Kourkoulis, R. S. et al. Géotechnique [http://dx.doi.org/10.1680/geot.12.P.

179]

Suction caisson foundations for offshore wind turbines subjected to wave

and earthquake loading: effect of soil–foundation interface
R . S . KO U R KO U L I S  , P. C . L E K K A K I S † , F. M . G E L AG OT I  a n d A . M . K AY N I A ‡

The response of wind turbines founded on suction caissons and subjected to lateral monotonic, cyclic
and earthquake loading is studied with due consideration of the role of soil–sidewall adhesion, using
non-linear three-dimensional finite-element analyses. In the case of monotonic and slow cyclic lateral
loading it is shown that imperfect interface bonding could reduce the moment capacity and may lead
to foundation detachment or even uplifting in the case of shallowly embedded caissons. A preliminary
comparison of two caisson alternatives has shown that increasing the caisson diameter while
maintaining the embedment ratio is more efficient in terms of material resources than increasing the
skirt length while keeping the diameter constant. The second part of the study evaluates the response
of a soil–foundation–wind turbine interacting system subjected to earthquake shaking. Contrary to an
often prevailing impression that seismic effects are insignificant, apparently originating from evaluat-
ing the seismic behaviour on the basis of spectral characteristics, it is illustrated that the system
kinematics may prove crucial for the response of large wind turbines subjected to simultaneous
environmental and seismic loads. Although not instantly catastrophic, the accumulation of foundation
rotation could lead to the turbine reaching serviceability limits early during its operation.

KEYWORDS: earthquakes; offshore engineering; soil/structure interaction

INTRODUCTION significant amount of research (e.g. Bransby & Randolph,

Construction of off-shore wind-turbines of massive dimen- 1998; Andersen & Jostad, 1999; Randolph & House, 2002;
sions has recently become a major activity in the field of Gourvenec & Randolph, 2003; Kelly et al., 2006; Gourve-
civil engineering as a result of European Union (EU) nec, 2007; Bransby & Yun, 2009); these studies have offered
demands to produce 20% of its electricity by means of valuable insight into the behaviour of such foundations.
renewable energy sources by 2020. Uniquely, the foundation However, most of these results refer to foundations of
design of wind turbines is characterised by the challenging limited skirt depth (low depth to diameter ratio, L/D)
combination of relatively low weight and large horizontal pertaining to off-shore oil platforms. As such, design of
loading due to the wind and wave currents, which produce deeper caissons based on these results may be affected by
large overturning moment at their bases (Houlsby & Byrne, the following uncertainties.
2000; Byrne & Houlsby, 2003).
Currently, several types of foundations are implemented for (a) Effect of soil non-uniformity: a widely used approach for
offshore wind turbines, depending on the site conditions and accounting for soil profiles of linearly increasing strength
water depth. In waters of medium depth, the monopile option is that proposed by Gourvenec & Randolph (2003) and
dominates the industry, with the alternative being a recently Gourvenec (2007). It suggests, in principle, that the
introduced scheme termed ‘suction caisson’, which was ori- failure envelopes of a skirted foundation may be
ginally proposed for the foundation of off-shore oil platforms. calculated on the basis of those of a shallow footing
It comprises a shallow footing whose capacity is enhanced by with equal dimensions resting at the skirt tip level of the
means of peripheral embedded skirts which confine the inter- original profile. This approach has been proven to provide
nal soil, thereby creating a soil plug. Ease of installation is the correct failure envelopes when referring to shallow skirt
main advantage of this foundation type. The process consists embedment, but its accuracy reduces in the case of longer
of floating the caisson to its location, where it is driven into skirts (i.e. L/D . 0.2) when the participation of the latter
the seabed under the action of its self-weight and pumping of in sustaining the moment demand is much more signifi-
water trapped within the skirts. The differential pressure due cant. Indeed, centrifuge tests results by Tani & Craig
to pumping creates suction which attracts the caisson lid (1995) and Watson & Randolph (1997) suggest this
downwards until it attains full contact with the soil. principle is reasonable for embedment depths less than
Over the last decade, the design of skirted foundations around 30% of the foundation diameter.
and particularly suction caissons has been the subject of a (b) Effect of interface conditions: most studies on suction
caisson foundations on clay tend to consider full contact
between the foundation and the soil, in other words where
the caisson can only detach from the surrounding soil if
Manuscript received 23 November 2012; revised manuscript accepted the strength of the latter is fully mobilised. This
29 November 2013. introduces two main assumptions, as follows.
Discussion on this paper is welcomed by the editor.
 Laboratory of Soil Mechanics, National Technical University of (i) As a consequence of the installation method, the
Athens (NTUA), Greece. foundation is able to develop tensile capacity when
† Stanford University, California, USA; formerly Laboratory of Soil the footing is subjected to uplift, owing to the
Mechanics, NTUA, Greece. negative excess pore pressures between the founda-
‡ Norwegian Geotechnical Institute (NGI) and Norwegian University tion lid and the confined soil. A series of experi-
of Science and Technology (NTNU), Trondheim, Norway. mental studies have been conducted to assess the

1
2 KOURKOULIS, LEKKAKIS, GELAGOTI AND KAYNIA
tensile capacity achieved by suction caissons
(Clukey & Morrison, 1993; Colliat et al., 1995;
Nacelle
Mello et al., 1998; Andersen & Jostad, 1999; Watson
et al., 2000; Houlsby et al., 2005; Kelly et al., 2006),
showing that their tensile resistance may be Rotor
Wind
comparable to their bearing capacity in compression, drotor
although this could depend on a number of factors
(e.g. cavitation, rate of uplifting etc.).
(ii) Full contact is ensured between the skirts and the V
surrounding soil. Yet, experimental evidence has
recently become available suggesting that this may H0
not always be valid. House & Randolph (2001) R
conducted centrifuge tests on large L/D ratio suction Waves
caissons concluding to a friction ratio of 0.4 between hw
the skirts and soil acting inside and outside the
Su
caisson. Andersen & Jostad (2002) suggest that an L
adhesion factor of 0.65 is applicable for a long D k
period after the installation of suction anchors, while 1
Jeanjean et al. (2006) have estimated a friction ratio
of 0.85 at their interface. Gourvenec et al. (2009)
experimentally measured a soil/skirt interface fric-
tion ratio as low as 0.3 during installation of low L/D (a)

ratios suction caissons; of course, these values refer

M
to installation conditions in relatively lightly over- V
consolidated soil and are referenced here only as an H (waves ⫾ wind)
indication of how low the friction ratio may
potentially become.
(c) Load reference point: traditionally, the bearing capacity
of embedded foundations subjected to combined loading
is tackled by means of interaction diagrams produced by
application of displacement on the base of the foundation. (b)
Yet, loading from wind turbines (stemming either from
earthquake or simply from the wind, sea waves or Fig. 1. (a) The problem under consideration and adopted nota-
currents) is transmitted from the turbine tower to the tion: a wind turbine founded on suction caisson in either
caisson top, and separation of the latter from the soil may homogeneous or inhomogeneous soil, whose strength is linearly
modify the amount of loading imparted to its base. increasing with depth with gradient k. (b) Schematic diagram
(d ) Response of wind turbines to earthquake: various studies explaining the possibility of foundation detachment from
surrounding soil when considering realistic shaft–soil interfaces
on the seismic response of on-shore wind turbines (i.e.
Bazeos et al., 2002; Lavassas et al., 2003; Ritschel et al.,
2003; Witcher, 2005; Haenler et al., 2006; Zhao &
Maisser, 2006), which were performed from a strictly on the parameter under investigation, two turbine assemblies
structural engineering perspective, have highlighted the (2 MW and 3.5 MW) were examined; their characteristics are
limited vulnerability of small-sized (i.e. 1 MW output or displayed in Table 1. Both homogeneous and linearly in-
lower) wind turbines to earthquake. However, less creasing soil profiles were considered. In the former case a
attention has been given to the role of kinematic stressing constant, undrained shear strength su ¼ 60 kPa was assumed
transferred to the superstructure due to possible founda- with Young’s modulus Estat ¼ 30 MPa and  ¼ 0.49, whereas
tion displacement. This effect may become of significant in the non-uniform deposit, the undrained shear strength sum
importance when examining modern mega-sized turbines. was assumed equal to 30 kPa at the seabed level and linearly
Despite their dimensions (and hence large eigenperiods) increasing with depth with a gradient of k ¼ 3 kPa/m. The
these structures will, during the earthquake, be subjected undrained shear strength of su ¼ 60 kPa corresponds to a
to the concurrent unidirectional wind and current loading. relatively stiff clay formation allowing focus on the caisson
This, although not disastrous (because wind turbine design response while avoiding exceptional soil yielding during the
codes impose very conservative limits to the maximum subsequent dynamic analyses. The soil has been assumed to
allowable rotation), could lead to the turbine reaching be a single-phase material with submerged unit weight of
serviceability limits very early during its operation. ª9 ¼ 10 kN/m3 : The caisson diameter and embedment ratio
was varied parametrically, as explained in the ensuing sec-
Based on this background, the scope of the current paper is tions.
to investigate the response of wind turbines founded on suction The problem is analysed using a three-dimensional (3D)
caissons subjected to monotonic lateral, cyclic and earthquake finite-element (FE) model taking account of soil inelasticity.
loading. Non-linear three-dimensional finite elements will be The developed FE model, taking advantage of problem
used in order to parametrically evaluate the effects of a linearly
increasing strength profile and of soil–sidewall interfaces;
emphasis will also be placed on the response of such structures Table 1. Geometric properties of the examined wind turbines
to moderate and strong earthquake loading.
R: m t: m H0 : drotor : Nacelle + rotor
m m mass: tn
GEOMETRY AND MODELLING 3.5 MW 2 0.023 80 90 220
The present research concerns offshore wind turbines 2 MW 2 0.023 60 60 200
founded on clay sites by suction caissons (Fig. 1). Depending
 SUCTION CAISSON FOUNDATIONS FOR OFFSHORE WIND TURBINES 3
pffiffiffi
geometry symmetry, is displayed in Fig. 2. Soil is modelled  y ¼ 3su (3)
with eight-noded hexahedral continuum elements. An initial
sensitivity analysis revealed that the foundation diameter has while
to be discretised into at least 15 elements to effectively  0 ¼  y  Æs ¼  y  E=ªr (4)
reproduce the mechanism of bearing capacity failure (due to
vertical loading) as well as the mechanisms of soil yielding Therefore, for the full description of the constitutive
developed during its lateral loading. model, only three parameters need to be determined: the
strength su , the E/su ratio and the parameter ªr : In other
words, for a given su value, proper selection of the E/su ratio
Constitutive model and the ªr term allows for simulation of a wide range of
Non-linear soil behaviour is modelled through a simple materials of varying dynamic properties (G–ª and –ª
kinematic hardening model with Von Mises failure criterion curves). Through the introduction of an external subroutine,
and associated flow rule, which is available in the Abaqus the rate of increase of kinematic hardening may also be
(2008) library and has been validated by Anastasopoulos et varied with respect to the level of plastic strain, which in
al. (2012) as to its capacity to simulate soil–structure inter- turn offers additional versatility to the model.
action systems under cyclic or seismic loading. The model The model adopted herein has been calibrated against
is appropriate for the simulation of clayey materials, the measured data by Raptakis et al. (2000) referring to a clayey
behaviour of which under undrained conditions may be con- material of su  60 kPa located at a depth of 10–20 m. Fig.
sidered as normal-pressure-independent. The evolution of 3(b) plots the result of the calibration procedure: the measured
stress is defined by G–ª (shear stiffness–shear strain) and –ª (damping–shear
 ¼  j0 þ º (1) strain) curves are compared to those derived numerically by
subjecting a soil element to cyclic simple shear loading at
where  j0 corresponds to the stress at zero plastic strain, various amplitudes of strain ª.
and º is a backstress parameter. The latter determines the Soil–foundation interfaces are simulated using special
kinematic evolution of the yield surface in the stress space contact elements. The latter allow sliding and detachment
through a function F where arising from the tensionless interface behaviour. Such ele-
F ¼ f (  º)   0 (2) ments have been used in all vertical interfaces, that is
between skirt and external and encaged (internal) soil. By
with f (  º) being the equivalent Mises stress with respect appropriately adjusting the interface properties, it is possible
to the backstress Æ and  0 is the size of the yield surface. A to model a variety of contact conditions of practical interest
one-dimensional representation of the non-linear stress low ranging from the fully bonded to totally tensionless regime.
is provided in Fig. 3(a). At large plastic strains, when  Imperfect contact between the caisson and the sidewalls may
approaches  y , the magnitude of Æ becomes equal to be caused by a number of factors related either to the
Æs ¼ E=ªr and (  Æ) tends to  0 (Fig. 3), where E is the installation process (i.e. some amount of shearing during
initial kinematic hardening modulus and ªr is a parameter driving of the skirt) or to the multitude of loading cycles
determining the rate of decrease of the kinematic hardening during the lifetime of a wind turbine, which could limit the
with increasing plastic deformation. available soil–foundation adhesion. In view of these, two
When considering clay materials, the maximum yield assumptions have been made regarding the maximum shear
stress  y is controlled by the undrained shear strength of the resistance that may be offered at the soil–foundation inter-
material su as face: (a) Æ ¼ 1 corresponding to fully bonded interface

Elastic beam

z
y
x Full bonding

L ⫽ 0·2D
Shell elements
L ⫽ 0·5D
D
Interface (τmax)
2·5D
τ
τmax ⫽ αsu
[ux, uy]*
3D brick elements: nonlinear soil δ

(b)
[ux, uy, uz]* ⫽ 0
*Boundary conditions not applicable to earthquake loading
6D
(a)

Fig. 2. (a) Finite-element mesh; (b) two assumptions are made for the caisson–soil interface: full adhesion (Æ 1) and reduced
adhesion (Æ 0.3)
4 KOURKOULIS, LEKKAKIS, GELAGOTI AND KAYNIA
σ

σy

σ0
σ|0

σ|0

σ|0 αs ⫽ E/γr
α

εpl
(a)

1·2 30
Measured behaviour
(Raptakis et al., 2000)
1·0 25
Calibration result

0·8 20
G/G0

ξ: %
0·6 15

0·4 10

0·2 5

0 0
1·0 ⫻ 10⫺6 ⫺5
1·0 ⫻ 10⫺4 ⫺3
1·0 ⫻ 10⫺2 ⫺1 1·0 ⫻ 10⫺6 1·0 ⫻ 10⫺4 1·0 ⫻ 10⫺2
1·0 ⫻ 10 1·0 ⫻ 10 1·0 ⫻ 10 1·0 ⫻ 10⫺5 1·0 ⫻ 10⫺3 1·0 ⫻ 10⫺1
γ γ
(b)

Fig. 3. (a) One-dimensional representation of the non-linear soil model. (b) Soil calibration result: each solid markers refers to
the calculated G/G0 ratio (or %) when a single soil element is subjected to a cyclic simple shear test at a fixed strain
amplitude, while the solid black line represents the target response

conditions or (b) Æ ¼ 0.3 reflecting a significantly reduced top) when the turbine is founded on a caisson with diameter
adhesion at the interface, which has been deliberately chosen D ¼ 20 m. Two embedment depths with L/D ¼ 0.2 and 0.5
in order to more clearly demonstrate the potential effects of are compared, and the plots are presented in non-dimensional
imperfect interface conditions. While it is also possible to terms so as to maintain compatibility with the prevailing
model the contact conditions at the soil/caisson lid interface, literature. However, attention is needed when comparing the
it is assumed that the latter maintains perfect contact with results of uniform with non-uniform soil because the actual
the underlying soil, warranted by its tensile capacity as reference strength is different. To avoid misinterpretations,
explained previously. the actual shear strength su,0 at the skirt tip level for each
case is highlighted on the diagrams.
Figure 4(a) compares the moment–rotation curves pro-
MONOTONIC LOADING: EFFECT OF SOIL duced for the shallow caisson (L/D ¼ 0.2) in the homoge-
INHOMOGENEITY AND INTERFACE CONDITIONS neous compared with the linearly increasing profile. When
An initial set of analyses is presented in this section which considering the uniform profile, it is evident that, despite
refers to displacement-controlled monotonic lateral loading their limited length, the skirts add significantly to the
applied at the centre of mass of the 3.5 MW turbine moment capacity, which reaches M0:2 ¼ ADsu0 : (where A is
(assumed to lie at Hc ¼ 80m from the caisson lid) until it the caisson base surface), that is, an increase of the order of
attains its ultimate capacity, while considering the tower 50% compared to that of a surface foundation (Msur ¼
absolutely rigid so that focus is on the foundation response. 0.67ADsu0 ). Yet, perhaps contrary to the reader’s intuitive
Note that for the cases examined here, the combination of a anticipation, consideration of linearly varying soil generates
very low V/Vult (vertical load over vertical capacity) ratio slightly higher moment capacity, despite its lower average
with a very high lever arm (moment over horizontal force strength along the skirt. (This holds true in non-dimensional
M/H) makes the moment capacity practically equal to that terms; of course, the actual value of the moment capacity in
under pure moment loading M0 under zero vertical loading the inhomogeneous soil case is lower than that of the
(V ¼ 0). Therefore, results of these sections are expected uniform soil stratum.) Indeed, failure is governed by a scoop
also to hold for taller wind turbines (e.g. 5 MW or 7 MW), mechanism underneath the skirt, contained within soil of
as the ones currently implemented or planned. Results are higher strength than that at the skirt tip level (su0 ), thus
plotted in terms of moment–rotation and settlement–rotation producing higher non-dimensional capacity.
diagrams calculated on the base of the tower (i.e. foundation Consideration of the reduced-adhesion scenario inevitably
 SUCTION CAISSON FOUNDATIONS FOR OFFSHORE WIND TURBINES 5
Linearly increasing profile
Homogeneous profile
1·2 1·2

1·0 1·0

0·8 0·8

M /ADsu0
M /ADsu0

0·6 c.m 0·6 c.m

δ δ

0·4 D ⫽ 20 m 0·4 D ⫽ 20 m

⬃M0 ⬃M0
0·2 0·2
L/D ⫽ 0·2 su0 ⫽ 60 kPa L/D ⫽ 0·2 su0 ⫽ 42 kPa
0 0
0 0·05 0·10 0·15 0 0·05 0·10 0·15
θ: rad θ: rad

Full contact

Low adhesion (α ⫽ 0·3)

(a)

1·6 1·6

1·2 1·2
M /ADsu0
M /ADsu0

c.m c.m
0·8 δ 0·8 δ
D ⫽ 20 m D ⫽ 20 m

0·4 ⬃M0 0·4 ⬃M0

su0 ⫽ 60 kPa
su0 ⫽ 60 kPa
L/D ⫽ 0·5 L/D ⫽ 0·5
0 0
0 0·02 0·04 0·06 0·08 0·10 0 0·02 0·04 0·06 0·08 0·10
θ: rad θ: rad

(b)

Fig. 4. Dimensionless moment–rotation curves obtained during monotonic lateral loading of the caisson in homogeneous and
linearly increasing soil for the two interface scenarios: (a) shallow caisson: D/B 0.2, (b) deep caisson: D/B 0.5; c.m, centre of
mass of the wind turbine

results in annulation of the shearing resistance offered by that focus can be on the effect of lateral loading. Consider-
the part of the sidewalls opposite to the direction of dis- ing the full contact case (solid black lines) the graphs reveal
placement. As such, moment may be transmitted to the soil that non-uniformity tends to limit the caisson’s settlement
through the soil–lid interface (considered fully bonded independently of embedment depth.
throughout this paper) as well as the normal and shear To explain this behaviour the failure mechanisms are
stresses developed at the part of the skirt lying in the same portrayed in Fig. 6. When the homogeneous profile is con-
direction as the loading, which is therefore unable to detach; sidered and full contact is ensured, the caisson internal soil
naturally, this results in lower resistance than the scenario system merges into one quasi-solid caisson whose reaction
under full contact conditions. results in the formation of a deep scoop failure mechanism
The same pushover curves for the case of deeper embed- similar to the one identified by Bransby & Yun (2009). In
ment (L/D ¼ 0.5) are plotted in Fig. 4(b), reflecting the the linearly varying profile (Fig. 6(b)), shear failure zones
crucial role of the deeper sidewalls in offering resistance. tend to form higher than in the uniform soil case due to the
The capacity of the deeply embedded caisson relies signifi- lower available strength in shallower strata. The nature of
cantly on the contribution of the skirts, which apparently this profile generates partial mobilisation of the internal soil
reduces as the adhesion factor decreases (leading to an plug strength, thus producing an inverted scoop mechanism
overall capacity reduction by 40% for the homogeneous soil) within the skirts in accord with findings by Yun & Bransby
while it drops dramatically when the lower adhesion is (2007), Bransby & Yun (2009) and Mana et al. (2010,
superimposed, with the skirts being contained within soil of 2013). The combination of a shallower external scoop, with
reduced strength, as is the case of the linearly varying the sliding of the foundation along the inverted internal
profile (where an overall capacity reduction by 65% is scoop, limits the tendency of the foundation to settle, as
observed). Still, even when adopting the low-adhesion sce- evidenced in Fig. 5.
nario, the dimensionless moment reaches a value which is When the non-uniform profile is superimposed with the
higher than that of the surface foundation by almost 35% in low adhesion interface (Fig. 6(c)), the very low sidewall
the homogeneous profile. resistance gives rise (in the extreme scenario examined) to
The settlement–rotation curves of the two soil profiles are sliding or even detachment of the caisson from the soil,
compared in Fig. 5. Notice that the initial static settlement thereby practically cancelling the formation of the external
(due to the weight of the structure) has been subtracted so scoop, and may even result in foundation uplifting.
6 KOURKOULIS, LEKKAKIS, GELAGOTI AND KAYNIA
Homogeneous profile Linearly increasing profile
0·010 0·010
δ δ

D ⫽ 20 m D ⫽ 20 m
θ θ
0·005 0·005
w su0 ⫽ 60 kPa w su0 ⫽ 42 kPa
L/D ⫽ 0·2 L/D ⫽ 0·2
w /D

w /D
Uplift Uplift
0 0
Settlement Settlement

⫺0·005 ⫺0·005
0 0·05 0·10 0·15 0 0·05 0·10 0·15
θ: rad θ: rad

Full contact

Low adhesion (α ⫽ 0·3)

(a)
0·001 0·001
Uplift Uplift
0 0
Settlement Settlement

δ δ
w/B
w/B

⫺0·002 ⫺0·002

D ⫽ 20 m D ⫽ 20 m
θ θ
w su0 ⫽ 60 kPa su0 ⫽ 60 kPa
⫺0·004 ⫺0·004 w
L/D ⫽ 0·5 L/D ⫽ 0·5

0 0·025 0·050 0·075 0·100 0 0·025 0·050 0·075 0·100

θ: rad θ: rad
(b)

Fig. 5. Dimensionless settlement–rotation curves obtained during monotonic lateral loading of a caisson in homogeneous and linearly
increasing soil for the two interface scenarios: (a) shallow caisson: D/B 0.2 (b) deep caisson: D/B 0.5
Tower

Tower
Tower

⬃M0 su ⬃M0 su ⬃M0 su

z z z
Fully bonded Fully bonded Low adhesion (α ⫽ 0·3)

⬃M0 ⬃M0 ⬃M0

(a) (b) (c)

Fig. 6. Failure mechanism of a suction caisson of L/D 0.5 under (almost) pure moment loading (i.e. H/Hult  1, V/Vult  1):
(a) homogeneous soil profile and fully bonded interface; (b) linearly increasing profile and fully bonded interface; (c) linearly increasing
profile and low-adhesion interface
 SUCTION CAISSON FOUNDATIONS FOR OFFSHORE WIND TURBINES 7
The results of non-linear foundation response are revisited the amplitude of the imposed displacements. For the low-
in more detail in the next section, which examines the amplitude cycles, loops are quite negligible (nearly elastic
behaviour of the caisson subjected to cyclic loading. behaviour). For the high-amplitude cycles, the loops tend to
swell, echoing the inelastic response of the system and
revealing the enhanced energy dissipation. Consideration of
CYCLIC LOADING: EFFECT OF IMPERFECT the low-adhesion interface tends to decrease energy dissipa-
INTERFACE tion, while in all cases the deeply embedded caisson offers
In this section, the example caissons will be subjected to substantially increased moment capacity compared to the
cyclic loading consisting of constant-amplitude displacement shallow alternative. Note that in the full contact case (Fig.
cycles, as explained below, attempting to derive a prelimin- 7(a)), the diagrams follow the monotonic curve, which
ary manifestation of the possible impact of earthquake understandably is not the case when full contact is not
loading. As such, they may serve as a preface for the warranted.
ensuing dynamic analysis section. Again, the analyses pre- Figure 7(b) depicts the cyclic moment–rotation graphs for
sented in this section refer to displacement-controlled lateral the ‘low-adhesion’ interface. For the low-amplitude cycles,
loading cycles applied at the nacelle level of the 3.5 MW the loops retain their original shape, owing to the fact that
turbine (assumed to lie at H0 ¼ 80 m from the caisson lid), the system still remains approximately elastic and only
considering the tower as absolutely rigid and the caisson minor separation between caisson and soil takes place. The
diameter D ¼ 20 m. difference is more conspicuous for the higher amplitude
Two sets of slow cyclic analyses were performed, each rotation of Ł ¼ 0.05, when the shape of the loop past the
containing five equal-amplitude cycles. Both embedment first cycle deviates from the monotonic curve and degener-
ratios (L/D ¼ 0.2 and 0.5) and interface conditions were con- ates to a more pinched shape.
sidered, while the soil was assumed homogeneous (su ¼ This is attributed to the effect of not only material non-
60 kPa). The first set (small cyclic loading) of analyses linearities, but also geometric ones, manifesting themselves
entailed application of displacements generating only minimal in the form of an irrecoverable gap behind the foundation
rotations of approximately 0.01 rad, which (according to the (Fig. 8): as the caisson rotates during the first cycle towards
monotonic results) roughly correspond to the state of incipi- one direction, it produces plastic deformation of the resisting
ent soil yielding (Fig. 7, thin black line). The second set of soil (see point A), which may not be recovered once the
analyses (large cyclic loading) consisted of imposing displa- direction of loading is reversed. Consequently, during
cements that produced rotations of 0.05 rad (Fig. 7, bold the second cycle of loading towards the same direction, the
black line), in order to stimulate excessive non-linearities and existence of the gap reduces the overall resistance until
engender a yielding dominated response. Note that the un- the gap closes. Thus, the ability of the system to meet
loading paths always obey Masing’s principle – an inherent moment demands is severely weakened in between the
assumption of the adopted constitutive law. extreme rotations, as evidenced by the flattened shape of the
The area of the loops (Fig. 7) indicates the energy loop.
dissipated during cyclic loading, which is dependent upon Naturally, in case of low L/D (Fig. 7(b)), even for the

Full bonding Low adhesion (α ⫽ 0·3)

1·50 1·50
L/D ⫽ 0·2 L/D ⫽ 0·2 Large cyclic (θmax ⫽ ⫾0·05)
Small cyclic (θmax ⫽ ⫾0·01)
0·75 0·75 Monotonic
M /ADsu

M /ADsu

0 0 δ
m

h ⫽ 80 m
M/ADsu
⫺0·75 ⫺0·75
L/D ⫽
θ ⫽ δ/h Su ⫽ 60 kPa
⫺0·05 ⫺0·01 0·2 or 0·5
⫺1·50 ⫺1·50
20 mm
⫺0·08 ⫺0·04 0 0·04 0·08 ⫺0·08 ⫺0·04 0 0·04 0·08

2 2
L/D ⫽ 0·5 L/D ⫽ 0·5

1 1
M /ADsu

M /ADsu

0 0

⫺1 ⫺1

⫺2 ⫺2
⫺0·08 ⫺0·04 0 0·04 0·08 ⫺0·08 ⫺0·04 0 0·04 0·08
θ: rad θ: rad
(a) (b)

Fig. 7. Dimensionless moment–rotation loops obtained during slow cyclic lateral loading of the shallow (L/D 0.2, top) and the deeply
(L/D 0.5, bottom) embedded caisson in homogeneous soil for the two interface scenarios: (a) full bonding; (b) low cohesion (numerical
results refer to a 3.5 MW turbine (H0 80 m) founded on a suction caisson with D 20 m)
8 KOURKOULIS, LEKKAKIS, GELAGOTI AND KAYNIA
2
tonic curve, while it tends to progressively increase with the
number of applied cycles of loading and mobilisation of
more soil. Not surprisingly, for the high-amplitude cycles,
Point A
1 excessive soil yielding is responsible for higher rotations and
settlements. A marginally lower settlement is experienced by
Point C
the L/D ¼ 0.5 caisson, which offers higher resistance.
The effect of low adhesion on the w–Ł response is
M /ADsu

0 explored in Fig. 9(b). Again, for low-amplitude loading the

response does not substantially differ from the full contact
case: independently of embedment ratio, settlement keeps
Point B accumulating during each loading cycle. Discrepancies are
⫺1
much more pronounced with high-amplitude loading. For
Positive
Negative
θ1 L/D ¼ 0.2, the diagrams for the ‘low-adhesion’ interface
θ1
(Fig. 9(b)) indicate a rocking-dominated response of the
⫺2 caisson–soil system. Under the action of lateral loading, part
⫺0·08 ⫺0·04 0 0·04 0·08 of the foundation uplifts towards the loading direction while
θ: rad the pole of rotation instantaneously shifts towards its edge.
Rocking of the foundation inevitably causes some soil yield-
ing underneath it (due to the instantaneous reduction of its
Point A effective area), which results in the system finally experien-
Gap at left cing some settlement during each loading–unloading–
Positive
θ1
reloading cycle. Yet, the rate of accumulation of settlement
remains remarkably lower than that under fully bonded
conditions.
For the L/D ¼ 0.5 foundation (bottom plots), as discussed
already, the larger area of the skirts provides increased
resistance along the periphery and hence limits the uplifting
ability of the caisson. Thus, settlement – due to soil yield-
ing – keeps accumulating from the first cycle.

Point B
RESPONSE UNDER SERVICE LOADS: ROLE OF
Gap at left INTERFACES AND SOIL NON-UNIFORMITY
Negative Having identified the mechanisms governing the response
θ1
of suction caissons to displacement-controlled cyclic loading,
this section is devoted to the behaviour of a typical 3.5 MW
wind turbine on suction caissons to service loads. To better
represent the loading conditions, the analyses were per-
formed in force-controlled mode. The scope of this investi-
gation is twofold

(a) to assess quantitatively the importance of interfaces and

soil inhomogeneity
Point C
Gap at both sides (b) preliminarily to weigh the balance of benefit between
increasing the diameter D (while maintaining L/D
constant) over increasing the depth of embedment (while
keeping D constant).

The design of suction caissons was performed on the basis

of loads acting on the turbine and allowable foundation
rotation according to Houlsby & Byrne (2005). Three foun-
dation configurations have been parametrically examined

(a) D ¼ 20 m, L/D ¼ 0.2

Fig. 8. Illustration of effect of detachment of caisson from (b) D ¼ 20 m, L/D ¼ 0.5
surrounding soil on resulting moment–rotation loop (c) D ¼ 25 m, L/D ¼ 0.2.

Apparently, the D ¼ 20, L/D ¼ 0.2 set-up would be unac-

ceptable but it is examined herein as an example of un-
weak interface scenario, the shape of the M–Ł loop is more conservative design; the remaining two alternatives may be
rounded because the foundation response is controlled by considered as rational choices. Based on literature data
the mobilised strength at the caisson base, while the lateral (Table 1), the adopted amplitude values of wind and wave
resistance and thus the possible formation of a gap cannot loads acting on the 3.5 MW turbine were taken as
substantially modify the response in subsequent cycles of
loading. (a) wind load: 1 MN, acting on the level of the rotor–nacelle
Quite similar conclusions may be drawn from the settle- assembly (80 m from mud-line)
ment–rotation diagrams (Fig. 9). Under the bonded interface (b) wave load: 1  2 MN, acting on a height of approxi-
regime (Fig. 9(a)) the initial settlement follows the mono- mately 7.8 m from the mud-line.
 SUCTION CAISSON FOUNDATIONS FOR OFFSHORE WIND TURBINES 9
Full bonding Low adhesion (α ⫽ 0·3)
4 4
L/D ⫽ 0·2 L/D ⫽ 0·2 Large cyclic (θmax ⫽ ⫾0·05)
2 2 Small cyclic (θmax ⫽ ⫾0·01)
Uplift
Uplift Monotonic
0 0
w /D: ⫻10⫺3

w /D: ⫻10⫺3
Settlement

⫺2 ⫺2 Settlement δ
m

⫺4 ⫺4 h ⫽ 80 m
θ
⫺6 ⫺6
L/D ⫽
0·2 or 0·5 w/D Su ⫽ 60 kPa
⫺8 ⫺8
⫺0·10 ⫺0·05 0 0·05 0·10 ⫺0·10 ⫺0·05 0 0·05 0·10 20 mm
θ: rad θ: rad

0 0
L/D ⫽ 0·5 L/D ⫽ 0·5

⫺2 ⫺1
w /D: ⫻10⫺3

w /D: ⫻10⫺3

⫺4 ⫺2

⫺6 ⫺3

⫺8 ⫺4
⫺0·10 ⫺0·05 0 0·05 0·10 ⫺0·10 ⫺0·05 0 0·05 0·10
θ: rad θ: rad
(a) (b)

Fig. 9. Dimensionless settlement–rotation curves obtained during slow cyclic lateral loading of the shallow (L/D 0.2, top)
and the deeply embedded (L/D 0.5, bottom) caisson in homogeneous soil for the two interface scenarios: (a) full bonding;
(b) low adhesion (Æ 0.3) (numerical results refer to a 3.5 MW turbine (H0 80 m) founded on a suction caisson with
D 20 m)

Response to environmental loading Consideration of the low-adhesion interface (Fig. 10(b))

The first step of the analyses entailed the application of produces augmented rotations in all cases, albeit the increase
dead loads to the model. This was followed by a second ratio is proportional to the original rotation value under full
loading step consisting of the application of wind load, contact conditions, which means that the relative efficiency
modelled as a constant horizontal force on the level of the between the cases examined is preserved. The increase is
rotor, and a third step containing ten cycles of pseudo- about 45% for the D ¼ 25 m footing, 57% for D ¼ 20 m and
statically imposed wave force. L/D ¼ 0.5 and 68% for D ¼ 20 m and L/D ¼ 0.2.
In the ensuing analysis, serviceability capacity will be
evaluated on the basis of current code limitations regarding
the allowable deformations as follows EARTHQUAKE LOADING
Settlement: wmax ¼ 0:05B, where B is the caisson diameter It is well known that construction of off-shore wind farms
Rotation: Łmax ¼ 0.001 rad (DNV, 2001; Houlsby et al., is being planned with increasing intensity worldwide, not
2005). excluding the seismically active regions (i.e. California,
Figure 10 plots the evolution of foundation rotation with Japan, Italy and Greece). In general, wind turbines are low-
increasing number of cycles for the cases examined. The frequency structures, and as such their structural systems are
same trends are in general observable in both the homoge- relatively insensitive to earthquake loading. Indeed, the first
neous and the linearly increasing profile, although rotations two eignefrequencies of the investigated 3.5 MW turbine
appear to be larger in the latter. Among the three types of have been numerically estimated at f0 ¼ 0.275 Hz and
foundations, the B ¼ 20 m with L/D ¼ 0.2 clearly exhibits f1 ¼ 2.75 Hz respectively. These values agree with those
the largest values for rotation as well as incremental rota- analytically calculated following the Van der Tempel (2006)
tions, exceeding the serviceability rotation limit SL (despite formula.
the very limited number of loading cycles); a fact that, Understandably, these values will be even higher for larger
expectedly, renders it insufficient to support a typical turbines, as the ones currently planned (e.g. 5 MW), confirm-
3.5 MW wind turbine. The remaining two caissons maintain ing the expectation for limited vulnerability of their structur-
almost constant rotation despite the increasing number of al systems to seismic loading. However, the focus of this
cycles. An interesting deduction is that by increasing the section will be on the investigation of the soil–foundation–
foundation’s diameter while maintaining a low embedment superstructure interaction, which may be responsible for
ratio (D ¼ 25 m, L/D ¼ 0.2), a more favourable response is additional kinematic loading being imposed on the system.
achieved than by even substantially increasing the skirt As already stated, the model turbines examined in this
length (i.e. from L/D ¼ 0.2 to L/D ¼ 0.5 in the D ¼ 20 m paper are assumed to be founded on an su ¼ 60 kPa soil,
case). Bearing in mind the fact that manufacturing of the corresponding to an intermediate category D soil according
two alternatives requires the same quantity of steel, it to the Eurocode 8 subsoil classification (CEN, 2004), which
becomes evident that the former solution constitutes a more represents a layer of predominantly soft-to-firm cohesive
efficient foundation (assuming that the installation cost re- soil. The imposed earthquake scenario will be typical of a
mains constant for both alternatives). moderate-to-strong seismicity region of Europe, correspond-
10 KOURKOULIS, LEKKAKIS, GELAGOTI AND KAYNIA

1 MN 3·5 MW

G ⫹ wind Application of wave G ⫹ wind Application of wave

1 ⫾ 2 MN
4 4
θ
Homogeneous Linearly increasing su

3 3
θ: ⫻10⫺3

θ: ⫻10⫺3
2 2

1 Serviceability 1
limit (SL)

0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Number of cycles (a) Number of cycles
4 4
D ⫽ 20,
L/D ⫽ 0·2
3 D ⫽ 20, 3
L/D ⫽ 0·5
θ: ⫻10⫺3

θ: ⫻10⫺3
D ⫽ 25,
2 L/D ⫽ 0·2 2

1 Serviceability 1
limit (SL)

0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Number of cycles Number of cycles
(b)

Fig. 10. Evolution of foundation rotation with increasing number of cycles of the 3.5 MW turbine foundation under environmental loads
in both homogeneous and linearly increasing soil profiles assuming: (a) fully bonded interface; (b) low-adhesion interface

ing to a peak ground acceleration, PGA ¼ 0.36g. The corre- each level through the application of kinematic constraints
sponding Eurocode 8 design spectrum is displayed in between the node on the central axis and each peripheral
Fig. 11. The load combinations (i.e. of environmental and node), while dashpot elements have been used at the base of
seismic loads) follow the norms set by the International the model to correctly reproduce radiation damping. The
Electrotechnical Commission (IEC) standard 61400/2005 properties of the base ‘substratum’ are shown in Table 2.
(IEC, 2005) and are presented in Table 3. The damping coefficient of the dashpots is given by
The time histories applied to the base of the model are
C ¼ rV s Ad (3)
defined as modified Takatori and Rinaldi records (Fig. 11(a))
as they have originated after mathematically manipulating where r is the material density, V s the shear wave velocity
the Takatori and Rinaldi accelerograms recorded during the and Ad the effective area of the dashpot (which depends on
Kobe (Japan), 1995 and Northridge (USA), 1994 earth- the mesh density at each region). A value of (Rayleigh)
quakes respectively. As can be seen, the imposed spectral damping of s ¼ 3% was adopted for the soil stratum in
acceleration at the region of the dominant period (T ¼ 3.6 s) order to ensure viscoelastic response under even low strain
of the turbine is compatible with the Eurocode 8 spectrum amplitude. Based on the IEC 61400/2005 wind turbine de-
for both scenarios examined. sign code the adopted tower damping was equal to t ¼ 1%.
The turbine is modelled as a 1-degree-of-freedom (dof) Loading is again imposed in three steps. During the first
system consisting of a beam and a concentrated mass at the step, the dead loads are applied to the model. This is
rotor–nacelle level. The beam section has been allocated the followed by a second step consisting of the application of
proper tower geometry and density so that the tower re- environmental loads, and a third dynamic step during which
sponse is correctly represented. As such, the tower has been the time history analysis is conducted. Performance is
modelled as a steel pipe section of radius 2 m and thickness assessed on the basis of the previously identified serviceabil-
t ¼ 2.3 cm rendering a bending stiffness of EI  120 GNm2 : ity limit which, in terms of rotation, is Łmax ¼ 0.001 rad.
The turbine has been assumed to be founded on the two
foundation alternatives that are regarded as acceptable fol-
lowing the reasoning of the previous section: D ¼ 20 m with Response to ground shaking
L/D ¼ 0.5 and D ¼ 25 m with L/D ¼ 0.2 (Fig. 11(c)). The The acceleration–time histories recorded on the tower top
initial elastic modulus over shear strength ratio, E0 /su , was are displayed in Fig. 12(a). Observe that, independently of
taken as equal to 1800 following the calibration test pre- the shaking scenario and the foundation type, the turbine
sented previously. Proper kinematic constraints have been response is maintained within controllable limits with the
assumed at the lateral boundaries of the FE model to maximum experienced acceleration at 0.25g. As anticipated,
simulate free-field response (i.e. by creating a rigid disc at its oscillation is invariably out of phase with the excitation–
 SUCTION CAISSON FOUNDATIONS FOR OFFSHORE WIND TURBINES 11

Modified Rinaldi Modified Takatori

5·0 5·0
0·33g

Acceleration: m/s2:
Acceleration: m/s2

2·5 2·5
Acc
0 0

⫺2·5 ⫺2·5
0·4g
⫺5·0 ⫺5·0
0 5 10 15 20 0 5 10 15 20 Dashpots
t: s t: s
(a)
3 3
Spectral acceleration: g

Spectral acceleration: g
Seismic excitation
(c)
2 2
Eurocode 8 Eurocode 8
(A ⫽ 0·36g, Soil D) (A ⫽ 0·36g, Soil D)
1 1

0 0
0 1 2 3 3·5 MW 0 1 2 3 3·5 MW
T: s T0 ⫽ 3·6 s T: s T0 ⫽ 3·6 s
(b)

Fig. 11. (a,b) The modified Rinaldi and Takatori time histories used as input for the dynamic analyses along with their
acceleration response spectra. (c) Boundary conditions allowing proper simulation of earthquake loading: dashpots at the bottom
of the numerical model and kinematic constraints at the peripheral nodes

Table 2. Dashpot properties Despite this observation, both foundations tend to accu-
mulate rotation during each cycle, even during the mild
su : kPa E0 /su r: tn /m3 Vs: m/s modified Rinaldi shaking (Fig. 12(c)). Although a paradox
according to initial spectral-based estimates, the long period
120 1800 2 190 of the 3.5 MW turbine is inadequate to render it insensitive
to ground shaking when taking account of the whole founda-
tion–structure system. On the contrary, the residual rotation
may be increased to approximately six to seven times its
initial value depending on the seismic scenario. This reveals
time history. The turbine oscillates mainly in its first eigen- a quite augmented detrimental effect of environmental (wind
mode; yet excitation of its second mode is also evidenced and wave) loading acting concurrently with the earthquake
by the ‘curly’ shape of the produced time histories. This and hence producing unidirectional rotation, as explained in
intense vibration in small periods could cause dysfunction of the next section and may be regarded as the direct analogue
the mechanical and electrical components of the rotor– of a rigid body sliding along a slope.
nacelle assembly or affect the vibration of the blades. Consistently with the conclusion drawn in the previous
The most interesting results, however, stem from the sections, consideration of the reduced-adhesion interface
examination of the foundations’ response. Both foundations provokes a further 40–60% increase in rotation amplitude,
offer quite large safety factors against overturning moment although the experienced acceleration on the tower top as
(FSM  7) and thus they are able to guarantee roughly well as the general response pattern remain unchanged (Fig.
equivalent fixity conditions on the tower base – a fact 13).
reflected in the tower’s response. Consistently with this Rotation build-up during each cycle is reasonably more
observation, the developed moments at the tower base (Fig. intense in case of the more severe (modified Takatori)
12(b)) are practically the same for the two different embed- loading scenario, but does also takes place under the milder
ment ratios. The fact that no significant discrepancy may be excitation scenario. Such accumulation of rotation is not
noticed between the two earthquake scenarios is primarily directly threatening the safety of the structure, yet it consti-
due to the large flexibility of the tower and is an additional tutes irrecoverable damage to the serviceability of the
indication of the limited sensitivity of the turbine to earth- turbine and might be responsible for curtailing the service
quake shaking. life of the facility.

Table 3. Load and structure configuration combinations used in the earthquake loading
analyses

Foundation geometry Loads Foundation response

D: m L/D Wind: kN Wave: kN Ł0 3 103 M/Mult ¼ FSM

3.5 MW 20 0.5 1000 2000 0.26 7.0

25 0.2 0.22 9.0
2 MW 17 0.2 350 1150 0.25 8.5
12 KOURKOULIS, LEKKAKIS, GELAGOTI AND KAYNIA
Modified Rinaldi Modified Takatori
5·0 D ⫽ 20 m, L/D ⫽ 0·5 D ⫽ 20 m, L/D ⫽ 0·5
D ⫽ 25 m, L/D ⫽ 0·2 D ⫽ 25 m, L/D ⫽ 0·2

2·5
Acceleration: m/s2

⫺2·5

⫺5·0
(a)
200
D ⫽ 20 m, L/D ⫽ 0·5 D ⫽ 20 m, L/D ⫽ 0·5
D ⫽ 25 m, L/D ⫽ 0·2 D ⫽ 25 m, L/D ⫽ 0·2
150
M: MNm

100

50

0
(b)
3
D ⫽ 20 m, L/D ⫽ 0·5 D ⫽ 20 m, L/D ⫽ 0·5
D ⫽ 25 m, L/D ⫽ 0·2 D ⫽ 25 m, L/D ⫽ 0·2

2
θ: ⫻10⫺3 rad

1 SL

0
0 5 10 15 0 5 10 15 20 25
t: s t: s
(c)

Fig. 12. Response of 3.5 MW turbine on the two foundation alternatives subjected to the two shaking scenarios assuming
fully bonded conditions: (a) acceleration–time history at the tower top; (b) bending moment–time history at the tower
base; (c) foundation rotation–time histories. The serviceability limit of Ł 1 3 1023 is marked on the diagrams

The role of inertial loading tions as well (Table 3). Indeed, the 2 MW tower, having
Although already implied by the previous discussion, it is higher natural frequencies, experiences significantly larger
here attempted to unveil the potential fallacy of predicting acceleration levels at its top. The most interesting results,
the turbine seismic response based solely on spectral indica- however, once more appear when examining the foundation
tions. To this end, Fig. 14 plots the rotation–time history on rotation plots (Fig. 14(b)). When neglecting environmental
the foundation of the 3.5 MW turbine subjected to modified loads, the rotations experienced by the 2 MW foundation
Rinaldi where the action of wind and waves are neglected. during shaking are larger than those of the 3.5 MW founda-
Noticeably, foundation rotation does take place; however, the tion. Yet, they are totally recoverable after the end of shaking.
rotation experienced during one cycle is recovered during the Observe that when considering the wind and current forces,
next, resulting in only negligible residual distortion – a the developed rotation is not only irrecoverable, but may
response practically consistent with the anticipation engen- exceed that of the 3.5 MW turbine. Remarkably, while the
dered by examining the modified Rinaldi spectrum. In the 2 MW foundation keeps accumulating rotation during each
same context, the figure also plots the response of a 2 MW cycle, the foundation of the low-frequency 3.5 MW turbine
superstructure whose higher eigenfrequencies ( f0 ¼ 0.39 Hz, acquires the most part of its residual rotation immediately
f1 ¼ 4.75 Hz) are expected to make it more prone to earth- after the strong pulse at about 6 s. This is a consequence of
quake-induced distortion. In order to obtain comparable the substantially larger environmental loads acting on it,
results with the 3.5 MW case, the 2 MW turbine is consid- which compel the caisson to rotate although the structural
ered to be founded on a D ¼ 17 m, L/D ¼ 0.2 foundation, vibration is practically insensitive to shaking.
which renders a safety factor against overturning moment, Of course, significantly larger turbines are currently being
FS2MMW ¼ 8.5 (recall FSM 3:5 MW
¼ 9). Its elastic rotation under planned or implemented than the ones examined herein; the
the action of environmental loads is Ł20 MW ¼ 2 3 103 , greater flexibility of these is expected to result in limited
which is also very similar to that of the 3.5 MW turbine, thus vulnerability to seismic shaking. However, owing to their
rendering the two systems comparable under elastic condi- larger diameters, they are expected to carry greater concurrent
 SUCTION CAISSON FOUNDATIONS FOR OFFSHORE WIND TURBINES 13
Modified Rinaldi Modified Takatori
5·0 Full adhesion 5·0 Full adhesion
Low adhesion Low adhesion

2·5 2·5
Acceleration: m/s2

Acceleration: m/s2
0 0

⫺2·5 ⫺2·5

⫺5·0 ⫺5·0
0 5 10 15 0 5 10 15 20 25
t: s t: s
(a)
4 4
Full adhesion Full adhesion
Low adhesion Low adhesion
3 3
θ: ⫻10⫺3 rad

θ: ⫻10⫺3 rad
2 2

1 SL 1

0 0
0 5 10 15 0 5 10 15 20 25
t: s t: s
(b)

Fig. 13. Effect of cohesive interface on earthquake response of 3.5 MW turbine on the D 20 m, L/D 0.5 caisson.
Comparison of: (a) acceleration–time history at the tower top and (b) foundation rotation–time history assuming both
perfect (Æ 1) and imperfect (Æ 0.3) soil–foundation interface. SL, serviceability limit

3·5 MW wind turbine 2·0 MW wind turbine

5·0 5·0
0·35g

2·5 2·5
Acceleration: m/s2

Acceleration: m/s2

0 0

⫺2·5 0·25g ⫺2·5

⫺5·0 ⫺5·0
0 5 10 15 0 5 10 15
t: s t: s
(a)

With environmental loads With environmental loads

2·0 2·0
Without environmental loads Without environmental loads
θ ⫺ θ0: ⫻10⫺3 rad

θ ⫺ θ0: ⫻10⫺3 rad

1·0 1·0

0 0

Zero residual rotation

⫺1·0 ⫺1·0
0 5 10 15 0 5 10 15
t: s t: s
(b)

Fig. 14. Illustration of the role of unidirectional environmental loads for a 3.5 MW turbine (on a suction caisson with
D 25 m, L/D 0.2) and for a 2 MW turbine (on a caisson of D 17 m and L/D 0.2): (a) acceleration–time history at the
nacelle level of each turbine and (b) foundation rotation–time history. Both turbines are excited by the modified Rinaldi
accelerogram
14 KOURKOULIS, LEKKAKIS, GELAGOTI AND KAYNIA
wind-induced moment that may prove important for the C damping coefficient of dashpots
foundation performance. D diameter of the caisson
drotor turbine diameter
E initial kinematic hardening modulus
Estat Young’s modulus under static loading
CONCLUSIONS EI bending stiffness of steel pipe section in turbine tower
This paper has investigated the response of a wind turbine model
founded on suction caissons subjected to monotonic lateral, F function
cyclic and earthquake loading while parametrically investi- FSM factor of safety against overturning moment
gating the role of soil–sidewall interface strength and soil G shear stiffness
non-uniformity. Consistent with previous studies, it was H/Hult horizontal load over horizontal capacity
found that in the non-uniform profile, shear zones tend to Hc height of centre of mass of turbine
form higher than in the uniform soil case, owing to the H0 height of the centre of mass of the nacelle measured from
lower available strength in shallower strata, resulting in the seabed
smaller settlements. k gradient of linearly increasing strength profile
L caisson length or depth of embedment
Consideration of imperfect interface conditions allows M/H moment over horizontal force ratio
sliding or even detachment of the caisson from the soil, thus R radius of the wind turbine tower
producing decreased foundation moment capacity. When su undrained shear strength
considering a low-adhesion interface scenario, the lower sum undrained shear strength at seabed level
shear resistance that may be offered on the peripheral side- su,0 shear strength at skirt tip level
walls results in an inverted scoop mechanism mobilised T period of turbine
within the skirts, thus producing larger plastic deformations t thickness of steel pipe section in turbine tower model
and slightly increased settlements, even when deeply em- tn total mass of the nacelle and the rotor
bedded caissons are considered. Under cyclic loading, the V/Vult vertical load over vertical capacity
V s shear wave velocity
moment–rotation loops at high amplitudes of imposed dis- wmax settlement
placement tend to develop a pinched shape due to gap Æ adhesion ratio
formation. The low sidewall adhesion combined with its ª9 submerged unit weight
inadequate depth give rise to an uplifting-dominated re- ªr parameter determining rate of decrease of kinematic
sponse of the low L/D caisson when subjected to high hardening with increasing plastic deformation
amplitudes of displacement. On the other hand the deeply Ł foundation rotation
embedded L/D ¼ 0.5 caisson responds through accumulation Łmax maximum rotation
of settlement. Ł0 elastic rotation under action of environmental loads
A preliminary comparison of two caisson alternatives, each º backstress parameter
demanding the same amount of steel to manufacture, has shown v Poisson ratio
 hysteretic damping ratio
that by increasing the foundation’s diameter while maintaining s value of (Rayleigh) damping for soil stratum
a low embedment ratio (D ¼ 25m, L/D ¼ 0.2), a more favour- t adopted tower damping
able response is achieved than by even substantially increasing  normal stress
the skirt length (i.e. from L/D ¼ 0.2 to L/D ¼ 0.5).  j0 stress at zero plastic strain
The response of the wind turbine to seismic shaking was  0 size of yield surface
assessed by subjecting it to a milder and a moderately strong  y maximum yield stress
earthquake scenario. It was shown that due to its large r material density of dashpot
flexibility, the superstructure is generally insensitive to
ground shaking. However, in terms of foundation response,
the caisson tends to accumulate rotations during each load-
ing cycle in both loading scenarios, resulting in significant REFERENCES
Abaqus (2008). ABAQUS: theory and analysis user’s manual,
residual rotation values. The latter may not be threatening
version 6.8-3. Providence, RI, USA: Dassault Systèmes Simulia
the safety of the structure but constitute irrecoverable distor- Corp.
tion that may question the serviceability of the turbine. Abaqus (2010). Standard user’s manual, Abaqus 6.10. Providence,
Although limited in population and therefore inadequate to RI, USA: Dassault Systèmes Simulia Corp.
culminate in quantitative outcomes, the results of the ana- Anastasopoulos, I., Gelagoti, F., Kourkoulis, R. & Gazetas, G.
lyses reveal the significance of properly accounting for (2012). Simplified constitutive model for simulation of cyclic
earthquake loading combined with the environmental loads response of shallow foundations: Validation against laboratory
acting on the turbine: contrary to the misconception appar- tests. J. Geotech. Geoenviron. Engng, ASCE 137, No. 12, 1154–
ently originating from evaluating the seismic behaviour on 1168.
Andersen, K. H. & Jostad, H. P. (1999). Foundation design of
the basis of inertial characteristics, it is proven that it is the
skirted foundations and anchors in clay. Proceedings of the 31st
system kinematics that ultimately govern the response. offshore technology conference, Houston, TX, paper OTC 10824.
Andersen, K. H. & Jostad, H. P. (2002). Shear strength along
outside wall of suction anchors in clay after installation. Pro-
ACKNOWLEDGEMENT ceedings of the 12th international offshore and polar engineer-
The financial support for this paper has been provided ing conference, Kitakyushu, Japan. Mountain View, CA, USA:
under the research project ‘DARE’, which is funded through International Society of Offshore and Polar Engineers (ISOPE).
the European Research Council’s (ERC) ‘IDEAS’ Pro- Bazeos, N., Hatzigeorgiou, G. D., Hondros, I. D., Karamaneas, H.,
gramme, in Support of Frontier Research – Advanced Grant, Karabalis, D. L. & Beskos, D. E. (2002). Static, seismic and
stability analyses of a prototype wind turbine steel tower. Engng
under contract/number ERC-2-9-AdG228254-DARE to Pro-
Structs 24, No. 8, 1015–1025.
fessor G. Gazetas. Bransby, M. F. & Randolph, M. F. (1998). Combined loading of
skirted foundations. Géotechnique 48, No. 5, 637–655, http://
dx.doi.org/10.1680/geot.1998.48.5.637.
NOTATION Bransby, M. F. & Yun, G. J. (2009). The undrained capacity of
A caisson base area skirted strip foundations under combined loading. Géotechnique
Ad effective area of dashpot 59, No. 2, 115–125, http://dx.doi.org/10.1680/geot.2007.00098.
 SUCTION CAISSON FOUNDATIONS FOR OFFSHORE WIND TURBINES 15
Byrne, B. W. & Houlsby, G. T. (2003). Foundations for offshore suction anchors: double-wall, over-consolidated clay, and layered
wind turbines. Phil. Trans. R. Soc. London 361, No. 1813, soil profile. Proceedings of the offshore technology conference,
2909–2930. Houston, TX, paper OTC 18007.
CEN (Comité Européen de Normalisation) (2004). EN 1998-1: Kelly, R. B., Houlsby, G. T. & Byrne, B. W. (2006). Transient
Eurocode 8: Design of structures for earthquake resistance – vertical loading of model suction caissons in a pressure cham-
Part 1: General rules, seismic actions and rules for buildings. ber. Géotechnique 56, No. 10, 665–675, http://dx.doi.org/
Brussels, Belgium: CEN. 10.1680/geot.2006.56.10.665.
Clukey, E. C. & Morrison, J. (1993). A centrifuge and analytical Lavassas, I., Nikolaidis, G., Zervas, P., Efthimiou, E., Doudoumis,
study to evaluate suction caissons for TLP applications in the I. N. & Baniotopoulos, C. C. (2003). Analysis and design of the
Gulf of Mexico. In Design and performance of deep founda- prototype of a steel 1-MW wind turbine tower. Engng Structs
tions: Piles and piers in soil and soft rock, ASCE Geotechnical 25, No. 8, 1097–1106.
Special Publication, vol. 38, pp. 141–156. Reston, VA, USA: Mana, D. S. K., Gourvenec, S. & Randolph, M. F. (2010).
American Society of Civil Engineers. A numerical study of the vertical bearing capacity of skirted
Colliat, J.-L., Boisard, P., Andersen, K. & Schroder, K. (1995). foundations. Proceedings of the 2nd international symposium
Caisson foundations as alternative anchors for permanent moor- on frontiers in offshore geotechnics (ISFOG), Perth, pp. 433–
ing of a Process Barge Offshore Congo. Proceedings of the 27th 438.
offshore technology conference, Houston, TX, paper OTC 7797, Mana, D., Gourvenec, S. & Martin, C. (2013). Critical skirt spacing
pp. 919–929. for shallow foundations under general loading. J. Geotech.
DNV (Det Norske Veritas) (2001). Guidelines for design of wind Geoenviron. Engng, ASCE 139, No. 9, 1554–1566.
turbines. Copenhagen, Denmark: Det Norske Veritas. Mello, J. R. C., Moretti, M. J., Sparrevik, P., Schroder, K. &
Gourvenec, S. (2007). Failure envelopes for offshore shallow foun- Hansen, S. B. (1998). PI9 and P26 moorings at the Marlim field.
dations under general loading. Géotechnique 57, No. 9, 715– The first permanent taut leg mooring with fibre rope and suction
728, http://dx.doi.org/10.1680/geot.2007.57.9.715. anchors. Proceedings of the FPS ’98 conference, London, UK,
Gourvenec, S. & Randolph, M. F. (2003). Effect of strength non- P19.
homogeneity on the shape of failure envelopes for combined Randolph, M. F. & House, A. R. (2002). Analysis of suction
loading of strip and circular foundations on clay. Géotechnique caisson capacity on clay. Proceedings of the annual offshore
53, No. 6, 575–586, http://dx.doi.org/10.1680/geot.2003.53.6.575. technology conference, Houston, TX, paper OTC 14236.
Gourvenec, S., Acosta-Martinez, H. E. & Randolph, M. F. (2009). Raptakis, D., Chávez-Garcı́a, F. J., Makra, K. & Pitilakis, K. (2000).
Experimental study of uplift resistance of shallow skirted foun- Site effects at EUROSEISTEST Part I. Determination of the
dations in clay under transient and sustained concentric loading. valley structure and confrontation of observations with 1D
Géotechnique 59, No. 6, 525–537, http://dx.doi.org/10.1680/ analysis. Soil Dynam. Earthquake Engng 19, No. 1, 1–22.
geot.2007.00108. Ritschel, U., Warnke, I., Kirchner, J. & Meussen, B. (2003). Wind
Haenler, M., Ritschel, U. & Warnke, I. (2006). Systematic model- turbines and earthquakes. Proceedings of the 2nd world wind
ling of wind turbine dynamics and earthquake loads on wind energy conference, Cape Town, South Africa. Bonn, Germany:
turbines. Proceedings of European wind energy conference and World Wind Energy Association (CD-ROM).
exhibition 2006, Athens, Greece, pp. 1–6. Brussels, Belgium: Tani, K. & Craig, W. H. (1995). Bearing capacity of circular
European Wind Energy Association. foundations on soft clay of strength increasing with depth. Soils
Houlsby, G. T. & Byrne, B. W. (2000). Suction caisson foundations Found. 35, No. 2, 37–47.
for offshore wind turbines and anemometer masts. Wind Engng Van der Tempel, J. (2006). Design of support structures for offshore
24, No. 4, 249–255. wind turbines. PhD thesis, Delft University of Technology, Delft,
Houlsby, G. T. & Byrne, B. W. (2005). Calculation procedures for the Netherlands.
installation of suction caissons in clay and other soils. Proc. Watson, P. G. & Randolph, M. F. (1997). Vertical capacity of
Instn Civil Engrs – Geotech. Engng 158, No. 2, 75–82. caisson foundations in calcareous sediments. Proceedings of the
Houlsby, G. T., Kelly, R. B., Huxtable, J. & Byrne, B. W. (2005). 7th international offshore and polar engineering conference,
Field trials of suction caissons in clay for offshore wind turbine Honolulu, vol. 1, pp. 784–790. Mountain View, CA, USA:
foundations. Géotechnique 55, No. 4, 287–296, http://dx.doi.org/ International Society of Offshore and Polar Engineers (ISOPE).
10.1680/geot.2005.55.4.287. Watson, P. G., Randolph, M. F. & Bransby, M. F. (2000). Combined
House, A. R. & Randolph, M. F. (2001). Installation and pull-out lateral and vertical loading of caisson foundations. Proceedings
capacity of stiffened suction caissons in cohesive sediments. of the offshore technology conference, Houston, TX, paper OTC
Proceedings of the 11th international offshore and polar engi- 12195.
neering conference, Stavanger, Norway, vol. 2. Mountain View, Witcher, D. (2005). Seismic analysis of wind turbines in the time
CA, USA: International Society of Offshore and Polar Engineers domain. Wind Energy 8, No. 1, 81–91.
(ISOPE). Yun, G. & Bransby, M. F. (2007). The undrained vertical bearing
IEC (International Electrotechnical Commission) (2005). IEC capacity of skirted foundations. Soils Found. 47, No. 3, 493–
61400-1: Wind turbines – Part 1: design requirements. Geneva, 505.
Switzerland: IEC. Zhao, X. & Maisser, P. (2006). Seismic response analysis of wind
Jeanjean, P., Znidarcic, D., Phillips, R., Ko, H.-Y., Pfister, S., turbine towers including soil–structure interaction. Proc. Instn
Cinicioglu, O., & Schroeder, K. (2006). Centrifuge testing on Mech. Engrs, Part K: J. Multi-Body Dynam. 220, No. 1, 53–61.