Engineering Structures 75 (2014) 388–401

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Effects of soil–structure interaction on real time dynamic response

of offshore wind turbines on monopiles
M. Damgaard a,⇑, V. Zania b, L.V Andersen c, L.B. Ibsen c
a
Technology and Engineering Solutions, Vestas Wind Systems A/S, Hedeager 42, 8200 Aarhus, Denmark
b
Department of Civil Engineering, Technical University of Denmark, 2800 Lyngby, Denmark
c
Department of Civil Engineering, Aalborg University, Sohngårdsholmsvej 57, 9000 Aalborg, Denmark

a r t i c l e i n f o a b s t r a c t

Article history: Offshore wind turbines are highly dynamically loaded structures, their response being dominated by the
Received 26 December 2013 interrelation effects between the turbine and the support structure. Since the dynamic response of wind
Revised 2 June 2014 turbine structures occurs in a frequency range close to the excitation frequencies related to environmen-
Accepted 3 June 2014
tal and parametric harmonic loads, the effects of the support structure and the subsoil on the natural
Available online 2 July 2014
vibration characteristics of the turbine have to be taken into account during the dynamic simulation of
the structural response in order to ensure reliable and cost-effective designs. In this paper, a computa-
Keywords:
tionally efficient modelling approach of including the dynamic soil–structure interaction into aeroelastic
Aeroelasticity
Analytical solution
codes is presented with focus on monopile foundations. Semi-analytical frequency-domain solutions are
Dynamic soil–structure interaction applied to evaluate the dynamic impedance functions of the soil–pile system at a number of discrete fre-
HAWC2 quencies. Based on a general and very stable fitting algorithm, a consistent lumped-parameter model of
Lumped-parameter model optimal order is calibrated to the impedance functions and implemented into the aeroelastic nonlinear
Offshore wind turbine foundation multi-body code HAWC2 to facilitate the time domain analysis of a wind turbine under normal operating
mode. The aeroelastic response is evaluated for three different foundation conditions, i.e. apparent fixity
length, the consistent lumped-parameter model and fixed support at the seabed. The effect of soil–struc-
ture interaction is shown to be critical for the design, estimated in terms of the fatigue damage 1 Hz
equivalent moment at the seabed. In addition, simplified foundation modelling approaches are only able
to capture the dynamic response reasonably well after tuning of the first natural frequency and damping
within the first mode to those of the integrated model. Nevertheless, significant loss of accuracy of the
modal parameters related to the second tower modes is observed.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction from the self-weight of the rotor, nacelle, tower and mean wind,
(b) stochastic loads from the wind turbulence and the irregular
As a consequence of the increasing reliability of offshore wind sea states, (c) transient loads from the start, stop and emergency
technology, wind energy has become a competitive source of breakdown procedures, and finally (d) cyclic loads from the rotor
renewable energy. Currently, turbines with rotor diameters and frequency 1P generated by mass imbalance in the blades and the
tower heights of more than 100 m are in production. Optimisation 3P frequency due to shadowing effects from the wind each time
of the turbine blades and the towers leads to slender and therefore a blade passes the tower. The operational speed of the rotor of
extremely flexible structures. Consequently, the first modes of res- state-of-the-art wind turbines is typically about 7–12 rounds per
onance of the total structure, including the foundation, tower, hub, minute (RPM), corresponding to a 1P frequency of 0.12–0.20 Hz.
nacelle and blades, are close to the excitation frequencies related to In addition, waves related to sea states with high rate of concur-
environmental loads from wind and waves. Thus, a modern wind rency typically have wave frequencies of 0.10–0.20 Hz [1]. There-
turbine may undergo large deformations, not only during extreme fore, in order to avoid large–amplitude stress variations in the
weather conditions but also during power production. In general, wind turbine structure, three design philosophies are defined and
offshore wind turbines are exposed to: (a) steady quasi-static loads illustrated in Fig. 1a [2]:

 Soft–soft design, where the resonance frequency of the turbine

⇑ Corresponding author. Tel.: +45 41188639. is lower than the rotor frequency 1P and the frequencies related
E-mail address: [email protected] (M. Damgaard). to the dominant wave actions.

http://dx.doi.org/10.1016/j.engstruct.2014.06.006
0141-0296/Ó 2014 Elsevier Ltd. All rights reserved.
 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401 389

Nomenclature

Scalars: pn polynomial coefficients of P(ix)

An residues corresponding to the real poles sn Q(ix) polynomial and denominator of the rational filter Z^ r ðixÞ
a0n real-valued coefficient in a partial fraction-form qp density of monopile
a1n real-valued coefficient in a partial fraction-form qs density of soil
b0n real-valued coefficient in a partial fraction-form q0 density of cylinder
b1n real-valued coefficient in a partial fraction-form r0 radius of cylinder
c1 real-valued constant selected so that the singular part of rp radius of monopile
the impedance function provides the entire stiffness in . real-valued coefficient for a discrete-element model
the high-frequency range 1 heuristic parameter
d1 modal damping in terms of the logarithmic decrement sn N complex roots of Q(ix)
related to the lowest eigenmode sn N complex conjugated pairs of 2N roots of Q(ix)
Dj moment width of the jth load cycle nf,j Sij(x) dynamic impedance function
e real-valued parameter for weighted least-squares fitting S0ij ðxÞ static stiffness component of the dynamic impedance
Ep Young’s modulus of monopile function Sij(x)
f physical frequency 3P harmonic frequency equal to three times the rotor fre-
fc frequency constant quency
f1 undamped eigenfrequency related to the lowest eigen- tp thickness of monopile
mode t time
F object function T simulation time
G constraint for weighted least-squares fitting # real-valued parameter for weighted least-squares fitting
c real-valued coefficient for a discrete-element model t0 wave velocity of the ground
H depth of soil layer ts shear wave velocity of soil
j real-valued coefficient for a discrete-element model vmean normal turbulent longitudinal mean wind speed
k1 real-valued constant selected so that the singular part of x circular frequency
the impedance function provides the entire stiffness in w weight function
the high-frequency range Zij(x) normalised dynamic impedance component
Lp length of monopile Zs(x) singular part of the normalised dynamic impedance
L0 height of cylinder component Z(x)
Meq,1 fore–aft fatigue damage 1 Hz equivalent moment Zr(x) regular part of the normalised dynamic impedance
Meq,2 side–side fatigue damage 1 Hz equivalent moment component Z(x)
M1 fore–aft moment as a function of time t Z^ r ðixÞ rational filter
M2 side–side moment as a function of time t fs hysteretic soil damping
m Wöhler exponent
M order of the rational filter Z^ r ðixÞ
Vectors and matrices:
nfj jth identified load cycle
r complex amplitude vector of forces Qi and moments Mi
mp Poisson’s ratio of monopile
S impedance matrix
ms Poisson’s ratio of soil
u complex amplitude vector of translational degrees of
1P harmonic frequency equal to the rotor frequency
freedom Vi and rotational degrees of freedom Hi
P(ix) polynomial and numerator of the rational filter Z^ r ðixÞ
p(t) pulse load

 Soft–stiff design, where the resonance frequency is in the range The soft–stiff design is often chosen, since the high wave load-
between the rotor frequency 1P and the blade passing fre- ing may induce wave fatigue for the soft–soft design and a stiff
quency 3P. foundation is required for a stiff–stiff design leading to a costly
 Stiff–stiff design, where the resonance frequency is higher than design. In order to identify potential sources of resonance and
the blade passing frequency 3P. the safe region for the soft–stiff design, a Campbell diagram is

Fig. 1. Design approaches for an offshore wind turbine: (a) 1P and 3P harmonic excitation (after [2]), (b) sparse Campbell diagram for a soft–stiff design (after [3]).
390 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401

useful in order to illustrate the relation between resonance fre- by several studies carried out in the literature the past three dec-
quency and excitations, see Fig. 1b. As indicated, the intersection ades [21–36]. When it comes to analytical solutions of pile vibra-
of the eigenfrequency and excitations take place outside the oper- tion to harmonic loading, two approaches of modelling the soil
ational range meaning that no significant resonance is expected. It response have prevailed: (a) linear, viscoelastic continuum [21–
should be noted that during a start-up sequence, resonance will 24] and (b) linear Winkler type medium [25–27], where the soil
occur for very limited time. is considered as a horizontally layered stratum via vertically dis-
Mostly, the gravitational and monopile foundation concepts tributed and independent linear springs [27]. Following the
have been employed so far for offshore wind turbines. However, approach of a beam on a nonlinear Winkler foundation, originally
foundation concepts such as jacket structures, tripods and bucket formulated by Winkler [27], several studies have suggested soil–
foundations might be suitable for wind farms located in deeper pile interaction models that account for soil nonlinear response
water. Nevertheless, independent of the foundation type, offshore [28–33]]. The aforementioned models increase in complexity as
wind turbine structures are prone to dynamic excitations, where they may account for different energy dissipation conditions for
the dynamic soil–structure interaction strongly affects the modal the near and far soil fields [29,30] or separation and gap formation
parameters of the structure [4]. Since a range of structural frequen- between the soil and the pile [32,33]. The near field element
cies is excited by stochastic loads with varying amounts of energy, usually includes also the effect of gaping and friction along the
special demands must be put on the computational modelling of soil–pile interface, while soil plastic behaviour, hysteretic and radi-
complex wind turbine foundations for real-time simulations. It ation damping is introduced by the aforementioned static or cyclic
may not seem sufficient to calibrate the fixity depth and the stiff- p–y curves and dashpots.
ness of a monopile foundation to the lowest natural frequency of For more advanced investigations of the dynamic soil–pile
the wind turbine structure (a common practice in the current interaction, the finite-element method (FEM) coupled with infinite
design procedure [5]), since the subsoil has a significant impact elements and the boundary-element method (BEM) are useful
on the impedance terms of the soil–foundation system in the fre- [34–37]. The dynamic impedances estimated according to the
quency range relevant for wind turbines. Moreover, the aeroelastic abovementioned studies are limited to linear elastic soil response.
multibody code HAWC2 [6] offers the inclusion of soil–pile interac- Capturing the nonlinear dynamic response of a soil–pile system in
tion using p–y and t–z curves as proposed by the design guidelines a 3D numerical model requires the prior application of adequate
[7]. However, the approach may be questioned, since the genera- constitutive models [38,39]. However this advanced computational
tion of soil damping is incorporated in a simple manner via fre- effort may not be justified for offshore monopile structures, since
quency-dependent Rayleigh damping that requires knowledge of the cyclic strain level in the subsoil is in the elastic range for nor-
the modal parameters prior to the simulation. Furthermore, this mal operating wind turbines. Based on a three-dimensional quasi-
type of modelling neglects the contribution of the soil mass to static analysis of a gravity base foundation carried out in ABAQUS
the dynamic response. [40], previous work by the authors indicates that even for an emer-
The design of offshore wind turbines requires the computation gency shutdown of an onshore wind turbine, the maximum cyclic
and analysis of thousands of load cases [8]. Hence, computation shear strain is so low that a viscoelastic model represents the soil
speed is of paramount importance and the ideal approach of behaviour reasonably. Hence, the analytical solutions seem an
including the interrelation effects between wind field, wind tur- adequate alternative even though their application is limited to a
bine, foundation and subsoil in one closed software and solve the linear viscoelastic soil response.
entire system of equations is not feasible. Some few attempts using This in turn led the authors to derive an alternative approach of
full integrated models are given by Vorpahl et al. [9] and Böker including the dynamic soil–pile interaction, in which the geomet-
[10]. The last-mentioned used a shared variable interface to couple rical damping by wave radiation into the subsoil is included. Based
a linear elastic finite-element (FE) model of a lattice offshore sup- on a uniform, linearly elastic monopile perfectly attached to a lin-
port structure with the aeroelastic code FLEX5 [11], where no soil– ear viscoelastic soil layer overlying rigid bedrock, dynamic imped-
foundation interaction was considered. To improve the numerical ance functions of the soil–pile system are derived that facilitate the
efficiency of the time domain simulations and ensure that the calibration of a consistent LPM. An appropriate set of constraint
foundation only adds few degrees of freedom to the model of the equations, which describes the relationship between the degrees
wind turbine, different kinds of modelling or reduction strategies of freedom in the LPM and the aeroelastic code HAWC2, is formu-
can be applied. For a fully coupled approach [12–14], where the lated. This makes it possible to evaluate the dynamic vibration
entire system of equation for the foundation and turbine is solved characteristics of a 5.0 MW offshore wind turbine for different
in the aeroelastic code, static condensation methods [15], classical environmental conditions and foundation properties. The novelty
spatial reduction methods [16], cone models or lumped-parameter of the present study is exemplified by comparing the dynamic
models (LPMs) [17] may be useful. Whereas static or modal con- response between the present method (coupled LPM) and time
densation methods follow the principle of reducing the global domain simulations of the investigated wind turbine with: (a)
mass, damping and stiffness matrices from a rigorous FE model, fixed boundary conditions at seabed (reference model) and (b)
an LPM is based on a parallel coupling of so-called discrete- fixed boundary conditions at a given depth called apparent fixity
element models with few degrees of freedom fitted directly to length (AFL) that is calibrated to fit the eigenfrequency and the
the frequency response of a rigorous FE or boundary-element damping related to the lowest eigenmode of the wind turbine
(BE) model. Based on transfer matrices for a layered half-space, structure (equivalent model), cf. Fig. 2. Hence, the importance of
originally proposed by Haskell [18] and Thomson [19], Damgaard including dynamic soil–pile interaction is evaluated in this paper
et al. [20] calibrated a consistent LPM and performed successful along with the impact of the adopted foundation modelling
aeroelastic simulations in HAWC2 of an offshore wind turbine approach on the dynamic response of offshore wind turbines.
installed on a surface foundation. The main objective of the paper Following the introduction, Section 2 contains a general and
was to establish a method for time efficient calibration of an LPM very stable fitting algorithm for the construction of consistent
that could be applied in aeroelastic codes for reliable evaluation LPMs. Further, a short introduction to semi-analytical solutions
of the dynamic structural response of wind turbines. for piles in horizontal, vertical and torsional vibrations is pre-
In the case of soil–pile systems, the dynamic impedances sented. In this regard, a three–dimensional coupled BE/FE model
required for the calibration of LPMs can be established by the [41] is applied and compared with the semi-analytical dynamic
application of analytical or numerical methodologies as proposed impedance functions for horizontal pile vibrations. Section 3 con-
 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401 391

amplitudes of forces and moments r in the coordinate system

(x1, x2, x3) as
3 2
2 32 3
Q1 S22 0 0 0 S24 0 V1
6Q 7 6 0 S22 0 S24 0 0 76 V 7
6 27 6 76 2 7
6 7 6 76 7
6Q 7 6 0 0 S33 0 0 0 76 V 3 7
6 37 6 76 7
rðxÞ ¼ SðxÞuðxÞ ) 6 7¼6 76 7 ð1Þ
6 M1 7 6 0 S24 0 S44 0 0 76 H1 7
6 7 6 76 7
6 7 6 76 7
4 M 2 5 4 S24 0 0 0 S44 0 54 H2 5
M3 0 0 0 0 0 S66 H3

Each component of the impedance matrix S may be expressed as

~Sij ðxÞ ¼ S0 Z ij ðxÞ, where S0 ¼ Sij ð0Þ is the static stiffness component
ij ij
and Zij(x) is the normalised dynamic impedance component. Omit-
ting the indices i and j, the normalised dynamic impedance compo-
nent Zij(x) is decomposed into a singular part Zs(x) and a regular
part Zr(x) as suggested by [17], i.e.,
ZðxÞ ¼ Z s ðxÞ þ Z r ðxÞ; ð2Þ
where Zs(x) and Zr(x) are given by
Fig. 2. Aeroleastic models with different foundation approaches: (a) reference
model, (b) coupled LPM, (c) equivalent model. 1 PðixÞ
Z s ðxÞ ¼ k þ ixc1 Z r ðxÞ  Z^ r ðixÞ ¼ ; ð3aÞ
QðixÞ
As indicated in Eq. (3a), the regular part Zr(x) is obtained by fitting
siders the implementation of the consistent LPMs in the aeroelastic
a rational filter to the result
code HAWC2 with special focus on the importance of including the
soil–pile interaction. Finally, the general conclusions and remarks Z r ðxÞ ¼ ~SðxÞ=S0  Z s ðxÞ; ð3bÞ
on the dynamic response of the wind turbine are given in Section 4.
where ~SðxÞ is the impedance function obtained by a rigorous solu-
tion and P(ix) and Q(ix) are two polynomials advantageously writ-
ten as
2. Rational approximation of dynamic impedance functions
1 2 M1
PðixÞ ¼ 1  k þ p1 ðixÞ þ p2 ðixÞ þ . . . þ pM1 ðixÞ ; ð3cÞ
For the dynamic analysis of offshore monopiles placed in a stra-
tum, the soil flexibility is important to investigate, since it modifies Y
N Y
MN

the stiffness of the foundation system and generates energy QðixÞ ¼ ðix  sn Þðix  sn Þ  ðix  sn Þ; ð3dÞ
n¼1 n¼Nþ1
dissipation through wave radiation and material damping. This
effect can be taken into account by a set of complex frequency- where M is the order of the rational filter Z^ r ðixÞ and sn, n = 1, 2, . . .,
dependent impedance functions that relate harmonic excitation N, are N complex roots of Q(ix) with sn defining their complex con-
forces and resulting displacements at the pile head. The real part jugates. The remaining M  2N roots are real. As indicated in Eq. (3),
of the dynamic impedance functions represents the restoring stiff- the rational approximation Z ^ r ðixÞ ! 0 for x ? 1, and Z ^ r ð0Þ ! 1.
ness, whereas the imaginary part represents the damping. The Thereby, S0Zs(x) represents the entire dynamic impedance in the
value of the last-mentioned strongly depends on the excitation fre- high-frequency limit. Even though the highest content of energy
quency, since a cutoff frequency (eigenfrequency of the soil layer) related to wind and wave loads typically is below 1 Hz, an accu-
exists, below which, no radiation damping takes place [21,36]. rately determination of S0Zs(x) is needed. Breaking waves or emer-
In contrast to the case of a surface footing on a homogeneous or gency stops may excite a wide range of structural frequencies and
layered ground, a significant coupling between horizontal sliding evidently, it must be ensured that the consistent LPM is numerical
and rocking exists for embedded foundations, see for instance stable in the high-frequency range providing a physical correct
[42,43]. Therefore, an LPM that properly simulates the impedance behaviour. Finally, it should be noted that the order M of the
characteristics of a monopile structure should also account for rational approximation Z ^ r ðixÞ must be high enough to ensure an
the coupling of sliding–rocking-vibration modes. In this paper, accurate fit to the target solution Z r ðxÞ ¼ ~SðxÞ=S0  Z s ðxÞ and low
the computational model of a wind turbine installed on a monopile enough to avoid wiggling in the relatively high frequency range.
consists of two parts coupled to achieve the calculation of the inte- By utilising a weighted least squares approach with the weight
grated response: an aeroelastic model of the wind turbine and an function w(x), the rational approximation Z ^ r ðixÞ is fitted to the
LPM of the monopile structure that reproduces the dynamic soil– target solution Zr(x). In this regard, the polynomial coefficients
pile interaction. The formulation is based on three steps: pn and the poles sm are identified as the optimisation variables,
i.e.,
1. Rigorous semi-analytical frequency-domain solutions are used
X
J
2
to evaluate the dynamic impedance functions for a monopile Fðpn ; sm Þ ¼ wðxj ÞðZ^ r ðixj Þ  Z r ðxj ÞÞ : ð4Þ
at a number of discrete frequencies. j¼1
2. A rational filter is fitted to the semi-analytical impedance
functions and a consistent LPM is calibrated. The object function in Eq. (4) is minimised with respect to one
^ r ð0Þ ¼ 0,
equality constraint which ensures that Z
3. A coupling between the wind turbine represented by an
aeroelastic model in HAWC2 and the LPM accounting for the Y
N Y
MN
soil–pile interaction is established. G0 ¼ 1  ðsn Þðsn Þ  ðsn Þ ¼ 0; ð5aÞ
n¼1 n¼Nþ1

In general, an axisymmetric monopile has three translational and M inequality constraints,

and three rotational degrees of freedom. In the frequency domain,
these are related to the impedance matrix S and the complex Gk ¼ Rðsk Þ þ e < 0 k ¼ 1; 2; . . . ; N; ð5bÞ
392 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401

Gk ¼ sk þ e < 0 k ¼ N þ 1; 2; . . . ; M  N; ð5cÞ 2.2.1. Dynamic impedance for piles in horizontal vibrations

Gj ¼ #Rðsk Þ þ Iðsk Þ < 0 j ¼ M  N þ k; k ¼ 1; 2; . . . ; N; ð5dÞ In the early pioneering studies by Nogami and Novak [21] and
Nogami and Novak [22] rigorous semi-analytical solutions were
where R(sk) and I(sk) are the real and the imaginary parts of the derived and presented of the interaction between a linear, visco-
complex roots sn. #  10–100 and e  0.01 are two real-valued elastic soil layer overlying rigid bedrock (fixed at the bottom) with
parameters. Eqs. (5b) and (5c) are introduced, since a stable solution hysteretic material damping and a vertical, uniform and linearly
^ r ðixÞ all reside in the second and third
only is obtained if the poles of Z elastic pile vibrating horizontally. Even though no displacements
quadrants of the complex plane, i.e. the real parts of the poles must all occur at the bottom of the soil layer and the pile is clamped or pin-
be negative. In addition, Eq. (5d) prevents the imaginary parts of the ned at the tip, these assumptions may seem reasonable for many
complex poles to become much greater than the real parts, since R offshore wind turbine locations, where the lower part of the mono-
(sn) = 0 will lead to a second-order pole I(sn)2 that is real and positive. pile is placed in a relatively stiff soil layer that prevents pile deflec-
Evidently, this will cause instability in the time-domain. tions during normal turbine operation. In addition, it must be
Now, recasting the rational approximation, given by Eq. (3), into highlighted that the semi-analytical solutions of the dynamic
a partial-fraction form leads to physical insight of the curve fitting. impedance functions are limited to soil profiles with constant stiff-
With N complex conjugate pairs of poles in Z ^ r ðixÞ; the total
ness along the depth, nonlinear soil behaviour is disregarded, and
approximation of the dynamic impedance coefficient Z(x) can be the separation and the sliding along the soil–pile interface is not
written as accounted for. Notice that the viscoelastic soil properties may be
X
N determined by standard geotechnical and geophysical exploration
^ xÞ ¼ k1 þ ixc1 þ b0n þ b1n ix
Zði based on low-strain field and laboratory tests [44].
a
n¼1 0n þ a1n ix þ ðia0 Þ2 As an example, the normalised dynamic impedance coefficients
X
MN
An Zij(f) for a clamped horizontally vibrating monopile with the pile
þ ; ð6aÞ length Lp = 30 m, pile radius rp = 3 m, pile thickness tp = 0.06 m,
n¼Nþ1
i a0  sn
Young’s modulus Ep = 210 GPa, Poisson’s ratio mp = 0.25 and mass
where An are the residues corresponding to the real poles sn, density qp = 7850 kg/m3 embedded in a homogeneous viscoelastic
n ¼ N þ 1; N þ 2; . . . ; M  N. The real coefficients a0n, a1n, b0n and soil layer with the depth H = 30 m, Poisson’s ratio ms = 0.35, mass
b1n are given by density qs = 1700 kg/m3 and hysteretic damping ratio fs = 0.025
are shown in Fig. 4 in terms of their magnitudes and phase angles.
a0n ¼ Rðsn Þ2 þ Iðsn Þ2 ; a1n ¼ 2Rðsn Þ; ð6bÞ For the physical frequencies f relevant for wind turbine vibrations,
the dynamic impedance functions are compared with a three-
b0n ¼ 2RðAn ÞRðsn Þ þ 2IðAn ÞIðsn Þ; b1n ¼ 2RðAn Þ: ð6cÞ dimensional coupled BE/FE model [41] for two different shear
wave velocities, ts = 100 m/s and ts = 400 m/s. The BE/FE model
According to Eq. (6a), the total approximation of the dynamic consists of a linear elastic FE section of the monopile utilising shell
impedance consists of a constant/linear term, M  2N first-order elements and a BE domain inside and outside the pile. In the FE
terms and N second-order terms. These terms may be interpreted region only half the model needs to be analysed when a plane of
as frequency–response functions for so-called discrete-element symmetry exists. The degrees of freedom in the plane of symmetry
models including frequency independent spring and damping con- are simply eliminated in the system of equations in order to satisfy
stants as well as point masses [17]. Since as many complex conju- the conditions at the interface between the modelled and non-
gate pairs as possible reduce the number of internal degrees of modelled part. As discussed in details by Andersen and Jones
freedom in the consistent LPM to a minimum for a given order [45], the procedure for introducing a plane of symmetry in the
M, N = M/2 has been used in the current study. Fig. 3 shows the BE region is complex, hence further description of the implementa-
discrete-element model of the constant/linear term and the tion in the code will not be provided here. For details, the inter-
second-order terms. ested reader should refer to the work performed by Andersen
[46], where a verification of the BE/FE model is also available.
2.1. Rigorous semi-analytical solutions for soil–pile interaction The numerical model of the monopile and subsoil contains 105
finite elements and 342 boundary elements, and the mesh of the
Several rigorous expressions for the impedance functions of the free surface has been truncated at a distance of 80 m. Based on a
soil–pile system exist in the literature. However, to the best of the horizontal nodal force and two opposite vertical nodal forces
author’s knowledge none of these expressions have been imple- applied at the pile cap for generating a rocking moment for each
mented into aeroelastic codes in order to evaluate the fully coupled frequency of interest, the complex displacements and rotations
dynamic vibration response of offshore wind turbines installed on are computed and used to evaluate the dynamic impedance func-
monopiles. In the following, the dynamic impedance functions in tions. In addition to this, it should be noted that the computation
horizontal, vertical and torsional vibration, which are employed time per frequency on an i5 3.2 GHz computer is around 17 min
in the current study for the calibration of the consistent LPMs, or roughly 18 h for 64 frequencies. Finally, in order to account
are introduced. for the soil columns inside the pile, an equivalent bending stiffness
and mass of the pile have been used for the analytical derivations
of the dynamic impedance functions. As indicated in Fig. 4, a rea-
sonable fit is observed between the two methods—especially for
ts = 100 m/s, where the evolution of progressive waves above the
resonance with the soil layer is well described. However, indepen-
dent of the shear wave velocity ts, a slightly higher resonance fre-
quency is obtained by the coupled BE/FE model. Contrary to the BE/
FE model, in which the resonances of the soil layer is altered by the
pile rigidity, the soil response is not affected by the pile stiffness in
Fig. 3. Discrete-element models: (a) constant/linear term, (b) alternative second- the semi-analytical solution. Hence, the observed resonance fre-
order term with one internal degree of freedom. quencies in Fig. 4 for the semi-analytical solution correspond to
 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401 393

Fig. 4. Normalised dynamic impedance coefficients for horizontal pile vibrations obtained by a rigorous solution proposed by Nogami and Novak [21] and Nogami and Novak
[22] and a three-dimensional coupled BE/FE model. A homogeneous viscoelastic soil layer with the shear wave velocity ts = 100 m/s and ts = 400 m/s is considered.

the resonance frequency of the lowest mode for a single soil layer model, while the wave propagation in the enclosed part of the soil
Hx/ts = p(2n  1)/2. Moreover, the vertical soil displacements are also modelled. Nevertheless, despite of discrepancies between
induced by horizontal vibrations are considered in the BE/FE the two approaches, the rigorous semi-analytical solution for hor-
394 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401

izontal pile vibrations still provides a reliable estimate of the for torsional vibrations ~
S66 according to the coupled BE/FE model
dynamic soil–pile interaction capable of capturing radiation effects shows almost identical values in the physical frequency range
and resonance/anti-resonance phenomena. f e [0;8] Hz. Therefore, instead of fitting a consistent LPM to the
non-varying normalised dynamic impedance coefficient Z66(f), cf.
Fig. 5, the static torsional stiffness S066 has been used hereafter.
2.1.2. Dynamic impedance for vertical and torsional vibrations Keep in mind that this is only possible, since the torsional motion
h3(t) as well as the vertical motion v3(t) are decoupled from the
In order to obtain a realistic vertical and torsional vibration remaining degrees of freedom according to Eq. (1).
response of an offshore wind turbine, the dynamic soil–structure
interaction for these degrees of freedom need to be identified
and used for the calibration of a consistent LPM. Based on a verti- 2.2. Validation of rational approximation
cal, elastic end bearing monopile embedded in linearly viscoelastic
soil layer with hysteretic material damping overlying rigid bed- Before the LPM is implemented into the aeroelastic code, it
rock, the vertical impedance function ~ S33 is found according to must be ensured that it behaves correctly in the frequency and
Nogami and Novak [23] in a similar way as described for the hor- time domain compared to the target solution. Based on two differ-
izontal pile vibrations. Fig. 5 shows the comparison of the norma- ent linearly viscoelastic soil layers overlying a rigid bedrock, Soil 1
lised dynamic impedance coefficient Z33(f) between the semi- and Soil 2, the consistent LPM is studied in terms of its ability to
analytical solution and the BE/FE approach as described in the pre- reproduce the dynamic impedance functions ~ Sij of a uniform mono-
vious section for identical soil conditions. As indicated reasonable pile with length Lp = 30 m, radius rp = 3 m and thickness
agreement is obtained for both soil types. A comparison between tp = 0.06 m. The material properties of the two soil profiles and
the static torsional stiffness S066 (equivalent stiffness of pile and soil the monopile are listed in Table 1. Fig. 6 shows the target solution
column inside the pile [47]) and the dynamic impedance function of the normalised impedance coefficients Z33(f), Z22(f), Z44(f) and

Fig. 5. Normalised dynamic impedance coefficients for vertical pile vibrations obtained by a rigorous solution proposed by Nogami and Novak [23] and a three-dimensional
coupled BE/FE model. In addition, normalised dynamic impedance function for torsional pile vibrations obtained by a three-dimensional coupled BE/FE model are shown. A
homogeneous viscoelastic soil layer with the shear wave velocity ts = 100 m/s and ts = 400 m/s is considered.
 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401 395

Table 1
Material and geometrical properties of subsoil and monopile used in the computation.

Material () Type () ts (m/s) m () q (kg/m3) f () H (m) Lp (m) rp (m) tp (m)
Soil 1 Sand 100 0.35 1700 0.05 30 – – –
Soil 2 Sand 250 0.35 1700 0.03 30 – – –
Monopile Steel 3271 0.25 7850 0 – 30 3 0.06

Fig. 6. Normalised dynamic impedance coefficients for the embedded monopile in a sandy soil layer according to Soil 1, cf. Table 1, obtained by rigorous semi-analytical
expressions and LPM with a filter order of M = 6.

Z24(f) in terms of their magnitude and phase angle as a function of 1

wðf Þ ¼ 1 13 : ð7Þ
the physical frequency f. As indicated by Fig. 6, a consistent LPM ð1 þ ð11 f Þ 2 Þ
with filter order M = 6 provides a reasonable fit for Soil 1. Since
the LPM is based on second-order discrete-element models with As indicated in Fig. 6, a high weight on the squared errors is applied
a point mass, see Fig. 3b, the LPM for each individual component in the frequency range before resonance with the soil layer occurs.
of the normalised impedance matrix Z has three internal degrees Since the target solutions of the dynamic impedances are based on
of freedom. Numerical experiments have shown that the weight hysteretic damping, they provide a complex stiffness in the static
function w(f) has a strong influence on the fit of Z ^ r ðif Þ to Zr(f) in limit, i.e. for x ? 0. This is in contrast to the LPM that contains vis-
the low-frequency range. Weights in the order of w = 103–105 at cous dashpots, which makes the choice of the heuristic parameters
low frequencies and unit weight at higher frequencies, as sug- 11 , 12 and 13 very important. 11 ¼ 7, 12 ¼ 40 and f3 = 0.1 have been
gested by Wolf [17], simply not allow for a reasonable fit in all sit- used for the vertical impedance term, 11 ¼ 65, 12 ¼ 45 and
uations. Therefore, a fairly general weight function with heuristic 13 ¼ 0:07 for the horizontal term, 11 ¼ 61, 12 ¼ 45 and 13 ¼ 0:08
parameters 11 , 12 and 13 is introduced: for the rocking term and 11 ¼ 20, 12 ¼ 45 and 13 ¼ 0:08 for the cou-
396 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401

pling term. A slight change in the heuristic parameters has been Newmark b-scheme [48]. Note that this is possible, since the global
made for the LPM fitting for Soil 2 that contains a higher shear wave mass, stiffness and damping matrices of the soil-pile system are
velocity ts and a lower hysteretic damping ratio fs, cf. Table 1. Here, easily derived from the LPM that consists of a parallel coupling
11 ¼ 2; 12 ¼ 8 and 13 ¼ 1 have been used for the vertical impedance, of discrete-element models. For both soil layers, the LPMs provide
11 ¼ 20; 12 ¼ 40 and 13 ¼ 0:08 for the horizontal term, useful results, i.e. the prediction of the maximum response as well
11 ¼ 4; 12 ¼ 40 and 13 ¼ 0:1 for the rocking and coupling terms as as the geometrical damping is captured in a reasonable manner.
shown in Fig. 7. Compared to Fig. 6, resonance with the soil layer However, for both soil layers, a small phase lag between the LPMs
for Soil 2 occurs at a higher frequency in Fig. 7. Nevertheless, a rea- and the target solution achieved by inverse Fourier transformation
sonable fit is obtained with a filter order of M = 6. The best filter of the dynamic impedance function for horizontal vibrations ~ S22 is
approximation is obtained for the rocking impedance, while only present. A higher filter order M seems not to change this discrep-
a slight deviation is observed in the horizontal impedance at higher ancy and may be caused by the fact that a high weight on the
frequencies. squared errors is applied in the low-frequency range. It is finally
The LPMs are further tested in the time domain by comparing noted that the off-diagonal terms that enter the impedance matrix
the response to a transient load with the results obtained by S according to Eq. (1) cannot be disregarded in the development of
inverse Fourier transformation of the target solutions ~ Sij carried the LPMs. This is clearly shown in Fig. 8, where an extremely poor
out in the frequency domain. A pulse load is applied in the form representation is obtained when the sliding–rocking coupling is
p(t) = sin(2pfct)sin(pfct/2) for 0 < t < 2/fc and p(t) = 0 otherwise. A neglected.
frequency constant fc = 2.5 Hz is utilised. The time histories of the Table 2 shows the RMS error between the target solutions and the
load and the response for Soil 1 and Soil 2 are given in Fig. 8. A rigid LPMs for each degree of freedom for the frequency constant fc = 2.5 -
cylinder with a mass density q0 = 2000 kg/m3, a radius r0 = 3 m and Hz. As indicated, the error is low for both soil layers. Evidently, the
a height L0 = 58 m has been put on the top of the monopile, and the highest RMS error is obtained for the rotational degree of freedom
response has been integrated over time by application of the h1(t) with no coupling between sliding and rocking. The LPMs have

Fig. 7. Normalised dynamic impedance coefficients for the embedded monopile in a sandy soil layer according to Soil 2, cf. Table 1, obtained by rigorous semi-analytical
expressions and LPM with a filter order of M = 6.
 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401 397

Fig. 8. Transient response for Soil 1 and Soil 2, cf. Table 1, obtained by inverse discrete Fourier transformation of the results from the rigorous semi-analytical expressions and
LPMs with filter order of M = 6.

Table 2
RMS error between the rigorous time domain solution and the LPMs with filter order of M = 6 according to Fig. 8. The reported values are presented as Soil 1/Soil 2 according to the
material properties listed in Table 1.

Degree of freedom fc = 0.2 Hz fc = 2.5 Hz fc = 10 Hz

v3 (N) 8.20e6/1.50e5 7.17e6/7.78e6 9.29e6/2.38e5
v2 (N) 2.08e2/9.75e4 1.17e2/1.24e2 5.53e4/6.46e4
h1 (Nm) (with coupling) 1.22e2/3.84e3 1.19e2/2.07e2 8.83e4/1.53e3
h1 (Nm) (without coupling) 2.91e1/1.34e1 1.42e1/2.98e1 1.21e2/2.87e2

further been validated in the time domain for a frequency constant is analysed using the nonlinear multi-body code HAWC2 coupled
fc = 0.20 Hz and fc = 10 Hz. The RMS errors obtained for these excita- with the LPMs derived in the previous section. The coupling is
tion frequencies are in the same low range as the corresponding ones achieved by an appropriate set of constraint equations that
for fc = 2.5 Hz indicating that the LPM provides accurate results for a describes the relationship between the degrees of freedom in the
wide range of excitation frequencies. HAWC2 model and the LPM. Using the Lagrange Multiplier Method
[50], an external Dynamic Link Library (DLL) is developed. Here,
unconstrained equations of motion of the external system and con-
7. Time–domain simulations of a 5.0 MW wind turbine straints required to couple the consistent LPM to HAWC2 are
structure defined. A detailed explanation may be found in [20].
For a mean water depth of 25 m, the significant wave height
Based on the material and geometrical properties of strata and and peak period of wind generated waves are determined by the
foundation listed in Table 1, aeroelastic simulations of the NREL SPM method [51] with the modifications suggested by Hurdle
5.0 MW reference turbine [49] installed on a monopile foundation and Stive [52]. The two parameters are used to generate the
398 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401

reasonable that the highest damping value related to this mode

is observed for Soil 1, since it accounts for a higher hysteretic
damping ratio compared to Soil 2. In contrast, the value of the
eigenfrequency f1 is lowest for Soil 1 due to a relatively small shear
wave velocity ts compared to Soil 2. Since the eigenfrequency and
damping related to the lowest eigenmode for the LPM and the
equivalent model are matched, it is interesting to observe the
implications of the simplified model on the higher modes. It is
observed that the eigenfrequency and damping related to the sym-
metrical 2nd tower modes for the coupled LPM and the equivalent
models differ significantly from each other, which may have an
effect on the dynamic response of the offshore wind turbine struc-
ture depending on the frequency content of the applied load.
The obtained modal parameters related to the lowest eigen-
Fig. 9. JONSWAP wave spectra for different normal turbulent longitudinal wind mode based on the LPMs have further been compared with the
speeds and a mean water depth of 25 m.
recently proposed analytical solution by Zania [53]. The obtained
eigenfrequency for Soil 1 and Soil 2 were 0.248 Hz and 0.259 Hz,
respectively, while the corresponding damping ratios were equal
parameterised JONSWAP wave spectrum with a peak enhancement to 0.33% and 0.24% (calculated after 0.21% damping ratio for the
coefficient that controls the sharpness of the spectral peak equal to reference model). The coupled LPMs estimate slightly larger damp-
3.3. Fig. 9 shows the applied JONSWAP wave spectrum for a normal ing ratios (0.5% and 0.3% for Soil 1 and Soil 2, respectively), while
turbulent longitudinal wind speed vmean = 8 m/s, vmean = 12 m/s the eigenfrequency is only marginally higher than the analytically
and vmean = 20 m/s. calculated. Note that disregarding the cross-coupling stiffness
In the following, three different foundation modules are consid- terms in the analytical expressions [53] would result to an overes-
ered, cf. Fig. 2: (a) a model with fixed boundary conditions at sea- timated eigenfrequency equal to 0.262 Hz for Soil 1 and 0.266 Hz
bed with no inclusion of soil–pile interaction (reference model), (b) for Soil 2.
an LPM coupled with HAWC2 (coupled LPM) and finally (c) an
apparent fixity length (AFL) of the tower calibrated to the
9. Normal turbine operation
undamped eigenfrequency f1 and damping d1 related to the lowest
eigenmode of the coupled LPM (equivalent model). Note that the
For each foundation model and soil layer according to Fig. 2 and
definition of the last model is not straightforward, since it requires
Table 1, a total of four load cases are investigated:
the prior estimation of the natural vibration characteristics of the
wind turbine, preferably by taking into account the soil–structure
 Load case 1: Normally operating turbine with normally turbu-
interaction (as in the current study). In addition, it should be men-
lent wind speed vmean = 8 m/s and wind-wave alignment.
tioned that the reference model is included in the paper in order to
 Load case 2: Normally operating turbine with normally turbu-
quantify the effects of neglecting the soil–pile interaction. This in
lent wind speed vmean = 12 m/s and wind–wave alignment.
turn provides a clear idea of how conservative (non-conservative)
 Load case 3: Normally operating turbine with normally turbu-
the modelling approach is.
lent wind speed vmean = 20 m/s and wind–wave alignment.
 Load case 4: Normally operating turbine with normally turbu-
8. Eigenvalue problem lent wind speed vmean = 8 m/s and 90° wind–wave
misalignment.
Table 3 shows the comparison of the lowest eigenmodes
between the three different models for the investigated soil layers, A total of six random seeds with time duration T = 300 s are per-
cf. Table 1. As expected, the undamped eigenfrequency f1 and formed for generating the turbulent wind and wave field for each
damping d1 in terms of the logarithmic decrement related to low- load case. Based on a Rainflow cycle counting method to decom-
est eigenmode decreases and increases, respectively, when the pose the fore–aft and side–side moments M1(t) and M2(t) into fati-
soil–pile interaction is taken into account. Further, it may seem gue cycles nf,j, the fore–aft and side–side fatigue damage 1 Hz

Table 3
Undamped eigenfrequency fi and logarithmic decrement di for different foundation models and soil conditions according to Fig. 2 and Table 1.

Mode Reference model Coupled LPM Soil 1 Equ. model Soil 1 Coupled LPM Soil 1 Equ. model Soil 2
fi (Hz) di (%) fi (Hz) di (%) fi (Hz) di (%) fi (Hz) di (%) fi (Hz) di (%)
1. Tower side–side 0.276 1.367 0.249 3.128 0.249 3.126 0.261 1.899 0.260 1.896
1. Tower fore–aft 0.278 1.321 0.251 3.138 0.250 3.137 0.262 1.881 0.262 1.880
1. Drivetrain torsion 0.587 11.986 0.582 12.156 0.582 11.993 0.584 12.107 0.584 12.005
1. Blade asym. flapwise yaw 0.610 3.057 0.610 3.097 0.609 2.882 0.610 2.941 0.610 2.950
1. Blade asym. flapwise pitch 0.641 2.989 0.638 2.993 0.638 2.924 0.640 2.976 0.640 2.944
1. Blade collective flap 0.678 3.827 0.676 3.792 0.676 3.766 0.677 3.798 0.677 3.790
1. Blade asym. edgewise pitch 1.001 3.033 1.000 3.107 1.000 3.032 1.000 3.031 1.000 3.031
1. Blade asym. edgewise yaw 1.013 3.092 1.011 3.235 1.011 3.079 1.012 3.083 1.012 3.084
2. Blade asym. flapwise yaw 1.583 6.376 1.582 6.868 1.577 6.503 1.583 6.379 1.580 6.332
2. Blade asym. flapwise pitch 1.693 7.067 1.814 18.436 1.493 8.878 1.623 7.730 1.623 6.357
2. Blade collective flap 1.837 8.378 1.842 10.933 1.851 9.150 1.834 8.444 1.834 8.319
2. Tower side–side 2.148 6.723 2.460 19.115 2.461 17.590 3.301 11.259 1.810 7.466
2. Tower fore–aft 2.228 7.494 2.855 11.266 1.814 8.775 3.563 15.997 1.944 8.142
 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401 399

Fig. 10. Fore–aft and side–side fatigue damage 1 Hz equivalent moments Meq,1 and Meq,2 evaluated from MSL to seabed level along the monopile structure for different soil
conditions and load cases.

equivalent moments Meq,1 and Meq,2 are computed using a Wöhler Meq,2 evaluated from mean sea level (MSL) to seabed level along
exponent m = 4: the monopile structure. All values are normalised with respect to
!1=m Meq,1 and Meq,2 at MSL from the reference model. Firstly, indepen-
Xn
Dm
j nf;j dent of the load case and soil type, the dynamic behaviour of the
M eq;i ¼ : ð14Þ
T reference model differs significantly from the other models
j¼1
indicated by Meq,1 and Meq,2 that differ with more than 7% and
Here, Dj is the moment width of the jth identified cycle nf,j of the 80%, respectively. This implies that the soil-pile interaction should
local time series. A mean value of the six random seeds is used to not be disregarded in the dynamic analysis of an offshore wind tur-
represent the fatigue damage 1 Hz equivalent moments for each bine. Neglecting soil–pile interaction appears to result to both over-
load case. Fig. 10 shows the 1 Hz equivalent moments Meq,1 and and underestimation of the equivalent moments, depending on the
400 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401

load case and the soil type considered in the analysis. The highest  The fitting goodness of the rational filter approximation is
deviation of the fatigue damage 1 Hz equivalent side–side moment highly sensitive to the choice of the weight function used in
Meq,2 is obtained for Soil 1 independently of the load case, and can the fitting algorithm. Since the consistent LPM makes use of lin-
be attributed to low aerodynamic damping in the side–side direc- ear viscous damping in contrast to a hysteretic damping model
tion, which provides a dynamic structural behaviour sensitive to applied for the derivations of the rigorous semi-analytical
the modal parameters. Secondly, the material properties of the sub- impedance functions, a high weight should be considered in
soil have strong influence on the dynamic response of the wind tur- the low-frequency range.
bine structure. This is supported by comparing Meq,1 and Meq,2 for  Coupling between horizontal sliding and rocking must be con-
the LPMs that show a maximum deviation of approximately 9% sidered by the consistent LPMs to avoid significant loss of
and 57% at seabed level for Load case 4, respectively. The reason is accuracy.
properly caused by the higher eigenfrequency and lower damping
related to the 1st tower mode for Soil 2 compared to Soil 1, cf. Secondly, based on successful coupling between the nonlinear
Table 3. Nevertheless, the effect of the soil type on Meq,1 is not con- multibody code HAWC2 and the consistent LPMs, it is found that:
sistent for all the load cases with the strongest impact being
observed for Soil 1 in Load case 1 and Load case 4. When it comes  The inclusion of soil–pile interaction has a high impact on the
to the comparison of the foundation models, it becomes apparent fatigue damage 1 Hz equivalent moment at seabed. In this
that the implications of the adopted foundation modelling depend regard, the dynamic response of the wind turbine is sensitive
also on the examined load case and the soil type. It is shown that to the soil type, since the modal parameters are changed.
Load case 3 appears to be the one giving the largest deviation in  Low aerodynamic damping in the side–side direction implies
Meq,1 and Meq,2 obtained after the LPMs and the equivalent models, that the side–side fatigue damage 1 Hz equivalent moment at
where differences of around 3% and 7% are present for Meq,1 and seabed is highly sensitive to a small change in the modal
Meq,2, respectively. Finally, it should be noticed that even though parameters related to the 1st tower mode.
the eigenfrequency and damping related to the 2nd tower modes  Equivalent foundation models with a fixity depth calibrated to
deviate between the LPMs and the equivalent models, Meq,1 and the eigenfrequency and damping related to the lowest eigen-
Meq,2 agree quite well. This tendency may be explained by Fig. 9, mode of the coupled consistent LPMs seem to capture the
which clearly indicates that the applied JONSWAP wave spectra dynamic response of the wind turbine reasonable well. This is
have almost no energy content for frequencies above 0.25 Hz. Note- the case even though the modal parameters related to the 2nd
worthy is the fact that the equivalent model has been calibrated tower mode differ significantly from each other in the two
after the modal parameters of the LPM, hence its validity is strongly models.
dependent on the adequate selection of the equivalent fixity length.
It is finally noted that despite of comparable results of the fati-
gue damage 1 Hz equivalent moments between the equivalent
10. Discussion and conclusions foundation models and the coupled consistent LPMs, physical
insight of the eigenfrequency and the soil damping contribution
The paper investigates the natural vibration characteristics and prior to the simulation is needed for the application of the equiva-
dynamic response of a 5.0 MW offshore wind turbine installed on a lent foundation models. This is not the case for the coupled consis-
monopile foundation. In this regard, rigorous semi-analytical tent LPMs that in a direct manner take the effects on stiffness and
expressions for the determination of the dynamic impedance func- the energy dissipation effects of the dynamic soil–pile interaction
tions of a linear uniform monopile embedded in a linear viscoelas- into account. This in turn makes the presented modelling approach
tic soil layer are used in order to evaluate the dynamic soil–pile of soil–pile interaction very attractive and may advantageously be
interaction. Compared to a three–dimensional coupled BE/FE used for other pile foundations like the jacket or tripod founda-
model, the semi-analytical methods provide a fast and reliable tions, taking into account any pile-to-pile interaction effects when-
evaluation of the dynamic impedance functions. The frequency- ever it is considered relevant.
dependent impedance characteristics may not seem straightfor-
ward to implement into aerolastic codes, since a dramatic increase References
in the number of degrees of freedom in the aeroelastic model
should be avoided. However, lumped-parameter models (LPMs) [1] Fugro GEOS. Wind and wave frequency distributions for sites around the
are powerful tools for solving this problem. Based on a particular British Isles. Offshore Technology Report 2001/030, Health and Safety
Executive, Southampton Oceanography Centre; 2001.
parallel coupling of the so-called discrete-element models consist- [2] Fischer T. Offshore foundation and support structures. Technical report,
ing of springs, dashpots and masses with frequency-independent Upwind project. Funded by the commission of the European communities;
coefficients extracted from a rational filter approximation of the 2006.
[3] Petersen B, Pollack M, Connell B, Greeley D, Davis D, Slavik C, Goldman B.
rigorous semi-analytical impedance functions, fully coupled aero- Evaluate the effect of turbine period of vibration requirements on structural
hydro-elastic simulations can easily be evaluated in which flexibil- design parameters. Technical report. Applied Physical Sciences; 2010.
ity and wave radiation of the subsoil are taken into account. Hence, [4] Damgaard M, Ibsen LB, Andersen LV, Andersen JKF. Cross–wind modal
properties of offshore wind turbines identified by full scale testing. J Wind
after the calibration of the LPM for a representative range of phys- Eng Ind Aerodyn 2013;116:94–108.
ical frequencies, the wave radiation problem is adequately taken [5] Jonkman J, Butterfield S, Passon P, Larsen T, Camp T, Nichols J, et al. Offshore
into account. code comparison collaboration within IEA Wind Annex XXIII: Phase II Result
regarding Monopile Foundation Modelling, National Renewable Energy
Firstly, regarding the rational filter approximation of the rigor- Laboratory, Report Number NREL/CP-500-42471; 2008.
ous semi-analytical impedance functions of a monopile embedded [6] Larsen TJ, Hansen A. How to HAWC2, The user’s manual. Technical Report Risø-
in two different linear viscoelastic soil layers, the following obser- R-1597(en), Risø National Laboratory, Technical University of Denmark; 2007.
[7] DNV-OS-J101. Design of offshore wind turbines. Det Norske Veritas, Norway;
vations can be made:
2004.
A rational filter approximation of the order m = 6 provides use- [8] International Electrotechnical Commission, IEC 61400-3. Wind turbines – part
ful results in the frequency range f 2 [0;8] Hz. Hence, models with 3: design requirements for offshore wind turbines, Technical Standard,
three internal degrees of freedom per non–zero component of the European Committee for Electrotechnical Standardization; 2009.
[9] Vorpahl F, van Wingerde A, Blunk M, Busmann HG, Kleinhansl S, Kossel T, et al.
impedance matrix of the monopile foundation suffice in the fre- Validation of a finite element based simulation approach for offshore wind
quency and time domain. turbines within IEA wind annex XXIII – Simulation challenges and results for a
 M. Damgaard et al. / Engineering Structures 75 (2014) 388–401 401

5–MW turbine on a tripod substructure, Proceeding of 19th ISOPE conference, pile and soils. In: Luan M, Zen K, Chen G, Nian T, Kasama K, editors, Recent
Osaka, ISOPE; 2009. development of geotechnical and geo–environmental engineering in Asia, p.
[10] Böker C. Load simulation and local dynamics of support structures for offshore 138. PT: B; CT: 4th Asian Joint Symposium on Geotechnical and Geo-
wind turbines. PhD thesis, Institute for Steel Construction, Leibniz University Environmental Engineering (JS–Dalian 2006); CY: NOV 23-25, 2006; CL:
of Hannover, Germany; 2009. Dalian, PEOPLES R CHINA; UT: ISI:000279628900023.
[11] Øye S. FLEX4 – simulation of wind turbine dynamics. State of the art of [33] Memarpour MM, Kimiaei M, Shayanfar M, Khanzadi M. Cyclic lateral response
aeroelastic codes for wind turbine calculations. Denmark: Lyngby; 1996. p. of pile foundations in offshore platforms. Comput Geotech 2012;42:180–92.
71–76. [34] Blaney GW, Kausel E, Roesset JM. Dynamic stiffness of piles. In: Proceedings of
[12] Passon P. Aero-elastic design tool for bottom-mounted offshore wind turbines. 2nd international conference on numerical methods in
Upwind deliverable D4.3.2. Stuttgart: Endowed Chair of Wind energy, geomechanics. USA: Blacksburg; 1976. p. 1001–12.
University Stuttgart; 2008. [35] Kuhlemeyer RL. Vertical vibrations of piles. J Geotech Eng Div, ASCE
[13] Kaufer D. Integrated analysis of the dynamics of offshore wind turbines with 1979;105:273–87.
complex support structures. Diploma thesis at the Endowed Chair of Wind [36] Gazetas G, Dobry R. Horizontal response of piles in layered soils. J Geotech Eng
energy, University Stuttgart; 2008. 1984;110(1):20–40.
[14] Seidel M, Ostermann F, Curvers A.P.W.M., Kühn M, Kaufer D, Böker C. [37] Medina C, Aznárez JJ, Padrón LA, Maeso O. Effects of soil–structure interaction
Validation of offshore load simulations using measurement data from the on the dynamic properties and seismic response of piled structures. Soil
downvind project. European Offshore Wind, Stockholm, Sweden; 2009. Dynam Earthquake Eng 2013;53:160–75.
[15] Guyan RJ. Reduction of stiffness and mass matrices. AIAA J 1965;3:380. [38] Cheng Z, Jeremic B. Numerical modelling and simulation of pile in liquefiable
[16] de Klerk D, Rixen DJ, Voormeeren NS. General framework for dynamic soil. Soil Dynam Earthquake Eng 2009;29:1405–16.
substructuring: history, review, and classification of techniques. AIAA J [39] Assimaki D. Varun. A generalized hysteresis model for biaxial response of pile
2008;46:1169–81. foundations in sands. Soil Dynam Earthquake Eng 2012;32:56–70.
[17] Wolf J. Foundation vibration analysis using simple physical [40] Dassault Systèmes Simulia Corp. Abaqus 6.12 Analysis user’s manual.
models. Englewood Cliffs, NJ: Prentice-Hall; 1994. providence, RI, USA: Dassault Systèmes Simulia Corp; 2012.
[18] Haskell N. The dispersion of surface waves on multi–layered medium. Bull [41] Andersen LV, Jones CJC. A computer program for boundary analysis of soil and
Seismol Soc Am 1953;73:17–43. three-dimensional structures. ISVR Technical Memorandum 868, Institute of
[19] Thomson W. Transmission of elastic waves through a stratified solid medium. J Sound and Vibration Research, University of Southampton; 2001.
Appl Phys 1950;21:89–93. [42] Veletsos AS, Wei YT. Lateral and rocking vibration of footings. J Soil Mech
[20] Damgaard M, Andersen LV, Ibsen L. Computationally efficient modelling of Found Eng Div, ASCE 1971;97:1227–48.
dynamic soil–structure interaction of offshore wind turbines on gravity [43] Bu S, Lin CH. Coupled horizontal–rocking impedance functions for embedded
footings. Renew Energy 2014;68:289–303. square foundations at high frequency factors. J Earthquake Eng
[21] Novak M, Nogami T. Soil–pile interaction in horizontal vibration. Earthquake 1999;3:561–87.
Eng Struct Dynam 1977;5(3):263–81. [44] Kramer ST. Geotechnical earthquake engineering. Belgium: Prentice-Hall
[22] Nogami T, Novak M. Resistance of soil to a horizontally vibrating pile. Upper Saddle River; 1996.
Earthquake Eng Struct Dynam 1977;5(3):249–61. [45] Andersen LV, Jones CJC. Three-dimensional elastodynamic analysis using
[23] Nogami T, Novak M. Soil–pile interaction in vertical vibrations. Earthquake multiple boundary element domains. ISVR Technical Memorandum 867,
Eng Struct Dynam 1976;4:277–93. Institute of Sound and Vibration Research, University of Southampton; 2001.
[24] Novak M, Howell JF. Torsional vibrations of pile foundations. J Geotech Eng [46] Andersen L.V. Wave propagation in infinite structures and media. PhD thesis,
Div–ASCE 1977;103(4):271–85. Aalborg University; 2002.
[25] Novak M, Howell JF. Dynamic–response of pile foundations in torsion. J [47] Randolph MF. Piles subjected to torsion. J Geotech Eng Div–ASCE
Geotech Eng Div–ASCE 1978;104(5):535–52. 1981;107(8):1095–111.
[26] Novak M. Dynamic stiffness and damping of piles. Can Geotech J [48] Newmark NM. A method of computation for structural dynamics. J Eng Mech
1974;11:574–98. Div, ASCE 1959;85:67–94.
[27] Winkler E. Die Lehre von Elasticzitat und Festigkeit (On Elasticity and Fixity), [49] Jonkman J, Butterfield S, Musial W, Scott G. Definition of a 5-MW reference
Prague; 1867. wind turbine for offshore system development, Technical Report NREL/TP-
[28] Nogami T, Otani J, Konagai K. Nonlinear soil/pile interaction/model for 500-38060, National Renewable Energy Laboratory; 2009.
dynamic lateral motion. J Geotech Eng/ASCE 1992;118(1):89–106. [50] Betsekas DP. Nonlinear programming. Cambridge, MA: Athena Scientific;
[29] El Naggar M, Novak M. Nonlinear lateral interaction in pile dynamics. Soil 1999.
Dynam Earthquake Eng 1995:14. [51] US Army Corps of Engineers, Shore protection manual, Technical Report 1,
[30] El Naggar M, Novak M. Nonlinear analysis for dynamic lateral pile response. Coastal Engineering Center, Department of the Army Waterways Experimental
Soil Dynam Earthquake Eng 1996;15:223–44. Station, Corps of Engineers; 1984.
[31] El Naggar M, Bentley KJ. Dynamic analysis for laterally loaded piles and [52] Hurdle DP, Stive RJH. Revision of SPM 1984 wave hindcast model to avoid
dynamic p–y curves. Can Geotech J 2000;37(6):1166–83. inconsistencies in engineering applications. Coast Eng 1989;12. 339–151.
[32] Kong D, Luan M, Ling X, Qiu Q. A simplified computational method of lateral [53] Zania V. Natural vibration frequency and damping of slender structures
dynamic impedance of single pile considering the effect of separation between founded on monopiles. Soil Dynam Earthquake Eng 2014;59:8–20.