Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

J. Wind Eng. Ind. Aerodyn.

116 (2013) 94–108

Contents lists available at SciVerse ScienceDirect

Journal of Wind Engineering


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

Cross-wind modal properties of offshore wind turbines


identified by full scale testing
M. Damgaard a,n, L.B. Ibsen b, L.V. Andersen b, J.K.F. Andersen a
a
Vestas Wind Systems A/S, Aarhus, Denmark
b
Department of Civil Engineering, Aalborg University, Aalborg, Denmark

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

Article history: According to the Danish wind turbine industry cross-wind vibrations due to wave loading misaligned
Received 3 August 2012 with wind turbulence often have a significant influence on the fatigue lifespan of offshore wind
Received in revised form turbine foundations. The phenomenon is characterised by increasing fatigue loads compared to the fore-
18 February 2013
aft fatigue and a small amount of system damping since almost no aerodynamic damping
Accepted 2 March 2013
from the blades takes place. In addition, modern offshore wind turbines are flexible structures with
Available online 13 April 2013
resonance frequencies close to environmental loads and turbine blades passing the tower. Therefore, in
Keywords: order to avoid conservatism leading to additional costs during the load calculation and the design
Fast Fourier Transformation phase, the structural response must be analysed using reliable estimations of the dynamic properties of
Eigenfrequency
the wind turbines. Based on a thorough investigation of “rotor-stop” tests performed on offshore
Free vibration decay
wind turbines supported by a monopile foundation for different wind parks in the period 2006–2011, the
Modal soil damping
Offshore wind energy structures paper evaluates the first natural frequency and modal damping of the structures. In addition,
Operational modal identification fitting of theoretical energy spectra to measured response spectra of operating turbines is
p–y curve method presented as an alternative method of determining the system damping. Analyses show distinctly
Scour time-dependent cross-wind dynamic properties. Based on numerical analysis, the variation is believed to
Spectral analysis be caused by sediment transportation at seabed level and varying performance of tower oscillation
Winkler approach dampers.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction engineering and science. Overall, the price of offshore wind


energy must decrease in order to be competitive with present
In the last 30 years, the need of alternative energy sources alternatives. As a consequence, offshore wind turbines have
has been given great attention. During this period, wind increased significantly in size with larger rotors, more power-
turbines have grown from producing less than 100 kW to ful generators and only a small increase in weight. Thus,
more than 6 MW. At this writing, the majority of wind modern wind turbines are slender and flexible structures
turbines are located onshore due to lower construction costs. showing dynamic response – even at low frequencies. The
However, the population density and existing buildings limit structural resonance frequency is close to the excitation
suitable wind turbine locations on land in many regions of the frequencies related to turbine blades passing the tower and
world. This justifies the development of offshore wind energy. waves, which means that a small change in the total system
Unlike onshore wind turbines, the size of offshore wind stiffness may result in a great change in the response. Conse-
turbines is not limited by the logistics problems during quently, provision of sufficient damping in the entire structure
transportation of large structural components. The amazing is crucial in order to decrease the fatigue damage accumula-
growth in the installed capacity of onshore and offshore wind tion during the lifetime of the wind turbine structure.
turbines, however, causes many challenges within civil In general, the dynamic behaviour of civil engineering
structures is characterised in terms of their modal parameters.
The modal representation can be determined analytically.
n
Corresponding author. Tel.: þ45 41188639. However, in cases where the structural system cannot be
E-mail addresses: mdamg@vestas.com (M. Damgaard),
lbi@civil.aau.dk (L.B. Ibsen), la@civil.aau.dk (L.V. Andersen),
represented by a lumped mass-spring system, a numerical
jakfa@vestas.com (J.K.F. Andersen). approximation is often made by discretising the structure in a

0167-6105/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jweia.2013.03.003
M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108 95

finite number of physical coordinates. The structural design Despite of the above-mentioned advanced loading conditions
often relies on the numerical model. However, a high number for a wind turbine, Hansen et al. (2006), Tcherniak et al. (2010)
of degrees of freedom is needed, which leads to higher and Ozbek and Rixen (2011) have all used operational modal
calculation time and unchanged inherent modelling accuracy analysis to estimate modal parameters of operating wind
limitations related to nonhomogeneous elements, complex turbines.
materials, boundary conditions etc. Experimental modal ana- In order to evaluate the cross-wind modal damping of
lysis addresses these limitations and is able to validate and offshore wind turbines, Tarp-Johansen et al. (2009),
improve the numerical model. Versteijlen et al. (2011) and Devriendt et al. (2012) have
Traditional experimental modal analysis is characterised by investigated “rotor-stop” tests, where aeroelastic damping
forced vibrations, where the input excitation is measured. can almost be neglected when the blades are pitched out of
Determination of resonance frequencies, damping ratios and the wind. Devriendt et al. (2012) showed that the system
mode shapes of structures is well established for this techni- modal damping was in agreement with the findings from Cook
que. The structure to be identified is artificially excited and Vandiver (1982) and Tarp-Johansen et al. (2009). However,
with a forcing function in a specific point, and its response a large deviation is observed when comparing the estimated
to this excitation is measured together with the forcing modal soil damping from Tarp-Johansen et al. (2009) with the
signal. Fourier Transformation of the time signal makes it work performed by Versteijlen et al. (2011). This in turn calls
possible to calculate all the frequency response functions for even more comprehensive research in the area, which
between the response and the forcing signals, i.e. the fre- justifies this paper.
quency response matrix. This matrix contains all the informa- Based on five offshore wind parks, experimental modal
tion to determine the dynamic properties of the structure identification of offshore wind turbine structures is considered
(Cantieni, 2004). in the present work, which is an updated and revised version
Impact excitation has been widely used for smaller civil of the conference paper by Damgaard et al. (2012). Simple
engineering structures and offers the advantages of quick set- estimation of crossing times and least-squares fitting to the
up time, mobility and the ability to excite a broad range of natural logarithm of the vibration extremes from “rotor-stop”
frequencies. Askegaard and Mossing (1988), Agardh (1991), tests, make it possible to evaluate the cross-wind modal
Wood et al. (1992), Aktan et al. (1992), Green and Cebon parameters. Similarly, it is attempted to determine the struc-
(1994) and Pate (1997) have all excited full scale bridge tural dynamic properties within true boundary conditions and
structures by impact excitation tools, either by dropping an actual force and vibration level by using spectral analysis on
impact weight or by a bolt gun with impulsive load. However, the response during power production. In general, cross-wind
by using this method, the wave form of the excitation is not dynamic properties of offshore wind turbines depend on
controlled. Consequently, a more common method of exciting several parameters. Tidal variation, soil-structure interaction,
civil engineering structures is by electrodynamic shakers. wind speed and tower damper performance all have an
Modal tests performed by Shepherd and Charleson (1971), influence on the eigenfrequency and the corresponding modal
Kuribayashi and Iwasaki (1973), Ohlsson (1986), Cantieni and damping. Based on numerical simulations, qualified reasons
Pietrzko (1993), Deger et al. (1993), Deger et al. (1994), for the variation in the measured cross-wind dynamic proper-
Miloslav et al. (1994) and Caetano et al. (2000) have used ties are stated in the paper.
shakers to excite bridge structures. Also, forced shaker excita- The outline of the paper is as follows; firstly, full scale
tion has been used for dam structures in order to obtain the testing regarding “rotor-stop” tests is described. Turbine char-
modal properties, see Duron (1995a), (1995b), Cantieni (2001) acterisation and procedures of determining the dynamic
and Nuss et al. (2003). Besides, Cantieni et al. (1998) have properties are listed. Secondly, numerical investigations of
estimated modal parameters of an office building by using natural frequency and modal soil damping are considered for
shaker excitation. different soil conditions for a specific turbine. The chapter
Full scale measurements of offshore support structures forms the basis of evaluating to what extent cross-wind
within the oil and gas sector have been reported by Cook dynamic properties of offshore wind turbines change during
and Vandiver (1982) and Jensen (1990). In those publications, time. Thirdly, an explanation of the theory and procedure of
spectral analyses have been used to estimate natural frequen- analysing cross-wind modal damping of operating turbines by
cies and modal damping ratios. For wind turbine structures, spectral analysis is given followed by a thorough documenta-
the presence of rotational loads and considerable aeroelastic tion of results. Finally, a brief summary of the main findings of
effects makes it difficult to use traditional experimental modal the present work are listed with a description of what the
analysis techniques. Using this method, the experimentally findings can be used for in future applications.
determined vibrations are not the pure modes of the turbine.
The shaker excitation simply ignores the ambient loads acting
on the turbine during operation. However, some few attempts 2. Turbine stop analysis—free decay
to excite parked wind turbines by measurable excitation forces
have been done, see Carne et al. (1988), Molenaar (2003), Wind turbine structures are characterised by closely spaced
Hansen et al. (2006) and Osgood et al. (2010). modes occurring at nearly identical frequencies, see for
A complementary technique to the traditional modal ana- instance Ibsen and Liingaard (2006) and Ibsen (2008). In this
lysis technique is operational modal identification, also regard, a coupling of the two lowest eigenmodes is possible,
denoted output-only modal identification, which is based on which means that pure fore-aft and side–side bending modes
measuring only the responses of test structures (Batel, 2002). cannot be assumed. Vibrational energy will be transferred
96 M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108

0.8
0.6

Fore-Aft Acceleration ay [m/s2]


0.6

0.4 0.4
Fore-Aft Acceleration ay [m/s2]

0.2
0.2
0.0

0.0 -0.2

-0.4
-0.2
-0.6

-0.4 -0.8
0 20 40 60 80 100 120
Time t [s]
-0.6

-0.8
-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 90

Side-Side Acceleration ax [m/s2] 80

70
Fig. 1. Fore-aft acceleration ay as a function of side–side acceleration ax during a

Blades Pitch θb [°]


“rotor-stop”. 60

50

40

30

20

z 10

0
0 20 40 60 80 100 120
y x Time t [s]

Fig. 3. Raw output signal: (a) fore-aft tower acceleration ay as function of time t,
(b) blade pitch angle θb as function of time t.

ax
example, Fig. 3 shows the raw output time domain signal. The
fore-aft tower acceleration ay and blade pitch angle θb as a
ay function of time t are shown. A sampling frequency of 10 Hz
has been used.

2.1. Wind turbine structures and site conditions

The foundation concept for the four wind parks is the well-
proven monopile concept. The concept consists of a tubular
Fig. 2. Monitoring system: (a) acceleration transducers placed in the top of the
tower in order to measure the fore-aft acceleration ay and the side-side accelera- steel pile section with a grouted transition piece. The soil
tion ax, (b) right hand coordinate system. profiles consist mainly of cohesionless soil in the top layers
followed by cohesive soils. For each turbine, an oscillation
damper is built into the top of the tower just beneath the
from the highest to the lowest damped mode, which may give nacelle. It consists of a pendulum partly immersed in high
wrong estimates of the modal damping, see Hansen et al. viscous oil and able to oscillate in the two horizontal direc-
(2006) and Magalhães et al. (2010). However, it turns out that tions. In Table 1, the overall characterisation of the four wind
“rotor-stop” tests of wind turbines are beneficial in order to parks is listed. Due to numerical simulations of a selected wind
achieve pure modal vibrations from one single mode due to turbine structure located at Wind Park I, detailed geometry
low influence of external forces. Fig. 1 shows the fore-aft and soil conditions are presented in Fig. 4 and Table 2 for this
acceleration ay as a function of the side–side acceleration ax structure.
during a “rotor-stop”, in which it is clearly observed that only
the fore-aft mode is excited.
In the period 2006–2011, more than 1500 “rotor-stop” tests 2.2. Procedure for estimation of the first eigenfrequency
have been investigated at four offshore wind parks. By use of
two accelerometers placed in the top of the tower, the tower In order to find the first eigenfrequency f1 of each wind
acceleration in the fore-aft direction y and the side–side turbine structure, two methods are considered in the paper.
direction x have been measured. Fig. 2a and b shows the The first method is based on how much information each
monitoring system and coordinate system, respectively. As an measured signal contains at different frequencies by use of a
M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108 97

Table 1
Overall characterisation of the four wind parks.

Tower height Monopile Soil Average water


[m] diameter conditions depth
[m] [-] w.r.t. LAT [m]

Park I 60.0 4.3 Dense 6/8


sand/Firm clay
Park II 58.0 4.8 Dense 13/18
sand/Stiff clay
Park III 54.1 4.7 Fine 15/27
sand/Stiff clay
Park IV 53.0 5.0 Dense sand/Stiff 15/20
clay

Fig. 5. Frequency and time domain: (a) power energy spectrum of the fore-aft
tower acceleration ay, (b) total damped fore-aft acceleration decay ay.

Fast Fourier Transformation (FFT). To obtain the power in the


signal at each frequency, the square of the absolute value of
the FFT coefficients is considered. The Nyquist sampling
theorem is used in order to avoid aliasing. As an example,
the power spectrum of the fore-aft tower acceleration ay is
shown in Fig. 5a. The dominant frequency peak in Fig. 5a
corresponds to the first damped eigenfrequency fd,1 of the
wind turbine structure. The width of the spectral bell indicates
the modal damping, which can be determined by the Half-
Band Width method (Nielsen, 2004). However, a high resolu-
tion frequency must be considered for this method. As an
alternative method, the first damped eigenfrequency fd,1 has
been determined by making least-squares fitting to the cross-
ing times in the time domain. The eigenfrequency f1 can then
Fig. 4. Geometry of the selected wind turbine structure at Wind Park I decomposed be determined by the following relation:
into: (a) monopile foundation, (b) transition piece, (c) tower. All dimensions are in
millimetres. f d,1
f 1 ¼ qffiffiffiffiffiffiffiffiffiffi
ffi, ð1Þ
1−ς21

where ζ1 is the modal damping ratio for the first eigenmode


Table 2
Φ(1). Linear interpolation has been used to find the
Detailed soil conditions for the selected wind turbine structure at Wind Park I. crossing times.

Depth [m] γ′ [kN/m3] φk′ [deg.] cuk [kPa]


2.3. Procedure for estimation of the first modal damping
Loose sand −5.0 7.0 30.0 –
Dense sand −6.5 8.0 33.8 – In general, during a “rotor-stop”, the oscillatory deformation
Dense sand −7.5 8.0 30.0 –
of the wind turbine structure is attenuated. This may be due to a
Dense sand −8.4 7.0 30.0 –
Firm clay −9.4 8.0 – 25.0 combination of geometrical damping, i.e. radiation of energy into
Dense sand −11.8 9.0 36.0 – the subsoil, sea, air and material damping due to conversion of
firm clay −20 9.0 – 90.0
mechanical energy into heat. The last-mentioned is a measure of
Firm clay −30 10.0 – 110.0
inherent damping of the structural system. In order to estimate
98 M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108

the damping from the modal vibrations, least-squares fitting of a


linear function to the natural logarithm of the rate of decay of the
transient response has been used. The absolute value of the
negative amplitudes is included in order to achieve a reliable
least-squares fit to the amplitude peaks. Each time history is
divided into bins with approximately 10 maxima and 10 minima
of the acceleration from which the logarithmic decrement δ1 can
be estimated. As indicated in Fig. 3a, the tower behaves almost as
a single-degree-of-freedom (SDOF) system with linear viscous
damping, and the corresponding damping coefficient cv is so low
that the system is undercritically damped. The logarithmic
decrement δ1 is defined by
  
2 A0  
δ1 ¼ ln An  , ð2Þ
n
where A0 and An are two extremes in the response placed with
the time interval nTd. Hence, the logarithmic decrement δ1 can be
found by least squares fitting on nδ1 and 2ln(|An|). For low levels
of damping, the following relation between the damping ratio ζ1
and the logarithmic decrement δ1 can be stated:
δ1 δ1
ζ 1 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ≈ for δ1 5 1 ð3Þ
4π 2 þ δ1 2 2π
Quadratic interpolation has been used to find the extreme
values of the acceleration decay. It should be noted that the
Fig. 6. Numerical representation of the selected wind turbine structure at Wind
decay of the extremes depend on the acceleration level. Hence,
Park I.
the system cannot be characterised by an unambiguous value
of the logarithmic decrement δ1. So, in principle, the presump-
structural z-axis. Epy is the modulus of subgrade reaction
tion within Eq. (2) is not fulfilled. Still, within the period of
which depends on both y and z. In Eq. (4), it is assumed that
30 s (corresponding to the time frame in which the least-
the soil has no shear stiffness, but only lateral stiffness
squares fit data are determined), the damped acceleration
represented by nonlinear springs based on semi-empirical
decay agrees well with Eq. (2). In order to avoid aerodynamic
relations between the soil force per unit length p acting
affects from the rotor blades when they pitch out of the wind,
against the monopile wall and the lateral deflection y of the
the first 15 s after initiation of the “rotor-stop” are neglected.
structure, see Fig. 6.
As an example, the damped acceleration decay is shown in
For modal analysis of offshore wind turbines, it is common
Fig. 5b. The exponential function that specifies the decrease of
to estimate the spring stiffness Enpy by linearising the nonlinear
the vibration amplitude with time t is included.
p–y curves suggested by DNV (2011). Theoretically, the
linearising can be done by use of either the initial soil stiffness
3. Computational model from the nonlinear uncoupled soil springs or from the
secant and tangent soil stiffness when the turbine is producing
A numerical representation of a selected wind turbine power at the nominally rated output level. However,
structure at Wind Park I, cf. Fig. 4 and Table 2 is performed studies regarding cyclic loading of piles made by Klinkvort
in order to estimate the dynamic properties for different et al. (2011) and Roesen et al. (2012) indicate that the
environmental conditions. By use of a Winkler foundation unloading-reloading path almost follows the initial stiffness
model, the first eigenfrequency f1 and the corresponding soil Enpy of the virgin curve. Despite decreasing secant stiffness
damping δsoil can be evaluated for different seabed conditions. during cyclic loading, it seems sensible for a modal analysis of
offshore wind turbines to determine the spring stiffness
3.1. Winkler approach from the initial stiffness of the nonlinear p–y curve
formulation, cf. Fig. 6.
Laterally loaded monopiles are traditionally designed based
By differentiating the p–y curve relationship for piles in
on a Winkler approach. Assuming that the pile and tower act
cohesionless soils according to DNV (2011), the initial stiffness
as
Enpy is given by
a Bernoulli–Euler beam, the governing differential equation is
  
given by ∂p  ∂ kzy 
Enpy ¼  ¼ Apu tanh  ¼ kz, ð5Þ
4 ∂y y ¼ 0 ∂y Apu y¼0
d y
Es I s −Epy y ¼ 0, ð4Þ
dx4 where pu is the ultimate lateral resistance per unit length at a
where y is the lateral deflection of the structure at a point z depth below soil surface, k is the initial modulus of subgrade
along the pile, EsIs is the bending stiffness of the structure, reaction determined from the peak angle of internal friction φk′,
where Es is the modulus of elasticity and Is is the moment of y is the lateral deflection and A is a factor accounting for cyclic
inertia around the horizontal x-axis perpendicular to the or static loading conditions. For cyclic loading, A is 0.9. For
M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108 99

clayey soils, DNV (2011) recommends to linearise the nonlinear of the elastic energy that is converted into heat (Andersen,
p–y curves from the discretisation point given by the relative 2006). However, due to the fact that soils dissipate elastic
displacement y/yc ¼0.1 with ordinate value p/pu ¼0.23, where energy hysteretically by the slippage of grains with respect to
yc ¼2.5ε50D. D is the pile diameter and ε50 is the strain each other, their dissipation characteristics are insensitive to
corresponding to a stress of 50% of the ultimate stress in a frequency. Hence, for a given deformation level and constant
laboratory stress–strain curve. frequency the soil material damping can be approximated to
The Winkler approach is implemented in MATLAB in order an equivalent viscosity. In this paper, a simple method of
to find the eigenfrequencies fk and corresponding eigenmodes determining the soil material damping contribution of the
Φ(k) of the selected offshore wind turbine at Wind Park I. The selected offshore wind turbine at Wind Park I is presented
differential equation for undamped vibrations of a multi- based on the p–y curves proposed by the design regulations.
degree-of-freedom (MDOF) system is solved. By performing a static deformation analysis with a load level
€ þ Ku ¼ 0;
Mu ð6Þ based on the measured wind speed during the “rotor-stop”
test, the irreversible deformations in the soil are a measure of
where M and K are the mass and stiffness matrices and u(t) is energy dissipation in the first cycle after the “rotor-stop” takes
the generalised displacement and rotational vector. In order to place. The method assumes that the first eigenmode Φ(1) is
find the eigenfrequency fk for the kth eigenmode Φ(k), a equal to the static deflection of the wind turbine structure. The
harmonic function is applied as a solution to Eq. (6). procedure is stated below:
uðtÞ ¼ ReðΦðkÞ eiωk t Þ; ð7Þ
1. A 10-minute transient simulation of the operating wind turbine
where it is used that the kth angular eigenfrequency ωk of the structure is conducted by use of the aeroelastic code FLEX (Øye,
harmonic motion u(t) is given by 1996). A wind speed level equal to the measured value just
before the “rotor-stop” takes place is included in order to
ωk ¼ 2πf k ð8Þ
determine the load level when the “rotor-stop” sequence starts.
Inserting Eq. (7) into Eq. (6) makes it possible to find the kth 2. By performing a static deformation analysis based on the
eigenfrequency fk and corresponding eigenmode Φ(k) by sol- Winkler approach with nonlinear p–y curves, the horizontal
ving the frequency condition pile deformation in each node point below the soil surface is
determined. The initial slope of the curves has a large influence
detðK−ω2k MÞ ¼ 0 ð9Þ on the deformations, and thus, the nonlinear p–y curves for the
cohesive soil stratification are discretised and approximated by
The structural mass contribution to the global mass matrix
a piece-wise linear curve drawn between the discretisation
M from monopile, transition piece and tower is calculated
points according to DNV (2011).
according to the Bernoulli–Euler beam theory. The mass of the 3. Based on a load–displacement cycle, as indicated in Fig. 7a, the
nacelle and rotor is added as a concentrated mass in the top area captured within the curve is found by Simpson integra-
node. Likewise, masses of the tower flanges and internal tower tion. The area is used to obtain the hysteresis loop for the
equipment are added as concentrated masses. To account for damping force Fd as shown in Fig. 7b. In other words, based
the increased mass and stiffness in the presence of the grout on the soil material damping in Fig. 7a, an equivalent viscous
annulus between the pile and the transition piece, an equiva- damping model is considered.
lent steel wall thickness is used. The monopile is assumed 4. At each node point below the soil surface, the damping
flooded. Further, the model is fixed in the x-direction, which constant c can be found from the hysteresis loop. Hence, the
seems admissible for a monopile structure excited at its first global damping matrix C is determined.
5. Based on the theory of MDOF systems, the soil damping ratio
resonance frequency. It should be noted that Simpson integra-
ζsoil is determined from the global damping matrix C, the
tion has been used to estimate the element soil stiffness
angular eigenfrequency ω1, the eigenmode Φ(1) and the modal
matrix by determining the initial soil stiffness Enpy in each mass M1 according to Eq. (10).
integration point. Hence, the global stiffness matrix K is a
combination of structural and soil stiffness. Φð1ÞT CΦð1Þ
ζ soil ¼ , ð10Þ
In order to take seabed sediment erosion due to current 2ω1 M 1
actions into consideration in the numerical model, the vertical where the modal mass M1 is given by
effective stress p0 is reduced linearly with depth to a depth
M 1 ¼ Φð1ÞT MΦð1Þ ð11Þ
equal to 3D below the base of the scour hole.

3.2. Modal soil material damping—hysteresis loop method

Besides tower oscillation dampers, soil damping is believed


to contribute with the greatest amount of the total system
damping for offshore wind turbines on a monopile foundation,
when considering “rotor-stop” sequences with vanishing aero-
dynamic damping. For a “rotor-stop” test, cyclic motion in the
soil takes place leading to material damping and geometric
damping. However, Andersen (2008) concludes that the geo-
metric damping can be neglected for frequencies below 1 Hz. Fig. 7. Hysteresis Loop Method (Nielsen, 2004): (a) load–displacement curve for
the first load cycle after the “rotor-stop”, (b) hysteresis loop implied by viscous
Within the earthquake engineering field it is common to damping in a harmonic motion with the amplitude A and the first angular
implement a viscoelastic model to represent the dissipation eigenfrequency ω1.
100 M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108

The nonlinear model in Fig. 7a assumes a loading phase where the


soil-pile interaction is described by the p–y curve formulation

Mean Power Energy Spectrum [m/s2]2


7
given by DNV (2011). The unloading phase follows the initial
stiffness Enpy of the virgin curve. Due to irreversible deformations, a 6

gap between the pile and the soil will be created. The pile will 5
then move towards the initial position and into the opposite soil
4
face. In this phase, the soil resistance to the lateral movement of
the pile is expressed by a shear drag pdrag developed along the side 3
of the pile. Based on Zang et al. (2005), the shear drag pdrag can be 2
given by
1
pdrag ¼ Dτmax , ð12Þ
0
0.26 0.28 0.30 0.32 0.34 0.36 0.38 0.40 0.42 0.44 0.46
where τmax is the maximum shear resistance. According to Ovesen Frequency [Hz]
et al. (2006), the maximum shear resistance τmax for a steel
monopile placed in cohesionless soils can be expressed as

Mean Power Energy Spectrum[m/s2]2


τmax ¼ 0:6s′v , ð13Þ 6 2 DOF Response Fit
Measured
where s′v is the vertical effective stress. For cohesive soils, the 5
maximum shear resistance τmax is given by
4
τmax ¼ 0:7cu ð14Þ
3
Whether the gap will take place for cohesionless soils is difficult to
derive from the literature. According to El-Naggar et al. (2005), 2

cohesionless soils will cave in and close the gap. However,


1
centrifuge tests carried out for a pile in sand made by Klinkvort
(2009) show that the soil will not cave in completely. This in turn 0
0.26 0.28 0.30 0.32 0.34 0.36 0.38 0.40 0.42 0.44 0.46
may justify the application of the nonlinear material model in
Frequency [Hz]
Fig. 7a for both cohesionless and cohesive soils.
Fig. 8. Spectral analysis on the response during power production: (a) mean power
energy spectrum for all generator speeds as a function of frequency, (b) measured
and theoretical 2 DOF energy spectra.
4. Spectral analysis on response during power production

As an alternative method of estimating the dynamic prop-


erties of offshore wind turbines from experimental testing,
spectral analysis on side–side accelerations during power 4.2. Procedure of damping estimation
production is presented in this paper. Contrary to evaluating
The following procedure of determining the first modal
the dynamic parameters from “rotor-stop” sequences, this
damping δ1 from side–side accelerations ax has been used:
method has the advantages of being valid for operating
turbines, i.e. the dynamic properties are investigated within
1. Each side–side acceleration time series is divided into sequences
true boundary conditions and actual force and vibration level. with duration of approximately 200 s (2048 samples at a sample
For two offshore turbines, a total of 27 measurements, each rate of 10 Hz). The sequences are divided into bins with similar
with a time duration of approximately 12 h, have been conditions in terms of rotor speed.
investigated in order to estimate the first modal damping δ1. 2. Fast Fourier Transformation of each time sequence into an
The measurements have been performed in the period from energy spectrum is performed. The rotational rotor frequency
1 May to 31 June 2009. Fig. 2a and b shows the monitoring 1P is filtered out since this excitation has nothing to do with
system and coordinate system, respectively. the system response.
3. Considering the spectral bell for the first damped eigen-
frequency fd,1, makes it possible to determine mean energy
4.1. Wind turbine structure and site condition spectra of at least 30 spectra for each bin, cf. item 1 and
Fig. 8a.
The two considered wind turbine structures are part of an 4. Assuming a damped 2 degree-of-freedom (DOF) system, see
offshore wind park with a total of 30 wind turbines installed on a Fig. 9, the global mass, stiffness and damping matrices are
monopile foundation. The soil profiles contain mainly frictionless found by fitting an energy spectrum to each measured mean
spectrum, cf. Fig. 8b.
soil. In Table 3, the overall characterisation of Wind Park V can
5. Solving the nonlinear eigenvalue problem for a damped 2 DOF
be seen.
system makes it possible to determine the modal damping δ1
for the first eigenmode Φ(1).
Table 3
Overall characterisation of Wind Park V.

Tower Monopile Soil conditions Average water The theoretical fitting of the measured energy spectrum is
height [m] diameter [m] [-] depth w.r.t. obtained by minimising the object function given by
LAT [m]
n
Park V 60.0 4.3 Dense sand/Firm clay 6/8 E ¼ ∑ jTðωÞ−SðωÞj2 , ð15Þ
i¼1
M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108 101

k1 k2 0.999
0.997
0.990
m1 mv 0.980
0.950
c1 c2 0.900

Probability [-]
0.750

x2(t) 0.500
x1(t)
0.250
Fig. 9. Equivalent dynamic system with two degree of freedoms representing the
offshore wind turbine with a mechanical damper installed. 0.100
0.050
0.020
0.010
where S(ω) is the measured energy spectrum and T(ω) is the 0.003
theoretical energy spectrum given by 0.001

TðωÞ ¼ HðωÞFðωÞ ¼ ð−ω2 M þ iωC þ KÞ−1 FðωÞ, ð16Þ 0.000


-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15

where H(ω) is the response function and F(ω) is the input force Side-Side Acceleration ax [m/s2]
derived from Gaussian white noise. The mass matrix M, the Fig. 10. Normal probability plot of a selected time serie. A mean value of
damping matrix C and the stiffness matrix K for an equivalent −0.0002 m/s2, a skewness of 0.06 and a kurtosis of 2.9 are present.
damped 2 DOF system represent the turbine with a mechan-
ical damper according to Fig. 9. The matrices are given by
" #
1
μ 0
M ¼ μm1 , ð17Þ
0 1
"ζ #
1 ð1 þ μÞ
2ωref μm1 μ þζ 2 −ζ 2
C¼ , ð18Þ
1þμ −ζ 2 ζ2

 ref 2 " #
ð1 þ μÞ2
ω μm1 μ þ1 −1
K¼ , ð19Þ
ð1 þ μÞ2 −1 1

where it has been utilised that the mass ratio μ, the angular
reference frequency ωref, the damping constants c1 and c2 are
determined by
m1
μ¼ ð20Þ
m2
sffiffiffiffiffiffiffi
ref k1
ω ¼ ω1 ¼ ð21Þ
m1
Fig. 11. “Rotor-stop” sequences for Wind Park I with an R-square value of 0.95:
c1 ¼ 2ζ 1 ω1 m1 ð22Þ (a) Damping histogram, (b) Eigenperiod histogram, (c) Campbell diagram in terms
of the logarithmic decrement δ1 as a function of the fore-aft acceleration ay,
μ (d) Campbell diagram in terms of the eigenfrequency f1 as a function of the fore-aft
c2 ¼ 2ζ 2 ω2 m2 ¼ 2ω1 m1 ζ ð23Þ
1þμ 2 acceleration ay.

By keeping the mass ratio μ constant, the optimum of the


object function in Eq. (15) is found by varying the parameters precisely normal distributed, for practical purpose are a
m1, ωref, ζ1 and ζ2. Finally, solving the following nonlinear realisation of a zero-mean Gaussian process.
eigenvalue problem gives the first modal damping δ1: Further, as mentioned earlier, closely spaced modes occur-
ring at nearly identical frequencies are present for wind
detðMλ2 þ Cλ þ KÞ ¼ 0 ð24Þ turbines. In this regard, the method assumes that no coupling
It should be noticed that the method relies on the assump- between the two first eigenmodes exists, i.e. pure fore-aft and
tion that the input forces are derived from Gaussian white side–side modes are assumed. Stochastic Subspace Identifica-
noise and excite in multiple points. This assumption implies tion will be used to validate this assumption (Brincker and
that each time serie should be investigated for Gaussianity by Andersen, 2006).
calculating the mean value, the skewness and kurtosis. If the
time serie is a Gaussian process the skewness and kurtosis
theoretically are 5. Interpretation of “rotor-stop” sequences
0 and 3, respectively. In Fig. 10, a normal probability plot of a
selected time serie is shown with a mean value of −0.0002 m/ Based on 665 “rotor-stop” tests on 30 turbines for Wind
s2, a skewness of 0.06 and a kurtosis of 2.9. Overall, it can be Park I, the structural dynamic properties are presented in
concluded that the investigated time series, though not Fig. 11 with an R-square value of at least 0.95, meaning that the
102 M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108

fit explains 95% of the total variation in the data about the 140 200
average. An upper limit for the first eigenfrequency f1 of

"Rotor-Stop" Sequencies [-]

"Rotor-Stop" Sequencies [-]


X=0.159 180 X=2.844
120 S=0.047 S=0.078
160
offshore wind turbines exists due to the fact that the value 100 140

never exceeds the first eigenfrequency f1 for a fixed turbine. 80 120


100
Hence, lognormally distributed eigenperiods may be expected 60 80
40 60
due to its reciprocal description of the eigenfrequency. Fig. 11a 40
20
and b show the histogram of the first modal damping 0
20
0
in terms of the logarithmic decrement δ1 and eigenperiod T1 0 0.10 0.20 0.30 0.40 2.4 2.6 2.8 3.0 3.2
Logarithmic Decrement δ1 [-] Eigenperiod T1 [s]
for the tests with a mean value of 0.15 and 3.024 s and a
standard deviation of 2.5% and 4.8% respectively. Instead of 60 60

"Rotor-Stop" Sequencies [-]

"Rotor-Stop" Sequencies [-]


X=0.163 X=2.837
representing the parameters by use of a histogram, Fig. 11c and 50 S=0.050 50 S=0.078
d show the modal damping δ1 and eigenfrequency f1 as a 40 40
function of the fore-aft acceleration level ay. A relatively high 30 30
scatter is seen on the modal damping δ1. Increasing the R- 20 20
square value to 0.99 decreases the number of “rotor-stop” tests 10 10
from 665 to 298 and also the scatter of the modal damping δ1 0 0
0 0.10 0.20 0.30 0.40 2.5 2.7 2.9 3.1
to some extent. However, scatter is still observed, cf. Fig. 12. Logarithmic Decrement δ1 [-] Eigenperiod T1 [s]
The modal damping δ1 decreases when the acceleration level
ay increases. This indicates that the performance of the tower 30 30

"Rotor-Stop" Sequencies [-]

"Rotor-Stop" Sequencies [-]


X=0.161 X=2.806
oscillation damper is significantly reduced. Same tendency is 25 S=0.043 25 S=0.074

observed for the eigenfrequency f1, which might be caused by 20 20

distinct nonlinear soil behaviour for high acceleration levels. 15 15

This reduces the secant stiffness and thereby changes the 10 10

boundary conditions. 5 5

The distribution of the first modal damping δ1 and eigen- 0


0 0.10 0.20 0.30 0.40
0
2.5 2.7 2.9 3.1
period T1 for Wind Park II, III and IV are shown in Fig. 13 with Logarithmic Decrement δ1 [-] Eigenperiod T1 [s]
an R-square value of 0.95. Mean values and standard devia-
Fig. 13. “Rotor-stop” sequences for Wind Park II, III and IV with an R-square value
tions of the modal damping δ1 and eigenperiod T1 for the three of 0.95: (a) damping histogram for Wind Park II, (b) eigenperiod histogram for
wind parks correspond very well with the findings for Wind Wind Park II, (c) damping histogram for Wind Park III, (d) eigenperiod histogram
for Wind Park III, (e) damping histogram for Wind Park IV, (f) eigenperiod
Park I. Using a lognormal probability density function, a 5%
histogram for Wind Park IV.
quantile of the modal damping δ1 and the eigenperiod T1 for
all four wind parks can be seen in Table 4. As indicated,
sensible similarities between the wind parks are seen. Finally,
it should be mentioned that Damgaard et al. (2012) have
performed a similar study of an offshore wind turbine Table 4
5% quantile of the first modal damping δ1 and eigenperiod
with a monopile foundation located in the North Sea.
T1 for all four wind parks.
Based on 10 “rotor-stop” tests a mean value of 0.14 of the
logarithmic decrement δ1 with a standard deviation of 1.51% δ5% T1
[-] [s]
was observed.
Fig. 14 shows a locally weighted, scatter smooth plot of the Park I 0.11 2.95
eigenfrequencies and modal damping values for the four wind Park II 0.09 2.71
Park III 0.10 2.70
parks. The higher deviation of the eigenfrequency estimations
Park IV 0.10 2.69
might be caused by different soil conditions for the four wind
parks. Better agreement is observed for the modal damping
estimations due to the installation of the tower oscillation
damper in all the investigated turbines. Despite varying soil

Wind Park I Wind Park I


Logarithmic Decrement δ1 [-]

0.30 Wind Park II 0.40 Wind Park II


Natural Frequency f1 [Hz]

0.30 0.37 Wind Park III 0.39 Wind Park III


Logarithmic Decrement δ1 [-]

0.25 Wind Park IV 0.38 Wind Park IV


0.36
Eigenfrequency f1 [Hz]

0.25
0.20 0.37
0.35 0.36
0.20
0.15 0.35
0.34
0.15 0.34
0.33 0.10
0.33
0.10 0.32
0.32 0.05
0.31
0.05 0.31 0 0.30
0 0.2 0.4 0.6 0.8 1.0 1.2 0 0.2 0.4 0.6 0.8 1.0 1.2
0 0.30
0 0.2 0.4 0.6 0.8 1.0 1.2 0 0.2 0.4 0.6 0.8 1.0 1.2 Fore-Aft Acceleration ay [m/s2] Fore-Aft Acceleration ay [m/s2]
Fore-Aft Acceleration ay [m/s2] Fore-Aft Acceleration ay [m/s2]
Fig. 14. Local weighted linear regression to smooth out the dynamic properties for
Fig. 12. “Rotor-stop” sequences for Wind Park I with an R-square value of 0.99 the four wind parks shown in a Campbell diagram: (a) logarithmic decrement δ1 as
shown in a Campbell diagram: (a) logarithmic decrement δ1 as a function of fore-aft a function of fore-aft acceleration ay, (b) eigenfrequency f1 as a function of fore-aft
acceleration ay, (b) eigenfrequency f1 as a function of fore-aft acceleration ay. acceleration ay.
M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108 103

damping between the wind parks, it must be noticed that the is calculated in accordance with EN 1991-1-4 (2005).
tower oscillation damper has a higher contribution to the
measured system damping. ζ aero ≈0:12% ð27Þ

5.1. Selected turbine investigation

It can be concluded that there is relatively high scatter in


the dynamic properties, cf. Figs. 11–13. Different reasons can
explain this tendency. Each wind turbine structure has a
unique geometry designed for specific soil stratification and
environmental conditions. However, it has been observed that
the mass pendulum of the oscillating tower damper during
some “rotor-stop” tests hits the surfaces of the oil basin.
Hence, the time duration of pitching the blades out of the
wind and the acceleration level in the turbine, together with
the initial position of the mass pendulum before the “rotor-
stop” takes place, have a high impact on the tower damper
performance. In some “rotor-stop” tests, the movement of
the mass pendulum correlates with the movement of
the wind turbine tower and creates almost no additional
damping. In other tests, the movement of the two systems is
out of phase causing a high damping value from the tower
damper. Hence, a more proper way of comparing the dynamic
properties is by investigating each individual turbine with the
following requirements:

 Ensuring same slope of generator speed when the blades pitch


out of the wind.
 Ensuring same acceleration level during the “rotor-stop” test.

Fig. 15 shows comparison of the measured 10 min wind


speed and tidal variation together with the dynamic properties
of three turbines at Wind Park I. For each turbine, the “rotor-
stop” tests are collected with almost same acceleration level
and slope of decreasing generator speed. As indicated, the first
eigenfrequency f1 and the modal damping δ1 now only have a
slight deviation for each turbine.

5.2. Linear combination of damping contributors – an example of


modal soil damping estimation

Based on the measured system damping ζ1 for the first


eigenmode Φ(1), it is possible to estimate the soil damping for
a specific “rotor-stop” test of the selected wind turbine placed
at Wind Park I, cf. Fig. 4 and Table 2. In general, the measured
system damping ζ1 can be approximated as a linear combina-
tion of the following parts due to low levels of damping:
ζ 1 ¼ ζ steel þ ζ aero þ ζ water þ ζ tower þ ζ soil , ð25Þ
where ζsteel is the steel hysteretic damping ratio, ζaero is the
aerodynamic damping ratio from the tower, ζwater is the wave
making radiation and viscous hydrodynamic damping ratio
due to the presence of water, ζtower is the damping ratio from
the tower oscillation damper and ζsoil is the soil damping ratio.
According to EN 1991-1-4 (2005), the material damping for the
monopile and tower steel ζsteel can be estimated to
ζ steel ≈0:19% ð26Þ
With a measured averaged wind speed vwind of 13.4 m/s Fig. 15. Detailed investigation of three selected turbines at Wind Park I with same
acceleration level ay and slope of generator speed: (a) wind and tidal variation as a
and a first eigenfrequency f1 of 0.323 Hz for the investigated function of measurement time, (b) logarithmic decrement δ1 and eigenfrequency f1
turbine, the aerodynamic damping on the tower structure ζaero as a function of measurement time.
104 M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108

Theoretical estimation of ζwater has been presented by


Leblanc and Tarp-Johansen (2010) based on research per- Scour
Backfilling
formed by Tarp-Johansen et al. (2009). Due to small relative 0.336

Eigenfrequency f1 [Hz]
0.334
velocities for an offshore wind turbine structure, the viscous
0.332
hydrodynamic damping ratio is small compared to damping 0.330
created from wave radiation. According to Leblanc and Tarp- 0.328
Johansen (2010), the damping ratio from wave radiation only 0.326
0.324
is approximately 0.12% when considering an offshore wind 0.322
turbine with a first eigenfrequency f1 of 0.3 Hz, a pile diameter 0.320
of 4.7 m and a water depth of 20 m. The damping value is 0.318
assumed to represent the investigated wind turbine at Wind 0.316
0 1 2 3 4 5 6
Park I, i.e., Scour Depth/Backfilling Height [m]
ζ water ≈0:12% ð28Þ

For the specific turbine at Wind Park I, the lowest modal Scour
Backfilling
damping ζ1 has been measured as 1.43, cf. Fig. 12a. Assuming 0.070

Soil Damping δsoil [-]


that the tower oscillation damper does not perform any 0.065
additional damping, i.e. ζtower ¼0, the soil damping required 0.060
to make the total theoretical damping equal to the total 0.055
experimental damping is determined to 0.050
0.045
ζ soil ≈1:00%, ð29Þ
0.040
which is equal to a logarithmic decrement δsoil of approxi- 0.035
mately 0.06. Notice that Damgaard et al. (2012) have per- 0.030
0 1 2 3 4 5 6
formed a similar study of another offshore wind turbine. A
Scour Depth/Backfilling Height [m]
modal soil damping in terms of the logarithmic decrement δsoil
of 0.036 was estimated where the soil stratification primarily Fig. 16. Numerical analysis of the selected wind turbine structure at Wind Park I,
(a) first eigenfrequency f1 as a function of scour depth and backfilling height,
consisted of cohesive soils with a 3 m loose sand layer placed (b) soil damping in terms of the logarithmic decrement δsoil as a function of the
in the top. scour depth and backfilling height.

5.3. Variation of modal parameters with time

Despite of collecting “rotor-stop” tests with the same This statement is supported by research performed by Larsen
acceleration level and slope of generator speed for each et al. (2005).
individual turbine at Wind Park I, the dynamic properties are A much more vital phenomenon is the erosion of soil
time-dependent, cf. Fig. 15. Assuming that the tower oscillat- particles near the foundation. Based on experimental tests,
ing damper contributes with the same damping value, the Sumer et al. (1992) determined that the mean value of the
time-dependency might be caused by the following: equilibrium scour depth for a vertical cylinder in steady
current is given by 1.3D, where D is the diameter of the
 Temperature dependent dynamic properties cylinder. However, for combined current and wave conditions,
 Tidal variation it becomes more complicated as wave action will tend to
 Moveable seabed and scour around the foundation reduce the scour depth (Høgedal and Hald, 2005). As a
consequence, the scour depth will change with time and
The mechanical properties of steel depend on temperature thereby also the eigenfrequency f1 and modal damping δ1.
level, which have an influence of the structural stiffness and Based on the previously mentioned Winkler model, a numer-
damping. However, due to a small damping value of structural ical parameter study of the first eigenfrequency f1 and corre-
material damping, it is assumed that the variation caused by sponding soil damping in terms of the logarithmic decrement
temperature changes has a negligible impact on the measured δsoil has been performed for different scour depths and backfill
system damping. According to Engineering ToolBox (2012), a heights for the selected turbine at Wind Park I, cf. Fig. 4 and
temperature change from −731 to 93 1C causes a reduction of Table 2. A maximum scour depth equal to the equilibrium
Young's modulus of elasticity E with approximately 5%. scour depth proposed by Sumer et al. (1992) has been applied
Using a Winkler approach and assuming no scour develop- and a wind speed of 13 m/s is assumed. The cohesionless
ment, it corresponds to a change in the first eigenfrequency f1 backfill material is assumed to have a friction angle φk′ of 281.
of 0.5%. Fig. 16a shows as expected a decreasing eigenfrequency f1 for
One-year measurements of tide levels at Wind Park I show increasing scour depth. The opposite is observed for increasing
a maximum difference between highest and lowest astronom- backfill height. According to Fig. 16b, the soil damping δsoil
ical tide of approximately 2 m. However, according to Eq. (28), tends to decrease for increasing scour depth. However, only to
the damping ratio from wave radiation is only 0.12% for a a certain depth, hereafter an increasing soil damping δsoil is
water depth of 20 m. It might then be concluded that the tidal observed, and for large scour depths the soil damping δsoil
variation has almost no influence on the dynamic properties. decreases again. The opposite is observed for increasing
M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108 105

32

Spectral Density ax[dB|m/s2/Hz]


0.346
Eigenfrequency f1 [Hz]

0.344 40000 22
0.342
0.340

Time [s]
0.338
12
0.336
0.334 20000
0.332
-2
0.330
0.328
28 30 32 34 36 38 40
Friction Angle φk' [°] 0 -8
0 0.30 0.60 0.90
Frequency [Hz]

Fig. 18. Spectrogram of side-side accelerations ax for an operating offshore wind


turbine at Wind Park V. The rotational rotor frequency 1P, the first damped
0.10
eigenfrequency fd,1 and the blade passing frequency 3P are clearly represented in
Soil Damping δsoil [-]

0.09 the plot.

0.08

0.07

0.06

0.05 gated offshore wind turbine might be caused by scour


0.04 development.
28 30 32 34 36 38 40
Friction Angle φk' [°]
6. Interpretation of power production sequences
Fig. 17. Numerical analysis of the selected wind turbine at Wind Park I, (a) eigen-
frequency f1 as a function of friction angle φk′ of backfill material, (b) soil damping
in terms of the logarithmic decrement δsoil as a function of the friction angle φk′ of Modal identification of an operating wind turbine is diffi-
backfill material. cult to perform due to complicated loading forces. Time-
invariance of the structure during the test is a general
requirement in modal testing, i.e. the structure under test
remains the same during the test. This is not the case for an
operational wind turbine structure. The nacelle revolves
backfill height. In order to understand this tendency, Fig. 7 around the tower, the rotor rotates around its axis and the
must be considered. The plastic soil deformations, which are a blades pitch depending of the wind speed and rotor speed
measure of soil material damping, highly depend on the pile (Damgaard, 2011). However, dividing time sequences into bins
deflection and the initial soil stiffness Enpy for each soil layer, i.e. with similar conditions in terms of rotor speed and identifying
for increasing scour depth, a higher pile deflection at seabed the rotational rotor frequency 1P and separating it from the
and a higher initial soil stiffness Enpy create a higher soil structural modes, spectral analysis on side-side accelerations
damping contribution. However, due to the fact that the pile ax during power production may be an alternative way of
deflection at the base of the scour hole only increases to a estimating modal parameters. Fig. 18 shows the time-varying
certain scour depth and the initial soil stiffness Enpy for each soil spectral representation of a measurement carried out for one
layer depends on the scour depth, it becomes complicated. The of the operating wind turbines at Wind Park V. Using short-
pile deflection and the initial soil stiffness Enpy will be able to time Fourier Transformation of the side–side acceleration
increase or decrease relative to each other for increasing signal, Fig. 18 indicates that the energy related to the rotational
scour depth. rotor frequency 1P and the first damped eigenfrequency fd,1
According to Sørensen et al. (2010) compaction of the only to some extent vary with time.
backfilled material will take place during time due to the Using spectral analysis to determine the first modal damp-
presence of waves inducing depth compaction. Hansson et al. ing δ1 from side–side accelerations ax of operating wind
(2005) have reported friction angles above 401 for top soil turbines requires that the shape of the first two closely spaced
layers. Using the Winkler approach, Fig. 17 shows the eigen- eigenmodes only show pure bending in the fore-aft and side–
frequency f1 and the modal soil damping δsoil as a function of side direction, i.e. an elliptical trace of each eigenmode is not
the strength of the backfill material. allowed. The requirement has been verified through a Sto-
Overall, the numerical investigations of the scour develop- chastic Subspace Identification (SSI) technique. The theory
ment and backfill material show a variation of the first within the method will not be outlined in this paper. However,
eigenfrequency f1 of around 8%. The soil damping δsoil is in by using the SSI algorithm, it can be concluded that the two
the range of 0.05–0.08 logarithmic decrement. Comparing closely spaced eigenmodes only show pure bending in the
these results with the experimental findings in Fig. 15, fore-aft and side–side direction, respectively. It turns out that
this paper shows that time-dependent natural frequencies the modal damping of the fore-aft eigenmode is three
and corresponding modal damping values of the investi- times higher than the side–side eigenmode, which might be
106 M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108

use of a Winkler approach. Several interesting observations


0.25 can be made:
Logarithmic Decrement δ0 [-]

0.20  “Rotor-stop” tests of offshore wind turbine structures allow


determination of the modal parameters from the free vibration
0.15
decay. Considering more than 1,500 tests divided between four
wind parks, reliable and similar mean values of the first modal
damping in terms of the logarithmic decrement δ1 are observed
0.10
to be in the range of 0.15–0.16.
 “Rotor-stop” tests of offshore wind turbines induce distinctly
0.05
time-dependent eigenfrequencies and modal damping values
with a relatively high scatter, even for ideal fits. During a “rotor-
0
0 0.05 0.10 0.15 stop”, the initial position of the mass pendulum of the tower
Max Side-Side Acceleration Level ax [m/s2]
oscillating damper has a high impact on the damping contribu-
tion. In some tests, the mass pendulum follows the motion of
the main system and in other tests, it tends to move in the
opposite direction. Moreover, it has been observed that for
0.25 some tests the mass pendulum hits the side surfaces in the oil
Logarithmic Decrement δ0 [-]

basin. Sorting tests for each turbine with the same acceleration
0.20 level and slope of generator speed when the blades pitch out of
the wind produce much lower scatter.
0.15  Eliminating the varying tower damper performance, the
observed time-dependent modal parameters of offshore wind
0.10 turbines might be caused by sediment transportation at seabed.
A numerical investigation of this phenomenon shows that
0.05 scour development and backfilling change the eigenfrequency
f1 and the modal damping δ1 of the wind turbine structure.
0
Good agreement between experimental and numerical modal
0 0.05 0.10 0.15 0.20 0.25 0.30 parameters is observed.
Max Side-Side Acceleration Level ax [m/s2]  Qualified estimations of each damping contribution to the
measured system damping for a specific turbine make it
Fig. 19. Damping estimation for the two considered wind turbines at Wind Park V,
possible to determine modal soil damping δsoil in terms of the
(a) logarithmic decrement δ1 as a function of max side–side acceleration ax for
turbine 1, (b) logarithmic decrement δ1 as a function of max side–side acceleration
logarithmic decrement of 0.06. The paper has shown that using
ax for turbine 2. a Winkler approach with an implemented hysteresis loop
method predicts reasonable estimates of modal soil material
damping, even though the soil-structure interaction is mod-
elled by p–y curves designed for slender piles with a slender-
the reason for the observed characteristics of the two ness ratio of L/D o34.4.
structural modes.  Spectral analysis on side–side accelerations of operating off-
In Fig. 19, the damping estimation in terms of the logarith- shore turbines may serve as an alternative way of determining
the dynamic properties. True boundary conditions and actual
mic decrement δ1 is shown for the two considered turbines at
force and vibration level make this approach beneficial, which
Wind Park V. Mean values of 0.18 and 0.16 are observed, which
also produce a more stable performance of the tower damper. A
corresponds very well to the mean damping values from the mean value of the modal damping in terms of the logarithmic
“rotor-stop” tests. Moreover, it can be concluded that much decrement δ1 is in the range of 0.16–0.18.
smaller scatter is present when analysing the modal damping
from operating turbines. This might be caused by a more In this paper it has been assumed that the system damping
stable performance of the installed oscillating damper, i.e. the of an offshore wind turbine is a linear combination of different
damper is designed for actual vibration conditions, which contributors. Moreover, no viscous effects from pore water
characterise this type of tests. dissipation have been taken into account in the soil damping
estimation. The two topics will in the near future be
investigated into more details by the authors. In addition,
7. Conclusions model updating, in which a fully integrated model is refined
by measured data of an offshore wind turbine, will be
In order to assess the fatigue damage accumulation during examined.
the lifetime of offshore wind turbine structures, a correct Overall, significant research is needed for determining
estimation of the structural modal parameters is necessary. dynamic properties of offshore wind turbine structures. In this
In this regard, experimental and numerical studies of the regard, two major research directions have to be considered.
cross-wind dynamic properties of offshore wind turbines have Firstly, additional experimental investigations of the modal
been presented in this paper. By use of free vibration tests and parameters are required in order to draw strong conclusions of
normal production measurements of several offshore wind the magnitudes. This is especially important for cross-wind
turbine structures, reliable estimates of the structural dynamic vibrations, where the aerodynamic damping is negligible.
properties have been stated. Qualified reasons for the variation Secondly, developing, validating and updating numerical
of the eigenfrequency and modal damping are evaluated by modal analysis and numerical integrated models based on
M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108 107

the experimental findings are crucial in order to achieve cost- Duron, Z.H., 1995a. Seven Mile Vibration Testing Results from the First Series of Test
Performed February 20-07. Research Report Prepared for B.C Hydro and Power
effective designs. Especially cyclic pile–soil interaction models Authority, Canada.
must be given high priority in order to ensure a correct Duron, Z.H., 1995b. Seven Mile Vibration Testing Results from the Second Series of
Test Performed August 14–20. Research Report Prepared for B.C Hydro and
behaviour of dynamically loaded offshore turbines.
Power Authority, Canada.
El-Naggar, M.H., Shayanfar, M.A., Kimiaei, M., Aghakouchak, A.A., 2005. Simplified
bnwf model for nonlinear seismic response analysis of offshore piles with
nonlinear input ground motion analysis. Canadian Geotechnical Journal 42,
Acknowledgements 365–380.
EN 1991-1-4, 2005. Eurocode 1: Actions on Structures—Part 1-4: General Actions—
Wind Actions. European Committee for Standardization, Brussels.
The work presented within this paper is supported by the Engineering ToolBox, 2012. 〈http://www.engineeringtoolbox.com〉 (accessed
16.02.13).
research projects Cost Effective Monopile Design funded by Green, M.F., Cebon, D., 1994. Dynamic response of highway bridges to heavy vehicle
Energiteknologiske Udviklings- og Demostrationsprogram loads: theory and experimental validation. Journal of Sound and Vibration 170,
51–78.
(EUDP) and Cost Effective Deep Water Foundations Hansen, M.H., Thomsen, K., Fuglsang, P., Knudsen, T., 2006. Two methods for
for Large Offshore Wind Turbines funded by the Advanced estimating aeroelastic damping of operational wind turbine modes from
Technology Foundation. The financial supports are sincerely experiments. Wind Energy. 9, 179–191.
Hansson, M., Hjort, T.H., Thaarup, M., 2005. Data Report 0408 Fredrikshavn Sand.
acknowledged. Technical Report. Aalborg University.
Høgedal, M., Hald, T., 2005. Scour Assessment and Design for Scour for Monopile
Foundations for Offshore Wind Turbines. Copenhagen Offshore Wind,
Copenhagen.
Ibsen, L.B., 2008. Implementation of a new foundations concept for offshore
wind farms. In: Proceedings of Nordisk Geoteknikermøtte nr. 15. Sandefjord,
References Norway.
Ibsen, L.B., Liingaard, M., 2006. Prototype Bucket Foundation for Wind Turbines—
Agardh, L., 1991. Modal analysis of two concrete bridges in Sweden. Structural Natural Frequency Estimation. DCE Technical Report No. 9. Aalborg University.
Engineering International 1, 35–39. Jensen, J.L., 1990. Full-Scale Measurements of Offshore Platforms. Technical Report.
Aktan, A.E., Zwick, J., Miller, R.A., Sharooz, B.M., 1992. Nondestructive and Aalborg University.
destructive testing of a decommissioned RC slab highway bridge and associated Klinkvort, R.T., 2009. Laterally Loaded Piles—Centrifuge and Numerical Modelling.
analytical studies. Transportation Research Record. Master's Thesis. Technical University of Denmark.
Andersen, L.V., 2006. Linear Elastodynamic Analysis. DCE Lecture Notes No 3. Klinkvort, R.T, Hededal, O., Svensson, M., 2011. Laterally cyclic loading of monopile
Aalborg University. in dense sand. In: Proceedings of the 15th European Conference on Soil
Andersen, L.V., 2008. Assessment of lumped-parameter models for rigid footings. Mechanics and Geotechnical Engineering. vol. 15. pp. 203–208.
Computers and Structures 88, 1333–1347. Kuribayashi, E., Iwasaki, T., 1973. Dynamic properties of highway bridges. In:
Askegaard, V., Mossing, P., 1988. Long term observation on RC-bridge using changes Proceedings of the 5th World Conference on Earthquake Engineering.
in natural frequencies. Nordic Concrete Research 7, 20–27. pp. 938–941.
Batel, M., 2002. Operational modal analysis—another way of doing modal testing. Larsen, T.J., Madsen, H.A., Hansen, A.M., Thomsen, K., 2005. Investigation of Stability
Journal of Sound and Vibration, 22–27. Effects of an Offshore Wind Turbine Using the new Aeroelastic Code HAWC2.
Brincker, R., Andersen, P., 2006. Understanding stochastic subspace identification. Copenhagen Offshore Wind, Copenhagen.
In: Proceedings of the 24th International Modal analysis Conference. St. Louis, Leblanc, C., Tarp-Johansen, N.J., 2010. Monopiles in Sand. Stiffness and Damping.
Missouri. Oral Presentation, EWEA 2011, Brussels.
Caetano, E., Cunha, A., Taylor, C.A., 2000. Investigation of dynamic cable-deck Magalhães, F., Cunha, Á., Caetano, E., Brincker, R., 2010. Damping estimation using
interaction in a physical model of a cable-stayed bridge part I: modal analysis. free decays and ambient vibration tests. Mechanical Systems and Signal
International Journal Earthquake Engineering and Structural Dynamics 29, Processing. 24, 1274–1290.
481–498. Miloslav, B., Vladimir, B., Michal, P.., 1994. Dynamic behaviour of footbridge by
Cantieni, R., 2001. Assessing a dam's structural properties using forced vibration analysis and test. In: Proceedings of the 13th International Modal Analysis
testing. In: Proceedings of IABSE International Conference on Safety, Risk and Conference. vol. 1. pp. 687–693.
Reliability—Trends in Engineering, Malta. Molenaar, D.P., 2003. Experimental Modal Analysis of a 750 kW Wind Turbine for
Cantieni, R., 2004. Experimental Methods Used in System Identification of Civil Structural Modal Validation. 41st Aerospace Sciences Meeting and Exhibit.
Engineering Structures. 2nd Workshop: Problemi di vibrazioni nelle strutture Reno, Nevada.
civili e nelle costruzioni meccaniche, Perugia. Nielsen, S.R.K., 2004. Linear Vibration Theory, first ed. Aalborg Tekniske Universi-
Cantieni, R., Pietrzko, S., 1993. Modal testing of a wooden footbridge using random tetsforlag, Denmark.
excitation. In: Proceedings of the 11th International Modal Analysis Conference. Nuss, L.K., Chopra, A.K., Hall, J.F., 2003. Comparison of Vibration Generator Tests to
vol. 2. pp. 1230–1236. Analyses Including Dam-Foundation-Reservoir Interaction for Morrow Point
Cantieni, S.J., Pietrzko, S., Deger, Y., 1998. Modal investigation of an office building Dam. The Commission International Des Grande Barrages. 20th Congress of
floor. In: Proceedings of the 16th International Modal Analysis Conference. Large Dams.
vol. 2. pp. 1172–1178. Ohlsson, S., 1986. Modal testing of the Tjorn bridge. In: Proceedings of the 4th
Carne, T.G., Lauffer, J.P., Gomez, A.J., 1988. Modal testing of a very flexible 110 m International Modal Analysis Conference, Florida. pp. 599–605.
wind turbine structure. In: Proceedings of the 6th International Modal Analysis Osgood, R., Bir, G., Mutha, H., 2010. Full-scale modal wind turbine tests:
Conference. Kissimmee, Florida. comparing shaker excitation with wind excitation. In: Proceedings of the 28th
Cook, M.F., Vandiver, J.K., 1982. Measured and predicted dynamic response of a International Modal Analysis Conference IMAC XXVIII. Jacksonville, Florida.
single pile platform to random wave excitation. In: Proceedings of the 14th pp. 113–124.
Offshore Technology Conference. Houston, Texas. pp. 637–646. Ovesen, N.K., Fuglsang, L., Bagge, G., 2006. Lærebog i Geoteknik, second ed.
Damgaard, M., 2011. An Introduction to Operational Modal Identification of Polyteknisk Forlag, Denmark.
Offshore Wind Turbine Structures. DCE Technical Memorandum No. 13. Aalborg Ozbek, M., Rixen, D.J., 2011. Optical measurements and operational modal analysis
University. on a large wind turbine: lessons learned. In: Proceedings of the Society for
Damgaard, M., Ibsen, L.B., Andersen, L.V., Andersen, J.K.F., 2012. Natural frequency Experimental Mechanics Series. pp. 257–276.
and damping estimation of an offshore wind turbine structure. In: Proceedings Øye, S., 1996. FLEX 5 User Manual, Lyngby.
of the 20th International Offshore and Polar Engineering Conference. Rhodos, Pate, J.W., 1997. Dynamic testing of a highway bridge. In: Proceedings of the 15th
Greece. pp. 300–307. International Modal Analysis Conference. pp. 2028–2037.
Deger, Y., Cantieni, R., Pietrzko, S., 1994. Modal analysis of an arch bridge: Roesen, H.R., Ibsen, L.B., Andersen, L., 2012. Small-scale testing rig for long-term
experiment, finite element analysis and link. In: Proceedings of the 12th cyclic loaded monopiles in cohesionless soil. In: Proceedings of the 16th Nordic
International Modal Analysis Conference. vol. 1. pp. 425–432. Geotechnical Meeting. Copenhagen, Denmark. pp. 435–442.
Deger, Y., Cantieni, R., Pietrzko, S., Ruecker,W., Rohrmann, R., 1993. Modal Shepherd, R., Charleson, A.W., 1971. Experimental determination of the dynamic
analysis of a highway bridge: experiment, finite element analysis and link. In: properties of a bridge substructure. Bulletin of the Seismological Society of
Proceedings of the 13th International Modal Analysis Conference. vol. 2, America 61, 1529–1548.
pp. 1141–1149. Sumer, B.M., Fredsøe, J., Christiansen, N., 1992. Scour around a vertical pile in
Devriendt, C., Jordaens, P.J., De Sitter, G., Guillaume, P., 2012. Damping estimation of waves. Journal of Waterway, Port, Coastal and Ocean Engineering, ASCE 117,
an offshore wind turbine on a monopile foundation. In: Proceedings of the 15–31.
EWEA 2012 Conference, Copenhagen. Sørensen, S.P.H., Ibsen, L.B., Frigaard, P., 2010. Experimental evaluation of
DNV, 2011. Design of Offshore Wind Turbine Structures. Det Norske Veritas backfill in scour holes around offshore monopiles. In: Proceedings of the 2nd
AS. International Symposium on Frontiers in Offshore Geotechnics. Perth, Australia.
108 M. Damgaard et al. / J. Wind Eng. Ind. Aerodyn. 116 (2013) 94–108

Tarp-Johansen, N.J., Andersen, L., Christensen, E.D., Mørch, C., Kallesøe, B., Frandsen, structure caused by its interaction with soil. In: Proceedings of the EWEA
S., 2009. Comparing Sources of Damping of Cross-Wind Motion. The European Offshore 2011 Conference, Amsterdam.
Offshore Wind Conference & Exhibition, Stockholm. Wood, M.G., Friswell, M.I., Penny, J.E.T., 1992. Exciting large structures using a bolt
Tcherniak, D., Chauhan, S., Rosseth, M., Font, I., Basurko, J., Salgado, O., 2010. gun. In: Proceedings of the 10th International Modal Analysis Conference.
Output-Only Modal Analysis on Operating Wind Turbines: Application to pp. 233–238.
Simulated Data. European Wind Energy Conference. Warsaw, Poland. Zang, L., Silva, F., Grismala, R., 2005. Ultimate lateral resistance to piles in
Versteijlen, W.G., Metrikine, A.V., Hoving, J.S., Smid, E., De Vries, W.E., 2011. cohesionless soils. Journal of Geotechnical and Geoenvironmental Engineering
Estimation of the vibration decrement of an offshore wind turbine support 131, 78–83.

You might also like