You are on page 1of 13

Estimation of the Linear Spring Constant for a Laterally

Loaded Monopile Embedded in Nonlinear Soil


Shivani Rani1 and Amit Prashant2

Abstract: In the case of large-diameter monopiles, pile deflection against lateral loads must be predicted accurately to determine the direct
influence on the superstructure. One-dimensional (1D) linear models of the pile-soil system, which are simple to analyze, are more commonly
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

used in practice than the sophisticated three-dimensional (3D) simulations. This study proposes an equivalent lateral strain concept to incor-
porate soil nonlinearity into an equivalent linear analysis of the pile-soil system. The spring constant is obtained by iterating over soil stiffness at
different strain levels and finding equivalent lateral strain from updated pile deflection at each iteration. It requires the maximum shear modulus
and modulus reduction curve for different layers of the soil profile. Correlation between the equivalent lateral strain and pile deflection has been
developed by comparing results from 1D and 3D simulations of long piles. The existing correlations between the lateral spring constant and
Young’s modulus of soil are too approximate to help in quantifying nonlinearity of soil accurately in this comparative study. Therefore, a more
accurate relationship between spring constant and Young’s modulus of soil for long piles has been proposed by incorporating relative stiffness of
soil and pile. DOI: 10.1061/(ASCE)GM.1943-5622.0000441. © 2014 American Society of Civil Engineers.
Author keywords: Soil-structure interaction; Spring constant; Strain; Pile; Nonlinear.

Introduction along the depth. However, finding representative p-y curves from the
available site data can often be difficult. The objective of this study is
A large-diameter pile (monopile) is often provided as a foundation to develop a simpler alternative approach of 1D simulation by
under structures such as bridges, flyovers, etc. It generally involves less performing an equivalent linear analysis of SPS.
construction time and cost in comparison with a pile group. The design In this study, 3D models of a monopile system with linear and
of a long monopile has relatively stricter limits on top deflection and nonlinear soils were created so that their responses could be used as
inclination of pile to check against misalignment with the superstruc- references to develop a procedure of 1D equivalent linear analysis.
ture. A pile load test in the field can provide a good estimate of pile For the 1D linear model, the spring constant ks could be calculated
deflection. However, such tests can only be performed in a limited using one of its correlations with the Young’s modulus of soil, Esoil , as
number during a project because of the associated huge cost and time, suggested in the literature (Terzaghi 1955; Muskhelishvili 1963;
besides requiring skilled personnel to handle them. In addition to these Poulos 1971; Matlock 1970; Scott 1980), based on experimental find-
tests, pile response is generally estimated through one-dimensional (1D) ings. While developing a correlation between nonlinear 3D analysis
or three-dimensional (3D) simulation of the soil-pile system (SPS) or by and equivalent linear analysis, it was necessary to first predict well the
using simplified models (Matlock and Reese 1962; Broms 1964a, b; linear analysis part using a more precise relationship between ks and
Poulos 1971; Maheshwari et al. 2005; Basu and Salgado 2007). Esoil than the existing ones, with due consideration to key factors
Although 3D simulation of a SPS is expected to simulate reality involved in this relationship. Hence, a parametric study was performed
well, it is more complex and time-consuming in comparison with to develop a more precise relationship between ks and Esoil .
1D models (Desai and Appel 1976; Randolph 1981; Trochanis et al. A procedure for equivalent linear analysis is proposed in this
1991; Yang and Jeremic 2005). Addressing soil nonlinearity and study along with a new correlation between ks and Esoil . Using the
interface effects in 3D models can further add to the complexity of proposed method, one can simulate a field problem, provided that
simulations (Kim and Jeong 2011; Mardfekri et al. 2013). As the value of Esoil in different layers along the depth of soil can be
a result, a 1D model has been more popular in practice. For the 1D obtained. Although Esoil is a generic property of soil, it is often
model, a subgrade reaction approach (Winkler 1867) has been correlated with field tests such as the standard penetration test
commonly used, in which soil is replaced by a series of unconnected (SPT) and the cone penetration test (CPT) or calculated though
springs. Nonlinear p-y curves (Matlock 1970; Reese et al. 1974, geophysical methods.
1975) are also sometimes used to define the lateral reaction of soil

Description of 1D and 3D Models


1
Formerly, M.Tech. Student, Indian Institute of Technology Gandhi-
nagar, Chandkheda, Ahmedabad 382424, India. E-mail: shivani.rani@ A monopile of 2.4-m diameter (D) and 31.5-m length (L) and
iitgn.ac.in having 29.58-GPa Young’s modulus (Epile ) was considered in this
2
Associate Professor, Indian Institute of Technology Gandhinagar, study. A pile length of 30 m was embedded into the ground, with
Chandkheda, Ahmedabad 382424, India (corresponding author). E-mail:
1.5 m above the ground surface. In the analysis, both homogenous
ap@iitgn.ac.in
Note. This manuscript was submitted on December 15, 2013; approved and heterogeneous soil profiles were considered. The considered
on July 28, 2014; published online on August 29, 2014. Discussion period heterogeneous soil profile was from one of the sites in Ahmedabad,
open until January 29, 2015; separate discussions must be submitted for India, where the monopile is expected to be constructed. The site was
individual papers. This paper is part of the International Journal of reported to have sandy soil up to 13-m depth with the SPT value
Geomechanics, © ASCE, ISSN 1532-3641/04014090(13)/$25.00. varying from 13 to 30, which was overlaying alternate layers of

© ASCE 04014090-1 Int. J. Geomech.

Int. J. Geomech.
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Multilayer soil profile considered for prediction of maximum


shear modulus with SPT value

sand and clay, as shown in Fig. 1. The SPT blow count in the bottom
layers was observed to exceed 50 for the first 150-mm penetration,
which is considered as refusal. Thus, the SPT value for these layers
is taken as 100. Fig. 2. 1D linear model of pile with springs spaced at 0.25-m interval

Details of Pile-Soil Models


one side of the plane of symmetry was simulated, as shown in Fig. 3.
Winkler’s model (1867) was utilized for 1D modeling, in which the Half of the lateral load was applied to the pile at 0.5 m above the
monopile was represented through a beam-column element and the ground surface. Pile and soil nodes along the plane of symmetry of
surrounding soil was replaced by a series of unconnected linear the SPS were allowed to move in the direction of load application
springs. The behavior of both soil and pile was assumed to be elastic. and along the depth of the pile, but they were restricted against out-
Roller support was provided at the bottom of the pile, and the lateral of-plane deformation. The nodes at the periphery and the bottom of
load (P) was applied at 0.5 m above the ground surface. Although the soil domain were given fixed boundary condition. Similar to the
there was a small amount of resistance available at the base of the 1D model, sliding and separation at the soil-pile interface were not
bored pile, its contribution toward the pile’s lateral load capacity considered, i.e., the nodes of the soil and pile elements at the in-
was not taken into account because of the effects of improper base terface had perfect bonding. To analyze the effect of domain size,
contact with the soil in such a construction. One can account for a a number of analyses were carried out by extending the side soil
fraction of base resistance in the analysis by providing an additional domain up to 3D, 5D, 7D, 11D, 13D, and 15D from the center of pile,
spring at the pile base. Both soil and pile elements were permitted to where D is the diameter of the pile. Pile deflection decreased with an
deform only in the plane containing the direction of the applied load increase in soil domain size, and the results predicted from the 13D
and depth of pile. Sliding or separation at the soil-pile interface was and 15D soil domains were approximately the same. Hence, the 15D
not considered in the analysis. Mesh size sensitivity was analyzed by size of the soil domain was considered for the analysis. Analyses
providing soil springs at different intervals along the depth of the were also performed with the soil domain extending up to 2D, 4D,
pile, i.e., at intervals of 3, 2, 0.5, 0.25, and 0.1 m. Overall, 10% and 6D below the pile, which showed negligible variation in the
variation was observed in the predicted values of pile deflection predicted deflection. Mesh-size sensitivity analysis showed con-
among the considered cases. Because the models with intervals of vergence for the mesh size of 0:2 3 0:2 3 0:5 m with an increase of
0.25 and 0.1 m showed almost no difference in response, the springs size in the radial direction.
were provided at 0.25-m intervals along the depth of the pile in the For nonlinear 3D analyses, the soil domain was simulated using
final model, as shown in Fig. 2. a pressure-independent multiyield material model (Yang et al.
In the 3D model, the soil and pile were assumed to have linear- 2008). The model parameters were density, Poisson’s ratio, shear
elastic isotropic material. Eight-noded standard brick elements with modulus, and shear strain, whose values for different layers are given
trilinear isoparametric formulation, which requires only three in Table 1. The values for density, thickness of soil, and soil type
degrees of freedom to define, were used in this study. The size of the were taken from the site investigation report. The maximum shear
soil elements was kept fine at the soil-pile interface and increased in strain and Poisson’s ratio were assumed to be 10% (0.1) and 0.3,
the radial direction away from the pile. The size of the elements respectively. The maximum shear modulus of each layer was cal-
along the depth of the pile was kept constant. Because the results culated using its correction with SPT data suggested by Kramer
were expected to be symmetric for a plane passing through the center (1996) for sandy strata and Ohsaki and Iwasaki (1973) for clayey
line of the pile and load vector, only half of the pile-soil system on strata.

© ASCE 04014090-2 Int. J. Geomech.

Int. J. Geomech.
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. 3D finite-element model of the considered monopile embedded in a multilayer soil profile

Table 1. Parameters Considered for 3D Nonlinear Model Having Multi- model with Esoil 5 65,000 kN=m2 are shown in Fig. 5. Similarly, the
layer Soil Profile along Depth of Pile maximum bending moment predicted from the 3D simulation
Soil density, Maximum shear modulus, showed about 10–15% with respect to Randolph’s (1981) solution,
Depth (m) Soil type r (kg=cm3 ) Gr (kPa) which is reasonable considering all the approximations involved in
such analyses. Basu and Salgado’s (2007) analytical elastic solution
0–6 Medium sand 1.75 74,475
for a multilayer soil system was used to further verify the linear-
6–13 Medium sand 1.8 167,725
elastic 3D model. Basu and Salgado (2007) provided an example for
13–17 Dense clay 1.85 308,223
a pile with D 5 0:25 m, L 5 20 m, and Epile 5 25 GPa. The analysis
17–21 Dense sand 1.85 356,459
was performed for three types of soil profiles having different Esoil
21–28 Dense clay 1.85 361,136
(Table 2). The pile was subjected to a lateral load of P 5 1,000 kN at
28–32 Dense sand 1.85 356,459
the ground surface. The predicted pile deflections at the ground
32–39 Dense clay 1.85 361,136
surface from the two solutions have also been listed in Table 2, which
Note: Poisson’s ratio ðnÞ 5 0:3; maximum shear strain, g max ð%Þ 5 10. indicates a reasonable match between the corresponding results.
After verification of the 3D linear-elastic model, it was assumed that
the model would work for other similar geometries of the problems
Verification of Models considered in this study, including the nonlinear case.
In the current study, the 1D linear model was verified using Hetényi’s
(1946) closed-form solution for a homogenous soil profile with Deflection of Pile Embedded in Nonlinear Soil
constant modulus along the depth of the pile. Ground-level deflec-
tion was calculated for the monopile with D 5 2:4 m, L 5 30 m, and Deflection of a pile depends upon the diameter (D), length (L),
Epile 5 29:58 GPa, which was embedded in a homogenous soil Young’s modulus of pile (Epile ), Young’s modulus of soil (Esoil ),
profile and subjected to lateral load P of 2,000 kN at the ground Poisson’s ratio of pile and soil (npile and nsoil ), and lateral load (P)
surface. The analysis was performed for five different values of soil applied to the pile (Matlock and Reese 1962). The best-known method
Young’s modulus (Esoil ) covering a wide range of soil, i.e., soft/loose for estimating pile capacity is performing a full-scale test in the field.
to stiff/dense. The predicted pile deflection at the ground surface The results predicted through such field tests are expected to be
from the 1D linear model is compared with the solution from representative of the SPS response, provided that all the measurements
Hetényi’s (1946) theory in Fig. 4(a), which shows a good match. are taken without ambiguity and the test is carried out with due care.
Similarly, the 3D model was verified by comparing the results with Such field tests involve enormous cost and pertinent skills. As a result,
Randolph’s (1981) solution for a pile embedded in homogenous soil. field tests are performed in a limited number and most of the SPSs
As shown in Fig. 4(b), the predicted values of pile deflection at the at different locations in a project are analyzed through simulations.
ground level from both the solutions match well. Although 3D simulations are close to field conditions, they require high
The bending moments and shear forces predicted from the 1D computational effort and time. Hence, 1D simulation is more common
linear model were compared with Hetényi’s (1946) solution, and in design practice due to its simple procedure and time efficiency.
were also found to match closely. As an example, the results for the Because soil is a nonlinear material and its Young’s modulus reduces

© ASCE 04014090-3 Int. J. Geomech.

Int. J. Geomech.
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Predicted ground-level lateral deflection of pile for homogenous soil profile (a) Hetényi’s (1946) theory and 1D linear model; (b) Randolph’s
(1981) solution and 3D linear model

Fig. 5. Shear force and bending moment along the depth of the pile using Hetényi’s (1946) theory and 1D linear models

with an increase in strain level, nonlinear p-y curves are often used Many investigators, such as Seed and Idriss (1970), Imazu and
to define the response of soil springs in the 1D model. Several Fukutake (1986), and Zhang et al. (2005), have suggested modulus
researchers such as Matlock (1970), Reese et al. (1974, 1975), and reduction curves for different soil types. These curves were gen-
Reese and Welch (1975) have suggested that mathematical ex- erated by performing dynamic laboratory tests such as the cyclic
pressions for the shape of p-y curves be based on the results of triaxial test, bender element test, resonant column test, etc. These can
experiments performed in the field. Recently, Guo (2006, 2013) be used for static analysis as well, assuming that the so-called
proposed design charts for laterally loaded piles based on the es- backbone curve from the dynamic testing results will coincide
timation of p-y curves through elastic-plastic formulation in active with the static stress-strain curve. Fig. 6 shows a representative
and passive modes. These nonlinear p-y curves can be used for modulus reduction curve suggested by Imazu and Fukutake (1986)
simulations in a project after building confidence in the nature of for different types of soils. Using this curve, the modulus reduction
those curves by verifying them on the site with due consideration to factor can be obtained at different strain levels. One can also estimate
heterogeneity of soil strata. Alternatively, one can envisage using the modulus reduction by performing triaxial or simple-shear tests
a 1D linear-elastic model of SPS by incrementally modifying the soil on representative soil samples, but in this case the data may not be
properties with strain level. Such an equivalent linear model will be reliable at small strains (below 0.05% shear strain) due to mea-
easier to simulate and analyze than using p-y curves. surement constraints.

© ASCE 04014090-4 Int. J. Geomech.

Int. J. Geomech.
Table 2. Ground-Level Deflection of Pile Estimated from 3D Linear Model and Basu and Salgado’s (2007) Theory
Deflection at ground surface (mm)
Model number Soil profile Soil layer thickness (m) Shear modulus of soil (MPa) Basu and Salgado (2007) 3D model
1 Homogenous soil 0–40 10 26 26.63
2 Multilayer soil (Case I) 0–1 10 22 21.9
1–4 20
4–9 40
9–40 80
3 Multilayer soil (Case II) 0–1 10 19.5 21.2
1–9 40
9–40 80
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

Table 3. Existing Correlations for Calculation of Spring Constant from


Young’s Modulus of Soil
Coefficient of horizontal
Serial number Reference subgrade reaction
1 Terzaghi (1955) 0:74 Esoil =Da
2 Broms (1964b) ð0:48e0:90Þ Esoil =D
3 Muskhelishvili (1963) 2Esoil =D
4 Matlock (1970) 1:8 Esoil =D
5 Poulos (1971) 0:82 Esoil =D
6 Scott (1980) Esoil =D
a
Esoil is the secant Young’s modulus of soil for a strain of approximately 1%.

to Esoil . In this analysis, the pile was subjected to P 5 2,000 kN,


Fig. 6. Simplified modulus reduction curve depending on the type of and Esoil was varied from 13,000 to 520,000 kN=m2 , covering its
soil expected range of values in the field. The deviation between the
predicted deflections from the 1D and 3D models was minimized by
iterating the value of the spring constant factor ( fsc ). This factor was
The spring constant ks can be estimated from Esoil using one of found to be a function of the Young’s modulus of soil instead of
the correlations listed in Table 3. The SPS model was considered being a constant, as shown in Fig. 9. This data can be approximated
to estimate pile deflection at different soil shear modulus values using the power function
using all of these correlations. The results are shown in Fig. 7,
which indicates a variation of approximately 250% (1.45–3.78 mm) fsc ¼ 0:424 Es0:136 (1)
in the calculated pile deflection at the ground surface. This vari-
ation was not unexpected because the recommended correlations It is interesting to note in Fig. 8 that Muskhelishvili’s (1963) rec-
had that variation as well. The results of the 3D simulations have also ommendation of fsc 5 2:0 is nearly the average of the variation range
been plotted in the same Fig. 7, which indicates that Muskhelishvili’s of fsc predicted through the present analysis.
(1963) recommendation provides a relatively better comparison
with the 3D simulation. It is uncertain, though, how this recom-
mendation will respond to various other scenarios. Further, the Effect of Poisson’s Ratio (n) on Spring Constant Factor
comparison of the 1D and 3D responses to incorporate the nonlinear To study the impact of Poisson’s ratio (n) on fsc , the SPS was an-
response into the equivalent linear model requires a more accurate alyzed for five different values of n, i.e., 0.25, 0.30, 0.35, 0.40, and
estimate of the linear response first. Hence, the relevant parametric 0.45, by keeping the other parameters of pile and soil constant.
study has been performed and discussed in the following section. These cases were analyzed at two different values of Esoil , i.e., 65,000
(medium-dense soil) and 520,000 kN=m2 (very stiff soil). The lateral
Linear Spring Constant for 1D Model load P 5 2,000 kN acted at 0.5 m above the ground surface. The
value of fsc was obtained by iterating its value to match results
The 1D model will be considered efficient if its response matches from the 1D and 3D models. The variation of fsc with the n value is
well with that from the 3D model for a given relationship of shown in Fig. 10(a), which shows only 5% overall variation for each
ks 5 fsc 3 Esoil . Pile deflection depends on the pile’s diameter, case of the Esoil value. The results also indicate that Muskhelishvili’s
embedded length, and relative stiffness with the soil, along with its (1963) ks works relatively well at low Esoil (65,000 kN=m2 ), but not for
lateral load P (Matlock and Reese 1962). Hence, fsc can be a stiffer soils.
function of some or all of these parameters. This is discussed in
light of the results from the 1D and 3D simulations. Effect of Young’s Modulus of Pile (Epile ) on Spring
Constant Factor
Effect of Young’s Modulus of Soil (Esoil ) on Spring
The effect of Epile was also studied by considering the same two
Constant Factor
values of Esoil , i.e., 65,000 and 520,000 kN=m2 . Poisson’s ratio (n)
Pile deflection at the ground surface from both the 1D and 3D linear- was taken as 0.3 for both soil and pile. The deflection of the pile for
elastic models are shown in Fig. 8, in which ks is taken to be equal the load P 5 2,000 kN was calculated at three different values of

© ASCE 04014090-5 Int. J. Geomech.

Int. J. Geomech.
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 7. Comparison study of pile lateral deflection estimated using existing correlations of spring constant and Young’s modulus of soil

Epile , i.e., 25, 29.58, and 35 GPa. The observed variation in fsc with
Epile is presented in Fig. 10(b), which shows that Epile alone has no
significant impact on fsc .

Effect of Lateral Load Intensity (P ) on Spring


Constant Factor
The effect of lateral load intensity on fsc was studied by changing the
value of P in multiple steps between 500 and 5,000 kN. The values of
Epile , Esoil , and n were taken as 29.58 GPa, 65,000 kN=m2 , and 0.3,
respectively. The estimated pile deflection at the ground surface
increased linearly with load intensity in both the 1D and 3D models;
however, there was no impact of load intensity on fsc , as shown in
Fig. 10(c). This was expected because the materials were elastic and
there was no geometric nonlinearity involved in the analyzed Fig. 8. Predicted lateral deflection of pile at ground level through 1D
system. and 3D linear models

 0:15
Spring Constant Factor for Multilayer Soil System Esoil
fsc ¼ 5 (2)
Epile
A multilayer soil system, as discussed previously during the de-
scription of the model, was also analyzed in this study. Maximum
shear modulus of soil for each layer was calculated using SPT cor- Matlock (1970) and Muskhelishvili (1963) proposed constant values
relations suggested by Kramer (1996) and Ohsaki and Iwasaki of fsc as 1.8 and 2.0, respectively, as shown in Fig. 12. Because
(1973). Spring constant values were estimated using the factor Muskhelishvili’s (1963) recommendation is close to the average of
suggested in Eq. (1), and Epile and npile were taken as 29.58 GPa the fsc predicted from the current analysis, it can be conveniently
and 0.3, respectively. The predicted deflection of the pile at the used to estimate pile deflection with some approximation. For a more
ground surface is presented in Fig. 11, which shows that the results accurate estimate of pile deflection, the soil spring constant (ks ) can
from the 1D analysis were approximately 15% higher than those be computed by using the following proposed relationship:
from the 3D model analysis at all load intensities. Hence, it is evident  0:15
Esoil
that the variability of soil strata under consideration can have some ks ¼ fsc  Es ¼ 5  Esoil (3)
influence on the value of fsc . Epile

Proposed Spring Constant Factor for 1D Linear Model

The parametric study presented in the previous section concludes Proposed Method of 1D Equivalent Linear Analysis
that it was only Esoil that had a major influence on the value of fsc for Considering Nonlinear Soils
the 1D linear model. The fsc values from all the cases analyzed in this
study have been put into one graph and plotted against the relative Fig. 13 shows the predicted pile deflection from the 3D nonlinear
stiffness of the soil and pile (Esoil =Epile ), as shown in Fig. 12. The and 1D linear models described previously. The results matched
Epile did not have significant influence on fsc , but it is used here to well only at small loads. With the increasing lateral load intensity,
make the Esoil term nondimensional. The predicted values of fsc fall the pile deflection from the 3D nonlinear model increased at
in a narrow band and the following power function can be used to a higher rate than that from the 1D model, which is as expected.
approximate its value: In this analysis, the lateral loads ranged from 500 to 5,000 kN.

© ASCE 04014090-6 Int. J. Geomech.

Int. J. Geomech.
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. Spring constant factor estimation from existing correlations and present study

The value of ks for the 1D linear model was estimated using the generally expected to be small enough that the maximum Esoil can
proposed relationship in Eq. (3), and Epile and n were taken as be conveniently used to represent the soil stiffness. For convenience
29.58 GPa and 0.3, respectively. of design, one can assume this depth of significant strains to be up to
To perform the 1D equivalent linear analysis, one can use a rep- approximately 5D from the ground surface. Hence, ELS will rep-
resentative modulus reduction curve to estimate the soil modulus at resent an average value of strain magnitudes in the soil layers up to
different strain magnitudes, as discussed previously. The de- a depth of 5D from the ground surface at a particular load. After
formation pattern around the pile has significant variation of strain in a depth of 5D from the ground level, the strains were assumed to be
the radial direction as well as along the depth, as shown by small enough to ignore nonlinearity effects, and the maximum Esoil
Hajialilue-Bonab et al. (2013) through imaging of deformations was used to estimate ks . The ELS (g) for a 5D depth of soil was then
around a scaled-down model of the pile. Therefore, estimating the correlated with the pile deflection at ground level yg normalized by
representative lateral strain in the soil that can be used to correlate pile diameter D
with a reduction in spring stiffness is not an easy task. A concept of
equivalent lateral strain (ELS) is proposed as a simple way of in- yg
g ¼ aels (4)
corporating it into the design. D

Here, aels is referred to as the equivalent lateral strain coefficient.


ELS in Soil
Imazu and Fukutake’s (1986) modulus reduction curves (Fig. 6)
The magnitude of strain varies from point to point in the soil as the were used to calculate the reduced value of the soil modulus at
pile deforms at some lateral load. A representative value of spring different strain levels, which were again used to define the multiyield
constant can be calculated only if the average value of the reduced material model for the 3D simulation. In the 1D model, initial values
modulus at a depth can be estimated with due consideration to strain of the soil’s secant modulus were first used to obtain spring constants
variation in the radial and vertical directions. Hence, an average using Eq. (3). These values were then iterated to match the predicted
value of lateral strain in a particular soil layer needs to be de- pile deflection with the response from the 3D nonlinear model. From
termined, which is referred to as ELS. The ELS will be a function of the modulus reduction curve, the strain corresponding to the final
pile deflection that can be normalized by pile diameter (D). secant modulus was then assumed to represent ELS. These values of
Because strain varies in both vertical and radial directions, one ELS and pile deflection at the ground level were put in Eq. (4) to
can divide the soil profile into multiple thin layers and define ELS for calculate aels and have been plotted in Fig. 14 for all the cases
each of those layers. This can make the procedure of estimating ELS analyzed in this study. Each model analysis at each load contributed
a little complex, as explained later in Example 2. For design sim- one point in this plot. These cases include combinations of different
plification, a solitary value of ELS can be considered to characterize n (0.25, 0.30, 0.35, 0.40, and 0.45), Epile (25, 29.58, and 35 GPa), and
the soil deformation up to a certain depth that is generally expected magnitude of loads P (2,000, 3,000, 5,000, and 10,000 kN). Fig. 14
to have significant strain magnitudes and hence major nonlinear also includes the results of aels for the models with multilayer soil
deformations. The lateral load response of the pile is mainly con- profiles used in linear analysis, but this time it was simulated with
trolled by the spring constant taken for the soil at a shallow depth (up nonlinear soil, and it shows a small amount of scatter in the estimated
to 3De4D), as reported by Zhang et al. (2013) through a comparative values of aels for the considered values of soil and pile properties. For
study. Below a depth of approximately 4D to 7D in the case of long all practical purposes, these values can be reasonably approximated
monopiles (here D is diameter or width of pile), the strains are as a constant aels 5 0:3. Hence, Eq. (4) can be modified to

© ASCE 04014090-7 Int. J. Geomech.

Int. J. Geomech.
y 
g
g ¼ 0:3 (5)
D
A few points had aels values much larger than 0.3, which corre-
sponded to homogeneous soil with high Young’s modulus
(2,000,000 kN=m2 ) at small deformations. Deviation for soils with
such a high stiffness has been ignored while making the recom-
mendation in Eq. (5).

Procedure of 1D Equivalent Linear Analysis using the


Proposed Method
The procedure of calculating pile deflection and the corresponding
ELS at different lateral loads in the proposed method is given in the
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

following steps:
1. Select a modulus reduction curve to be used in the analysis
and identify the corresponding threshold strain, i.e., the
strain up to which the soil modulus remains constant
(modulus reduction factor 5 1) and reduces thereafter.
2. Estimate the spring constant (ks ) for each soil layer by
substituting the corresponding Esoil in Eq. (3). Use maximum
Esoil in the first iteration.
3. Calculate the pile deflection for the model using ks values from
Step 2.
4. Determine the ELS value g for the top 5D depth of soil by
substituting the predicted pile deflection at the ground surface yg
from Step 3 into Eq. (5), i.e., g 5 0:3ðyg =DÞ. If the computed
strain is less than the threshold strain value, then there is no need
to perform further analysis. If not, proceed to Step 5.
5. Estimate reduced soil modulus Esoil for the strain value g from
Step 4 by using the modulus reduction curve.
6. Repeat Steps 2–5 until the value of ELS converges within
a certain tolerance limit.
The predicted pile response after convergence of equivalent
lateral strain can be used for the design purpose.

Example Simulations

Example 1: Single ELS Value Representing Soil


Deformations up to 5D Depth

Fig. 10. Spring constant factor at different values of (a) Poisson’s ratio; This example assumes ELS to represent an average strain of up to
(b) Young’s modulus of pile; (c) lateral loads (Esoil 5 65,000 kN=m2 ) a depth of 5D below the ground surface. The soil below a depth of
5D is assumed to have relatively small strains that allow using

Fig. 11. Predicted lateral deflection of pile at ground level through 3D and 1D linear models at different lateral loads assuming a multilayer soil profile

© ASCE 04014090-8 Int. J. Geomech.

Int. J. Geomech.
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 12. Spring constant factor as a function of relative stiffness (Esoil =Epile ) for pile subjected to lateral load

Fig. 13. Estimated lateral deflection of pile at ground surface at various lateral loads through 3D nonlinear model and 1D linear model

maximum Esoil to compute spring constants. This design simplifi- the lateral strain, the next iteration had to be performed by con-
cation is not necessary, but it makes the design process easier. sidering lower values of ks corresponding to the reduced modulus of
The multilayer soil-pile model illustrated previously was con- soil. Using the curves from Fig. 6, the modulus reduction factor
sidered as an example for estimating deflection along the depth of obtained was 0.630 for an ELS value of 0.000266. With the new
the pile using the proposed method. Recommendations by Kramer values of ks , the ground-level deflection was predicted as 2.964 mm,
(1996) and Ohsaki and Iwasaki (1973) were used to estimate the for which ELS in the soil was estimated as 0.00037. Because the
maximum shear modulus of soil in the sand and clay layers, re- change in ELS value was still more than the tolerance limit, further
spectively. The modulus reduction curve was assumed according to iterations were performed to reach convergence of ELS. After five
Imazu and Fukutake’s (1986) recommendations (Fig. 6). The pile such iterations, the value of ELS converged with the deflection at
was subjected to a 2,000 kN lateral load at a height of 0.5 m above ground level and the value of ELS in the soil were 3.29 mm and
the ground surface. 0.000412 (0.0412%), respectively. The deflected shape of the pile
In the first iteration, the magnitude of ELS, modulus reduction for all the iterations has been plotted in Fig. 15. All the ELS values
factor, and tolerance limit for ELS convergence were taken as 10206 , and other key information are summarized in Table 4.
1, and 62% of the strain level, respectively. Deflection of the pile Bending moment and shear force values were also estimated
was calculated for the ks value estimated from maximum Esoil using along the depth of the pile because these are important design
Eq. (3). Deflection at the ground surface was 2.13 mm in this it- parameters. The maximum bending moment was found to be at
eration. The ELS value was estimated as 0.000266 using Eq. (5), 4.5 m below the ground surface and its magnitude was approxi-
g 5 0:3ðyg =DÞ. Because ELS was greater than the threshold value of mately 15% higher than that from the 3D nonlinear simulations, as

© ASCE 04014090-9 Int. J. Geomech.

Int. J. Geomech.
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 14. Strain coefficient of a pile subjected to lateral load considering the variation of soil and pile properties

shown in Fig. 16. Hence, the 1D model prediction is slightly con- introduced in predictions by the approximations in the earlier model
servative for the design. The point load was applied at an external is not expected to be significant enough to worry about it in design
node at 0.5 m above the ground surface. Its lateral distribution within practice. This statement makes further sense when one considers the
the pile of 3D simulation had influence up to a depth of 1 m below the expected errors in predictions due to the uncertainties associated
ground surface, which was not expected to feature in the 1D sim- with soil and pile properties. Hence, it is recommended to use the
ulation. The shear forces below a depth of 1 m matched well for both proposed method with equivalent strain considering a depth of 5D
of the simulations. and a strain coefficient of 0.3 because it is relatively simple and easy
to implement.
Example 2: ELS Defined in Multiple Layers to Represent
Strain Variation along Depth
In this example, the entire soil profile was divided into a number of Strengths and Limitations of Proposed Methods
thin layers with 2-m thickness and then ELS was defined for each
layer independently. This model would represent a generalized re- Using this study, 3D soil-pile model behavior can be analyzed ap-
sponse of the pile-soil system subjected to lateral loads. The soil proximately by simulating a 1D linear model for long piles. Hence,
profile and pile were kept the same as in Example 1 to compare the the involved complexity, computation effort, and the time needed in
results from the two approaches. Because the definition of ELS has developing 3D models can be avoided. One can reasonably predict
changed in this example, from it representing 5D depth to individual deflection of a pile subjected to lateral load by calculating the mod-
thin layers, it is expected that the ELS coefficient (aels ) may change ulus of soil estimated using SPT/CPT correlations or geophysical
as well. First, to compare the results with Example 1, the same aels of methods and by adopting a suitable modulus reduction curve from
0.3 was considered in the analysis. Iterations for ELS convergence the literature. This can be of great use when there is a lack of high-
were performed for pile deflection at the ground level using the same quality field data.
procedure as in Example 1. The final pile deflection at the ground In this study, the models were validated using theoretical so-
surface after iterations was found to be 3.10 mm, which was slightly lutions available in the literature. It will be interesting to compare the
less than the 3D results (3:38 mm) as well as the result from 3D simulation results with pile-load test data at low deformations,
Example 1 (3:29 mm). This does not mean that the variable strain where the expected response is linear elastic. Experiments at large
consideration is erroneous. Considering the expected change in aels enough loads, which induce significant nonlinearity in the soil, will
due to redefined ELS, the value of aels was increased to 0.35 and then bring out the effect of common construction practices on the de-
to 0.40 for two more cases, and the pile deflection was predicted for formation pattern. This is important because the construction-related
each case by following the same procedure as explained previously. issues may have some impact on the response of the pile-soil system
The calculated deflection of the pile at the ground level and ELS in to future loads. The effect of the soil-pile interface, water table, and
soil layers are listed in Table 5. It was found that an aels value of inclination of soil layers were not considered in this study to reduce
0.4 predicted the deflection approximately the same as the deflection the complexity of the problem. Pile shape and construction method
from the 3D simulation. By analyzing a number of cases for different may also influence the results to some extent. The proposed method
soil and pile properties with variable strain considerations, a strain is limited to long piles with large diameters, and it may not directly
coefficient of approximately 0.4 may be used one day to perform apply to pile groups.
1D linear analysis. In that case, there could be a choice to use either Ideally, the predicted response from the proposed model should
the modified model with equivalent lateral strain in multiple layers of match the response of a full-scale field test. However, there are
the pile or the earlier model having a single ELS coefficient for the uncertainties associated with Young’s modulus of soil and mod-
top 5D of the pile length. ulus reduction curves of different layers of soil strata, which are
The model with variable strain consideration might be able to mostly determined using empirical correlations (Seed and Idriss
produce more accurate predictions than the earlier model, depending 1970; Imazu and Fukutake 1986; Zhang et al. 2005). It will be useful
on the pile and soil properties and geometry. However, the error to calibrate these two properties by comparing the predictions with the

© ASCE 04014090-10 Int. J. Geomech.

Int. J. Geomech.
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 15. Lateral displacement along the depth of pile at different iterations performed in the analysis of Example 1

Table 4. Equivalent Lateral Strain and Other Parameters at Different Iterations Performed in Analysis of Example 1
Iteration number Considered strain Strain factor Displacement at ground surface Calculated lateral strain
1 0.000001 1 0.00213 0.00027
2 0.00027 0.63 0.00296 0.00037
3 0.00037 0.564 0.00321 0.0004
4 0.0004 0.547 0.00328 0.00041
5 0.00041 0.543 0.00329 0.00041

Table 5. Pile Deflection at Ground Surface in Example 2 Considering Variation of Equivalent Lateral Strain in Soil along the Depth of Pile
Lateral strain in layer
Equivalent lateral Deflection at ground
strain equation level (m)a 1 2 3 4 5 6
b
0:40y=D 0.0033866 0.00057 0.00033 0.00016 0.0000564 0.00000146 0.0000164
0:35y=D 0.0032441b 0.00048 0.00028 0.00014 0.0000448 0.00000020 0.0000143
0:30y=D 0.0031025b 0.00039 0.00023 0.00011 0.0000347 0.00000130 0.0000122
a
Lateral load 5 2,000 kN.
b
3D deflection 5 0:003326 m.

© ASCE 04014090-11 Int. J. Geomech.

Int. J. Geomech.
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 16. Lateral deflection, bending moment, and shear force of laterally loaded pile embedded in multilayer soil system (Example 1)

field test response. The same model then can be used to predict pile the support from IIT Gandhinagar. The authors are also thankful
response at other similar locations in the same project. to MEGA, Gujarat for providing inspiration to conduct this study.

Conclusion Notation
Three types of models—1D linear, 3D linear, and 3D nonlinear— The following symbols are used in this paper:
were developed using OpenSees 2.3.2 to simulate long monopiles aels 5 equivalent lateral strain coefficient;
embedded in the soil. A parametric study was performed by com- D 5 diameter of pile;
paring the results from 3D and 1D linear models, which considered
Epile 5 Young’s modulus of pile;
variation of the Young’s modulus of soil, Poisson’s ratio of soil,
Esoil 5 Young’s modulus of soil;
Young’s modulus of the pile, and the applied lateral loads. The ratio
fsc 5 spring constant factor;
of the spring constant and Young’s modulus of soil was significantly
influenced by the stiffness of the soil. Influence of other parameters ks 5 lateral spring constant of soil;
such as the Poisson’s ratio of soil, stiffness of the pile, and lateral L 5 length of pile;
load were relatively insignificant. Based on these observations, P 5 lateral load applied on pile;
a new equation has been proposed to compute the spring constant yg 5 ground deflection of pile;
factor from the relative stiffness of the soil and pile (Esoil =Epile ). To g 5 expected equivalent lateral strain in soil;
consider soil nonlinearity in 1D equivalent linear analysis, a concept gmax 5 maximum shear strain; and
of ELS has been introduced as an average strain value representing n 5 Poisson’s ratio.
deformation pattern around the pile. The ELS is proposed to have
a correlation with normalized pile deflection through a constant
strain coefficient. In a simplified method, ELS was considered as References
a single value for soil up to a depth of 5D below the ground surface.
This ELS was then used to compute reduced stiffness of soil springs Basu, D., and Salgado, R. (2007). “Elastic analysis of laterally loaded pile
based on the lower modulus of soil at higher strains. For this def- in multi-layered soil.” Geomech. Geoeng., 2(3), 183–196.
inition of ELS, the associated strain coefficient was found to be Broms, B. B. (1964a). “Lateral resistance of piles in cohesionless soils.”
approximately 0.3 based on the results from a number of simulations. J. Soil Mech. and Found. Div., 90(3), 123–158.
Finally, an iterative procedure of 1D equivalent linear analysis is Broms, B. B. (1964b). “Lateral resistance of piles in cohesive soils.” J. Soil
proposed to estimate lateral load response of the large diameter long Mech. and Found. Div., 90(2), 27–64.
piles embedded in nonlinear soils. Desai, C. S., and Appel, G. C. (1976). “3D analysis of laterally loaded
structures.” Proc., 2nd Int. Conf. on Numerical Methods in Geomechanics,
Vol. 1, ASCE, Reston, VA, 405–418.
Acknowledgments Guo, W. D. (2006). “On limiting force profile, slip depth and response of
lateral piles.” Comput. Geotech., 33(1), 47–67.
The authors gratefully acknowledge Mr. Raj Mashruwala for en- Guo, W. D. (2013). “Pu -based solutions for slope stabilizing piles.” Int. J.
couraging this research through partial financial support besides Geomech., 10.1061/(ASCE)GM.1943-5622.0000201, 292–310.

© ASCE 04014090-12 Int. J. Geomech.

Int. J. Geomech.
Hajialilue-Bonab, M., Sojoudi, Y., and Puppala, A. J. (2013). “Study of Randolph, M. F. (1981). “The response of flexible piles to lateral loading.”
strain wedge parameters for laterally loaded piles.” Int. J. Geomech., Géotechnique, 31(2), 247–259.
10.1061/(ASCE)GM.1943-5622.0000186, 143–152. Reese, L. C., Cox, W. R., and Koop, F. D. (1974). “Analysis of laterally
Hetényi, M. (1946). Beams on elastic foundation, University of Michigan loaded piles in sand.” Proc., 6th Offshore Technology Conf., Vol. 2,
Press, Ann Arbor, MI. Offshore Technology Conference, Houston, 473–483.
Imazu, M., and Fukutake, K. (1986). “Dynamic shear modulus and damp- Reese, L. C., Cox, W. R., and Koop, F. D. (1975). “Field testing and analysis
ing ratio of gravel materials.” Proc., 21st Japan National Conf. on Soil of laterally loaded piles in stiff clay.” Proc., Offshore Technology Conf.,
Mechanics and Foundation Engineering, Japanese Society of Soil Me- Offshore Technology Conference, Houston.
chanics and Foundation Engineering, Tokyo, 509–512. Reese, L. C., and Welch, R. C. (1975). “Lateral loading of deep foundations
Kim, Y., and Jeong, S. (2011). “Analysis of soil resistance on laterally in stiff clay.” J. Geotech. Engrg. Div., 101(7), 633–649.
loaded piles based on 3D soil–pile interaction.” Comput. Geotech., Scott, R. F. (1980). “Analysis of centrifuge pile tests; Simulation of pile-
38(2), 248–257. driving.” Rep. through September 30, 1979 Research Program for
Kramer, S. L. (1996). Geotechnical earthquake engineering, Prentice Hall, American Petroleum Institute OSAPR Project 13, California Institute of
Upper Saddle River, NJ. Technology, Pasadena, CA.
Maheshwari, B. K., Truman, K. Z., Gould, P. L., and El Naggar, M. H. Seed, H. B., and Idriss, I. M. (1970). “Soil moduli and damping factors for
(2005). “Three-dimensional nonlinear seismic analysis of single piles
Downloaded from ascelibrary.org by New York University on 05/15/15. Copyright ASCE. For personal use only; all rights reserved.

dynamic response analysis.” Rep. EERC 70-10, Earthquake Engineering


using finite element model: Effects of plasticity of soil.” Int. J. Geomech., Research Center, Berkeley, CA.
10.1061/(ASCE)1532-3641(2005)5:1(35), 35–44. Terzaghi, K. (1955). “Evaluation of subgrade reaction.” Géotechnique, 5(4),
Mardfekri, M., Gardoni, P., and Roesset, J. M. (2013). “Modeling laterally 279–326.
loaded single piles accounting for nonlinear soil-pile interactions.” J. Trochanis, A. M., Bielak, J., and Christiano, P. (1991). “Three-dimensional
Eng., 2013, 243179. nonlinear study of piles.” J. Geotech. Engrg., 10.1061/(ASCE)0733-
Matlock, H. (1970). “Correlations for design of laterally loaded piles in 9410(1991)117:3(429), 429–447.
soft clay.” Proc., 2nd Offshore Technology Conf., Vol. 1, Offshore Winkler, E. (1867). Die Lehre von der Elasticitaet und Festigkeit, Dominicus,
Technology Conference, Houston, 577–594. Prague.
Matlock, H. M., and Reese, L. C. (1962). “Generalized solutions for laterally Yang, Z., and Jeremic, B. (2005). “Study of soil layering effects on lateral
loaded piles.” Trans. Am. Soc. Civ. Eng., 127(1), 1220–1247. loading behavior of piles.” J. Geotech. Geoenviron. Eng., 10.1061/
Muskhelishvili, N. I. (1963). Some basic problems of the mathematical (ASCE)1090-0241(2005)131:6(762), 762–770.
theory of elasticity: Fundamental equations, plane theory of elasticity, Yang, Z., Lu, J., and Elgamal, A. (2008). OpenSees soil models and solid-
torsion, and bending. Trans. J. R. M. Radok, P. Noordhoff, Groningen, fluid fully coupled elements: User’s manual, Version 1.0, Univ. of
Netherlands. California at San Diego, San Diego.
Ohsaki, Y., and Iwasaki, R. (1973). “On dynamic shear moduli and Pois- Zhang, J., Andrus, R. D., and Juang, C. H. (2005). “Normalized shear
son’s ratios of soil deposits.” Soils Found., 13(4), 61–73. modulus and material damping ratio relationships.” J. Geotech. Geo-
OpenSees 2.3.2 [Computer software]. Berkeley, CA, Pacific Earthquake environ. Eng., 10.1061/(ASCE)1090-0241(2005)131:4(453), 453–464.
Engineering Research Center. Zhang, L., Zhao, M., and Zou, X. (2013). “Behavior of laterally loaded piles
Poulos, H. G. (1971). “Behavior of laterally loaded piles: I—Single piles.” in multilayered soils.” Int. J. Geomech., 10.1061/(ASCE)GM.1943-
J. Soil Mech. and Found. Div., 97(5), 711–731. 5622.0000319, 06014017.

© ASCE 04014090-13 Int. J. Geomech.

Int. J. Geomech.

You might also like