As the pile deflection below the strain wedge is quite small, the
where A is the ratio of the equivalent pile face stress, p/D, to the subgrade reaction modulus is consequently assumed unchanged
horizontal stress change, Drh, in the strain wedge, and is written for a lateral load applied to the pile. For cohesionless soils, it is rea-
as the following equation by substituting Eqs. (7)–(9) into sonable to assume that the modulus of the subgrade reaction var-
Eq. (31): ies linearly with the depth [8,35]. According to Terzaghi [35] and
Reese et al. [8], the variation of the subgrade reaction modulus
2ðh xi Þ tan bm;i tan um;i 2S2 K r0v 0;i tan us;i
A i ¼ S1 1 þ þ ð32Þ with the depth can be expressed by:
D Drh;i
Es ¼ nh z ð37Þ
Given that Drh = Ee in Eq. (31), pi can be rewritten as:
where nh is the constant of the subgrade reaction, and z is the var-
pi ¼ Ai DDrh;i ¼ Ai DEi e ð33Þ
iation of the depth. Again, the size effect of the pile diameter or
where E is the secant soil modulus. width should be taken into account for Es (see Eq. (35)).
Substituting Eqs. (28) and (33) into Eq. (30), Es at sublayer i Terzaghi [35] recommended values of nh for dry or moist sands
yields: above groundwater level and submerged sands, respectively (see
Fig. 6). Research results show that the values of nh recommended
pi Ai DEi e D
Es;i ¼ ¼ ¼ Ai Ws Ei ¼ N i Ei ð34Þ by Terzaghi [35] are too low [8] and may only be representative
yi dðh xi Þ ðh xi Þ of the ultimate (near failure) conditions [36]. According to Zhang
[15], the recommended correlations are shown in Fig. 6 for deter-
where Ni equals AWsD/(h–xi). The value of Ni becomes increasingly
mining the initial constant of the horizontal the subgrade reaction,
larger as xi approaches h. When the finite element grid is very small,
nh, based on the relative density.
the value of Ni increases sharply in a couple of sublayers near the
end of the strain wedge. This sharp increase is caused by the linear
approximation of the pile deflection pattern, which assumes that 3.4. Flow chart of the proposed procedure
the rotation of the pile at the bottom of the strain wedge is larger
than its actual value. Therefore, the value of Ni is forced to equal Based on Section 3, a computer code, SWPILE, was developed in
Ni–1 in the previous sublayer to avoid sharp increases in Ni when Fortran to analyze the laterally loaded pile. A flow chart is shown in
(h–xi)/D is lower than the critical value Cn. According to our re- Fig. 7.
search, Cn is in the range of 0.5–1 and is taken as the average value,
0.75, in this paper for all piles. 4. Verification process-case histories
To account for the size effect of the pile diameter or width on
the subgrade reaction modulus, Carter [29] and Ling [30] 4.1. Case histories: Reese et al. (1974)
suggested that the modulus of the subgrade reaction is linearly
proportional to the pile diameter or width. Thus, in this paper, Reese et al. [8] reported a series of tests performed on two steel
the subgrade reaction modulus is expressed by the following pipe piles with an outside diameter (O.D.) of 0.61 m and a wall
equation: thickness of 9.5 mm. The pile (pile 1 in the tests) was driven to a
L.-Y. Xu et al. / Computers and Geotechnics 51 (2013) 60–71 65
Little and Briaud reported six piles tested under a lateral mono-
Fig. 7. Flow chart for analysis of a laterally loaded pile in sandy soils.
tonic loading in sand [39]. The selected pile in this paper is a steel
pipe pile (pile No. 2) with an O.D. of 0.61 m and a wall thickness of
penetration of 21 m below the ground surface. A standard penetra-
15.875 mm. Concrete piles also were tested in this site. However,
tion test was conducted on the test site, which is considered to be
the concrete piles had previously been stressed or cracked during
composed of two layers according to the blow counts N1: (1) the
the vertical load tests and correctly selecting the properties of
upper layer with N1 = 18 from 0 to 6 m; and (2) the lower layer
the piles with concrete material was very complicated [39]. Hence,
with N1 = 40 from 6 to 21 m [30]. The relative density of the sands
the steel pipe pile only was analyzed in this paper. Based on boring
in the two layers is calculated from Eqs. (20) and (21). The horizon-
at the test site, the site in this paper is simplified as having four lay-
tal load level was 0.26 m above the ground surface. The pile and
ers: the first layer is underlain by a 7 m thick deposit of sand
Table 1
Pile properties used in SWPILE program.
Case histories Pile type Diameter (m) Embedded length (m) Flexural rigidity (kN m2) Distance* (m) Shape
Reese et al. (1974) Steel pipe 0.61 21 167,168 0.26 Round
Mohan and Shrivastava (1971) Steel pipe 0.1 5.25 313.6 0 Round
Little and Briaud (1988) Steel pipe 0.61 27.43 261,070 0.02 Round
Mansur and Alizadeh (1971) Pre-cast concrete, fluted pile (SB-2) 0.41 18.29 58,163 0 Round
Concrete filled steel pipe pile (P-3) 0.41 19.81 78,268 0 Round
Pre-cast concrete square pile (C-6) 0.41 18.29 98,740 0 Square
Cast-in-Place concrete pile (R-5A) 0.42 18.29 63,810 0 Round
*
Distance from line of loading to the ground surface.
66 L.-Y. Xu et al. / Computers and Geotechnics 51 (2013) 60–71
Table 2
Soil properties used in SWPILE program.
Case histories Layer Layer Unit weight Relative Constant of subgrade Internal friction
No. thickness (m) (kN/m3) density (%) reaction (kN/m3) angle (°)
Reese et al. (1974) 1 6 10.3 77.7 37.5 37.7
2 15 10.3 86.5 44 38.8
Mohan and Shrivastava (1971) 1 2.1 20.5 75* 69 37.4
2 1.2 10.2 75* 33 37.4
3 1.95 10.2 – 3.26 –
Little and Briaud (1988) 1 1.10 19.8 78.6 65 37.8
2 5.91 10 78.6 38 37.8
3 7.62 10 42.1 11 33.3
4 5.91 10 59.9 21 35.5
5 6.89 10 74.2 32 37.3
Mansur and Alizadeh (1971) 1 4.57a,b 19.66 60* 37.5 35.5
1.22c,d
2 1.60a,b 9.86 60 *
22.0 35.5
4.94c,d
3 12.13a,c,d 9.86 70* 29.5 36.8
13.65b
a
SB-2 pile.
b
P-3 pile.
c
C-6 pile.
d
R-5A pile.
*
The relative density is from reference.
300 300
250 250
Lateral load (P0), kN
Lateral load (P0), kN
200 200
150 150
Measured Norris (1986)
100 Reese (1974) 100 Measured
Mohr-Coulomb Reese (1974)
50 Norris (1986) 50 Duncan-Chang
Duncan-Chang Mohr-Coulomb
0 0
0 0.02 0.04 0 200 400 600
Ground line deflection (y0), m Max. moment (Mmax), kN·m
(a) (b)
Fig. 8. Comparison of the predicted and the measured (a) load–deflection curves and (b) load–maximum moment curves.
Fig. 10. Comparison of the predicted and the measured results of the defections and moments for a variation in depth under a lateral load of P0 = 4.9 kN.
200 250
200
Lateral load (P0), kN
150
150
100
100
Measured (SB-2) Measured (C-6)
50 LPILE PLUS LPILE PLUS
50
Duncan-Chang Duncan-Chang
Mohr-coulomb Mohr-coulomb
0 0
0 0.01 0.02 0.03 0.04 0 0.01 0.02 0.03 0.04
Ground line deflection (y0), m Ground line deflection (y0), m
(a) (b)
250 200
Lateral load (P0), kN
200
Lateral load (P0), kN
150
150
100
100
Measured (R-5A) Measured (P-3)
LPILE PLUS 50 LPILE PLUS
50
Duncan-Chang Duncan-Chang
Mohr-coulomb Mohr-coulomb
0 0
0 0.02 0.04 0.06 0.08 0 0.01 0.02 0.03 0.04
Ground line deflection (y0), m Ground line deflection (y0), m
(c) (d)
Fig. 12. Comparison of the predicted and the measured load-deflection curves for (a) SB-2; (b) C-6; (c) R-5A; and (d) P-3.
68 L.-Y. Xu et al. / Computers and Geotechnics 51 (2013) 60–71
predicts the deflections at a larger lateral load, which might be due strain in the strain wedge increases with an increase in the lateral
to the assumed average relative density used in the analysis. De- load. The shape of the curve is quite similar to that of load–deflec-
spite the discrepancy in the predictions using the p–y method for tion curve. In addition, the soil strain using the Mohr–Coulomb
cases SB-2, C-6 and R-5A, this method can give a good prediction model is slightly smaller than that using the Duncan–Chang model.
for case P-3. In general, the predicted results using the proposed This is why the predicted deflections and moments using the
method have a higher degree of accuracy than the results using Mohr–Coulomb model are slightly smaller.
the p–y method.
5.3. The strain wedge depth with a variation in the lateral load
5. Discussions
Fig. 15 shows the strain wedge depth for a variation in the lat-
5.1. The size effect of the pile diameter or width on the pile response eral load in the case of Reese et al. [8]. The strain wedge depths cal-
culated by the proposed method are close to those calculated by
Fig. 13 shows the size effect of the pile diameter or width on the Norris’ method. However, there are some differences between
pile response in the cases of Reese et al. [8] and Mohan and Shriv- the results using the Duncan–Chang model and those using the
astava [37]. For both cases, the results considering the size effect Mohr–Coulomb model. The strain wedge depth in the case of the
are found to be better than those not considering the size effect. Duncan–Chang model monotonically increases with an increase
However, there are large differences between the results consider- in the lateral load, while in the case of the Mohr–Coulomb model,
ing the size effect and those not considering the size effect, espe- the strain wedge depth first decreases and then increases with an
cially in the case of Mohan and Shrivastava [37] for the pile with increase in the lateral load. This is probably attributed to the stress
a smaller diameter. This is because the size effect in this study state in the strain wedge as well as to the discontinuous change of
not only influences the subgrade reaction modulus, but also influ- the stress–strain relationship in the Mohr–Coulomb model. At the
ences the soil strain in the strain wedge. Accordingly, the soil strain beginning of the loading, the soil strain in the strain wedge is very
in these two cases turns out to be larger when considering the size small (see Fig. 14) and not all sublayers reach the ultimate stress
effect; this can in turn reduce the secant soil modulus in the strain state. According to the second assumption on calculating the strain
wedge. Consequently, the subgrade reaction modulus is largely re- wedge depth, the amount of volumetric compression in the strain
duced when considering the size effect. It is expected that the wedge equals that calculated from the BNEF analysis (see Sec-
smaller is the pile diameter or width, the bigger is the influence tion 3.2). As a result, the strain wedge depth (h1 = yFEM0Ws/e1) re-
of the size effect on the pile response. quired for a small soil strain at the beginning of the loading
might be deeper compared to that under a subsequent load in
5.2. Soil strain in the strain wedge with a variation in the lateral load the case of the Mohr–Coulomb model. However, as the growth of
the lateral load continues, all sublayers approach the ultimate
Fig. 14 shows the soil strain in the strain wedge model for a var- stress state, and the strain wedge depth at this time must increase
iation in the lateral load in the case of Reese et al. [8]. The soil to meet the requirement of this assumption, even though the soil
Fig. 13. Size effect of the pile diameter or width on the pile response in the case of: (a) Reese et al.; and (b) Mohan and Shrivastava.
L.-Y. Xu et al. / Computers and Geotechnics 51 (2013) 60–71 69
300 sublayer in the strain wedge will not easily reach the ultimate
stress state if the lateral load applied to the pile is not very large.
250 Thus, the stress state in the strain wedge in the case of the Dun-
can–Chang model may be closer to the actual stress state than that
Lateral load (P0), kN
200 in the case of the Mohr–Coulomb model, even though the pre-
dicted deflections and moments using these two models have the
150 same accuracy.
300 300
250 250
Lateral load (P0), kN
Lateral load (P0), kN
200 200
150 150
Rf Rf
100 0.75 100 0.75
0.9 50 0.9
50
1 1
0 0
0 0.02 0.04 0 200 400 600
Ground line deflection (y0), m Max. moment (Mmax), kN·m
300 300
250 250
150 150
Cn Cn
100 100
0.0 0.0
50 0.5 50 0.5
Duncan-Chang Duncan-Chang
1.0 1.0
0 0
0 0.02 0.04 0 200 400 600
Ground line deflection (y0), m Max. moment (Mmax), kN·m
300 300
150 150
100 100
Ψs Ψs
50 Duncan-Chang 1.74 50 1.74
Duncan-Chang
1.82 1.82
0 0
0 0.02 0.04 0 200 400 600
Ground line deflection (y0), m Max. moment (Mmax), kN·m
Ws on the pile deflection (yFEM0 = h1e1/Ws, see Section 3.2) is insig- (4) The soil strain in the strain wedge increases with an increase
nificant. Consequently, it is taken as the average value in all case in the lateral load. The shape of this curve is quite similar to
histories. The calculations show that the same conclusion can also that of the load–deflection curve.
be drawn to the other cases. (5) The stain wedge depth in the case of the Duncan–Chang
model increases with an increase in the lateral load. How-
6. Conclusions ever, in the case of the Mohr–Coulomb model, the strain
wedge depth decreases with an increase in the lateral load
A modified strain wedge model is developed for laterally loaded until it reaches a critical value. This is attributed to the stress
single piles in sandy soils. The Duncan–Chang model and the state in the strain wedge as well as to the discontinuous
Mohr–Coulomb model are introduced to describe the stress–strain change of the stress–strain relationship in the Mohr–Cou-
behavior in the strain wedge. The input soil property for sandy lomb model.
soils only requires a relative density. The depth of the strain wedge (6) The predicted results are insensitive to the two parameters
is determined by an iterative procedure that assumes that the pile Cn and Ws. Consequently, the average values of the two
deflection at the ground level in the strain wedge equals that cal- parameters are taken in all case histories.
culated from the BNEF analysis and that the amount of volumetric (7) It can be concluded that the proposed method can provide
compression in the strain wedge equals that, from the pile at the an efficient and effective way for laterally loaded single piles
ground level to the first zero-deflection depth, calculated from in sandy soils.
the BNEF analysis. The subgrade reaction modulus below the strain
wedge is assumed to increase linearly with depth, but it does not
change with a lateral load applied to the pile. In addition, the sub-
grade reaction modulus is assumed to be linearly proportional to Acknowledgements
the pile diameter or width to account for the size effect of the pile
diameter or width on the subgrade reaction modulus. From the The first author thanks the China Scholarship Council for its
analysis of the seven case histories, the following conclusions can financial support and for a scholarship award. This research was
be obtained: partly supported by the National Natural Science Foundation of
China (Grant No. 51121005) and the Specialized Research Fund
(1) Good agreements are found between the predicted and mea- for the Doctoral Program of Higher Education of China (Grant No.
sured results of full scale tested piles. 20100041110003).
(2) The predicted deflections and the moments using the Dun-
can–Chang model are almost the same as those using the
Mohr–Coulomb model. References
(3) The size effect of the pile diameter or width on the subgrade
reaction modulus should be considered. [1] Randolph MF. The response of flexible piles to lateral loading. Géotechnique
1981;31:247–59.
L.-Y. Xu et al. / Computers and Geotechnics 51 (2013) 60–71 71
[2] Basu D, Salgado R, Prezzi M. Analysis of laterally loaded piles in multilayered [23] Norris G. Theoretically based BEF laterally loaded pile analysis. In: Proceedings
soil deposits. Technical report of joint transportation research program. Purdue of the 3rd international conference on numerical methods in offshore piling,
University; 2008. Paris, France; 1986. p. 361–86.
[3] Matlock H, Reese LC. Generalized solutions for laterally loaded piles. J Soil [24] Ashour M, Norris G, Pilling P. Lateral loading of a pile in layered soil using the
Mech Found Div 1960;86:63–91. strain wedge model. J Geotech Geoenviron Eng 1998;124:303–15.
[4] Broms BB. Lateral resistance of piles in cohesive soils. J Soil Mech Found Div [25] Duncan JM, Chang CY. Nonlinear analysis of stress and strain in soils. J Soil
1964;90:27–63. Mech Found Div 1970;96:1629–53.
[5] Broms BB. Lateral resistance of piles in cohesionless soils. J Soil Mech Found [26] Duncan JM. The role of advanced constitutive relations in practical
Div 1964;90:123–56. applications. In: Proceedings of 13th international conference on soil
[6] Francis A. Analysis of pile groups with flexural resistance. J Soil Mech Found mechanics and foundation engineering, vol. 5. New Delhi; 1994. p. 31–48.
Div 1964;90:10–32. [27] Zhang G, Robertson PK, Brachman RW. Estimating liquefaction-induced lateral
[7] Matlock H. Correlations for design of laterally loaded piles in soft clay. In: displacements using the standard penetration test or cone penetration test. J
Proceedings of 2nd annual offshore technology conference, Houston, Texas, Geotech Geoenviron 2004;130:861–71.
vol. 1; 1970. p. 577–94. [28] Japan Road Association. Specifications for highway bridges: Part V: Seismic
[8] Reese LC, Cox WR, Koop FD. Analysis of laterally loaded piles in sand. In: design. 5th ed. Tokyo; 2004 [in Japanese].
Proceedings of 6th offshore technology conference, Houston, TX, Paper No. [29] Carter DP. A non-linear soil model for predicting lateral pile response. Rep. No.
2080; 1974. p. 473–83. 359. Civil Engineering Dept., Univ. of Auckland, New Zealand; 1984.
[9] Reese LC, Cox WR, Koop FD. Field testing and analysis of laterally loaded piles [30] Ling LF. Back analysis of lateral load test on piles. Report No. 460. University of
in stiff clay. In: Proceedings of 7th offshore technology conference, Richardson, Auckland, Auckland, New Zealand; 1988.
TX, Paper No. 2312; 1975. p. 671–90. [31] Briaud JL, Smith T, Mayer B. Laterally loaded piles and the pressuremeter:
[10] Budhu M, Davies TG. Analysis of laterally loaded piles in soft clays. J Geotech comparison of existing methods. Laterally loaded deep foundation: analysis
Eng Div 1988;114:21–39. and performance. STP835, ASTM, West Conshohocken, PA; 1984. p. 97–111.
[11] Budhu M, Davies TG. Nonlinear analysis of laterality loaded piles in [32] Material Models Manual of PLAXIS, Delft, Netherlands; 2011.
cohesionless soils. Can Geotech J 1987;24:289–96. [33] Brinkgreve RBL, Engin E, Engin HK. Validation of empirical formulas to
[12] Reese LC. Analysis of laterally loaded piles in weak rock. J Geotech Geoenviron derive model parameters for sands; 2010. <http://kb.plaxis.nl/publications/
Eng 1997;123:1010–7. validation-empirical-formulas-derive-model-parameters-sands> (cited 24.
[13] Wu D, Broms BB, Choa V. Design of laterally loaded piles in cohesive soils using 09.12).
p-y curves. Soils Found 1998;38:17–26. [34] Skempton AW. Standard penetration test procedures and the effects in sands
[14] Kim Y, Jeong S, Lee S. Wedge failure analysis of soil resistance on laterally of overburden pressure, relative density, particle size, ageing and
loaded piles in clay. J Geotech Geoenviron Eng 2011;137:678–94. overconsolidation. Geotechnique 1986;36:425–47.
[15] Zhang L. Nonlinear analysis of laterally loaded rigid piles in cohesionless soil. [35] Terzaghi K. Evaluation of coefficient of subgrade reaction. Geotechnique
Comput Geotech 2009;36:718–24. 1955;5:297–326.
[16] Wang ST, Reese LC. COM624P: Laterally loaded pile analysis for the [36] Bhushan K, Lee LJ, Grime DB. Lateral load tests on drilled piers in sand. In:
microcomputer, Version 2.0. Report FHWA-SA-91-048. U.S. Department of, Drilled Piers A, Caissons A, editors. Proceedings of a session sponsored by the
Transportation; 1993. geotechnical engineering division and the ASCE national
[17] Reese LC, Wang ST, Arrellaga JA, Hendrix J. A program for the analysis of piles convention. Missouri: St. Louis; 1981. p. 14–131.
and rilled shafts under lateral loads. Computer Program LPILE PLUS, ENSOFT [37] Mohan D, Shrivastava SP. Nonlinear behavior of single vertical pile under
Inc; 1997. lateral loads. In: Proceedings of the 3rd annual offshore technology conference,
[18] Poulos HG. Behavior of laterally loaded piles: I–single piles. J Soil Mech Found vol. 2. Houston, Texas; 1971. p. 677–86.
Div 1971;97:711–31. [38] Poulos HG, Davis EH. Pile foundation analysis and design. New York: Wiley;
[19] Brown DA, Shie CF. Some numerical experiments with a three dimensional 1980.
finite element model of a laterally loaded pile. Comput Geotech [39] Little RL, Briaud JL. Full scale cyclic lateral load tests on six single piles in sand.
1991;12:149–62. Final report, Miscellaneous Paper, GL-88-27; 1988.
[20] Fan CC, Long JH. Assessment of existing methods for predicting soil response of [40] Mansur CI, Alizadeh M. Soil and foundation investigation report, rush island
laterally loaded piles in sand. Comput Geotech 2005;32:274–89. plant. Prepared by McLelland Engineers Inc.; 1971.
[21] Chik ZH, Abbas JM, Taha MR, Shafiqu QSM. Lateral behavior of single pile in [41] Kumar S, Alizadeh M, Lalvani L. Lateral load-deflection response of single piles
cohesionless soil subjected to both vertical and horizontal loads. Eur J Sci Res in sand. Electronic Journal of Geotechnical Engineering, Southern Illinois
2009;29:194–205. University; 2000.
[22] Kim Y, Jeong S. Analysis of soil resistance on laterally loaded piles based on 3D
soil–pile interaction. Comput Geotech 2011;38:248–57.