Engineering and Computational Mechanics Proceedings of the Institution of Civil Engineers

http://dx.doi.org/10.1680/jencm.16.00006
Soil–monopile interactions for offshore Paper 1600006
Received 08/02/2016 Accepted 13/07/2016
wind turbines
Keywords: foundations/granular materials/offshore engineering
Cui and Bhattacharya

ICE Publishing: All rights reserved

Soil–monopile interactions for

offshore wind turbines
Liang Cui BE, PhD Subhamoy Bhattacharya PhD
Lecturer, Department of Civil and Environmental Engineering, University of Professor, Department of Civil and Environmental Engineering, University
Surrey, Guildford, UK (corresponding author: l.cui@surrey.ac.uk) of Surrey, Guildford, UK

Many offshore wind turbines are supported by large-diameter piles (known as monopiles) and are subjected to a
large number of cyclic and dynamic loads. There are evidences suggesting that foundation stiffness are changing
with cycles of loading and this may lead to changes in the natural frequency of the system with the potential for
unplanned system resonances. There are other consequences such as excessive tilt leading to expensive repair or
even complete shutdown. Therefore, it is vital to understand the long-term response of wind turbine foundation so
that a method to predict the change in frequency and long-term tilt could be established. This paper aims to present
the experimental work of small-scale physical modelling and discrete element modelling of the interactions between
a monopile and the surrounding soil. Changes in soil stiffness under cyclic loading of various strain amplitudes were
examined for both physical modelling and discrete element modelling. Micromechanics of soils underlying the soil
stiffness change was investigated using discrete element method. Variation of force distribution along the monopile
under cyclic loading was analysed to show the influence of monopile stability.

Notation
A area of hysteresis loop
AΔ area of the triangle representing elastic energy
D pile diameter
G shear modulus of soil
P lateral load
α hysteresis damping ratio
1. Introduction
1.1 Background
Offshore wind turbines (OWTs) are relatively new structures
and are providing increasing proportion of energy. This is due
to the fact that offshore sites are characterised by stronger and
more stable wind conditions than the corresponding land sites
and thus have a higher capacity factor when compared with
equivalent onshore turbines.
The design and construction of foundations for offshore tur-
bines are challenging due to the harsh environmental con-
ditions. Different types of foundations have been proposed:
monopile, gravity base, jacket, suction caisson and floating
anchoring systems. However, most of the offshore turbines cur-
rently in operation (UK Round 1 development) are supported
on driven monopiles with diameters ranging between 3·5 and
6 m. The choice of monopiles results from their simplicity of
installation and the proven success of driven piles in support-
ing offshore oil and gas infrastructures. The available methods
for designing monopiles for OWTs (e.g. the approach suggested
by DNV, 2007) are based on the methods originally developed
for the offshore oil and gas industry (API, 1993). Lombardi
et al. (2013) explain the obvious differences between offshore
platform piles and monopoles, which are summarised below:
Piles for offshore structures are typically 60–110 m long and
1·8–2·7 m diameter; monopiles for OWTs are commonly
30–40 m long and 3·5–6 m diameter. Degradation in the upper
soil layers resulting from cyclic loading is less severe for off-
shore piles which are significantly restrained from pile head
rotation, whereas monopiles are free headed and more vulner-
able to tilt. In any case, the non-linear behaviour of the struc-
ture could be taken into account using an approach similar to
that suggested by Comodromos et al. (2009) and Conte et al.
(2013, 2015). A design method using a beam on non-linear
Winkler springs (‘p − y’ method in API (1993) code or DNV
(2002, 2007) code) may be used to obtain pile head deflection
under cyclic loading, but its use is limited for wind turbines
due to the following.
& The widely used API (1993) model is calibrated
against response to a small number of cycles
(maximum 200 cycles) for offshore fixed platform
applications – for example Matlock (1970), Reese et al.
(1974), Reese et al. (1975). In contrast, for a real OWT,
107–108 cycles of loading are expected over a lifetime of
20–25 years.

Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
 Engineering and Computational Mechanics
& Under cyclic loading, the API (1993) or DNV (2002, 2007)
model always predicts degradation of foundation stiffness
in sandy soil. However, recent work by Bhattacharya and
Adhikari (2011), Cuéllar et al. (2012), LeBlanc (2009)
suggested that the foundation stiffness for a monopile in
sandy soil will actually increase as a result of densification
of the soil next to the pile.
& The ratio of horizontal load to vertical load is very high
in OWTs when compared with fixed jacket structures.
Therefore, monopiles can be considered as moment
resisting foundations.
1.2 Complexity of loads on the monopile
The loads acting on the wind turbine tower are ultimately
transferred to the foundation and can be classified into two
types: static or dead load due to the self-weight of the com-
ponents and the cyclic/dynamic loads arising from the wind,
wave, 1P and 3P loads (for further details see Arany et al.,
2014). However, the challenging part is the dynamic loads
acting on the wind turbine and the salient points are discussed
below.
& The rotating blades apply a cyclic/dynamic lateral load at
the hub level (top of the tower) and are a function of the
turbulence of the wind. The magnitude of the dynamic
component depends on the turbulent wind speed
component.
& The waves crashing against the substructure apply a lateral
load very close to the foundation. The magnitude of this
load depends on the wave height and wave period, as well
as the water depth.
Peak frequency/period of turbulence spectrum
Mudline bending moment
Mmax
Mmean
Mmin
0 20 40 60
Wind + wave load
Figure 1. Mudline moment (wind and wave load) acting on the
wind turbine tower
Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya
& The mass and aerodynamic imbalances of the rotor
(blades) generate vibration at the hub level and apply
lateral load. This load has a frequency equal to the
rotational frequency of the rotor (referred to as 1P loading
in the literature). Since most of the industrial wind turbines
are variable speed machines, 1P is not a single frequency
but a frequency band between the frequencies associated
with the lowest and the highest revolutions per minute
(rpm).
& The blade shadowing effects (referred to as 2P/3P in the
literature) also applied loads in the tower. This is a
dynamic load having a frequency equal to three times the
rotational frequency of the turbine (3P) for three-bladed
wind turbines and two times (2P) the rotational frequency
of the turbine for two-bladed turbines. The 2P/3P loading
is also a frequency band like 1P and is simply obtained by
multiplying the limits of the 1P band by the number of
turbine blades.
A calculation procedure is developed by Arany et al. (2014)
which can be easily carried out in a spreadsheet program. The
output of such a calculation will be relative wind and the wave
loads and an example is shown in Figure 1, where it is
assumed that the wind and the wave are perfectly aligned
which is a fair assumption for deeper water further offshore
projects (i.e. fetch distance is high).
From the Kaimal spectrum as suggested in the DNV (2002,
2007) code, the peak frequency of wind turbulence can be
obtained theoretically. In the absence of site-specific data,
and for the purpose of foundation design, the time period
for wind can be conservatively assumed to be acting at the
Simplified cyclic load scenario Peak
frequency/period
of wave spectrum
~10 s
~100 s
80 100 120 140 160 180 200
Time: s
Wind load
 Engineering and Computational Mechanics
hub level having a time period at about 100 s as shown in
Figure 1.
Therefore, one of the objectives of the paper is to study theor-
etically the soil–structure interaction on a monopile foundation
for an asymmetrical loading as shown in Figure 1.
1.3 Dynamic issues in OWT design
OWTs are characterised by a unique set of dynamic loading
conditions and are summarised as follows.
& Environmental dynamic loads arising from the wind and
waves. Figure 2 shows the plot of power spectral density
of wind and wave loading around the UK coastline
(particularly in the North Sea). The predominant wave
frequency is 0·1 Hz, which corresponds to 10 s wave
period.
& Rotor loading at a frequency which is commonly referred
to as 1P. Figure 2 shows the rotor frequency for a 3·6 MW
wind turbine having an operational range between 5 and
13 rpm – that is, 0·08–0·22 Hz. In the power spectral
density plot, the 1P frequency appears as a band.
& The blade passing frequency (3P or 2P for a three-bladed
or two-bladed turbine, respectively) is a forced loading
generated from the effect of wind deficiency that occurs
1P ±10% safety region
Kaimal
spectrum
(wind)
Rotor
frequency 1P
Power spectral density
JONSWAP
spectrum
(wave)
0 0·1 0·2
Soft–soft Soft–stiff
Figure 2. Forcing simplified power spectral density of the forcing
frequencies applied to typical three-bladed 3·6 MW OWT with
an operational interval in the range of 0·08–0·22 Hz (5–13 rpm)
Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya
as each blade passes through the shadow of the tower.
Figure 2 shows the blade passing frequency for the
3·6 MW wind turbine generator.
From Figure 2 it may be observed that in order to avoid the
resonance of the system, the designed frequency of the overall
system must be kept away from the frequency content of
applied loads. Specifically, DNV (2002) suggests that the
natural frequency of the wind turbine should be at least ±10%
away from the 1P and 2P/3P frequencies. Bearing these con-
siderations in mind, there are three possible slots where the
natural frequency of the system may lie. They correspond to
three different design approaches namely: soft–soft (natural
frequency <1P), soft–stiff (natural frequency between 1P and
2P or 3P) and stiff–stiff (natural frequency >2P or 3P).
The most common design, used, for example, in the Round 1
UK development, is ‘soft–stiff’, which implies that the natural
frequency lies between 1P and 3P. Few points may be noted as
follows.
& The natural frequency of three-bladed wind turbines in
soft–stiff design is very close to the forcing frequency due
to 1P and 3P.
& Any change in natural frequency either an increase or a
decrease will have adverse consequences on fatigue life of
the wind turbine system.
Typical frequency diagram
3P ±10% safety region
Blade passing frequency – 3P
0·3 0·4 0·5 0·6 0·7 0·8
Frequency: Hz
Stiff–stiff
 Engineering and Computational Mechanics
& It is well known that under the action of cyclic/dynamic
loads most soils change its properties. Depending on the
frequency and amplitude of loading, the soil may stiffen or
soften. Stiffness of a monopile to large extent is dependent
on the stiffness of the soil and therefore the impact of a
large number of cycles on the monopiles needs to be
evaluated.
The next section of the paper collates the research carried out
on dynamic soil–structure interaction along with field evidences.
2. Dynamic soil–structure interaction
in OWTs
2.1 Summary based on scaled model tests
Extensive research has been carried out by Bhattacharya et al.
(2013a, 2013b), Lombardi et al. (2013) and Yu et al. (2015) to
study the effects of cyclic loads on the first natural frequency
of wind turbines. Tests were carried out on different types of
foundations: monopiles, jackets, multiple pods. A typical test
consists of the application of cyclic loading for a particular
time interval (or for a certain number of cycles) and then
measuring the frequency and damping of the system by a free
vibration test. The cyclic loading was applied through an
actuator. However, during the free vibration test (also known
as a ‘snap back’ test in the literature), the actuator was discon-
nected from the tower and the tower was given a small ampli-
tude vibration and the acceleration of the system was recorded.
The cyclic lateral loading was applied at different frequencies
and for different lateral load magnitudes. This set of tests
10
0
% change in natural frequency
0 10 000
–5
–10
–15
–20
–25
–30
–35
–40
Figure 3. The observed change in natural frequency with the
number of cycles for different strain levels in the soil around
the pile
Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya
created a database of change of frequency and damping of the
wind turbine system for different values of
& strain field in the soil next to the pile. This is expressed in
Figure 3 by a non-dimensional group (P/GD 2), where P is
the lateral load, G is the shear modulus of the soil and D is
the pile diameter. Further details on the scaling relations
can be found in Bhattacharya et al. (2011) and Lombardi
et al. (2013)
& forcing frequency imposed by the different dynamic loads
& number of cycles of loading.
Figure 3 shows one such graph obtained from scaled model
test studies by Bhattacharya et al. (2013a) and Yu et al. (2015)
on monopiles where the observed change in natural frequency
is plotted with the number of cycles for various levels of strains
in the soil next to the pile. The main observations from the
tests are as follows.
& For strain-hardening sites (e.g. loose to medium dense
sand) where the stiffness of the soil increases with cycles of
loading, the natural frequency of the overall system will
increase possibly due to densification.
& For strain-softening sites (clay sites) where the stiffness of
the soil may decrease with cycles of loading, the natural
frequency of the overall system will also decrease
correspondingly. Of course, this depends on the strain level
in the soil next to the pile and the number of cycles.
Lombardi et al. (2013) established a link between the strain of
the soil next to the pile as observed in the mechanism (P/GD 2)
20 000 30 000 40 000 50 000 60 000 70 000 80 000
Saturated sand (strain 5·0 × 10–4)
Dry sand (strain 5·0 × 10–4)
Soft clay (strain 2·0 × 10–4)
Soft clay (strain 3·4 × 10–3)
Dry sand (strain 1·945 × 10–4)
Dry sand (strain 1·876 × 10–4)
Dry sand (strain 1·798 × 10–4)
Dry sand (strain 0·243 × 10–4)
Number of cycles
 Engineering and Computational Mechanics
with ‘threshold strains’ from element testing. This allows
designers to choose pile diameter to avoid stiffness degradation.
2.2 Field evidences of dynamic soil–structure
interaction
There are some field evidences of dynamic soil–structure inter-
action and they are briefly discussed below.
& Natural frequency was measured for a wind turbine
structure in the Hornsea wind farm where it was observed
that the natural frequency decreased from 1·23 to 1·13 Hz
after 3 months of operation (Lowe, 2012).
& It has been observed that there is an increasing scatter in
the observed natural frequency at higher wind speeds.
Higher wind speed will lead to higher strain in the soil next
to the pile and therefore higher P/GD 2 leading to higher
soil–structure interaction.
& Kuhn (2000, 2002) reported that the target design
frequency for Lely wind farm of 0·4 increased to 0·63 Hz
after 6 years of service.
While the field studies provides anecdotal evidence of the
change in system frequency and small-scale tests provides the
overall understanding, discrete element modelling of the prob-
lem is carried out to better understand the monopile–soil inter-
action. The next section of the paper describes the modelling
technique and the results obtained from the simulation.
3. Numerical analyses using discrete
element method modelling
3.1 Description of discrete element method model
Numerical simulations were performed to investigate the
underlying mechanism for soil stiffness changes surrounding
the wind turbine monopile. The discrete element method
(DEM) was found to be more appropriate than other numeri-
cal methods (e.g. finite-element method) as it allows direct
monitoring of change in soil stiffness, and more importantly it
offers a method to analyse the micromechanics, which
underlies the stiffness changes. Originally proposed by Cundall
and Strack (1979), DEM simulates granular materials as
assemblies of individual particles which respond to given load
conditions. The interactions between particles are simulated by
contact laws, where the normal and tangential contact forces
are dependent on the overlap and relative displacement
between two contact particles. In this study, the elastic Hertz–
Mindlin contact model (Mindlin and Deresiewicz, 1953) is
adopted. The contact forces, accelerations, velocities and dis-
placements of all particles are updated in each small time step
using the central difference time integration method. Stresses
and strains are then calculated from the contact forces within a
representative volume element or along a boundary. An open-
source DEM code modified and validated in previous studies
Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya
(Cui, 2006; Cui et al., 2007; O’Sullivan et al., 2008) was used
to perform the presented study.
To simulate the soil stiffness changes, a DEM model of a soil
tank (100 mm
100 mm
50 mm) was first created. The soil
tank was filled by about 13 000 spherical particles with radii in
the range of 1·1–2·2 mm. The particles were deposited under
gravity. The pile is 20 mm in diameter and was embedded to a
depth of 40 mm by removing particles located in the space
which was to be occupied by the pile. Particles were allowed to
settle down again following the installation of the pile. The
current work aims to obtain qualitative characteristics of soil
behaviours for investigations of micromechanics, not to repro-
duce quantitatively the model tests. Therefore, this study used
large particle sizes, which may cause size effect. Once the soil
particles were settled down in the soil tank, cyclically horizon-
tal movements were assigned to the pile to simulate the cyclic
movements of OWT monopile due to the cyclic loadings.
Translational movements rather than rotational movements
were assigned to the pile at this stage. Rotational movement
will be simulated in the future study to differentiate the effects.
Three different strain amplitudes, 0·1, 0·01 and 0·001%, were
chosen to examine the effects of strain levels. For each strain
amplitude, two types of cyclic loading were applied: symmetric
cyclic loading with stains in the range of (−0·1, 0·1%), (−0·01,
0·01%) and (−0·001, 0·001%) and asymmetric cyclic loading
with stains in the range of (0, 0·2%), (0, 0·02%) and
(0, 0·002%). As constrained by the computational costs,
500 cycles were simulated for strain amplitude of 0·1% and
1000 cycles were simulated for other strain amplitudes. The
current simulation for 0·1% strain amplitude required about
1 month in computational time. The simulation parameters are
listed in Table 1.
3.2 Stress–strain response and damping ratio
The resultant horizontal stress applied on the pile against the
horizontal strain of soil is illustrated in Figure 4. As shown
in Figure 4, the stress–strain curves form hysteresis loops,
Parameters Value
Soil particle density ρs: kg/m3 2650
Particle sizes: mm 1·1, 1·376, 1·651,
1·926, 2·2
Inter-particle frictional coefficient, μ 0·3
Particle-boundary frictional coefficient, μ 0·1
Gs (Hertz–Mindlin contact model): Pa 2·868
107
Poisson’s ratio 0·22
Initial void ratio, e 0·539
Table 1. Input parameters for DEM simulation
 Engineering and Computational Mechanics Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya

600 800
First cycle First cycle
Tenth cycle 600 Tenth cycle
400 100th cycle 100th cycle
500th cycle 500th cycle
400
Horizontal stress: Pa

Horizontal stress: Pa
200
200

0 0

−200
−200
−400
−400
−600

−600 −800
−0·10 −0·05 0 0·05 0·10 0 0·05 0·10 0·15 0·20
Horizontal strain: % Horizontal strain: %
(a) (b)

150 150
First cycle First cycle
Tenth cycle Tenth cycle
100 100th cycle 100 100th cycle
1000th cycle 1000th cycle
Horizontal stress: Pa

Horizontal stress: Pa

50 50

0 0

−50 −50

−100 −100

−150 −150
−0·010 −0·005 0 0·005 0·010 0 0·005 0·010 0·015 0·020
Horizontal strain: % Horizontal strain: %
(c) (d)

15 25
Firstcycle First cycle
Tenth cycle 20 Tenth cycle
10 100th cycle 100th cycle
1000th cycle 1000th cycle
15
5
Horizontal stress: Pa

Horizontal stress: Pa

10
0
5
−5
0
−10
−5

−15 −10

−20 −15
−1·0 −0·5 0 0·5 1·0 0 0·5 1·0 1·5 2·0
Horizontal strain: % x 10−3 Horizontal strain: % x 10−3
(e) (f)

Figure 4. Hysteresis loops formed by the stress–strain curves

during cyclic loadings: (a) strain (−0·1, 0·1%); (b) strain (0, 0·2%);
(c) strain (−0·01, 0·01%); (d) strain (0, 0·02%); (e) strain (−0·001,
0·001%); (f) strain (0, 0·002%)

Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
 Engineering and Computational Mechanics Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya

indicating the energy dissipations during cyclic loading. It is damping ratio for the strain amplitude of 0·01% oscillated
also observed that the areas of the hysteresis loops increase about a constant value. However, it is obvious that the
with strain amplitude, indicating greater energy dissipation. damping ratio for the strain amplitude of 0·1% decreased dra-
The hysteresis damping ratio, α, can be determined by the matically in the first 30 cycles and then oscillated at an
expression (Karg, 2007) approximately constant value. The damping ratios for asym-
metric cyclic loading are much lower than those for the corre-
A sponding symmetric cyclic loading, due to the higher elastic
1: α¼
4πAΔ energy stored in each cycle.

where A is the area of the hysteresis loop, representing the
energy dissipated and AΔ is the area of the triangle as indicated
in Figures 4(a) and 4(b), representing the elastic energy stored
in the soil during one load cycle. The variation of the damping
ratio during cyclic loading is illustrated in Figure 5. The
0·40
0·35
Hysteresis damping ratio, α
It is also interesting to observe that although the stresses in
the first half cycle for the asymmetric cyclic loading is positive,
it reduced to negative when the strain goes to zero. Following
a few cycles, the minimum negative stress approaches the
same magnitude as the maximum positive stress. Moreover,
the magnitude of stresses and the shape of the hysteresis loops
for both symmetric cyclic loading and asymmetric cyclic
loading are almost identical after many cycles. The system
Strain (–0·1, 0·1%)
Strain (0, 0·2%) under asymmetric cyclic loading behaves the same as the sym-

Strain (–0·01, 0·01%)

metric cyclic loading with the same strain amplitude after
0·30
many cycles, indicating that the strain amplitude, rather than
Strain (0, 0·02%)
0·25 the maximum strain, dominates the long-term cyclic
0·20 behaviour.

0·15
3.3 Evolution of stiffness
0·10
The secant Young’s modulus of soil in each cycle was calcu-
0·05 lated by determining the slope of a line connecting the
0 maximum and minimum points of each full loop. It is evident
0 200 400 600 800 1000 from Figure 4 that the secant Young’s modulus of soil
Number of cycles increased during cyclic loading. A clearer evolution of secant
Young’s modulus is shown in Figure 6. At a strain amplitude
Figure 5. Hysteresis damping ratio at the end of each cycle of 0·1%, Young’s modulus increases dramatically from 250 to
around 600 kPa for symmetric cyclic loading and from 130 to

1600 1600
Secant Young’s modulus: kPa

Secant Young’s modulus: kPa

1400 1400
1200 1200
1000 1000
800 800
600 600
400 Strain (–0·1, 0·1%) 400 Strain (0, 0·2%)
Strain (–0·01, 0·01%) Strain (0, 0·02%)
200 200
Strain (–0·001, 0·001%) Strain (0, 0·002%)
0 0
0 250 500 750 1000 0 250 500 750 1000
Number of cycles Number of cycles
(a) (b)

Figure 6. Secant Young’s modulus of soil at the end of each

cycle: (a) symmetric cyclic loading and (b) asymmetric cyclic
loading

Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
 Engineering and Computational Mechanics Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya

around 600 kPa for asymmetric cyclic loading. At a strain
amplitude of 0·01%, the Young’s modulus increases quickly in
the first few cycles and then only increases slightly to about
1100 kPa. At a strain amplitude of 0·001%, Young’s modulus
only increases initially then mobilises at a constant value at
1500 kPa. The initial stiffness of asymmetric cyclic loading is
lower than that of symmetric cyclic loading with the same
strain amplitude due to a higher maximum strain applied.
However, following a few cycles, the stiffness for both types of
cyclic loading approaches the same values, confirming the
observations from the stress–strain responses.
increase in soil stiffness in the first 50 cycles. The soil particle
displacements for a strain amplitude of 0·01% are less remark-
able, corresponding to less reduction in soil stiffness.
It is also evident that only soil particles in an inverted triangu-
lar region enclosing the monopile have noticeable convective
displacements. This is because that the vertical effective stress
in a soil body increases linearly against depth, which imposes
50
40

50 mm in height
The Young’s modulus against horizontal strain in the first half
cycle and in the 500th cycle for a strain amplitude of 0·1% are 30
shown in Figure 7. The stiffness–strain curve in the first cycle
20
displays the similar ‘S’ shape as expected for the shear
modulus–shear strain curve. Following cyclic loadings, stiffness 10
at different strain levels all increases significantly. At a strain
amplitude of 0·01 and 0·001%, stiffness also increases but at a 0
0 20 40 60 80 100
much smaller scale as indicated in Figure 6.
Soil tank − 100 mm in width
(a)
3.4 Convective granular flow
In the model test, ground settlements around the monopile 50
could be observed. To illustrate the ground settlement, plots
40
50 mm in height

of incremental soil particle displacements in the symmetric

cyclic loading with a strain amplitude of 0·1% in the first 30
50 cycles and in the next 50 cycles are given in Figure 8. The
soil particle displacements in the symmetric cyclic loading with 20
a strain amplitude of 0·01% were also illustrated for compari-
10
son. Each arrow in the plot starts from the original centre of a
particle and ends at the new centre at the end of a given cycle. 0
It is evident that the soil particles surrounding the pile moved 0 20 40 60 80 100
downwards, causing ground settlement. Soil densification Soil tank − 100 mm in width
around the pile is the main reason causing the increase of soil (b)
stiffness. It is also clear that the soil particle displacements are 50
only significant in the first 50 cycles, underlying the significant
40
50 mm in height

20 000 30
Secant Young’s modulus: kPa

First cycle
500th cycle strain (–0·1, 0·1%)
20
15 000 500th cycle strain (0, 0·2%)
10
10 000
0
0 20 40 60 80 100
5000 Soil tank − 100 mm in width
(c)
0
1 × 10–8 0·000001 0·0001 0·01 1
Figure 8. Incremental soil displacements at the end of the given
Horizontal strain
cycle (unit: mm): (a) strain (−0·1, 0·1%), 1st–50th cycles; (b) strain
(−0·1, 0·1%), 51th–100th cycles; (c) strain (−0·01, 0·01%),
Figure 7. Secant Young’s modulus against horizontal strain 1st–1000th cycles

Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
 Engineering and Computational Mechanics Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya

an increasing constraint on particle movement against depth. 650

First cycle
Cuéllar et al. (2012) performed physical tests of cyclic rotation 600 Tenth cycle
of a pile and observed the granular convective flow and soil 100th cycle
densification around the pile. The shape of the convective soil 550 500th cycle

volume observed by them is similar to the region where the 500

soil particle displacements were concentrated in the current

Stress: Pa
450
DEM simulations. In the current DEM simulations, samples
with same initial void ratio of 0·539 (medium dense sample) 400
all showed densification behaviour and stiffness increase under 350
cyclic loading. It would be interesting to investigate soil beha-
300
viours and stiffness evolutions for a wide range of initial void
ratios in the future study. 250

200
−0·10 −0·05 0 0·05 0·10
3.5 Contact stresses and forces Strain: %
(a)
3.5.1 Average radial stresses
The evolution of average radial stress on the pile at the end of 650
First cycle
each cycle is depicted in Figure 9. Due to soil densification, Tenth cycle
600
the radial stress increased significantly for a strain amplitude 100th cycle
500th cycle
of 0·1% under cyclic loading, reached a peak value and then 550
mobilised at about a constant value. Increased radial stress
Radial stress: Pa

500
would increase side friction and would improve shaft resist-
ance. The increase in radial stress is less remarkable for smaller 450
strain amplitudes due to smaller particle displacements. The
400
evolutions of average radial stress in representative cycles are
illustrated in Figure 10. The shapes of the radial stress–strain 350
curves are quite different for different strain amplitudes.
300
For symmetric cyclic loading with 0·01% strain amplitude
(Figure 10(c)), the radial stress increases at positive strain 250
0 0·05 0·10 0·15 0·20
values, but decreases slightly at negative strain values.
Strain: %
However, for 0·1% strain amplitude (Figure 10(a)), the radial
(b)
stress increases at both positive and negative strains, forming a
‘butterfly’ shaped curve. It is also interesting to observe that, 295
for asymmetric cyclic loading (Figure 10(b)), the stress–strain First cycle
290 Tenth cycle
100th cycle
285 1000th cycle
280
Radial stress: Pa

700
Average radial stress on pile: kPa

275
600
270
500 265
400 260
300 255

200 Strain (–0·1, 0·1%) Strain (0, 0·2%) 250

Strain (–0·01, 0·01%) Strain (0, 0·02%) 245
100 −0·010 −0·005 0 0·005 0·010
Strain (–0·001, 0·001%) Strain (0, 0·002%)
0 Horizontal strain: %
0 250 500 750 1000 (c)
Number of cycles

Figure 10. Evolution of average radial stresses in representative

Figure 9. Evolution of average radial stress on the pile at the end cycles: (a) strain (−0·1, 0·1%); (b) strain (0, 0·2%); (c) strain
of each cycle (−0·01, 0·01%)

Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
 Engineering and Computational Mechanics
response in the first few cycles is different from that in the cor-
related symmetric cyclic loading, However, it eventually
evolves to the same ‘butterfly’ shape after many cycles. It con-
firms again that the influence of different maximum strains of
cyclic loading can be eliminated after many cycles and the
dominant factor for cyclic soil response is the strain amplitude.
A comparison between the patterns of convective particle
flow was made; however, no noticeable difference can be ident-
ified between the particle flow patterns. In other words, par-
ticle flow is not the main reason contributing to the change
in the shape of the radial stress–strain curves for symmetric
cyclic loading. Further study will seek the underlying
mechanism from the perspective of microstructure/fabric of
particles.
Asymmetric response of radial stresses may suggest unba-
lanced radial stresses at both sides of the pile. Further investi-
gation on the unbalanced horizontal forces is provided as
follows.
3.5.2 Unbalanced horizontal force
As shown in the hysteresis loops formed by the cyclic stress–
strain curves in Figure 4, at the end of each cycle, the stress
does not return to zero but builds up gradually. The evolution
of unbalanced horizontal force at the end of each cycle is illus-
trated in Figure 11. It is evident that the unbalanced horizon-
tal force is more significant with increasing strain amplitude. It
can also be seen that the unbalanced force for asymmetric
cyclic loading is much larger in magnitude and is oriented in
the opposite direction compared with that for symmetric cyclic
loading. It is because that for asymmetric cyclic loading, the
pile only moves to the right side and compresses the soils on
the right side significantly; therefore, the horizontal force on
0·3
0·2
Unbalanced horizontal force: N
0·1
0 &
0 250 500 750
–0·1
Strain (–0·1, 0·1%)
–0·2
Strain (0, 0·2%)
–0·3 Strain (–0·01, 0·01%)
Strain (0, 0·02%)
–0·4
Strain (–0·001, 0·001%)
–0·5 Strain (0, 0·002%)
–0·6
Number of cycles
Figure 11. Evolution of unbalanced horizontal force on the pile
at the end of each cycle
10
Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya
the pile at the end of each cycle is oriented to the left.
However, for the symmetric cyclic loading, the pile compressed
the soils on both sides to the same strain level. At the end of a
full cycle, the residual horizontal force is oriented to the right.
By referring to Figures 4(a) and 4(b), it can be understood
clearly that the end of a symmetric loading cycle is in the
middle of a hysteresis loop, but the end of an asymmetric
loading cycle is at the left corner of the hysteresis loop, even
though the stress magnitudes for both hysteresis loops are
approximately the same. Therefore, although the asymmetric
cyclic loading does not cause different dynamic soil responses
(stress–strain curves) from the symmetric cyclic loading during
the cyclic loading, it does build up greater unbalanced hori-
zontal force, which will cause excessive tilt and undermine the
stability of the monopile more significantly in the long term.
In this study, monopile was driven to move by a predefined
constant velocity. It is more realistic to simulate the free
motions of monopile under the action of resultant external for-
ce/moment, including the interaction force/pressure between
the monopile and the soil. It will be the next objective of the
authors’ future work.
4. Conclusions
The DEM simulations and small-scale tests provide a good
understanding on the soil–structure interaction of foundations
of OWTs. Various features observed in model tests could also
be replicated in DEM studies and thus provides confidence in
small-scale physical model tests. The following conclusions
could be drawn from the study.
& Stiffness of granular soils increases under cyclic loading.
Therefore, the stiffness of monopiles founded on
granular material is expected to increase with cycles
of loading and this increase may cause a change in
natural frequency of the wind turbine system. It may be
concluded that a ‘soft–stiff’ will move towards the 3P
frequency. It is necessary for the designers to predict the
change in frequency which is essential to predict the
fatigue life.
The convective soil flow and the soil densification
1000
surrounding the monopile is the main reason underlying
the increase in stiffness. These phenomena are more
significant with the increase in strain amplitude.
& Due to soil densification, the average radial stress on the
pile, and thus the shaft resistance of the pile increases
under cyclic loading.
& Asymmetric cyclic loading applies larger maximum strain
to soil, which results in higher stress in the first cycle;
however, the difference in the magnitude of stresses and the
shape of the stress–strain curves between asymmetric cyclic
loading and symmetric cyclic loading eliminates quickly
under cyclic loading. The long-term governing factor for
 Engineering and Computational Mechanics
the dynamic stress–strain response is the cyclic strain
amplitude.
& The asymmetric cyclic loading builds up greater
unbalanced horizontal force, which will undermine
the stability of the monopile more significantly in the
long term.
Acknowledgements
The authors acknowledge the RWE Innogy and the University
of Surrey for collectively funding this project.
REFERENCES
API (American Petroleum Institute) (1993) Recommended
Practice for Planning, Designing, and Constructing Fixed
Offshore Platforms: Working Stress Design, RP2A-WSD,
20th edn. American Petroleum Institute, Washington, DC,
USA.
Arany L, Bhattacharya S, Macdonald J and Hogan SJ (2014)
Simplified critical mudline bending moment spectra of
offshore wind turbine support structures. Wind
Energy 18(12): 2171–2197.
Bhattacharya S and Adhikari S (2011) Experimental
validation of soil–structure interaction of offshore
wind turbines. Soil Dynamics and Earthquake Engineering
31(5–6): 805–816.
Bhattacharya S, Lombardi D and Muir Wood D (2011)
Similitude relationships for physical modelling of
monopile-supported offshore wind turbines.
International Journal of Physical Modelling in
Geotechnics 11(2): 28–68, http://dx.doi.org/10.1680/ijpmg.
2011.11.2.58.
Bhattacharya S, Cox J, Lombardi D and Muir Wood D (2013a)
Dynamics of offshore wind turbines on two types of
foundations. Proceedings of the Institution of Civil
Engineers – Geotechnical Engineering 166(2): 159–169,
http://dx.doi.org/10.1680/geng.11.00015.
Bhattacharya S, Nikitas N, Garnsey J et al. (2013b) Observed
dynamic soil–structure interaction in scale testing of
offshore wind turbine foundations. Soil Dynamics and
Earthquake Engineering 54: 47–60.
Comodromos EM, Papadopoulou MC and Rentzeperis IK
(2009) Effect of cracking on the response of pile test
under horizontal loading. Journal of Geotechnical
and Geoenvironmental Engineering ASCE 135(9):
1275–1284.
Conte E, Troncone A and Vena M (2013) Nonlinear
three-dimensional analysis of reinforced concrete piles
subjected to horizontal loading. Computers and
Geotechnics 49: 123–133.
Conte E, Troncone A and Vena M (2015) Behaviour of flexible
piles subjected to inclined loads. Computers and
Geotechnics 69: 199–209.
Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya
Cuéllar P, Georgi S, Baeßler M and Rücker W (2012) On the
quasi-static granular convective flow and sand densification
around pile foundations under cyclic lateral loading.
Granular Matter 14(1): 11–25.
Cui L (2006) Developing a Virtual Test Environment
for Granular Materials Using Discrete Element
Modelling. PhD thesis, University College Dublin,
Dublin, Ireland.
Cui L, O’Sullivan C and O’Neill S (2007) An analysis of the
triaxial apparatus using a mixed boundary three-
dimensional discrete element model. Geotechnique
57(10): 831–844, http://dx.doi.org/10.1680/geot.2007.57.
10.831.
Cundall P and Strack O (1979) A discrete numerical model for
granular assemblies. Géotechnique 29(1): 47–65,
http://dx.doi.org/10.1680/geot.1979.29.1.47.
DNV (Det Norske Veritas) (2002) Guidelines for Design of Wind
Turbines, 2nd edn. DNV/Riso, Hovik, Denmark.
DNV (2007) Offshore Standard: Design of Offshore Wind
Turbine Structures, DNV-OS-J101. Det Norske Veritas,
Hovik, Denmark.
Karg C (2007) Modelling of Strain Accumulation Due to Low
Level Vibrations in Granular Soils. PhD thesis, Ghent
University, Ghent, Belgium.
Kuhn M (2000) Dynamics of offshore wind energy converters
on mono-pile foundation experience from the Lely offshore
wind turbine. OWEN Workshop, CLRC Rutherford
Appleton Laboratory, Swindon, UK.
Kuhn M (2002) Offshore wind farms. In Wind Power Plants:
Fundamentals, Design, Construction and Operation
(Gasch R and Twele J (eds)). Springer, Heidelberg,
Germany, pp. 365–384.
LeBlanc C (2009) Design of Offshore Wind Turbine Support
Structures – Selected Topics in the Field of Geotechnical
Engineering. PhD thesis, Aalborg University, Aalborg,
Denmark.
Lombardi D, Bhattacharya S and Muir Wood D (2013) Dynamic
soil–structure interaction of monopile supported wind
turbines in cohesive soil. Soil Dynamics and Earthquake
Engineering 49: 165–180.
Lowe J (2012) Hornsea met mast – A demonstration
of the ‘twisted jacket’ design. Proceedings of the
Global Offshore Wind Conference, ExCel London,
London, UK.
Matlock H (1970) Correlations for design of laterally loaded
piles in soft clay. Proceedings of the 2nd Annual Technology
Conference, Houston, TX, USA, pp. 577–587.
Mindlin R and Deresiewicz H (1953) Elastic spheres in contact
under varying oblique forces. ASME Journal of Applied
Mechanics 20: 327–344.
O’Sullivan C, Cui L and O’Neil S (2008) Discrete element analysis
of the response of granular materials during cyclic loading.
Soils and Foundations 48(4): 511–530.
11
 Engineering and Computational Mechanics
Reese L, Cox WR and Koop FD (1974) Analysis of laterally
loaded piles in sand. Proceedings of the 6th Annual
Technology Conference, Houston, TX, USA, Paper no.
2079, pp. 473–483.
Reese L, Cox WR and Koop FD (1975) Field testing and analysis
of laterally loaded piles in stiff clay. Proceedings of the
12
Downloaded by [ UNIVERSITY OF CAMBRIDGE] on [23/09/16]. Copyright © ICE Publishing, all rights reserved.
Soil–monopile interactions for offshore
wind turbines
Cui and Bhattacharya
7th Annual Technology Conference, Houston, TX, USA,
OTC paper no. 2312, pp. 671–675.
Yu LQ, Wang LZ, Guo Z et al. (2015) Long-term dynamic
behavior of monopile supported offshore wind turbines in
sand. Theoretical and Applied Mechanics Letter 5(2):
80–84.
HOW CAN YOU CONTRIBUTE?
To discuss this paper, please email up to 500 words to the
editor at journals@ice.org.uk. Your contribution will be
forwarded to the author(s) for a reply and, if considered
appropriate by the editorial board, it will be published as
discussion in a future issue of the journal.
Proceedings journals rely entirely on contributions from
the civil engineering profession (and allied disciplines).
Information about how to submit your paper online
is available at www.icevirtuallibrary.com/page/authors,
where you will also find detailed author guidelines.