See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/301569514

Optimization Technique to Determine the p-y Curves of Laterally Loaded Stiff

Piles in Dense Sand

Article  in  Geotechnical Testing Journal · September 2016

DOI: 10.1520/GTJ20140257

CITATIONS READS

5 1,594

5 authors, including:

Jianfeng Xue Kenneth Gavin

University of New South Wales, Canberra, Australia Delft University of Technology
62 PUBLICATIONS   617 CITATIONS    169 PUBLICATIONS   2,234 CITATIONS   

SEE PROFILE SEE PROFILE

Gerry Murphy Paul Doherty

Mott MacDonald Group, Singapore University College Dublin
11 PUBLICATIONS   57 CITATIONS    44 PUBLICATIONS   690 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Multi-modal probabilistic analysis of slope stability View project

Smart Rail View project

All content following this page was uploaded by Gerry Murphy on 28 September 2016.

The user has requested enhancement of the downloaded file.

An optimization technique to determine the p-y curves of laterally loaded stiff piles in
dense sand

Jianfeng Xue1, Gerry Murphy2, Paul Doherty3, David Igoe2 and Kenneth Gavin2

1
School of Applied Sciences and Engineering, Monash University, Australia

Jianfeng.xue@monash.edu

2
School of Civil Engineering, University College Dublin

3
Gavin and Doherty Geosolutions, Dublin, formerly University College Dublin

Xue, J., Gavin, K., Murphy, G., Doherty, P., and Igoe, D., "Optimization Technique to Determine the p-
y Curves of Laterally Loaded Stiff Piles in Dense Sand," Geotechnical Testing Journal,
http://dx.doi.org/10.1520/GTJ20140257. ISSN 0149-6115

Volume 39, Issue 5 (September 2016)

1
ABSTRACT

Lateral loading is often the governing design criteria for piles supporting offshore wind
turbines and with the recent growth of this sector the reliability of traditional design
approaches is receiving renewed interest. To accurately assess the behaviour of a laterally
loaded pile requires a detailed understanding of the soil reaction mobilised as lateral
deflection of the pile occurs. Currently, the p-y curve method is widely adopted to model the
response of laterally loaded piles. The limitations of existing p-y formulations are widely
known and there is acceptance that load tests on large diameter stiff monopiles are urgently
required in order to formulate appropriate design methods. However, interpretation of the data
from instrumentation placed on stiff monopiles is not straightforward. This paper proposes an
optimization technique to derive the soil reaction profile along the shaft of instrumented piles,
from which the correlated p-y curves can then be obtained. The method considers force
equilibrium, pile deflection, and additional boundary conditions. A set of fourth order
polynomial equations are assumed to model the soil reaction profile under each load-step
during a monotonic load test. By minimizing the difference between the measured and
calculated bending moment and considering equilibrium of the shear forces acting on the pile,
the soil reaction profile and the concentrated tip resistance can be obtained simultaneously. A
stiff instrumented test pile installed in over-consolidated sand was load tested and the results
were used to test the performance of the proposed method. The results are compared with
other methods used in literature and practice. The method provides a consistent framework to
derive p-y curves from measured strain data. The results of the field test and derived p-y
curves confirmed that existing design methods do not accurately capture the lateral loading
response of piles in overconsolidated sand.

Keywords: optimization method, p-y curve, over-consolidated sand, laterally loaded piles.

2
Nomenclature
E Elastic stiffness of the pile material
F Applied lateral load
FEM Finite Element Modelling
GA Genetic algorithm
h Height above ground level of applied load
I Moment of inertia of the pile’s cross section
M Bending Moment
Mhead Calculated moment at the pile head
Mtip Calculated moment at the pile tip
N Soil stiffness variation with depth parameter
N Normal force at the tip of the pile
p Soil Reaction Force per metre length
pht Distributed force along the pile due to soil resistance acting from head to tip
pth Distributed force along the pile due to soil resistance acting from tip to head
Pu Ultimate soil resistance
R Soil stiffness parameter
Rs Total skin resistance
T Concentrated lateral stress at pile tip
V Shear Force
W Pile Weight
y Lateral Displacement
z Depth below ground level
z0 constant to normalize the depth
β Degree of nonlinearity parameter

3
1. Introduction

Standard industry practice for the design of laterally loaded piles is to adopt a Winkler spring
method, which allows the pile-soil reaction to be modelled by a series of uncoupled springs
with a simplified stiffness response, see Doherty and Gavin (2012). This form of analysis is
commonly referred to as a “p-y” approach where the soil spring is defined by the soil reaction,
p, mobilised by a set displacement, y. Research on the form of p-y curves to be implemented
has been ongoing for forty years, (McClelland and Focht 1958, Matlock and Reese 1962,
Reese et al. 1974, Reese et al. 1975; Reese and Welch, 1975; Dyson and Randoph, 2001).
However, determining accurate site specific “p-y” curves remains problematic particularly for
stiff piles. This paper addresses this problem by proposing an optimisation procedure to
derive p-y curves from strain gauge data measured on instrumented piles.

There are several limitations of the current p-y approaches as outlined by Doherty and Gavin
(2012) and others and therefore a larger database of measured field tests with derived p-y
values is required. The majority of research to date, such as Reese et al. (1974); Murff (1987),
Matlock (1970) Dunnavant and O’Neill (1989), focus on flexible piles typically used in
offshore oil and gas applications. The recent rapid expansion of the offshore wind industry
requires the construction of large diameter (up to 7.5 m) monopiles in water depths of up to
35 metres. Many of these sites designated for development within the North Sea, and are
composed of very dense sands with cone penetration test end resistance (qc) values of >
20MPa and very stiff to hard glacial till deposits, with undrained strength (Su) values often
higher than 400kPa. In order to improve the feasibility, safety and efficiency of future pile
designs it is imperative that further understanding of the behaviour of stiff piles in these soil
conditions is obtained.

A number of models have been proposed to derive p-y curves using the bending moment
profile calculated from the measured strains in the piles, see Reese et al., (1974); Mezazigh
and Levacher (1998) and others. Curve fitting methods, e.g. 4th or 5th order polynomials
Wilson (1998), Nip and Ng (2005) and McVay et al (2009) or piecewise curve fitting,
Matlock and Ripperger (1956) and Dunnavant (1986) are widely adopted to derive the
complete bending moment (M) profile with depth (z) from discrete strain gauges locations:

M  f ( z) (Eq.1)

4
The equation which described the bending moment profile can be double integrated to obtain
the pile deflection (y):

 M  (Eq.2)
y   dz dz
 EI 
Where: E is the stiffness of the pile material, and I is the moment of inertia of the pile’s cross
section. During integration, certain boundary conditions should be considered, typically the
deflection and rotation of the pile head are used. The soil reaction along the pile can also be
derived by double differentiation of (Eq.1):

d M2
(Eq.3)
p 2
dz
A detailed comparison of the methods for deriving the p-y curves acting on laterally loaded
flexible piles was presented by Yang and Liang (2005). They concluded that the interpreted p-
y curves were highly sensitive to even small errors in the strain gauge readings and that due to
the complexity of the problem, there is not a universally applicable method currently available.
A non-iterative method of calculating the p-y curves was proposed by Sinnreich and Ayithi
(2014) where the polynomial functions are modified with the addition of an exponential
multiplier. This method was validated using full-scale test data and shown to provide very
good predictions for the response of long slender composite (concrete in a steel casing) piles.

In recent years, soil-pile interaction can be modelled on personal computers using

commercially available 3D finite element (FEM) software such as PLAXIS 3D (Dao 2011;
Kim and Jeong 2011), FLAC 3D (Moreno et al. 2011) and ABAQUS 3D (Trochanis et al.
1991). However, the accuracy of such models is highly sensitive to the correct calibration of
soil stiffness and strength parameters. Recent research has also proposed using in-situ
measurements such as CPT or dilatometer test results to predict the p-y curves (Novello 1999;
Robertson et al. 1985; Robertson et al. 1989). But the applicability of the methods is very
much site dependant (Li et al. 2014).

This paper proposes an optimization technique to derive the lateral soil reaction per unit
length p, and therefore the p-y curve from an instrumented load test performed on a stiff
monopile. The proposed method uses a set of general fourth order polynomial functions (the
curve fitting parameters are variables) to describe the soil reaction profile at increasing load
levels. The equilibrium of the pile is analysed using functions fitted to the recorded bending
moment. An optimised curve is then selected by comparing the inputted bending moment

5
measured (using strain gauges) and the calculated bending moment from the optimisation
method. The solution of the polynomial functions was performed using the Genetic Algorithm
toolbox imbedded in the Matlab environment. The p-y curves, and the concentrated tip
resistance, can be solved simultaneously using the proposed method. The resulting p-y curves
were then compared to traditional design methods and when implemented in a commercial
code, the new method was proved to be more accurate at predicting the overall lateral load-
displacement response and by considering equilibrium throughout the loading process to
provide consistent predictions for the individual components of resistance considered.

2. Proposed method

The soil reaction profile (Eq.3) is highly sensitive to any small change of the bending moment
function (Eq.1) due to the differentiation process. As expected, small changes to the bending
moment function will have marginal effect on the integrated deflection-depth profile. For
example, if a fourth order polynomial equation is adopted to describe the profile of bending
moment the equation will take the following form:

M  az  bz  cz  dz  e
4 3 2
(Eq.4)

where a - e are fitted parameters. The deflection of the pile can be obtained using (Eq.2):

y  (az / 30  bz / 20  cz /12  dz / 6  ez / 2) / EI  fz  g
6 5 4 3 2
(Eq.5)

Where: f and g are the rotation and displacement of the pile at known locations (typically at
the ground surface or head of the pile). Higher order of polynomial equations can be fitted to
the bending moment profile, however these were found to make have a negligible effect on
the integrated displacement, y, profile. For piles with low slenderness ratios such as those
adopted in the offshore wind industry, EI tends to be high and penetration depth (z) small
causing the first part of the equation to be almost negligible. The deflected shape of the pile is
then controlled by the boundary conditions as the effect of bending is relatively small
compared to the total displacement of the pile. Therefore, for a stiff pile, there is only a weak
relationship between the bending moment profile and the final displacement, as the bending
moment only determines the curvature of the pile and the final location or deflection profile is
mainly influenced by the boundary condition as shown in Figure 1.

6
To overcome this problem and noting that the main source of errors in deriving p-y curves
arise during the double differentiation of the bending moment profile to obtain the distribution
of p, the authors propose to fit the p profile using an assumed profile. A 4th order polynomial
equation was found to provide a good fit to the experimental data:

𝑝 = 𝑥1,𝐹1..𝑚 𝑧 4 + 𝑥2,𝐹1..𝑚 𝑧 3 + 𝑥3,𝐹1..𝑚 𝑧 2 + 𝑥4,𝐹1..𝑚 𝑧 + 𝑥5,𝐹1..𝑚 (Eq.6)

Where: x1 to x5 are coefficients to be determined for the load curve function and F1..m
represents the applied lateral load during a given load step.

At any stage during a load test (i.e. under the action of each applied load) the equilibrium
equations for the pile, see Figure 2, can be written as:

∑ 𝐹𝑧 = 𝑁 − 𝑊 − 𝑅𝑠 = 0 (Eq.7)

𝐿 (Eq.8)
∑ 𝐹𝑦 = 𝐹 + 𝑇 + ∫ 𝑝 𝑑𝑧
0

(Eq.9)
∑ 𝑀ℎ𝑒𝑎𝑑 = 𝑇(𝐿 + ℎ) + ∬ 𝑝𝑡ℎ 𝑑𝑧 = 0
𝐿

(Eq.10)
∑ 𝑀𝑡𝑖𝑝 = 𝐹(𝐿 + ℎ) + ∬ 𝑝ℎ𝑡 𝑑𝑧 = 0
𝐿

Where: Fz is vertical force, N is the normal force at the tip of the pile, W is the weight of the
pile (in kN) and the Rs is the total skin resistance (in kN), Fy is lateral force, F is the applied
lateral force, T is the concentrated lateral resistance at the tip of the pile, Mhead is the
calculated moment at the head of the pile (in kN·m), pth is the distributed force acting along
the pile due to soil reaction in kN/m from the tip to the head of the pile. This force depends on
the depth, z and the length, L of the pile: pth=f(L-z), while pht is the distributed force acting
along the pile due to soil reaction in kN/m from the head to the tip of the pile, Mtip is the
calculated moment at the tip of the pile (in kN·m).

By integrating (Eq.8) to (Eq.10), a relationship between the applied load F, the horizontal tip
resistance T, and the soil reaction p can be obtained. However, the relationship is
indeterminate and cannot be solved analytically, as there are more unknowns in the
relationship than the number of independent functions.

7
To obtain the bending moment profile of the pile, strain gauges are normally installed at
discrete locations along the pile shaft in the bending axis. Using the recorded strains along the
pile shaft, the bending moment at the gauge locations can be obtained using:

M i  Ei Iii (i=1, 2, …, n) (Eq.11)

Where: Mi, EiIi and i are the respective measured moment value, section modulus and
curvature of the shaft at location i, as shown in Figure 3(b).

Figure 3(b) also shows the moment values at the tip and the head of the pile, which are zero
for non-fixed (free-head) piles, and the moment in the pile at ground surface, which can be
calculated directly using the applied load. For standard curve fitting approaches the minimum
number of locations, nmin, must be equal to the order of the polynomial used to fit the data,
plus one (e.g. for a fourth order polynomial nmin, = 5). The optimum arrangement of
measuring points can be determined by simulating the field tests using commercially available
software to predetermine a bending moment curve.

By considering moment equilibrium of the pile section above and below the ith (i=0, 1, 2 …, n)
measurement location (as shown in Figure 4) at depth Li, we have:

𝐿𝑖 (Eq.12)
𝑀ℎ𝑡𝑖 = 𝐹(ℎ + 𝐿𝑖 ) − ∫ ∫ 𝑝 𝑑𝑧
0

𝐿𝑖 (Eq.13)
𝑀𝑡ℎ𝑖 = 𝑇(𝐿 − 𝐿𝑖 ) + ∫ ∫ 𝑝 𝑑𝑧
𝐿

Where: Mhti and Mthi are measured moments at i, calculated using the sections above and
below a given cross section respectively, ∫p is the shear force (v). For i=0, at the ground
surface the moment can be directly calculated from the applied lateral load.

An optimization technique which minimises the difference between the moment calculated
(from the assumed soil reaction profile generated by the Genetic Algorithm using the applied
load as a constant) and the moment measured by the strain gauges is used to satisfy Equations
7 to 10. The solution then becomes, to minimize:

8
 (Eq.12)
𝑓(𝑥)𝐹1..𝑚 = √(∑(𝑀ℎ𝑡𝑖 − 𝑀𝑖 )2 ) + (∑(𝑀𝑡ℎ𝑖 − 𝑀𝑖 )2 )

by varying the parameters (x1,x2,…, x5, T), subject to the following constraints:

Fy 0 , M head  0 and M tip 0 (Eq.13)

as defined in (Eq.8) to (Eq.10). This is a nonlinear constrained optimization problem and is

solved with the Genetic Algorithm (GA) method. The GA method is a global optimisation
technique based on the principles of biological evolution. The method has proven successful
in solving several geotechnical problems (Cui & Sheng, 2005; Xue & Gavin, 2007; Gavin &
Xue, 2009). The genetic algorithm translates the optimisation parameters into chromosomes
with a data string (binary or real). A range of possible solutions is obtained from the variable
space and the fitness of these solutions is compared with some predetermined criteria (in this
case the bending moment profile). If a solution is not obtained, a new population is created
from the original (parent) chromosomes. This is achieved using ‘crossover’ and ‘mutation’
operations. Crossover involves gene exchange from two random (parent) solutions to form a
child (new solution). Mutation involves the random switching of a single variable in a
chromosome, and is used to maintain population diversity, as the process converges towards a
solution.

The problem is coded in a Matlab environment and solved with the built in optimization GA
toolbox. A flow chart for the problem is show in Figure 5. The soil reaction profile thus
derived will be compatible with the measured moment profile and satisfy the static
equilibrium of the pile during all stages of loading. Therefore the method is an advance on
previous curve fitting procedures in which the force equilibrium is neglected and curve-fitting
parameters are fixed for each load step.

3. Application

The optimisation method was tested using data measured from a series of field tests
performed on an instrumented steel monopile installed in dense over-consolidated sand.

9
3.1 Site description

The tests were conducted at the University College Dublin Geotechnical Research Group
(GRG) test site in Blessington Co. Wicklow, Ireland. The site is located at the base of a sand
quarry near Blessington, a small village located 25 km south of Dublin. The ground
conditions consist of a uniformly graded, horizontally bedded, heavily over-consolidated, very
dense sand deposit to great depth. The ground conditions comprise of very dense, glacially
deposited fine sand with a relative density close to 100%. Particle size distribution analyses
performed on samples taken from depths ranging from 0.7 m to 2 m below ground level (bgl)
indicated the mean particle size, D50, varied between 0.1 and 0.2 mm. The fines content
(percentage of clay and silt particles) of the sand is between 5 and 15 %. Samples typically
had less than 10% coarse grained particles (> 0.6 mm). The water table was approximately 13
m below ground level. Profiles of Cone Penetration Test (CPT) end resistance, qc values are
shown in Figure 6. A summary of some of the main geotechnical parameters measured from
field and laboratory tests are presented in Table 1. Further details on the geotechnical
properties at the site can be obtained in Gavin et al. (2009), Doherty et al. (2012) and Murphy
et al. (2014).

3.2 Test Procedure

The lateral load tests were conducted using a 245 mm diameter steel tube with a wall
thickness of 8 mm driven to a penetration depth of 1.5m. The pile was driven using a 5 tonne
Juntan hydraulic hammer with verticality set prior to driving and maintained using the piling
rig and constant checking using levels.

A number of important scale effects must be considered when performing pile tests at reduced
scale. Klinkvort and Hededal (2013) noted from a series of centrifuge tests that normalising
the pile resistance parameters in a non-dimensional format allows for comparison of results
across scale. The approach was also adopted by Byrne et al. (2015) in designing a large field
test programme in which lateral load tests were performed on piles with diameters of between
0.275m and 2m diameter. The geometry of the test pile considered in this paper (ratio of
embedded length, L to pile diameter, D = 6.1) was chosen to be representative of stiff
monopiles used in the offshore wind industry, see Doherty and Gavin (2011). The pile wall
thickness (t = 8 mm) results in a D/t ratio of 31, which is significantly lower than the value
typically adopted (D/t in the range 60-120) for offshore monopiles, see Byrne et al. (2015).

10
The test pile was installed near three 7 m long, 300 mm square reinforced concrete piles
which were used to provide reaction during the lateral load test series. The test pile was
laterally loaded using a 25 tonne hydraulic jack placed between the test pile and a steel
reaction beam spanning between two concrete piles. The lateral load was applied at a height
(h) of 400mm above ground level and measured using a 200 kN load cell. The lateral
displacement of the test piles was measured at four locations using linear displacement
transducers and the pile head rotation was measured using digital inclinometers. The test
configuration is presented in Figure 7 and a summary of the recorded test data is provided in
Table 2.

3.3 Test results

The pile was instrumented with 30 mm TML single element strain gauges fixed to the outside
of the pile shaft using cyanoacrylate glue at eight levels along the tension and compression
faces (see Figure 7). The calculated bending moments using the measured strain at the
instrument locations are shown in Figure 8. The deflection of the pile, see Figure 9, was
obtained using measurements from an inclinometer installed on the pile which was calibrated
using the bending moment diagram (i.e. the effect of pile translation during the load test was
included). The pile rotation was then used as the boundary condition for the determination of
the pile displacement at a given depth. Even though the pile was relatively stiff, the effects of
bending were considered when calibrating the point of rotation. This was achieved by
correcting the measured rotation at ground level using the rotation values at each gauge
location along the pile shaft determined by integrating the calculated moment profile. The
figure shows that the depth to the point of rotation of the pile decreased from 1.17 m below
ground level to 1.02 m as the applied load increased. This depth corresponded to between 70%
and 80% of the pile’s embedded length. The fitted moment curvature was then compared to
the recorded strain measurements by back calculating the strain values from the fitted moment
curves. Linear regression analysis were performed to measure of how close the data are to the
fitted regression line. The R-Squared value or coefficient of determination, varied between
0.72 and 0.97 (see Table 2) and tended to increase with each load step suggesting a very good
fit to the test data. The lower R-squared values in the earlier stages may be due to
measurement error coupled with the very small strains mobilised at low applied load levels.

11
4. Performance of proposed method

4.1 Soil reaction profile

Using the optimisation method proposed, the soil reaction profiles along the pile under the
applied loads were obtained, see Figure 10. By considering the force equilibrium in the
horizontal direction and the shear force at the pile tip, the tip resistance was obtained. The tip
resistance of the test pile at each load stage is shown in Figure 11. A small residual load is
present at the start of the test which is consistent with experience from instrumented pile tests
at this site (Gavin and Lehane 2007). The p-y curves calculated for the pile are plotted in
Figure 12. The figure shows that the pile resistance and the initial stiffness response increased
with depth, with the resistance at the pile toe (z = 1.5 m) being significantly higher than those
above the rotation point (at comparable lateral displacements). For locations close to the point
of rotation (i.e. z = 1.02 m to 1.17 m bgl) the lateral displacements are low, and the data is
therefore is subject to small errors.

4.2 Model Validation

The derived p-y curves were verified by back-calculating the ground line displacement of the
pile using the p-y curves to define the soil stiffness parameters as per the API design method.
The API guidelines recommend a Winkler beam approximation which defines the soil
resistance as a series of uncoupled springs. Each spring is characterised by a non-linear curve
describing the relationship between the soil reaction and displacement. The derived
experimental p-y curves were input into Oasys ALP, an analysis program for laterally loaded
piles. the programs in built finite difference solver was used to back-calculate the lateral
deflection of the pile at ground level for a given set of input pile head loads, corresponding to
those used on site.. The optimised p values and the calculated deflection profiles were
inputted as the soil strength parameters at depths corresponding to the strain gauge locations.
The solver then calculated the lateral displacement of the pile using the inputted p-y curves to
form the soil stiffness matrix. The measured load-displacement behaviour is compared to the
back-calculated values in Figure 13. Given that the programme discretises the pile in
segments that were much smaller than the distance between strain gauge level, the optimised
soil resistance profile produces a reasonably accurate estimate of the actual pile behaviour for
applied lateral loads of up to 40 kN. At higher load levels, the pile exhibited creep during
period of maintained loading. As settlement accumulation due to creep was not modelled in

12
the finite difference solver, the predicted displacement would be expected to be lower than
that measured at higher load levels.

5. Comparison of p-y curves with existing models

5.1 Comparison with the parabola soil reaction curves

Wesselink et al. (1988) studied the lateral load response of piles in Bass Strait calcareous sand
using centrifuge tests and proposed an expression to describe the soil reaction:

p = R * D *(z / z0 ) N *( y / D)b (Eq.14)

Where: p is the soil reaction per unit length, D is the diameter of the pile, z is the depth below
the ground surface, z0 is a constant to normalize the depth, R is a soil stiffness related
parameter and is site specific, N is a parameter considering the soil stiffness variation along
the depth, and β is used to describe the degree of nonlinearity of the curves. The following
equation was suggested for piles in calcareous sand:

p  R * D *( X )0.7 *( y / D)0.65 (Eq.15)

Where: X is the normalised depth below ground surface (z/z0). An iteration process was
recommended in the original paper to determine the coefficients in (Eq.14). In this paper, the
GA optimization toolbox was used to solve the problem by minimizing the calculated
moments and the measured moments at the instrumented locations. The following equation
was found for the pile tested:

p  1400* D *( X / 0.12)0.39 *( y / D)0.59 (Eq.16)

It can be seen that the coefficients are quite different than the values suggested by Wesselink
etc. (1988), this is at least in part due to differences in the geotechnical properties of the sand
at both sites.

The soil reaction profiles obtained from (Eq.16) are shown in Figure 14. By comparing the
figure with the results in Figure 10, we can see that the two methods predicted similar soil
reaction profile between the depth of 0.2 m and 0.8 m. The bending moment diagram was

13
then back analysed by integrating (Eq.16) considering the correct boundary conditions. The
bending moment diagrams corresponding to the applied load levels of 45, 30 and 15 kN are
compared in Figure 15 with the measured values obtained in this paper.

It can be seen that the fourth order polynomial equation proposed in this paper provides a
more satisfactory soil reaction profile when comparing the bending moments in the pile. This
may partly be due to the fact that, in the method proposed by Wesselink (1988), a unique
equation is used to describe the soil reaction at all load levels. However, the optimistaion
method proposed in this paper uses a general 4th order polynomial equation to be developed
for each load-step. Also in (Eq.16), the soil resistance at the ground surface is assumed to be
zero, which may be true for the remoulded calcareous sand for which the model was
developed. While for the test site described in this paper, the in-situ soil is over-consolidated
and large near-surface suctions result in an apparent cohesion in the sand deposit. Therefore,
the assumption of soil resistance at ground surface is not appropriate in this case.

The derived p-y curves are compared with the results obtained in this paper in Figure 16. The
p-y curves obtained from the fourth order polynomial equations developed herein suggest that
the pile capacity is almost fully mobilised (with evidence of asymptotic values for the active
forces). The tip resistance was obtained by considering the equilibrium of the pile. The
normalized tip resistance against tip displacement is shown in Figure 17. It can be seen that
the tip resistance obtained from Wesselink's (1988) method varies with pile displacement in a
non-uniform manner. This unreasonable results arises because of use of (fixed) average curve
fitting parameters for all stages of the load test and ignoring equilibrium for each loading
stage.

5.2 Comparison with API method

The API method proposed by American Petroleum Institute (2005) based on the method
proposed by O’Neill and Murchinson (1983) is widely used in analysing the behaviour of
lateral loaded sand. The API method uses a hyperbolic tangent function to describe the p-y
curves:

 Ez 
p  A  Pu  tanh   y
 A  Pu  [19]

14
Where: z is the depth below ground surface, y is the lateral displacement of the pile at the
depth z, A is a factor to account for cyclic or static loading conditions. For static loading, A is
the larger value of (3-0.8z/D) and 0.9), E = initial modulus of subgrade reaction, and Pu =
ultimate soil resistance. Three soil stiffness values (40 MPa, 60 MPa and 80 MPa) were
adopted in the analysis to study the sensitivity of the API method to soil stiffness for this
problem.

The p-y curves of the soil at the depth of 0.4 meters are compared in Figure 18 using the three
stiffness values. The curves show that the p-y curves converge at displacement levels of
between 5 mm and 10 mm. The p-y curves obtained from API method using E=60 MPa are
compared with the results obtained with the proposed method in

140
API at 0.2m Predicted 0.2m

120 API at 0.6m Predicted 0.6m

API at 1.5m Predicted 1.5m
Soil Resistance, p (kN/m)

100

80

60
depth, z = 0.2 m
40
depth z = 1.5 m
20
depth, z = 0.6 m

0
-0.02 -0.01 0 0.01 0.02 0.03 0.04
Displacement, y (m)

Figure 19. The results show that the API method does not perform well in predicting the soil
behaviour for this case. The results from API analysis further illustrates that assuming a soil
reaction force at ground level equal to zero, in this case, is not adequate. The API curves are
typically over-conservative when calculating the ultimate soil resistance, and the pile-soil
reaction as presented in Figure 19, is significantly less stiff than the derived p-y curves would
suggest.

15
6. Conclusion

In current practice, p-y curves are widely used to describe the soil reaction due to lateral pile
deflection. However, there is not a universally accepted method for the derivation of the
curves from instrumented field data. A number of workers have proposed methods for the
interpretation of p-y curves which involve direct differentiation of the bending moment curve
(obtained using various curve fitting procedures). However, for stiff monopiles such methods
can result in serious errors if the appropriate boundary conditions are not considered.

An optimization technique is proposed in this paper which considers force equilibrium of the
pile, the deflected shape and various boundary conditions. In the analysis, the general soil
reaction profile along the pile was assumed to be fit by a fourth order polynomial equation.
By considering the equilibrium of the pile, and minimizing the difference between the
calculated and measured bending moment diagrams, the soil reaction profile and the
concentrated tip resistance due to movement and rotation of the pile-tip were obtained
simultaneously.

The performance of the proposed optimisation procedure was compared to a number of

existing techniques using the results from field tests on a model, stiff monopile. It was found
that the proposed method predicted a realistic soil reaction profile and consistent p-y curves.
The results show that the fourth order polynomial provided a satisfactory description of the
soil reaction profile. The authors note that additional case histories of large diameter
instrumented pile tests are needed to further trial and validate this approach.

Acknowledgements

The works described in this paper was funded by the Irish Research Council and Science
Foundation Ireland.

16
References

Byrne, B.W., Mcadam, R.A., Burd, H.J., Houlsby, G.T., Martin, C.M., Gavin, K.,
Doherty, P., Igoe, D., Potts, D.M., Jardine, R.J., Sideri, M., Schroeder, F.C., Wood, A.M.,
Gretlund, J.S., 2015. "Field testing of large diameter piles under lateral loading for offshore
wind applications", Proceedings of the XVI ECSMGE Geotechnical Engineering for
Infrastructure and Development. Edinburgh, Scotland, pp. 1255–1260. doi:10.1680.60678

Cui L. J., Sheng D. Ch., 2005, "Genetic Algorithms in Probabilistic Finite Element
Analysis of Geotechnical Problems", Computers and Geotechnics, 32, 555-563.

Dao T.P.T., 2011, "Validation of PLAXIS Embedded Piles for Lateral Loading", Master
of Science Thesis, Delft University of Technology

Doherty, P., Gavin, K., 2012. "Laterally loaded monopile design for offshore wind
farms", in: Proceedings of the ICE - Energy 165 February 2012 Issue EN1. pp. 7–17.
doi:10.1680/ener.11.00003

Doherty, P., Kirwan, L., Gavin, K., Igoe, D., Tyrrell, S., Ward, D., O’Kelly, B., 2012.
"An Overview of the UCD Geotechnical Research Site at Blessington", in: Proceedings of
BCRI 2012, 6-7th September 2012, Dublin.

Dunnavant, T. W., 1986, "Experimental and Analytical Investigation of the Behaviour

of Single Piles in Overconsolidated Clay Subjected to Cyclic Lateral Loads", Ph.D.
Dissertation, University of Houston, TX.

Dunnavant, T. W., and O’Neill, M. W. ,1989, "Experimental p-y model for submerged,
stiff clay", J. Geotech. Engrg., ASCE, 115(1), 95–114.

Dyson G. J., Randolph M. F. ,2001, "Monotonic Lateral Loading of Piles in Calcareous

Sand", Journal of Geotechnical and Geoenvironmental Engineering, Vol. 127, No. 4, 346-
352.

Gavin, K.G. and Lehane, B.M. ,2009, "Base Load-Displacement Response of Piles in
Sand", Canadian Geotechnical Journal, Vol. 44, No. 9, September, pp 1053-1063

Gavin, K.; Xue, J. F., 2009, "Use of a Genetic Algorithm to Perform Reliability
Analysis of Unsaturated Soil Slopes", Géotechnique, 59 (6): 545-549

17
 Gavin, K. Adekunte, A and O'Kelly, B. ,2009, "A field investigation of vertical footing
response on sand". K.Gavin, A.Adekunte & B.O’Kelly, Proceedings of ICE, Geotechnical
Engineering, Vol.162, Issue GE5, pp 257-267,

Igoe, D, Gavin, K.G. and O'Kelly, B. ,2011, "The shaft capacity of pipe piles in sand",
ASCE Journal of Geotechnical and Geoenvironmental Engineering,(2011) Vol 137, No.10,
pp 903-912

Klinkvort, R. T. and Hededal, O. ,2013 "Lateral response of monopile supporting an

offshore wind turbine", Proceedings of the ICE - Geotechnical Engineering, 166(GE2), 147-
158.

Li, W, Igoe, D and Gavin, K., 2014, "Evaluation of CPT-based P–y models for laterally
loaded piles in siliceous sand", Geotechnique Letters, Vol 4, Issue April, pp 110-117

Matlock, H. and Ripperger, E. A., 1956, "Procedures and instrumentation for tests on a
laterally loaded pile", Proceedings, 8th Texas Conference on Soil Mechanics and Foundation
Engineering, Austin, TX.

Matlock, H. and Reese, 1962, "Generalised solutions for laterally loaded piles",
Transactions of the American Society of Civil Engineers, Vol. 127, p 1220-1248.

Matlock, H. (1970). "Correlations for design of laterally loaded piles in soft clay." Proc.,
2nd Annu. Offshore Technol. Conf., Houston.

McClelland, B. and Focht, J.,1958, "Soil Modulus for Laterally Loaded Piles,"
Transactions of the American Society of Civil Engineers, Vol. 123, p 1049-1086.

McVay, M., Wasman, S., Consolazio, G., Bullock, P., Cowan, D., and Bollmann,
H. ,2009, "Dynamic Soil–Structure Interaction of Bridge Substructure Subject to Vessel
Impact." J. Bridge Eng., 14(1), 7–16

Mezazigh S. and Levacher D. ,1998, "Laterally loaded piles in sand: slope effect on P–
Y reaction curves", Can. Geotech. J. 35: 433-441.

Moreno A. B., Mikalauskas L., Troya Díaz J. L., 2011, "Evaluation of Load-
Displacement Relationship and p-y Curves for Monopiles Submitted to Different Pressures",
M.Eng Thesis, , Aalborg University, Denmark

18
 Murff, J. D., 1987, "Pile capacity in calcareous sands: State of the art." J. Geotech.
Engrg., ASCE, 113(5), 490–507.

Murphy, G., Doherty, P., Gavin, K., Ward, D., 2014. Characterisation of two dense sand
sites in Ireland using in-situ testing, Proceedings of 3rd International Symposium on Cone
Penetration Testing, Las Vegas, Nevada, USA - 2014. Las Vegas, Nevada, USA, pp. 637–644.

Nip, D. C. N., Ng. C. W., 2005, "Back analysis of laterally loaded piles", The Institution
of Civil Engineers: Geotechnical Engineering, 158 (GE2), 63–73.

Novello, E.A., 1999, "From static to cyclic p–y data in calcareous sediments", in:
Proceedings of the Second International Conference on Engineering for Calcareous
Sediments. Bahrain, pp. 17–24.

O’Neill, M. W., and Murchinson, J. M., 1983, "An evaluation of p-y relationships in
sand." Report to American Petroleum Institute, Washington, D.C.

O’Neill, M. W., Townsend, F. C., Hassan, K. H., Buller, A., and Chan, P. S., 1996,
"Load Transfer for Drilled Shafts in Intermediate Geomaterials" FHWA Publication No.
FHWA-ED-9-172, Department of Transportation, Federal Highway Administration, McLean,
VA.

Reese, L. C., Cox, W. R., and Koop, F. D., 1974, "Analysis of Laterally Loaded Piles in
Sand" Proceedings of 6th Offshore Technical Conference, Houston, TX, Vol. 2, 473-483.

Reese, L. C., Cox,W. R., and Koop, F. D., 1975, "Field Testing and Analysis of
Laterally Loaded Piles in Stiff Clay", 7th Offshore Technology Conference, 671-690.

Reese, L. C. and Welch, R. C., 1975, "Lateral Loadings of Deep Foundations in Stiff
Clay", J. Geotech. Eng. Div., Am. Soc. Civ. Eng., ASCE, Vol. 101, No. 7, 633-649.

Robertson, P. K., Campanella, R. G., Brown, P. T., Grof, I., and Hughes, J. M., 1985,
‘‘Design of Axially and Laterally Loaded Piles Using in Situ Tests: A Case History.’’
Canadian Geotechnical Journal, 22, 518–527.

Robertson, P. K., Davis, M. P., and Campanella, R. G., 1989, "Design of laterally
loaded driven piles using the flat dilatometer", Geotechnical Testing Journal, 12, 30–38.

19
 Sinnreich, J. and Ayithi, A., 2014, "Derivation of p-y Curves from Lateral Pile Load
Test Instrument Data." Geotechnical Testing Journal, 10.1520/GTJ20130127, 20130127.

Trochanis A. M., Bielak J., Christiano P., 1991, "Three Dimensional Nonlinear Study of
Piles", Journal of Geotechnical Engineering, 117: 429-447

Wesselink, B. D., Murff, J. D., Randolph, M. F., Nunez, I. L., and Hyden, A. M., 1988,
"Analysis of centrifuge model test data from laterally loaded piles in calcareous sand"
Engineering for calcareous sediments, Jewell & Andrews (Eds), Vol. 1, Balkema, Rotterdam,
The Netherlands, 261–270.

Xue J. F., Gavin K., 2007, "Simultaneous determination of critical slip surface and
reliability index for slopes", Journal of Geotechnical and Geoenvironmental Engineering, 133
(7), 878-886.

Yang K., Liang R., 2005, "Methods for Deriving p-y Curves From Instrumented Lateral
Load Tests", Geotechnical Testing Journal, Vol. 30, No. 1, 1-8.

Youngho K., Sangseom J., 2011, "Analysis of Soil Resistance on Laterally Loaded Piles
Based on 3D Soil–Pile Interaction", Computers and Geotechnics, 38: 248-257

20
 V
M
H

p
Epy1

Epy1
y
p
Epy2

Epy2
y
p
Epy3

Epy3
y
p
Epy4

Epy4
y

Figure 1a: Winkler p-y model for lateral loading

Figure 1: Comparison of the lateral displacement of a stiff and flexible pile

21
22
Figure 2: Forces acting on a lateral loaded pile (assuming no applied axial load)

23
 Figure 3: (a) location of the instruments, (b) idealised moment profile at instrumented
locations.

24
 Figure 4: (a) force diagram of the pile section above the point i, (b) force diagram of the
pile section below the point i.

25
 Pre-process the test results

Determination of bending
moment and pile deflection

Select a trial polynomial

function for soil reaction

Establishment of target
function and constraints

Solution of the problem

with GA toolbox

Calculate the bending

moment and pile deflection

Measured and No
calculated bending
moment comparable

Yes

Output

Figure 5: Flow chart for the proposed method.

26
 CPT cone resistance, qc (MPa)
0 5 10 15 20 25
0

0.5

1.5

2
Depth, z (m)

2.5

3.5

4.5

Figure 6: CPT test result (tip resistance qc against depth) of the site

27
Figure 7: Pile test Configuration and Strain gauge locations

28
 Bending Moment (kNm)
0 10 20 30 40
0

0.2
Depth below ground level (m)

0.4

0.6

0.8

1.2

Applied load =
1.4 10, 15, 20, 25, 30,
40, 45 kN
1.6

Figure 8: Fitted bending moment profiles using 4th order polynomial equation

29
 Displacement (mm)
-40 -20 0 20 40
0
R² = 0.9951 R² = 0.9994

0.2

Applied load = R² = 0.9998

Depth below ground level (m)

0.4 10, 15, 20, 25, 30,

40, 45 kN

0.6

0.8

1.2
Calculated
Linear (30 kN)
1.4

1.6

Figure 9: Deflection of the pile at each loading stage (below ground surface).

30
 Soil reaction Force, P (kN/m)
-150 -100 -50 0 50 100
0

0.2
Applied load
Depth below ground level (m)

=10, 15, 20, 25, 30,

0.4 40, 45 kN

0.6

0.8

Applied load
1.2
= 10, 15, 20, 25,
30, 40, 45 kN
1.4

1.6

Figure 10: Soil reaction profile along the pile under different load levels.

31
 Tip Displacement (mm)
0 5 10 15 20
Pile Tip Lateral Resistance (kN) 0

-5

-10

-15

-20

-25

-30

Figure 11: Normalized tip resistance against tip displacement.

32
 160

140
z = 1.5 m
120
Soil Resistance, p (kN/m)

100

80
z = 1.2 m
60

40
Depth, z = 0.2, 0.4, 0.6,
0.8, 1.0 m
20

0
-0.02 0 0.02 0.04
Displacement, y (m)

Figure 12: p-y curves of the pile obtained from the optimization method.

33
 50

40
Lateral Load (kN)

30

20

Predicted
10
Measured

0
0 10 20 30 40
Displacement, y (mm)

Figure 13: Comparison of measured and predicted load-displacement response

34
 Soil reaction force (kN/m)
-200 -100 0 100 200
Depth below ground surface (m) 0

0.2

Applied load
0.4 = 10, 15, 20, 25, 30,
40, 45 kN

0.6

0.8

1.2

1.4

1.6

Figure 14: Soil reaction along the pile using the equation proposed by Wesselink et al.
(1988)

35
 Bending Moment (kNm)
0 10 20 30 40
0

15kN 30kN 45kN

0.2

0.4
Depth below ground level (m)

0.6

0.8

1.2 Wesselin et al.

Fitted Moment
1.4
Inputted moments

1.6

Figure 15: Comparison of the raw calculated bending moment (squares) with the fitted
moment results obtained in this paper (solid curves) and the method by Wesselink etc. (1988)
(dashed curves).

36
 200
180
160 Depth, z = 1.5m
Soil Resistance, p (kN/m)

140
120
100
80
60
40 Depth, z = 0.2, 0.4, 0.6,
20 0.8, 1.0 m

0
-0.02 -0.01 0 0.01 0.02 0.03 0.04
Displacement, y (m)

Figure 16: Comparison of p-y curves obtained in this paper (solid curves) and the
method by Wesselink etc. (1988) (dashed curves).

37
 Displacement (mm)
0 5 10 15 20
10
Pile Tip Resistance (kN)

8
Wesselink et al.
6
4
2
0
-2
-4

Figure 17: Normalized tip resistance against tip displacement using the equation
proposed by Wesselink et al. (1988)

38
 90
80
Soil reaction force (kN/m)

70
60
50
40 k = 40,000 kN

30 k = 60,000 kN

20 k = 80,000 kN

10
0
0 5 10 15 20 25
Displacement (mm)

Figure 18: Sensitivity of API method to soil stiffness at the depth of 0.4 meters.

140
API at 0.2m Predicted 0.2m

120 API at 0.6m Predicted 0.6m

API at 1.5m Predicted 1.5m
Soil Resistance, p (kN/m)

100

80

60
depth, z = 0.2 m
40
depth z = 1.5 m
20
depth, z = 0.6 m

0
-0.02 -0.01 0 0.01 0.02 0.03 0.04
Displacement, y (m)

Figure 19: Comparison of p-y curves obtained in this paper (dashed curves) and the
method by API (solid curves).

39
Table 1
Summary of geotechnical investigation results
Soil Property Units GRG Blessington Test Site
Ave Cone tip resistance, qc MPa 9.88
Peak friction angle, φ'p Deg 42o @ 0.5, 44o @ 2m
Constant volume friction angle, φ'cv Deg 36.5o
Relative Density, Rd % ≈100%
Coefficient of lateral earth pressure, Ko - 2.5 @ 0.5m, 1.5 @ 2m
Bulk unit weight, γ'b kN/m3 20
Overconsolidation Ratio, OCR - 20@0.5m, 10@ 2m
Small Strain Shear Stiffness, Go MPa 150
Moisture Content % ≈12%
Percentage fines % 4-13%
Soil classification - Dense silty SAND

40
 Table 2
Summary of recorded test data

Applied Load kN 10 15 20 25 30 40 45
GL Displacement mm 2.26 4.07 6.09 8.87 12.10 24.07 34.88
GL Rotation Deg 0.130 0.228 0.35 0.524 0.717 1.414 2.019
Strain με
GL 287.1 433.2 579.7 724.5 869 1160 1300
-0.2 429.7 569.0 872.4 1055 1244 1703 1937
-0.4 538.4 793.0 1047 1261 1482 2035 2307
-0.6 587.4 830.3 1091 1311 1535 2089 2352
-0.8 563.4 772.4 1002 1197 1392 1854 2075
-1.0 465.3 627.7 796.9 938 1075 1379 1538
-1.2 304.7 411.2 503.7 577 643 772.8 862.8
-1.4 1.52 0.874 -0.044 -0.29 -0.39 0.339 2.35
Fit of fitted R-
moment to Squared 0.82 0.93 0.95 0.95 0.94 0.96 0.97
recorded strain value

41

View publication stats