J. Wind Eng. Ind. Aerodyn.

120 (2013) 96–106

Contents lists available at ScienceDirect

Journal of Wind Engineering

and Industrial Aerodynamics
journal homepage: www.elsevier.com/locate/jweia

Experimental and computational damping estimation

of an offshore wind turbine on a monopile foundation
R. Shirzadeh n, C. Devriendt, M.A. Bidakhvidi, P. Guillaume
Acoustics and Vibration Research Group (AVRG), Department of Mechanical Engineering, Vrije Universiteit Brussel (VUB), Pleinlaan 2, 1050 Brussels, Belgium

art ic l e i nf o a b s t r a c t

Article history: In this study, Operational Modal Analysis (OMA) is used to identify the damping value of the
Received 14 September 2012 fundamental for-aft (FA) mode of an Offshore Wind Turbine (OWT) using both real life measurements
Received in revised form and simulations. Estimations of the total damping of an offshore wind turbine (taking into account the
1 July 2013
effects of the aerodynamic, hydrodynamic and soil loads) give a quantitative view of the stability
Accepted 10 July 2013
Available online 7 August 2013
characteristics of the wind turbine. Two different test cases including an overspeed stop and ambient
excitation have been considered. The experimental data has been obtained during a measurement
Keywords: campaign on an offshore wind turbine in the Belgian North Sea and the results are compared with the
Offshore wind turbine numerical simulations which have been carried out in HAWC2.
Offshore measurement
& 2013 Elsevier Ltd. All rights reserved.
Damping
Overspeed stop
Ambient excitation

1. Introduction exist to estimate the aeroelastic damping for stationary and rotating
wind turbines. However, experimental results are needed to verify
Many large-scale offshore wind farm projects use monopile and/or upgrade the analysis.
foundations to realize a cost effective design. During the design of James et al. estimated the modal damping using strain-gauge
these monopile structures fatigue due to combined wind and wave data from an operating wind turbine. The cross-correlation (or
loading is one of the most important problems to take into account. auto-correlation) functions have been applied on strain-gauge
Coincidence of structural resonances with wind turbine time histories. They then used the Polyreference method to extract
dynamic forces can lead to large amplitude stresses and subse- the modal parameters from correlation functions. The method has
quent accelerated fatigue. For this reason, the wind turbine rotor been verified for a vertical axis wind turbine. Hansen et al. (2006)
blades and support structure are designed to avoid resonance described two different experimental methods for estimating the
coincidence. In particular, the current practice is to design the aeroelastic frequencies and damping of the operational modes of
wind turbine support structure in such a way that the tower wind turbines from experiments. They compared the results with
fundamental resonance does not coincide with the fundamental theoretical predictions from the stability tool HAWCStab. Recently,
rotational (1P) and blade passing (3P for three-bladed turbines) Damgaard et al. (2013) have performed “rotor-stop” tests on
frequencies of the rotor (Van Der Tempel, 2006). monopile offshore wind turbines for different wind parks to
It is very crucial to correctly estimate the damping ratios for a evaluate the first natural frequency and modal damping of the
wind turbine as the amplitude of vibrations at resonance are structures.
inversely proportional to these ratios (Devriendt et al., 2012). The In this study the aeroelastic simulation of a Vestas V90 3 MW
overall damping of the first bending mode of an offshore wind wind turbine was carried out using HAWC2 aeroelastic code which
turbine consists of a combination of aerodynamic damping, is developed at Risø DTU. The simulations are performed for two
damping due to constructive devices, such as a tuned mass sets of test cases, namely the overspeed stop test and ambient
damper, and additional damping, e.g. structural, hydrodynamic excitation, and the results are compared with the experimental
and soil damping (Petersen et al., 2010). Compared to onshore data. For this paper, we will focus on identifying the frequency and
support structures, the additional damping is further influenced damping value of the first for-aft (FA) mode.
by effects such as soil damping and hydrodynamic damping
(http://www.marin.nl/web/events/events-2010/). Analytical methods
2. Offshore measurements

n
Corresponding author. Within the project two measurement campaigns have been
E-mail address: [email protected] (R. Shirzadeh). planned. The first short measurement campaign focused on

0167-6105/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jweia.2013.07.004
 R. Shirzadeh et al. / J. Wind Eng. Ind. Aerodyn. 120 (2013) 96–106 97

performing an overspeed test with the aim of obtaining a first above sea level. The interface level between the transition piece
estimate of the damping value of the fundamental for-aft vibration and the wind turbine is at 17 m above sea level. There are two
mode of the wind turbine. During the second long term measure- accelerometers mounted at the lower three levels and four at the
ment campaign we will continuously monitor the vibration levels top-level. The chosen configuration is primarily aimed at identi-
and the evolution of the frequencies and damping of several fication of tower bending modes. The two extra sensors on the
fundamental modes of the tower and foundation. Both the top-level are placed to capture the tower torsion. Accelerometers
resonance frequencies and damping values are crucial to quantify have been selected, which have a high sensitivity and are able to
the reliability and the lifetime of offshore wind turbines both in measure very low frequent signals. This is necessary considering
the design phase as during its life cycle. These parameters will also that the modal frequencies of interest, for the wind turbine
be analyzed to see if they can provide indications about the structure, are expected to be around 0.35 Hz, and the expected
current state of the soil and foundation characteristics for e.g. vibration magnitude is very low, especially during ambient
monitoring scour development (Zaaijer and Van der Tempel, excitation.
2004). The measurement campaigns are performed at the Belwind The data-acquisition system is mounted in the transition piece
wind farm, which consists of 55 Vestas V90 3 MW wind turbines. (green circle in Fig. 2). The project invested in a ‘multi-purpose
The wind farm is located in the North Sea on the Bligh Bank, 46 km monitoring system’ to support dedicated R&D projects in the field
off the Belgian coast (Fig. 1). of offshore wind energy. It can be used to monitor several
The hub-height of the wind turbine is on average 72 m above parameters, e.g. accelerations and strains, on existing offshore
sea level. Each transition piece has a height of 25 m and a weight wind turbines. There was the demand for a robust data-acquisition
of 120 ton. The tests are performed on the BBCO1-turbine that is system considering the harsh offshore conditions and any down-
located in the north of the wind farm directly next to the offshore time had to be avoided taking into account the high cost related
high voltage substation (OHVS). The wind turbine is placed on a with working offshore. Since the project aims at characterizing the
monopile foundation structure with a diameter of 5 m and a wall- dynamics of an operational turbine under various operating
thickness of 7 cm. The actual water depth at the location of BBCO1 conditions, it is also necessary that the data is acquired over a
is 22.9 m and the monopile has a penetration depth of 20.6 m. The long period of time. This requires the data-acquisition system to be
soil is considered stiff and mainly consists of sand. remotely monitored and capable of automatic startup in case of
The structures instrumented in this campaign are the tower power shutdowns. Bearing in mind the specific demands of the
and transition piece. Measurements are taken at 4 levels on project a Compact Rio system of National Instruments were used
9 locations using a total of 10 sensors. The measurement locations (Fig. 3).
are indicated in Fig. 2 by yellow circles. The locations are chosen The data acquisition software allows for the continuous mon-
based on the convenience of sensor mounting, such as the vicinity itoring of the accelerations. The data-acquisition system was
of platforms. The chosen levels are 67 m, 37 m, 23 m and 15 m programmed to acquire data with a sampling ratio of 5 kHz.

Fig. 1. A render of an offshore wind turbine at Belwind (left), location Belwind wind farm (middle) and park layout Belwind wind farm (right).

Fig. 2. Measurement locations on BBCO1. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this article.)
98 R. Shirzadeh et al. / J. Wind Eng. Ind. Aerodyn. 120 (2013) 96–106

Fig. 3. Data acquisition system based on NI Compact Rio System (left) and logger software (right).

1
degree 0
Probability Density Function

degree 90
0.8
degree 180
degree 240
0.6

0.4

0.2

0
0 1 2 3 4 5 6
Significant wave height, Hs, (m)
Fig. 4. Weibull distribution for the wave for 4 different directions versus significant wave height (left) wind and wave directional roses. Wind at hub height (72 m) (right).

Considering the frequency band of interest and in order to reduce 3. Damping estimation from measurements
the amount of data the recorded time series have been filtered
with a band-pass filter and re-sampled with a sampling frequency As we mentioned before, two sets of test cases have been
of 12.5 Hz. After the down sampling and filtering a coordinate considered to measure the frequencies and damping values of the
transformation was performed, because the accelerometers are fundamental modes of the offshore wind turbine. These test cases
mounted on the tower. Therefore, in order to measure the are:
vibrations along the axis of the nacelle, it is necessary to take
the yaw-angle into account by transforming them into the coordi- (i) An overspeed stop where the wind speed is the minimum
nate system of the nacelle (Chauhan et al., 2011). required (6.5 m/s). This allows the wind turbine to speed up until
The software measures continuously and sends data every 19.8 rpm. This is the speed at which the wind turbine is
10 min to the server that is installed onshore using a dedicated automatically stopped and the pitch angle is put on 88.2 degrees.
fiber that is running over the seabed. All data receives a time- (ii) Ambient excitation where the wind speed is always very low
stamp from an NTP timeserver in order to be able to correlate around 4.5 m/s and the pitch angle is around 80.5 degrees and
them with the SCADA and Meteo data. The measurements can be the rotor is slowly rotating (0.2 rpm). This permits us to assume
monitored real-time using the online scope-function. Finally, in that the contribution of aerodynamic damping is very low.
order to classify the operating conditions of the wind turbine
during the measurements SCADA data (power, rotor speed, pitch The overall characteristic of the SCADA and wave data for both sets
angel, nacelle direction) is gathered at a sample rate of 1 Hz. In of tests is listed in Table 1.
order to monitor also the varying environmental conditions, the The objective is to obtain an estimate of the overall first mode
ambient data (wind speed, wind direction, significant wave height, damping. This is the damping due to the aerodynamic, hydro-
air temperature) is being collected at 10-min intervals. dynamic, soil, and structural damping. Therefore both during the
The Weibull distribution for waves is measured at Belwind overspeed test and during the ambient excitation test the tuned
wind farm and a typical distribution is plotted in Fig. 4 for mass damper was turned off.
different selected directions. The wind and wave misalignment
has a great influence on the response of the offshore wind 3.1. Overspeed stop
turbines. Therefore this is very important to measure the mis-
alignment accurately. The wind and wave directional roses is The overspeed test is commonly used to accurately identify the
measured at the Belwind location and shown in Fig. 4. modal damping ratios. The damping ratios can be obtained by
 R. Shirzadeh et al. / J. Wind Eng. Ind. Aerodyn. 120 (2013) 96–106 99

fitting an exponential function to the relative maxima of the around 0.35 Hz. We can also identify some smaller peaks with a
decaying time series and extracting the damping ratio from the frequency higher than 1 Hz. These peaks are related to higher
parameters of the fitted expression. This method assumes that the tower modes and blade modes. We can also observe some small
decay has only the contribution of a single mode. When this peaks below the dominant frequency. These peaks are related to
assumption is not met it may result in a high scatter and might the waves. The significant wave height during the tests was 0.5 m.
give wrong estimates for the damping. This is especially the case The frequency domain identification algorithms (Parloo, 2003;
for closely spaced modes where it might not be possible to get a Verboven, 2002; Cauberghe, 2004) can now be applied to a matrix
decaying vibration with just the contribution of one mode with a single column containing the Fast Fourier Transformation of
(Magalhães et al., 2010). the free decays measured during the overspeed test. During this
During an overspeed stop the wind turbine speeds up until it analysis we used the data between 0.8 and 0.2 of the maximum
reaches 19.8 rpm. Then the wind turbine is automatically shut acceleration. An initial estimate of the damping ratios was
down. During this stop the pitch angle is changed from
2.5 to obtained with a least squares estimator in the frequency domain,
88.2 degrees in a couple of seconds. The thrust release due to this using polynomials with orders between 1 and 60 (Verboven,
sudden collective pitch variation excites the tower mainly in the 2002). The fitting was performed in the frequency range 0.1–
wind direction as can be seen in Fig. 5. Note that on the figure the 2 Hz. These results can be used to construct a stabilization chart
wind direction is shown from left to right. from which the user can try to separate the physical poles
(corresponding to a mode of the wind turbine) from the mathe-
matical ones. By displaying the poles (on the frequency axis) for an
3.1.1. Frequency domain analysis
increasing model order (i.e number of modes in the model), the
Instead of analyzing the data in the time domain, one can also
diagram helps to indicate the physical poles since, in general, they
perform the analysis in the frequency domain. The Fast Fourier
tend to stabilize for an increasing model order, while the compu-
Transformation of the decaying functions can directly be used as
tational poles scatter around. The construction of the stabilization
input for the analysis methods in the frequency domain (Bendat
chart is nowadays one of the requirements for a modal parameter
and Piersol, 1980). Fig. 6 shows the Fast Fourier Transformation of
estimation algorithm, and it has become a common tool in modal
the accelerations obtained on the 3 different levels in FA direction.
analysis. In Fig. 6 the stabilization diagram is displayed. In the
One can clearly identify the dominant peak from the first FA mode
stabilization diagram algorithm used in this paper we evaluate for
every pole, the distance to the nearest pole calculated with the
Table 1
Overall characterization of the SCADA and wave data for overspeed stop and
previous model order and we plot:
ambient excitation tests.
 a red s, if distance is smaller than 1%,
Parameter Overspeed stop Ambient excitation  a blue f, if only frequency variation is smaller than 1%,
 a purple d, if only damping variation is smaller than 5%,
Wind speed, m/s 6.5 4.5
Pitch angle of blades, deg
2:5-88:2 80.5
 a black o, if neither the pole, nor the frequency, nor the
Rotor speed, rpm 19.8 - 0 0.2 damping ratio stabilizes
Significant height (Hs), m 0.5 0.5
Wave period (Tp), s 3.5 3.2 In the stabilization diagram we can clearly see that the dominant
Nacelle wind misalignment, deg 0 0
Wind-wave misalignment, deg 20 20
mode around 0.35 Hz is well identified and results in a clear stable
line. The damping ratio of this mode was found to be 1.05%.

Fig. 5. Movement seen from above (left) and accelerations (g) on 3 levels in FA direction during overspeed stop (right).

Fig. 6. Fast Fourier transformation of the accelerations obtained on the 3 different levels in FA direction (left) and stabilization diagram after applying the least squares
frequency domain estimator (right).
100 R. Shirzadeh et al. / J. Wind Eng. Ind. Aerodyn. 120 (2013) 96–106

The peak below the dominant mode was identified with a The matrix containing the Fast Fourier transformations of the
much higher damping and has a frequency of 0.29 Hz. This auto and cross correlation functions between the accelerations at
perfectly coincides with the wave period of the waves with all measurement points and the accelerations at each chosen
a wave direction almost in line with the nacelle. During the reference point can be used as input for the analysis methods in
overspeed stop the wave period was around 3.5 s. Note that the frequency domain. In this paper the two sensors on the third
this confirms that waves can induce a significant oscillation and level (c.f. Fig. 2) in the FA and SS-direction were chosen as
accelerations in the turbine as was stated in http://www.marin.nl/ reference signals. Fig. 8 shows the spectra obtained from the
web/events/events-2010/presentations-offshore-wind-seminar- correlation functions, when using 40 min of data and using the
2010.htm. The wave frequency is close enough to the resonance first 512 positive time lags.
frequency of the fundamental mode of the wind turbine to have After applying the poly-reference least square estimator (Peeters
a dynamic amplification. Therefore, waves can have a significant et al., 2004) to the 6 by 2 matrix containing the correlation functions,
effect on the lifetime of the wind turbine and should definitely in the frequency band of 0.1–2 Hz, we can construct the stabilization
be taken into account when performing fatigue calculations. diagram as is shown in Fig. 8. This analysis identifies several modes
within the selected frequency band and moreover it results in 3 stable
poles around the dominant peak. The poly-reference least square
3.2. Ambient excitation
estimator is able to identify the 2 closely spaced poles that correspond
respectively with the first for-aft mode at 0.358 Hz and with the first
Ambient vibration tests have the strong advantage of being very
side-side mode at 0.365 Hz and another stable pole just below the
practical and economical, as they use the freely available ambient
dominant peak that has a frequency of 0.31 Hz. This last one
wind wave excitation. Furthermore, the data is collected during the
corresponds again with the wave period during the ambient test that
normal use of the structure and consequently the identified modal
was now slightly higher in comparison with the overspeed test.
parameters are associated with realistic vibration levels. An opera-
Table 2 gives the estimated damping values for the first FA mode and
tional modal analysis was performed with the rotor slowly rotating
the first SS-mode.
(0.2 rpm). In the 40 min of recorded data (just before the overspeed
The damping value of the first FA-mode corresponds well with
stop was performed) the wind had an average speed of around
the one that was found using the data from the overspeed test. In
4.5 m/s. The nacelle was put into the direction of the wind. Fig. 7
the SS direction we find a slightly higher damping. This might be
shows the movement of the tower in the FA and SS direction. One
explained due to the presence of some small aerodynamic damp-
can observe that the movement is mainly in the FA direction, i.e. the
ing effects in this direction considering the pitch angle of 80.5
direction of the wind, but there is also a contribution of the SS
degrees (Tcherniak et al., 2011). According to Hansen et al. (2006)
movement. The tower does not vibrate purely in the wind direction,
the aerodynamic forces are present even at standstill due to the
both the FA mode as the SS mode are present, resulting in an
additional movement perpendicular to the wind.
Table 2
Estimated damping ratios (ζ) on the levels both in FA and SS direction using
3.2.1. Correlation driven analysis in the frequency domain different band-pass filters.
In a similar way as with the decaying functions obtained during
the overspeed stop, the Fast Fourier Transformation of the positive Damping ratio FA mode (S1FA) Damping ratio SS mode (S1SS)

time lags of the correlation functions can directly be used as input 1.05% 1.27%
for the analysis methods in the frequency domain.

Fig. 7. Movement seen from above (left) and accelerations (g) on 3 levels in FA direction during ambient excitation (right).

Fig. 8. Fast Fourier Transformation of the correlation functions between all measurement points and the reference sensor in the FA-direction (right). Stabilization diagram
after applying the least squares frequency domain estimator to the Fast Fourier transformations of the correlations.
 R. Shirzadeh et al. / J. Wind Eng. Ind. Aerodyn. 120 (2013) 96–106 101

larger blade surface that interacts with surrounding air when the Table 3
tower vibrates in the SS direction. Summary of properties for the Vestas 3-MW wind turbine.

Description Value

4. Numerical simulations Rating 3 MW

Rotor position, rotor type Upwind, 3-bladed
This section describes the aeroelastic simulation of Vestas V90 Control Variable speed, collective pitch
Drivetrain High speed, multiple-stage gearbox
3MW wind turbine which is mounted atop a monopile with a Rotor, hub diameter (m) 90, 2.32
flexible foundation in 22.9 m water depth. The numerical simula- Hub height (above MSL) (m) 72
tions have been carried out using HAWC2 aeroelastic code devel- Cut-in, rated, cut-out wind speed (m/s) 3.5, 15, 25
oped at Risø DTU. HAWC2 uses a multi-body formulation, allowing Cut-in, rated, cut-out rotor speed (rpm) 8.6, 16.1, 18.4
Shaft tilt, coning (deg) 6, 4
the user to model each component of the turbine as a separate
Rotor weight (ton) 39.8
body. The implementation of each body is achieved adopting a Nacelle mass (ton) 70
finite element theory. The code is capable to simulate the struc- Tower mass (ton) 108
tural response of a pitch controlled horizontal axis wind turbine
(HAWT) subject to aerodynamic, hydrodynamic and soil loads. In
principle HAWC2 code is a code intended for calculating wind transition piece to the soil by a soil penetration length of 20.6 m.
turbine response in time domain. The Young's modulus and shear modulus are taken to be 210 GPa
The wind turbine structure is modeled by a number of bodies, and 80.8 GPa, respectively. Furthermore the effective density of
e.g. a tower body or shaft body, and these bodies are then the steel is 8500 kg/m3 to account for paints, bolts, welds and
connected by a set of constraint equations. The flexibility of the flanges that are not accounted for in the thickness data.
structure is modeled by Timoshenko beam elements with For the lack of accessibility to the distributed properties of the
6 degrees of freedom for each node. All inertia forces and external blades, we scaled down the data from NREL 5 MW reference wind
loading are based on the deformed shape of the structure. The turbine (Jonkman et al., 2009) to generate the input file. This can
Newmark beta new is used to solve the equations (Larsen et al., result in a small deviation on the identified frequency of the blade
2005). modes, but should have little or no influence on the frequency of
fundamental FA tower mode. The overall mass, first mass moment
4.1. Geometry of support structure and wind turbine of inertia, second mass moment of inertia, and nominal radial
CM location of each blade are is 8244 kg, 1.3  105 kg m2,
A schematic picture consists of the foundation geometry as 3.3  106 kg m2 and 15.9 m with respect to the blade root,
well as wind turbine can be seen in Fig. 9. respectively.
The specifications of the wind turbine are summarized in
Table 3. This wind turbine is a conventional three-bladed, upwind 4.2. Numerical modeling of ambient loads
variable speed, pitch-blade controlled turbine. The detailed speci-
fications of the blade aerodynamic properties; monopile founda- 4.2.1. Wind and aerodynamic model
tion, tower, nacelle and drivetrain structural properties are The aerodynamic model in HAWC2 is based on the Blade
provided as an input file for the HAWC2 code. The hydrodynamic Element Momentum theory (BEM) which is extended with models
and elastic properties of the offshore support are also collected. to handle the dynamic inflow, skew inflow, shear effect on
The wind inflow data and irregular wave kinematics are measured induction, effect from large blade deflections and tip loss. All
during the experiments and have been imported for the airfoil data including the Cl, Cd, Cm and α are provided as input files.
simulations. To capture the geometrical non-linearity, the blades are divided
The distributed tower properties are based on the base dia- into several bodies.
meter (4.307 m) and thickness (0.025 m), top diameter (2.40 m) Also two dynamic stall models namely Stig Øye model and
and thickness (0.013 m). The thickness and radius of the tower are modified Beddoes–Leishmann model are included in the code. It is
assumed to be linearly tapered from base to the tower top. possible to model both deterministic and stochastic wind in
The monopile foundation has a constant diameter (5 m) and HAWC2. The deterministic part of the wind includes mean wind
thickness (0.07 m) and is extended from the bottom of the velocity, sudden acceleration, linear trend, special gust events, and
special shears. The stochastic wind usually referred as turbulent
wind is not considered in this study. This assumption is reliable as
the wind speed is low for both test cases. Furthermore, the tower
shadow effects, which account for the wind condition changes
near the tower, are used in simulations.

4.2.2. Wave and water kinematics

The hydrodynamic loads in HAWC2 are based on the Morrison's
equation. The force per unit length for a cylinder is expressed by
Eq. (1).

π:D2 dU 1
F ¼ F m þ F d ¼ Cmρ þ C jU jU ð1Þ
4 dt 2 d
In the Morrison's equation the wave force consists of a part due to
the drag and an inertia part. In the equation this is expressed by
the introduction of a drag coefficient Cd and an inertia coefficient
Cm. Where U is the fluid velocity, dU=dt the acceleration of the
Fig. 9. Schematic drawing of the wind turbine mounted on a monopile foundation. fluid, ρ the water density and D the diameter of the cylinder.
102 R. Shirzadeh et al. / J. Wind Eng. Ind. Aerodyn. 120 (2013) 96–106

According to the reference (Veldkamp and Van Der Tempel, where A is a factor to account for cyclic or static loading condition
2005), the inertia component is dominant for offshore wind and can be evaluated by
turbine support structures with the currently characteristic dia- 8
>
< 0:9 for cyclic loading
meter of 3–5 m in intermediate water depths of 5–25 m. Several  
A¼ H
empirical calculation methods exist to find the appropriate drag : 3:0
0:8D ≥0:9 for static loading:
>
and inertia coefficients. For random wave fields used in fatigue
calculations, the design rules advise, for a vertical cylinder, Cm ¼2.0 k¼ initial modulus of subgrade reaction, (kN/m3). Determine from
and Cd ¼0.65 (Din). The drag coefficient could not be determined figure 6.8.7-1 of API standard (AP Institute, 2000) as function of
so accurately, but at least common values of C d ¼ 0:65–1:05 are not angle of internal friction, ϕ.
contradicted. For this study, the inertia and drag coefficients are y ¼lateral deflection (m)
selected to be 2.0 and 1.0, respectively. H ¼depth (m).
The water kinematics describes how the sea condition is pu ¼ultimate bearing capacity at depth H (kN/m),
considered to affect the structure. The wave kinematics is provided The ultimate lateral bearing capacity for sand is the smallest
through a defined DLL (Dynamic Link Library) interface, including value of two following equations:
both regular and irregular airy waves. Two different empirical (
wave spectra namely Pierson Moskowitz (PM) and JONSWAP are pus ¼ ðC 1 H þ C 2 DÞγH
pu ¼ min
implemented in HAWC2. For current analyses the JONSWAP pud ¼ C 3 DγH
spectrum has been used which defines the distribution of energy
with frequency within the sea. The significant wave height and C1, C2, C3 are constants determined from 6.8.6-1 of API standard
wave period were measured during the overspeed stop test and (AP Institute, 2000). D is the pile diameter, ϕ is the angle of
ambient excitation and were 0.5 m and 3.5 s, respectively. The internal friction of sand, (deg) and γ is the effective soil weight,
wave and wind misalignment has been measured during two sets (kN/m3).
of tests and is 20 deg. Based on the geotechnical data, the soil consists of three layers
of sand and stiff clay with different properties in terms of internal
friction angles. The full pile foundation is modeled in LPILE
4.2.3. Soil–pile interaction
software, a program for analyzing a pile under lateral loading,
For the pile foundation we use the distributed springs (DS)
and the data including the p–y curves at a 1 m spacing along the
model which idealizes the monopile with flexible foundation as a
length of the pile is provided as an input file in HAWC2 to model
free–free beam with lateral (Winkler-type) springs distributed
the soil. For this project we assume that the soil exerts the lateral
along the subsoil portion of the monopile. The beam uses the real
forces and the axial forces and torsional moment are negligible.
properties of the monopile both above and below the mudline.
Fig. 10 shows the soil profiles and properties for every layer which
The spring stiffness of the subsoil portion is calculated based on
is in accordance with the data from the Belwind wind farm. Also
the p–y model and is depth-dependent (Jonkman et al., 2009). Soil
the p–y curves are represented at different depths.
models for sand show a relatively simple non-linear behavior.
Such models are well described in certain standards and the
derivation is not very complicated (Passon).
5. Damping of offshore wind turbines
Based on the two soil properties effective unit weight and angle
of internal friction together with the pile diameter the API
The overall damping of the wind turbine support structures
standard (AP Institute, 2000) describes an easy procedure to
consists of the aerodynamic damping, damping due to construc-
derive the p–y curves over depth. The ordinate for p–y curves is
tive devices and additional damping, e.g. structural damping. The
soil resistance, p, and the abscissa is soil deflection, y. By iterative
additional damping is influenced by further effects, e.g. soft soil
procedures, a compatible set of load-deflection values for the pile-
and hydrodynamic damping (WindEnergie, 2004). Therefore, the
soil system can be developed. As the soil profile in this study is
overall measured system damping (Dtot ) can be approximated as a
mainly consist of sand, here we just demonstrate the relations for
linear combination of following damping sources:
sand. The lateral soil resistance–deflection (p–y) relationships for
sand are non-linear and may be approximated at any specific Dtot ¼ Dstruc þ Dsoil þ Daero þ Dhydro þ Dmass:damp ð3Þ
depth H, by the following expression:

 with Dstruc , material damping of structure; Dsoil , soil damping due
kH to inner soil friction; Daero , aerodynamic damping; Dmass:damp ,
P ¼ Apu tanh y ð2Þ
Apu tower oscillation damper; Dhydro ¼ Dradiation þ Dvis;hydro , hydrodynamic

Fig. 10. Different soil profiles for the monopile support (left) and the p–y curves at different depths (right).
 R. Shirzadeh et al. / J. Wind Eng. Ind. Aerodyn. 120 (2013) 96–106 103

damping which consists of two terms; Dradiation , damping from wave as relevant in case of wind and wave misalignment, no aerody-
creation due to structure vibration; Dvis;hydro , viscous damping due to namic damping is present (Design basis part c, 2009).
hydrodynamic drag. Although the wind speeds at both overspeed stop and ambient
For this study we will select the different damping contribu- excitation tests are very low and the wind turbine is in parked
tions based on best practice and engineering judgement in such a condition, the aerodynamic damping cannot be completely
way that the overall damping for the first FA-mode is around 1%, neglected. We have therefore determined the aerodynamic damp-
which is in agreement with the damping found in the two ing performing two different simulations. In the first simulation,
measurement campaigns, respectively the overspeed stop and the air density is set to zero to eliminate the aerodynamic loading
ambient excitation test, while the tuned mass damper was not on the wind turbine. For the second simulation, the air density is
active. In the next paragraph we will discuss the different damping 1.235 kg/m3. Both simulations have been performed for a wind
contributions and their selected value in more detail. turbine in parked condition at wind speed of 4.5 m/s and a pitch
angle of 80.5 deg. The total damping value for the first FA mode
5.1. Structural damping was identified for both cases and the difference in damping can be
considered as the contribution of the aerodynamic damping. The
Internal friction of a material transforms vibrational energy aerodynamic damping contribution has been evaluated to be
into heat. This phenomena is called structural damping. The 0.08% for the wind turbine at wind speed of 4.5 m/s.
structural damping model is included in the equations of motion
in order to assure dissipation of energy from the structural system. 5.4. Hydrodynamic damping
The structural damping is covered in many papers and stan-
dards. Typical values of structural damping ratios for steel struc- The hydrodynamic loading and damping have long been of
tures are in the order of 0.5%–1.5% (Design basis part c, 2009). special interest to offshore engineering. In fact, the forces exerted
Here, the lower value of 0.5% accounts for the pure material on bodies subjected to vortex induced vibrations and to the
damping while higher values might be achieved for structures springing and ringing response of compliant systems require a
with additional damping sources such as joints. Also, the support clear understanding of the flow in the range of Keulegan–Carpen-
structure in offshore wind turbines consists of a pile and transition ter numbers from about K ¼10
3 to 10 and frequency parameters
2
piece grouted together. It is expected that the grout material will βð ¼ fD =νÞ from about 1 to 108.
introduce additional damping in the system. Moreover, to estimate According to Eq. (3) there are two sources of the hydrodynamic
the structural damping value, one should perform measurements. damping: wave radiation and viscous damping due to hydrody-
Experience (Veritas, 2009) shows that the structural damping in namic drag. Wave radiation is directly proportional to velocities,
terms of the logarithmic decrement is usually of the order of 3% where the latter is proportional to the squares of velocities. As
(ζ≈0:48%) for the blades and of 5% (ζ≈0:8%) for the shaft and the described in the hydrodynamic model section, the drag forces are
tower. Taking to account all matters mentioned above, we have of less importance for offshore wind turbines. To validate this, we
adopted a 0.6% for the structural damping in the simulations. performed several time domain simulations with different drag
coefficients (Cd). This has been noticed that the drag coefficient has
a negligible influence on the total damping (around 0.004%). On
5.2. Soil damping
the other hand, the support structure deflections and relative
velocities are small and therefore wave radiation will give a larger
During the operation of an offshore wind turbine the displace-
contribution to hydrodynamic damping. According to Leblanc and
ment of the subsoil structure causes a cyclic motion of the
Tarp-Johansen (2011), the damping ratio from wave radiation is
surrounding soil, leading to a combination of geometrical dissipa-
approximately 0.12% when considering an offshore wind turbine
tion and material damping. Extensive studies of wind turbines on
with a first eigenfrequency of 0.3 Hz, a pile diameter of 4.7 m and a
a homogeneous or layered ground show that geometric dissipa-
water depth of 20 m.
tion is insufficient at frequencies below 1 Hz (Tarp-Johansen and
The contribution of hydrodynamic damping can be evaluated
DONG Energy, 2009). However, soil damping is typically of low
by time domain simulations for a parked wind turbine after
importance for monopile foundations (Design basis part c, 2009)
specifying the structural and soil damping as input values. For
comparing to other damping sources.
the current study, this gives a hydrodynamic damping value of
The soil damping is introduced using a Rayleigh kind of
0.07% which is in agreement with the literature (WindEnergie,
damping (k) in HAWC2 (Larsen and Hansen, 2007). We have
2004).
assessed the contribution of the soil damping in the total damping
for different k values. We selected a k value of 1%, which agrees
with a soil damping contribution of 0.25%. This value will allow us 6. Simulation results
to have an overall additional offshore damping of 1% as will be
shown later. The numerical simulations have been performed in HAWC2 for
an overspeed stop test and an ambient excitation test to compare
5.3. Aerodynamic damping the frequency and damping value of the first FA mode with the
values obtained from the measurements. The time domain results
As the wind turbine support structure vibrates, the wind are processed using the PolyMAX OMA approach, and the stabi-
turbine interacts with the air which results in damping of the lization diagrams will be given for both test cases.
system's vibrations. This aerodynamic damping effect is part of an
important physical phenomenon for wind turbines called aero- 6.1. Overspeed stop test
elasticity; essentially, the vibrations of the offshore wind turbine
components are strongly coupled to the system's aerodynamic For the overspeed stop test, the wind speed is 4 m/s at the
mechanisms. beginning and is increased to 6.5 m/s. At time 85 s, the pitch angle
A rather high aerodynamic damping is achieved when the wind changes from 0 to 88.2 degrees. The thrust due to sudden
turbine is in the production state. Nevertheless, in a non- collective pitch angle variation excites the tower mainly in the
production state as well as in perpendicular rotor direction, e.g. wind direction (FA mode). The top-level acceleration in the for-aft
104 R. Shirzadeh et al. / J. Wind Eng. Ind. Aerodyn. 120 (2013) 96–106

direction and movement seen from above are plotted in Fig. 11. Fig. 12 7. Conclusions
shows the Fast Fourier Transformation of the accelerations obtained
on the 3 different levels in FA direction. From output spectra different In order to extend the life time of an offshore wind turbine, a
peaks are detectable. The first dominant peak occurs at frequency correct estimation of the eigenfrequencies and damping values is
0.36 Hz is the first FA mode. A smaller peak can be seen before the crucial. In this paper, the overspeed stop test and ambient
first peak around 0.28 Hz which is related to the wave loads. excitation have been considered to estimate the first for-aft mode
Furthermore, some peaks with higher frequencies can be seen after frequency and damping ratio of the support structure using both
the first peak. These peaks correspond to the higher support structure real measurements and simulations. The structural properties of
and blade modes. The PolyMAX estimator is applied on stabilization wind turbine as well as the wave kinematics, wind condition and
chart to identify the dominant frequencies and damping values. The soil stiffness parameters are implemented in HAWC2. The results
pre-processing parameters were chosen according to the best practice are processed applying a state of the art OMA approach PolyMAX
to achieve good estimates with PolyMAX. Table 4 represents a to identify the frequency and damping ratio of the first FA mode.
comparison between simulation and measurement for the estimated We assume that the overall system damping is a combination
resonance frequency and damping ratio of the first FA mode. One can of different sources. For this study we selected the different
easily see that the first FA frequency and damping value compare very damping contributions based on best practice and engineering
well with the measurement. judgment in such a way that the additional offshore damping for
the first FA-mode is around 1%. Therefore, a detailed discussion on
6.2. Ambient excitation the different damping contribution has been given.
Simulation results for the damping ratio of FA mode are in good
For the ambient excitation, the wind speed is 4.5 m/s and the agreement with the measurement data obtained during the both
blade pitch angle is set to 80.5 degrees from the beginning of the overspeed stop and ambient excitation tests. The mass tune
simulation. The simulation is performed for 1 h. Here again the damper was not active during two sets of measurements and we
time history of the acceleration for 3 levels in for-aft direction and found the damping values of 1.05% and 1.27% for respectively the
top view of the tower are displayed in Fig. 13. In a similar way with first FA and the first SS modes.
the measurements, the tower does not vibrate purely in the wind Two topics will be investigated in the future by the authors.
direction and there is a contribution of the SS movement. Firstly, significant research is needed to determine the dynamics of
We noticed that the PolyMAX estimator; when applied to the
outputs power spectra due to ambient excitation; is able to
identify a stable mode at the vicinity of the dominant peak around Table 4
0.35 Hz (1st FA mode) similar to the overspeed stop test case. Overspeed stop test comparison between simulation and experimental results for
estimated natural frequency and damping ratio for the 1st FA mode.
Considering the 40 min of ambient data, the stabilization
diagram is plotted in Fig. 14. One can clearly identify several stable HAWC2 Measurements
modes in this diagram corresponding to the eigenfrequencies of
the offshore wind turbine. A comparison for the frequency and 1st FA frequency Damping ratio 1st FA frequency Damping ratio
damping ratio of the first FA mode is provided in Table 5. Similar to (Hz) (%) (Hz) (%)

the overspeed stop test, there is a good agreement between 0.3688 1.0568 0.3529 1.05
simulation and measurements.

Fig. 11. Simulation results for movement seen from above (left) and accelerations (g) on 3 levels in FA direction during overspeed stop test (right).

Fig. 12. Fast Fourier Transformation of the correlation functions in the FA-direction (left) and stabilization diagram obtained from overspeed stop simulation (right).
 R. Shirzadeh et al. / J. Wind Eng. Ind. Aerodyn. 120 (2013) 96–106 105

Fig. 13. Simulation results for movement seen from above (left) and accelerations (g) on 3 levels in FA direction during ambient excitation (right).

Fig. 14. Fast Fourier Transformation of the correlation functions in the FA-direction (left) and stabilization diagram obtained from 40 min data of ambient excitation
simulation (right).

Table 5 Cauberghe, B., 2004. Applied Frequency-Domain System Identification in the Field
Ambient excitation comparison between simulation and experimental results for of Experimental and Operational Modal Analysis. Ph.D. Thesis. Faculty Of
estimated natural frequency and damping ratio for the 1st FA mode. Engineering, Vrije Universiteit Brussel.
Chauhan, S., Tcherniak, D., Basurko, J., Salgado, O., Urresti, I., Carcangiu, C., Rossetti,
HAWC2 Measurements M., 2011. Operational modal analysis of operating wind turbines: application to
measured data. Rotating Machinery, Structural Health Monitoring, Shock and
Vibration 5, 65–81.
1st FA frequency Damping ratio 1st FA frequency Damping ratio
Damgaard, M., Ibsen, L.B., Andersen, L.V., Andersen, J.K.F., 2013. Cross-wind modal
(Hz) (%) (Hz) (%)
properties of offshore wind turbines identified by full scale testing. Journal of
Wind Engineering and Industrial Aerodynamics 116, 94–108, http://dx.doi.org/
0.3667 1.02 0.3565 1.05 10.1016/j.jweia.2013.03.003.
Design Basis Part C – Bligh Bank Offshore Wind Farm. Technical Report, Ramboll Wind.
Devriendt, C., Jordaens, P.J., De Sitter, G., Guillaume, P., 2012. Damping estimation of
an offshore wind turbine on a monopile foundation. In: EWEA 2012, Copenha-
gen, Denmark.
an offshore wind turbine subjected to wind and wave actions. It is Din, E., 19902:2008-07: Petroleum and natural gas industries-fixed steel offshore
structures (ISO 19902:2007), English version EN ISO 19902.
a part of ongoing work to implement a detailed study on the Hansen, M., Thomsen, K., Fuglsang, P., Knudsen, T., 2006. Two methods for
influence of different hydrodynamic and wind parameters includ- estimating aeroelastic damping of operational wind turbine modes from
ing the different wave angle, wave period, tidal level (wave height) experiments. Wind Energy 9 (1–2), 179–191.
〈http://www.marin.nl/web/events/events-2010/〉.
and wind speeds. Secondly, additional experimental investigations 〈http://www.marin.nl/web/events/events-2010/presentations-offshore-wind-semi
are required to evaluate the contribution of aerodynamic damping nar-2010.htm〉.
at different wind speeds for both operational and parked wind James, G., Carne, T., Veers, P., Damping measurements using operational data.
Journal of Solar Energy Engineering 118 (3), 190–193.
turbines. A detailed comparison between simulations and mea- Jonkman, J., Butterfield, S., Musial, W., Scott, G., 2009. Definition of a 5-MW
surements will allow to determine the aeroelastic behavior of Reference Wind Turbine for Offshore System Development. National Renew-
OWTs at different operational and ambient conditions. able Energy Laboratory.
Larsen, T.J., Hansen, A.M., 2007. How 2 HAWC2, the User's Manual, Risø National
Laboratory.
Larsen, T.J., Madsen, H.A., Hansen, A.M., Thomsen, K., 2005. Investigations of
Acknowledgements stability effects of an offshore wind turbine using the new aeroelastic code
HAWC2. In: Proceedings of Copenhagen Offshore Wind 2005, Copenhagen,
Denmark, pp. 25–28.
The research presented in this paper is conducted in the
Leblanc, C., Tarp-Johansen, N.J., 2011. Monopile in sand, stiffness and damping. In:
framework of the “Offshore Wind Infrastructure Application Lab” Oral Presentation, EWEA Conference, Brussels.
www.owi-lab.be. The authors gratefully thank the people of Bel- Magalhães, F., Cunha, Á., Caetano, E., Brincker, R., 2010. Damping estimation using
wind NV for their support before, during and after the installation free decays and ambient vibration tests. Mechanical Systems and Signal
Processing 24 (5), 1274–1290.
of the measurement equipment. Parloo, E., 2003. Application of Frequency-Domain System Identification Techniques
in the Field of Operational Modal Analysis, Ph.D. Thesis.
Passon, P., Memorandum: derivation and description of the soil-pile-interaction
References
models. IEA-Annex XXIIII Subtask 2.
Peeters, B., Van der Auweraer, H., Guillaume, P., Leuridan, J., 2004. The polymax
AP Institute, 2000. Recommended Practice for Planning, Designing, and Construct- frequency-domain method: a new standard for modal parameter estimation?
ing Fixed Offshore Platforms-Working Stress Design: Upstream Segment. API Shock and Vibration 11 (3–4), 395–410.
Recommended Practice 2A-WSD (RP 2A-WSD), API. Petersen, B., Pollack, M., Connell, B., Greeley, D., Davis, D., Slavik, C., Goldman, B.,
Bendat, J.S., Piersol, A.G., 1980. Engineering Applications of Correlation and Spectral Medley, L., 2010. Evaluate the Effect of Turbine Period of Vibration Require-
Analysis. Wiley-Interscience, New York 315 p. 1. ments on Structural Design Parameters: Technical Report of Findings.
106 R. Shirzadeh et al. / J. Wind Eng. Ind. Aerodyn. 120 (2013) 96–106

Tarp-Johansen, N.J., DONG Energy, 2009. Comparing sources of damping of cross- Verboven, P., 2002. Frequency-Domain System Identification for Modal Analysis. Ph.
wind motion. In: European Offshore Wind Conference (EOW). DONG Energy. D. Thesis.
Tcherniak, D., Chauhan, S., Hansen, M.H., 2011. Applicability limits of operational Veritas, D., 2009. Guidelines for design of wind turbines. DNV.
modal analysis to operational wind turbines. In: Structural Dynamics and WindEnergie, G.L., 2004. Guidelines for the certification of offshore wind turbines,
Renewable Energy, vol. 1. Springer, pp. 317–327. Final Draft, Edition.
Van Der Tempel, J., 2006. Design of Support Structures for Offshore Wind Turbines. Zaaijer, M., Van der Tempel, J., 2004. Scour protection: a necessity or a waste of
Ph.D. Thesis. Technical University of Delft. money? In: Proceedings of the 43 IEA Topixal Expert Meeting, pp. 43–51.
Veldkamp, H., Van Der Tempel, J., 2005. Influence of wave modelling on the
prediction of fatigue for offshore wind turbines. Wind Energy 8 (1), 49–65.