Mechanical Systems and Signal Processing 99 (2018) 285–305

Contents lists available at ScienceDirect

Mechanical Systems and Signal Processing

journal homepage: www.elsevier.com/locate/ymssp

Semi-active control of monopile offshore wind turbines under

multi-hazards
C. Sun
Department of Civil and Environmental Engineering, Louisiana State University, Baton Rouge, LA 70803, USA

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

Article history: The present paper studies the control of monopile offshore wind turbines subjected to
Received 16 February 2017 multi-hazards consisting of wind, wave and earthquake. A Semi-active tuned mass damper
Received in revised form 12 June 2017 (STMD) with tunable natural frequency and damping ratio is introduced to control the
Accepted 17 June 2017
dynamic response. A new fully coupled analytical model of the monopile offshore wind
turbine with an STMD is established. The aerodynamic, hydrodynamic and seismic loading
models are derived. Soil effects and damage are considered. The National Renewable
Keywords:
Energy Lab monopile 5 MW baseline wind turbine model is employed to examine the per-
Offshore wind turbine
Multi-hazards
formance of the STMD. A passive tuned mass damper (TMD) is utilized for comparison.
Damage Through numerical simulation, it is found that before damage occurs, the wind and wave
Semi-active control induced response is more dominant than the earthquake induced response. With damage
Vibration and damping presence in the tower and the foundation, the nacelle and the tower response is increased
dramatically and the natural frequency is decreased considerably. As a result, the passive
TMD with fixed parameters becomes off-tuned and loses its effectiveness. In comparison,
the STMD retuned in real-time demonstrates consistent effectiveness in controlling the
dynamic response of the monopile offshore wind turbines under multi-hazards and dam-
age with a smaller stroke.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction

Due to environmental concerns, wind energy production and consumption has experienced a remarkable growth world-
wide in the passed decade and is projected to grow more rapidly in the following decade. Under the circumstance, offshore
wind plants are becoming more attractive than its onshore counterparts because of advantages including the steadier and
higher wind speed, less visual impacts and less noise constraints. However, due to the combined multiple hazards including
wind, wave and earthquake, excessive vibration and fatigue load will result, adversely influencing the structural integrity
and service life. In this regard, structural vibration control, which has been successfully applied in civil structures, is being
studied to control the offshore wind turbines.
Three basic control strategies have been developed: passive, semi-active and active [1,2]. Passive control of offshore wind
turbines has been studied actively in recent years. Murtagh et al. [3] studied the control of the wind turbine along-wind
vibration using a passive tuned mass damper (TMD). The authors concluded that response reduction could be archievd when
the TMD was tuned to the predominant frequency. Lackner et al. [4] used dual passive TMDs placed in the nacelle to control
both the fore-aft and side-side vibration. It was found that the dual TMDs can reduce the structural response of offshore wind
turbines. Colwell and Basu [5] used the tuned liquid column damper to control the vibration of an offshore wind turbine.

E-mail address: csun@lsu.edu

http://dx.doi.org/10.1016/j.ymssp.2017.06.016
0888-3270/Ó 2017 Elsevier Ltd. All rights reserved.
286 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305

Research findings indicated that the tuned liquid column dampers could prolong the tower fatigue life. Although passive
control techniques provide mitigation when the parameters are finely tuned, they might lose their effectiveness as a result
of environmental or system variations. It was reported in [6] that most of the passive dampers deployed on highway sign
structures are ineffective due to the detuning effect.
In comparison, semi-active control devices are more applicable to systems with time-variant parameters. Semi-active
TMDs (STMDs) have been investigated and demonstrated effective in controlling vibration of linear and nonlinear systems
subjected to stationary and non-stationary excitations [7–9]. Weber [10] utilized a semi-active vibration absorber with real-
time adjusted magnetorheological damper (MR-SVA) to mitigate harmonic loading induced vibrations. It was found that the
MR-SVA outperforms the passive TMD. Huang and Arrigan et al. [11,12] explored the mitigation of wind turbine blades using
the STMDs retuned in real-time via a short time Fourier transform (STFT) based control algorithm. The authors found that the
STMDs could mitigate the blade responses under varying operational or environmental conditions. In Refs. [7,11,12], only the
frequency of the STMDs was tuned in real-time yet the damping ratio remained constant. Recently, Sun et al. [13] further
advanced the control algorithm by incorporating the tuning of damping ratio. The authors examined the performance of
the STMD tuned by the modified control algorithm for seismic protection and achieved improved response reduction.
In addition to passive and semi-active control, active control of vibrations have been studied and demonstrated effective
under operational or environmental variations [14,15]. Staino et al. [16] used active tendons mounted inside the blade to
control the edgewise vibration of wind turbine blades. The authors concluded that the proposed control scheme can signif-
icantly mitigate the response of the blade. Fitzgerald et al. [17] utilized an active tuned mass damper to control the in-plane
vibration of the blades. It was found that the active TMDs can provide better reduction than the passive TMDs.
Among different types of offshore wind turbines, the monopile fixed-bottom offshore wind turbines are widely employed
and under active investigation. In most existing literatures, soil effects were not considered, which is inappropriate when the
soil is relatively soft. Veletsos and Verbic [18] found that flexible soil underneath the foundation can increase the damping
and reduce the structural natural frequency. Furthermore, under cyclic wind, wave loading and strong earthquake strike,
damage will potentially occur to the foundation [19] and the tower. The accumulated damage will change the structural nat-
ural frequency and the response magnitude. Fitzgerald et al. [20] considered soil-structure-interaction and used an active
TMD to control the onshore wind turbines under wind loading. It was found that the active TMD was effective when soil
structure interaction is considerable. However, the dynamic characteristics of offshore wind turbines under combined
multi-hazards and potential damage have not been studied in Ref. [20] or other existing literatures.
To fill this gap, the present paper explores semi-active control of the monopile offshore wind turbines subjected to multi-
hazards consisting of wind, wave and earthquake where soil effects and damage are considered. Novelty of the present study
is twofold. On one hand, the dynamic characteristics of the offshore wind turbines under the combined effects of multi-
hazards, soil effects and damage will be studied. On the other hand, the effectiveness of the STMD in controlling the dynamic
response will be evaluated. To achieve this, a new mathematical model of the monopile offshore wind turbines coupled with
the STMD is established where the dynamic interaction between the blades and the tower and the STMD is modeled. Aero-
dynamic loading, wave loading, seismic loading and gravity loading are incorporated in the model. It is found that the
response dominant frequency decreases as the damage develops. Significant response amplification phenomenon under
the combined effects of multi-hazards and damage can be observed. When the STMD is utilized, it is shown that the STMD
can provide excellent effectiveness in controlling the tower and foundation dynamic response while the passive TMD
becomes off-tuned and ineffective.

2. Establishment of equations of motion

The equations of motion of a fully coupled three dimensional monopile offshore wind turbine model with and without the
STMD are established using the Euler-Lagrangian equation. There are 10 degree-of-freedom (DOF) in the uncontrolled sys-
tem and 11 DOF in the controlled system.

2.1. Model description

Fig. 1 illustrates the model of a monopile fixed-bottom offshore wind turbine subjected to wind, wave and seismic load-
ings. The global coordinate system originates at the intersection point of the tower center and the mean sea level (MSL). The
left portion of Fig. 1 shows the original model and right portion shows the deformed geometry of the wind turbine under the
combined loadings. Soil effect is represented by a translational spring with a stiffness coefficient kx and a rotational spring
with a stiffness coefficient k/ . The damping property of the soil is represented two dash-pot with damping coefficients of cx
and c/ . Parameters q9 ; q10 in Fig. 1 denote the translation and rotation coordinates of the foundation which will be used to
establish the equations of motion in the subsequent sections.
Fig. 2(a) and (b) illustrate the coordinates of the blades(edgewise and flapwise), the nacelle and the STMD. Parameters
q1 –q3 denote the edgewise coordinates of the three blades and q4 –q6 denote the flapwise coordinates. Displacement of
the 1st blade is shown in Fig. 2 as a general illustration. Variables u1e and u1f represent the edgewise and flapwise displace-
ments of an infinitesimal unit dr at a distance r from the blade root; variables /1e and /1f denote the edgewise and flapwise
fundamental mode shape. To formulate the motion of the blade, a local coordinate system x0 y0 z0  o0 originating at the center
 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305 287

Fig. 1. A monopile offshore wind turbine controlled by an STMD under wind, wave and seismic loadings. Left: original schematic model; right: simplified
model for analysis.

Fig. 2. Displacement of the turbine blades and the nacelle. (a): In-plane displacement; (b): Out-of-plane displacement.

of the hub front surface (static state) is established in Fig. 2 where uss
nac is the absolute side-side displacement of the nacelle

and ufa
nac is the absolute fore-aft displacement.
The blades rotating speed is X and the azimuthal angle wj ðtÞ of the jth blade can be expressed as:

2p
wj ¼ Xt þ ðj  1Þ; j ¼ 1; 2; 3 ð1Þ
3
It is worthy to mention that the edgewise and flapwise vibration are essentially coupled. In the present study, the cou-
pling effect is not included because the focus of this study is to control the response of the nacelle and the tower where the
coupling effect of edgewise and flapwise vibration has minimal effect.
288 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305

2.2. Euler-Lagrangian equation

The Euler-Lagrangian equation is used to derive the equation of motion.

~ðtÞ; q
d @Tðt; q ~_ ðtÞÞ @Tðt; q ~_ ðtÞÞ @Vðt; q
~ðtÞ; q ~ðtÞÞ
 þ ¼ Q i ðtÞ ð2Þ
dt @ q_ i ðtÞ @qi ðtÞ @qi ðtÞ

where T and V are the system kinetic and potential energy, q ~ðtÞ is the generalized coordinates vector, Q i ðtÞ is the generalized
force corresponding to the ith component of q~ðtÞ. Sign ðÞ_ denotes the first derivative with respect to time.

2.3. Kinetic energy

Let parameters q7 and q8 denote the relative coordinates of the nacelle in the fore-aft and side-side directions with respect
to the foundation. In terms of Figs. 1 and 2, the absolute displacement of the nacelle in the fore-aft direction ufa
nac and the side-
side direction uss
nac can be expressed as:

ufa
nac ¼ q7 þ q9 þ h tanðq10 Þ  q7 þ q9 þ hq10
ð3Þ
nac ¼ q8
uss

where h is the nacelle height with reference to the MSL.

The absolute velocity of the nacelle in the fore-aft direction v fanac and the side-side direction v ssnac can be obtained as:

v fa
nac ¼ q_ 7 þ q_ 9 þ hq_ 10
ð4Þ
v ss
nac ¼ q_ 8
The resultant absolute velocity of the nacelle v nac can be written as:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
v nac ¼ ðv fanac Þ þ ðv nac Þ
2
ss 2 ð5Þ
Let parameter q11 denote the relative coordinate of the STMD with respect to the nacelle. Similarly, the fore-aft velocity of
the STMD v stmd can be represented as:

v stmd ¼ v fanac þ q_ 11 ¼ q_ 7 þ q_ 9 þ hq_ 10 þ q_ 11 ð6Þ

Consider an infinitesimal unit dr of the jth blade in Fig. 2, the coordinate of the unit in the x0 y0 z0  o0 system can be rep-
resented as:

x0r ¼ uss
nac þ r sin wj þ uje cos wj

y0r ¼ ufa
nac þ ujf ð7Þ
z0r ¼ r cos wj  uje sin wj

Taking the first derivative of the coordinates gives the velocity component as:

x_0 r ¼ v ss _
nac þ Xr cos wj þ qj /1e cos wj  Xqj /1e sin wj

y_0 r ¼ v fa _ jþ3 /1f

nac þ q ð8Þ
_z0 r ¼ Xr sin wj  q_ j /1e sin wj  Xqj /1e cos wj

The absolute velocity magnitude of the unit dr is:

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
v bj ðr; tÞ ¼ x_02 þ y_02 þ z_02
r r r ð9Þ

The absolute velocity v tow ðz; tÞ of an infinitesimal unit dz of the tower can be expressed as:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
v tow ¼ ½q_ 8 /1t  þ ½q_ 7 /1t þ q_ 9 þ zq_ 10 
2 2
ð10Þ

where /1t denotes the fundamental mode shape (side-side and fore-aft) of the tower in the side-side and fore-aft directions.
Therefore, the kinetic energy of the entire wind turbine system can be expressed as:
Z Z
1X 3 R
1 1 h
1 1 1
T¼  v 2bj ðr; tÞdr þ M nac v 2nac þ
m Mv 2tow dz þ Mf q_ 29 ðtÞ þ If q_ 210 ðtÞ þ ms v 2stmd ðtÞ ð11Þ
2 j¼1 0 2 2 0 2 2 2

where M nac is the mass of the nacelle (including the hub mass); M f and If denote the mass and moment of inertia of the foun-
 and M denote the mass density per length of the blade and the tower; ms denotes the mass of the STMD.
dation; m
 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305 289

2.4. Potential energy

Potential energy of the blades is calculated considering the strain energy of the blades due to bending, the centrifugal
stiffening effect and gravity.
The total potential energy of the blades can be represented as:
3 h i
1X
Vb ¼ ðkeg þ kge;eg  kgr;eg cos wj Þq2j þ ðkfp þ kge;fp  kgr;fp cos wj Þq2jþ3 ð12Þ
2 j¼1

where
RR 2 RR 2
keg ¼ 0
EIeg ðrÞð/01e 0Þ dr; kfp ¼ 0 EIfp ðrÞð/01f 0Þ dr;
R RR R 2 R RR R 2
kge;eg 
¼ X2 0 r ½mðnÞndn ð/01e Þ dr; kge;fp ¼ X2 0 r ½mðnÞndn
 ð/01f Þ dr ð13Þ
R RR R 2 R RR R 2
kgr;eg 
¼ g 0 r ½mðnÞdn ð/01e Þ dr; kgr;fp ¼ g 0 r ½mðnÞdn
 ð/01f Þ dr

Ieg and Ifp are the moment of inertia in the edgewise and flapwise direction, X is the turbine rotating speed, g is the accel-
eration due to gravity. Sign ðÞ0 and ðÞ00 denote the first and second derivatives with respect to length r.
Hence, the total potential energy V of the wind turbine system is:
1 fa 1 ss 1 1 1
V ¼ V b þ kt q27 ðtÞ þ kt q28 ðtÞ þ kx q29 ðtÞ þ k/ q210 ðtÞ þ ks q211 ðtÞ ð14Þ
2 2 2 2 2
fa ss
where kt and kt denote the fore-aft and side-side stiffness of the tower, kx and k/ denote the translational and rotational
stiffness of the foundation, ks denotes the stiffness of the STMD.

2.5. System equation of motion

Substituting Eqs. (11) and (14) into Eq. (2) produces the equations of motion of the wind turbine, which can be written in
a matrix form as:
eq
M €~ þ C
eq~_ þ K
eq e wind þ Q
~¼Q e wv þ Q
e seismic ð15Þ
 
e are the system mass, damping and stiffness matrices. Two sets of system matrices for the uncontrolled
where M ; C and K
 
e seismic are the generalized force vectors corresponding to the
and controlled systems are derived. Variables Q wind ; Q wv and Q
wind, wave and seismic loading, which are derived in the following section.

3. Loading

This section presents the derivation of wind, wave and seismic loadings based on the Principle of Virtual Work.

3.1. Wind model

Wind velocity can be represented by a constant mean velocity and a turbulent component, i.e. v ðtÞ ¼ v
 þv
~ ðtÞ. In the pre-
sent study, the logarithmic wind profile is adopted to calculate the mean velocity v
 ðzÞ, i.e.

logðz=z0 Þ
v ðzÞ ¼ V ref ð16Þ
logðHref =z0 Þ
where V ref is the mean velocity at the reference height Href ¼ 90 m. Parameter z0 is the length of roughness and its value is
z0 ¼ 0:03.
The turbulent component of wind velocity v ~ ðtÞ is computed using the IEC Kaimal spectral model which is described by the
following equations:

4I2 Lc
Sv ðf Þ ¼ ð17Þ
ð1 þ 6fLc =v Þ
5=3

where Sv ðf Þ is the power spectral density function, f is the wind frequency in Hz; I is the wind turbulence intensity and Lc is
an integral scale parameter.
To account for the spatial dependency of wind velocity v ~ ðz; tÞ at different points, the cross spectra between two points i
and j are defined as:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Sij ðf Þ ¼ Cohði; j; f Þ Sii ðf ÞSjj ðf Þ ð18Þ
290 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305

where Sij is the cross spectra, Sii and Sjj are the auto spectra at points i and j, respectively.
Referring to the IEC spectral mode [21], the spatial coherence function is given as:
0
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
 2  2
fL 0:12L A
Cohði; j; f Þ ¼ exp@a þ ð19Þ
v hub Lc

where L is the distance between points i and j on the grid, a is the coherence decrement, and Lc is a coherence scale param-
eter, v
 hub is the mean hub-height wind speed.
In terms of IEC-64000-1 3rd ; a ¼ 12 and Lc ¼ 340:2 m are adopted in the present study.
Based on Eqs. (16)–(19), a three dimensional wind velocity field covering the domain of the rotor disk is generated using
the TurbSim program [21]. Matlab code has been developed to map the full wind field profile onto each span station of the
rotating blades. A 31  31 velocity grids is produced to represent the three dimensional wind field. Two wind conditions are
considered: one is the operational condition where the mean velocity at the hub height is v  ¼ 15 m/s and the turbulent
intensity is 12%; another is the cut-out wind condition where hub height mean velocity is v  ¼ 25 m/s and the turbulent
intensity is 20%. Fig. 3 shows the wind velocity at the center of the tip element of blade 1 of the cut-out wind condition.

3.2. Aerodynamic load

The Blade Element Momentum (BEM) theory is used to estimate the aerodynamic loading acting on the rotating blades.
The BEM method is a combination of the momentum theory and the blade element theory. Time series of the aerodynamic
loading are computed based on the momentum theory, the blade characteristics and the operational conditions. The input
parameters include the rotor geometry (number of blades, twist, chord distribution, and the airfoils), wind speed and the
blade rotational velocity [22].
Wind Velo. at blade 1 tip element(m/s)

35

30

25

20

15

10

0 10 20 30 40 50 60 70 80 90 100
Time(s)

Fig. 3. Wind velocity at the tip element of blade 1. (v

 ¼ 25 m/s at hub height and turbulence intensity I ¼ 20%).

Fig. 4. Turbine blade discretized into N blade elements for BEM analysis.
 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305 291

Fig. 4 illustrates a general turbine blade discretized into N elements for BEM analysis where R is the rotor radius and X
is the rotation velocity. It is assumed in the BEM theory that no radial dependency exists along the blade span and thus
the elements can be analyzed independently via performing the momentum theory. For a general case, the ith blade ele-
ment at a distance r is detailed in Fig. 4 where dr is the element span length and cðrÞ is the chord length at the element
mid-span.
An arbitrary blade element experiencing local velocities and aerodynamic loading is illustrated in Fig. 5. The relative wind
velocity V rel with reference to the the element shown in Fig. 5 can be expressed by combining the axial velocity v ð1  aÞ and
the tangential velocity Xrð1 þ a0 Þ as:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
V rel ¼ ½v ð1  aÞ2 þ ½Xrð1 þ a0 Þ2 ð20Þ

where a and a0 are the axial velocity and tangential velocity induction factors.
The flow angle / can be calculated by
v ð1  aÞ
/ ¼ tan1 ð21Þ
Xrð1 þ a0 Þ
The attack angle a between the relative velocity V rel and chord line is obtained as:
a¼/h ð22Þ
where h is the summation of the pitch angle and the twist which is predetermined by the airfoil.
In terms of the attack angle a, the lift and drag coefficients C l and C d can be determined from the airfoil data. The lift force
P L which is perpendicular and the drag force which is parallel to the relative velocity can be computed as:
1 1
PL ¼ qV 2rel cC l ; PD ¼ qV 2rel cC d ð23Þ
2 2
where q is the density of air and c is the chord length.
The normal and the tangential coefficients C N and C T are defined as:
C N ¼ cos/C l þ sin/C d ; C T ¼ sin/C l  cos/C d ð24Þ
The normal and tangential forces PN and P T can be calculated as:
1 1
PN ¼ qV 2rel cC N ; PT ¼ qV 2rel cC T ð25Þ
2 2
Eqs. (20)–(25) demonstrate the primary procedure to calculate the aerodynamic load. In real application, a and a0 are
unknown and need to be determined via iterations. Matlab code has been developed to calculate the time series of P N
and P T in the present study based on the algorithm proposed in [22]. Prandtl’s tip loss factor and Glauert correction are con-
sidered in the Matlab code. Next the Principle of Virtual Work is applied to calculate the generalized aerodynamic load.
Under P N and PT , the virtual work dW wl done by external wind load is:
3 Z
X R Z R 
dW wl ¼ PTj ðr; tÞ½/1e dqj þ duss
nac cosðwj Þdr þ P Nj ðr; tÞ½/1f dqjþ3 þ dufa
nac dr ð26Þ
j¼1 0 0

where P Tj ðr; tÞ and PNj ðr; tÞ denote the tangential and normal wind loading intensity per unit length on the jth blade.
In terms of the principles of work and energy, the generalized force Q j can be determined as:
@ðdW wl Þ
Qj ¼ ð27Þ
@ðdqj Þ

Fig. 5. Blade element section subjected to local velocity and aerodynamic load.
292 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305

Substituting Eq. (26) into Eq. (27) yields the wind induced generalized forces:
RR RR
Q j;wind ¼ P ðr; tÞ/1e dr;
0 Tj
Q jþ3;wind ¼ P ðr; tÞ/1f dr;
0 Nj
j ¼ 1; 2; 3
3 R
P 3 R
P
R R
Q 7;wind ¼ P ðr; tÞdr;
0 Nj
Q 8;wind ¼ P ðr; tÞdr cosðwj Þ
0 Tj
ð28Þ
j¼1 j¼1

Q 9;wind ¼ Q 7;wind ; Q 10;wind ¼ hQ 7;wind ; Q 11;wind ¼ 0

It is noted that the wind load acting on the tower is relatively small and thus is ignored. Based on Eq. (28), the generalized
wind loading can be computed. As an illustration, the first 120 s aerodynamic loading data of Q 1;wind  Q 6;wind corresponding
to v 0 ¼ 25 m/s and I ¼ 20% are presented in Fig. 6.

3.3. Wave excitation

Wave loading on circular cylindrical structural members of fixed offshore structures can be estimated using the Morison
equation [23]. For the monopile offshore wind turbine tower, the horizontal force dF acting on a strip of length dz can be
written as:

pD2 _ þ
q
dF ¼ C M qudz C D Dujujdz ð29Þ
4 2
where C M and C D are the mass and drag coefficients (C M ¼ 1:0 and C D ¼ 1:2 are adopted in the present paper); q is water
density (1025 kg/m3), D is the diameter of the tower and the monopile; u_ and u are the wave induced horizontal acceleration
and velocity of fluid particles.
In terms of Ref. [24], the JONSWAP spectrum as listed in Eq. (30) is used to generate wave time histories.
2  
!5 !4 3 ðxxp Þ2
exp  2 2
f 5 f
exp4 5ð1  0:287lncÞc 2r x
Sðf Þ ¼ 0:3125H2s T p p
ð30Þ
fp 4 fp

where T p is the wave period, Hs is the significant wave height, f p ¼ 1=T p ; r ¼ 0:07 for f 6 f p ; r ¼ 0:09 for f > f p . Variable c is
the JONSWAP peakedness parameter [24]:

4
x 10 (a)
Edgewise Aerodynamic Load(N)

3.5

2.5

2 Q1,wind Q2,wind Q3,wind

0 20 40 60 80 100 120
5
x 10 (b)
Flapwise Aerodynamic Load (N)

Q Q Q
4,wind 5,wind 6,wind
2.5

1.5

0 20 40 60 80 100 120

Fig. 6. Aerodynamic loading Q 1;wind  Q 6;wind . Wind velocity at hub height is v 0 ¼ 25 m/s, turbulence intensity I ¼ 20%.
 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305 293

8 pffiffiffiffiffiffi
>
<5 T p = Hs 6 3:6
pffiffiffiffiffiffi pffiffiffiffiffiffi
c ¼ expð5:75  1:15T p = Hs Þ 3:6 < T p = Hs 6 5:0 ð31Þ
>
: p ffiffiffiffiffi
ffi
1 T p = Hs < 5:0
Based on the spectrum representation method, the wave elevation gðtÞ, the fluid particle velocity u and the acceleration u_
can be expressed as:
X
N
gðtÞ ¼ Aj sinðwj t  kj x þ /j Þ ð32Þ
j¼1

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Aj ¼ 2Sðxj ÞDx ð33Þ

X
N
cosh½kðz þ dw Þ
u¼ xj Aj sinðxj t  kj x þ /j Þ ð34Þ
j¼1
T w sinhðkdw Þ

X
N
cosh½kðz þ dw Þ
u_ ¼ x2j Aj cosðxj t  kj x þ /j Þ ð35Þ
j¼1
T w sinhðkdw Þ

where k is wave number in m1, x is wave frequency in rad/s, /j is a random phase angle uniformly distributed from 0 to
2p; dw is the water depth, T w is the wave period in s, z is the vertical ordinate from mean water level.
Given water depth z, the parameters x and k are related by the dispersion equation [23]:

k tanh kz ¼ x2 =g ð36Þ
Hence, for any given water depth z, the wave number k can be determined via solving Eq. (36).
Virtual work dW wv done by the wave load along virtual displacement dutow of the tower can be written as:
Z g Z g
dW wv ¼ dFdutow ¼ dFð/1t dq7 þ dq9 þ zdq10 Þ ð37Þ
0 0

Substituting Eq. (37) into into Eq. (27) yields the generalized forces corresponding to wave:
Z " #
gðtÞ XNz
qpD2 ðzi Þ € q
Q 7;wav e ¼ /1t ðzÞdF ¼ /1t ðzi Þ _ i ; tÞjuðz
C M uðzi ; tÞDz þ C D Dðzi Þuðz _ i ; tÞjDz ð38Þ
dw i¼1
4 2

Z " #
gðtÞ X
Nz
qpD2 ðzi Þ q
Q 9;wav e ¼ dF ¼ € ðzi ; tÞDz þ
CM u _ i ; tÞjuðz
C D Dðzi Þuðz _ i ; tÞjDz ð39Þ
dw i¼1
4 2

Z " #
gðtÞ XNz
qpD2 ðzi Þ € q
Q 10;wav e ¼ zdF ¼ _ _
C M uðzi ; tÞDz þ C D Dðzi Þuðzi ; tÞjuðzi ; tÞjDz zi ð40Þ
dw i¼1
4 2

where N z denotes the number of segments that the wetted portion of the tower is divided and Dz is the segment length.
Two representative operational wave conditions, Hs ¼ 5:5 m; T ¼ 10:0 s and Hs ¼ 6:5 m; T ¼ 10:9 s are adopted. Based on
Eqs. (29)–(40), the wave induced modal force Q 7;wv ,the resultant force Q 9;wv and the bending moment Q 10;wv are determined
and other wave induced generalized forces are zero.

3.4. Seismic load

€ gx and u
Let u € gy denote the seismic acceleration components in x and y directions. Consider an infinitesimal unit dr of the
blade, the effective earthquake force components acting on dr in x and y directions are m u
€ gx dr and mu
€ gy dr. The virtual
work dW seismic;bl done by seismic loading on the blades can be written as:
3 Z
X R Z R 
dW seismic;bl ¼ u
m € gx ð/1f dqjþ3 þ dufa
nac Þdr þ
u
m € gy ð/1e dqj cos wj þ duss
nac Þdr ð41Þ
j¼1 0 0

Similarly, the virtual work done by seismic loading on the nacelle and the tower are:
€ gx M nac dufa
dW seismic;nac ¼ u €
nac  ugy M nac dunac
ss
ð42Þ
294 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305

Z h Z h
dW seismic;tow ¼  € gx Mðdq7 /1tow þ dq9 þ dq10 lÞdl 
u €gy Mdq8 /1tow dl
u ð43Þ
0 0

where M nac denotes the mass of the nacelle, M denotes the mass density of the tower in unit of kg/m, /1tow denotes the first
mode shape of the tower.
Therefore, the total virtual work done by seismic loading on the entire wind turbine system can be obtained as:
dW seismic ¼ dW seismic;bl þ dW seismic;nac þ dW seismic;tow ð44Þ
Substituting Eq. (44) into Eq. (27) yields the earthquake induced generalized forces:
€ gy ;
Q j;seismic ¼ m1e cosðwj Þu € gx ;
Q jþ3;seismic ¼ m1f u j ¼ 1; 2; 3
€ gx ;
Q 7;seismic ¼ ð3m0 þ Mnac þ M 1tow Þu Q 8;seismic ¼ ðm0 cosðwj Þ þ M nac þ M1tow Þu€ gy ð45Þ
€ gx ;
Q 9;seismic ¼ ð3m0 þ Mnac þ M 0tow Þu Q 10;seismic ¼ ð3hm0 þ hMnac þ M 2tow Þu€ gx

where
Z h Z h Z h
M0tow ¼ Mdl; M1tow ¼ M/1tow dl; M2tow ¼ Mldl ð46Þ
0 0 0

It is noted that Eq. (45) illustrates the generalized seismic loading in the fore-aft and side-side directions. Because the
focus of the present study is to attenuate the dynamic response in the fore-aft direction, the seismic component in the
€ gy ¼ 0; Q j;seismic ¼ 0; j ¼ 1; 2; 3; 8.
side-side direction is ignored, i.e. u
To model the seismic response characteristics of wind turbines installed in the California coastal region, the recorded seis-
mic acceleration of the 1994 Northridge earthquake Newhall 90 is utilized.

4. Control algorithm

A short-time Fourier transformation (STFT) based control algorithm presented in Ref. [13] is briefly introduced in this
section.

4.1. Short-time Fourier transformation

Short-time Fourier transform is a widely used tool to extract the time-frequency characteristics of non-stationary signals.
Principles and the associated equations are illustrated as follows.
Consider a signal xðsÞ , multiplying the signal by a window function hðs  tÞ yields:
^xðsÞ ¼ xðsÞhðs  tÞ ð47Þ
where ^
xðsÞ is a weighted signal; t is the fixed time and s is the running time.
Applying Fourier transformation to ^ xðsÞ gives the spectrum Sðt; xÞ at the fixed time t:
Z Z
1 1
Sðt; xÞ ¼ ejxt ^x ¼ ejxt xðsÞhðs  tÞ ð48Þ
2p 2p
The power spectral density Pðt; xÞ at time t is obtained as:

Pðt; xÞ ¼ jSðt; xÞj2 ¼ Sðt; xÞ  Sðt; xÞ ð49Þ

At the fixed time ti , the dominant frequency xid is calculated using Eqs. (50) and (51).
xinst ¼ fxjPðti ; xÞ ¼ maxfPðti ; xÞgg ð50Þ
Pi
k¼maxf1;imþ1g inst ðt k ÞmaxfPðt k ;
x xÞg
xid ¼ Pi ð51Þ
k¼maxf1;imþ1g maxfPðt k ; Þg x
where xinst is the instantaneous frequency and xid is the dominant frequency at t i calculated through averaging the values of
the instantaneous frequencies over m time steps, m ¼ 3 in the present study.

4.2. Real-time tuning of the STMD

In the present study, the nacelle fore-aft displacement is tracked and the power spectrum density Pðti ; xÞ is computed via
Eqs. (47)–(49) in real-time. In the beginning at t ¼ 0, the STMD circular natural frequency is xs ¼ f opt xn and damping ratio is
set to fs ¼ 0 where f opt is the optimal frequency ratio to be determined next and xn is the original circular natural frequency
of the nacelle/tower top.
 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305 295

Control algorithm

1. A moving window of n time points (n ¼ 500 in the present study) of signal at time ti is convolved with a Hamming win-
dow to compute the power spectral density Pðt; xÞ (Eqs. (47)–(49)) of the nacelle displacement, producing a vector P i . The
length Lh of the Hamming window is set to Lh ¼ 1024 and the vector Pi is of size NN  1, where NN ¼ Lh =2 þ 1.
2. The dominant frequency xid at time ti is calculated using Eqs. (50) and (51).
3. If jxid  xn j=xn <¼ d, the stiffness of the STMD is retuned such that xs ¼ xid ; else, it is set to xs ¼ f opt xn , where d ¼ 0:1.
4. The relative displacement of the STMD q11 is recorded in a vector Ds . If the current magnitude is less than that of the pre-
vious amplitude, i.e. jDs ðkÞj 6 jDs ðk  1Þj, the damping ratio of the STMD is set to be fs ¼ 2fopt ; else it is set to fs ¼ 0.

Parameter d is related to the characteristics of external loadings and independent of the system parameters. The value of
d ¼ 0:1 is determined via a heuristic study to achieve consistent reduction under most stationary and non-stationary exci-
tations. Parameters f opt and fopt are determined using the modal analysis approach illustrated in [13].
It is worthy to note that in real application, the time-variant frequency of the STMD can be realized via using variable
stiffness devices, e.g., the semi-active and independently variable stiffness device [8] or the adaptive passive stiffness device
[25]. The time-variant damping ratio can be realized by using MR dampers [2].

5. Results

The effectiveness of the STMD and the control algorithm is evaluated in this section.

5.1. Model parameters

The NREL 5 MW OC3 Mono-pile wind turbine model [26] is used. Details are presented in Table 1. The fundamental mode
shape(edgewise and flapwise) of the blade and the tower are formulated by Eq. (52) and illustrated in Fig. 7(a) and (b).
    
/1e ðr Þ ¼ 0:6952r 6 þ 2:3760r 5  3:5772r 4 þ 2:5337r 3 þ 0:3627r 2
    
/1f ðr Þ ¼ 2:2555r 6 þ 4:7131r 5  3:2452r 4 þ 1:7254r 3 þ 0:0622r 2 ð52Þ
    
 ¼ 0:6952h6 þ 2:3760h5  3:5772h4 þ 2:5337h3 þ 0:3627h2
/1t ðhÞ
 ¼ h=87:6 denote the normalized blade radius and tower height, separately.
where r ¼ r=61:5 and h

5.2. Soil effects and damage parameters

With reference to [27], value of parameters kx and k/ are obtained as: kx ¼ 3:89E9 N=m; k/ ¼ 1:14E11 Nm=rad to repre-
sent clay soil condition. Soil damping coefficients cx and c/ are selected such that the corresponding damping ratio are
fx ¼ f/ ¼ 0:6%.
Due to cyclic wind and wave loading and earthquake strike, damage will probably occur to the foundation and the tower.
Here it is assumed that the damage occurs at a given time and accumulates linearly with respect to time. Fig. 8(a) and (b)
show the foundation and tower damage development where it is assumed that the damage development has a duration of

Table 1
Parameters of the NREL 5-MW baseline wind turbine [26].

Gross properties Rating 5 MW

Rotor diameter 126 m
Hub height 90 m
Cut-in, rated, cut-out wind speed 3 m/s, 11.4 m/s, 25 m/s
Cut-in, rated rotor speed 6.9 rpm, 12.1 rpm
Blade Length 61.5 m
Mass 17,740 kg
Second moment of inertia 11,776 kg m2
1st edgewise mode natural frequency 1.08 Hz
1st flapwise mode natural frequency 0.68 Hz
1st mode damping ratio (edgewise and flapwise) 0:48%
Nacelle + hub Nacelle mass 240,000 kg
Hub mass 56,780 kg
Hub diameter 3m
Tower Height above ground 87.6 m
Overall (integrated) mass 267,650 kg
1st mode fore-aft(side-side) natural frequency 0.92 Hz
1st mode fore-aft(side-side) damping ratio 1%
296 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305

(a) (b)
1 1
Edgewise mode
0.8 Flapwise mode 0.8
Mode shape

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Normalized blade length Normalized tower height

Fig. 7. Fundamental mode shape of the blade and the tower. (a): mode shape of the blade (b): mode shape of the tower.

Fig. 8. Illustration of the gradually developed damage: (a) foundation damage development; (b) tower structural damage development, ts and t e denote the
time of starting and ending of damage.

50 s. To provide insight into the system under combined effects of damage and multi-hazards, three damage cases referred to
as DC1, DC2 and DC3 are studied. The three damage cases are: DC1, tower damage starts at ts ¼ 50 s (before earthquake);
DC2, tower damage starts at t s ¼ 110 s (during earthquake); DC3, tower damage starts at ts ¼ 150 s (after earthquake).

5.3. Results of uncontrolled wind turbine

In terms of [24], two representative operational wind and wave states and the 1994 Northridge Earthquake are adopted.
Five load cases combined from the selected wind, wave and the earthquake are illustrated in Table 2 where ‘‘LC” is the acro-
nym for ‘‘load case”. To provide preliminary insight into the system under the combined loading, earthquake strike is applied
at t ¼ 100 s as an initial trial.
Fig. 9 shows the nacelle and tower top displacement time histories under LC2, LC3 and LC5. It is found that the peak
response induced by the operational wind and wave state (LC2) is approximately as 4:5 times as large as that induced by
the 1994 Northridge(NWH90) earthquake, which was actually a strong level earthquake with a peak ground acceleration
(PGA) of 0.59 g and a moment of magnitude M w of 6:7. Several other recorded seismic data with equivalent or higher
PGA have been tested in the present study and similar results were obtained. The relevant results are not shown due to space
limitation. Therefore, it can be concluded that for most monoplie offshore wind turbines, the wind and wave induced
response dominates the total response when compared to seismic loading. Since the result in Fig. 9 is a preliminary trial,
seismic induced response is almost 180 out-of-phase with the wind-wave induced response. As a result, the total response
under wind, wave and seismic loading has smaller magnitude than that induced by wind and wave loading. To produce a
more adverse situation for control purpose, the earthquake strike is adjusted to happen at t ¼ 101:8 s hereafter.
To investigate the effect from damage to structural response under the combined loading, simulations were conducted
with respect to the predefined three damage cases and the results are compared in Fig. 10. The result in the case of DC2,
which exhibits similar characteristics as DC1, is not shown. In Fig. 10, the response in the case of no damage denoted as
‘‘ND” is illustrated as a reference. It is found that from t ¼ 10 s (foundation damage starts) to t ¼ 60 s (foundation damage
fully develops), responses in the three cases are almost identical, signaling that foundation damage has minimal effect on
structural responses. In comparison, after tower structural damage starts at t ¼ 50 s in damage case DC1, the response is
increased by around 150%. Furthermore, the response magnitude is escalated dramatically when the earthquake strike hap-
pens at t ¼ 101:8 s. Similarly, in damage case DC3 where tower damage starts at t ¼ 150 s, it can be observed that the struc-
tural response is increased dramatically by around 300% when compared to the case of ND.
Using the short time Fourier transform, contour plot of the time-frequency power spectrum density (PSD) of the nacelle
response corresponding to no damage and the three damage cases are computed and illustrated in Fig. 11(a)–(d). It is
 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305 297

Table 2
Load cases considering wind wave and seismic excitation.

Load case 1 (LC1) Wind speed v 0 ¼ 15 m/s at hub height H ¼ 90 m, Turbulence intensity I ¼ 12%
Significant wave height Hs ¼ 5:5 m, Wave period T p ¼ 10:0 s
Load case 2 (LC2) Wind speed v 0 ¼ 25 m/s at hub height H ¼ 90 m, Turbulence intensity I ¼ 20%
Significant wave height Hs ¼ 6:5 m, Wave period T p ¼ 10:9 s
Load case 3 (LC3) Wind turbine parked. 1994 Northridge NewHall fire station (NWH90, Mw ¼ 6:7; PGA ¼ 0:59 g)
Load case 4 (LC4) Seismic loading + operational wind-wave loading LC1
Load case 5 (LC5) Seismic loading + operational wind-wave loading LC2

3.5 Seismic
Nacelle/tower top fore−aft disp. (m)

Wind−wave
3
Wind−wave−seismic
2.5

1.5

0.5

−0.5

−1
90 100 110 120 130 140 150 160 170 180 190 200
Time(s)

Fig. 9. Tower top displacement under seismic, wind-wave and the combined loading.

5
ND
DC1
4 DC3
Nacelle/tower top fore−aft disp. (m)

−1

−2

20 40 60 80 100 120 140 160 180 200 220

Time(s)

Fig. 10. Comparison of tower top displacement under different damage scenarios.

indicated that the response dominant frequency stays constant in Fig. 11 (a) while the value starts to decrease at different
times in Fig. 11(b)–(d). The different times when the frequency starts to decrease essentially correspond to the instants when
the tower structural damage starts, e.g., t ¼ 50 s in (b), t ¼ 102 s in (c) and t ¼ 150 s in (d). Through comparing the color line
values among Fig. 11(a)–(d), one can find that the response is increased when structural damage results. The worst case is
Fig. 11(d) where the damage occurs after the earthquake.

5.4. Results of controlled wind turbine

Mitigation of the nacelle and tower top response is presented in this section. The same load and damage cases as defined
in Section 5.3 are utilized. To show the advantage of the STMD, a passive TMD designed using Warburton’s formula [28] is
adopted for comparison. Performance of the STMD and the TMD with different mass ratios from 1% to 5% are evaluated.
298 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305

(a) (b)
40 55

50
35
45
0.3 30 0.3
40
Frequency (Hz)

Frequency (Hz)
25 35

30
0.25 20 0.25
25
15 20

15
0.2 10 0.2
10
5
5

0.15 0.15
50 100 150 200 250 300 50 100 150 200 250 300
Time (s) Time (s)
(c) (d)
50 80

45
70
40
0.3 0.3 60
35
Frequency (Hz)

Frequency (Hz)

50
30

0.25 25 0.25 40

20
30
15
0.2 0.2 20
10
10
5

0.15 0.15
50 100 150 200 250 300 50 100 150 200 250 300
Time (s) Time (s)

Fig. 11. Time-frequency spectrum of tower top displacement under different damage scenarios: (a) no damage, (b) damage case DC1, (c) damage case DC2
(d) damage case DC3.

Due to space limitation, time history and response spectrum results under LC5 and LC4, damage case DC3 and a mass
ratio of 3% are presented as a representative. Results corresponding to other load cases, damage cases, and mass ratios exhi-
bit similar characteristics and thus are not shown.
Fig. 12 illustrates the identified time-varying dominant frequency of the nacelle and the real-time retuned frequency of
the STMD under LC5 with damage DC3. One can find in Fig. 12 that the dominant frequency has two significant decreases
from 10 s to 60 s and from 150 s to 200 s. Essentially, these two frequency decreases are caused by the foundation damage
and the tower structural damage of DC3. In Fig. 12, the frequency of the STMD is retuned in real-time to match the time-
varying dominant frequency. Therefore, the proposed control algorithm can track the time-varying dominant response fre-
quency and retune the STMD efficiently and timely.
First, the soil effect on the TMD and the STMD mitigation performance is examined. Natural frequency of the TMD is
determined without considering the soil effect while the STMD is tuned based on the control algorithm. Fig. 13(a) shows
the time history of the nacelle/tower top response controlled by the TMD and the STMD under LC5 without damage. It is
found that the STMD can improve the mitigation of the maximum response by around 30% when compared to the TMD.
Fig. 13(b) shows the response spectrum where it is indicated that the response spectrum peak in the case of the TMD can
be further mitigated by 60% by using the STMD.
Next the TMD and the STMD performance under the combined soil effect and damage is examined. Fig. 14(a) and (b)
show the time history and response spectrum of the nacelle and tower top response under LC5 and DC3. Through comparing
Figs. 14(a) and 13(a), one can find that the nacelle/tower top response in the case of the passive TMD is remarkably increased
by around 40% after the tower damage happens at t ¼ 150 s. To the contrary, the response controlled by the STMD remains a
low level regardless of the soil effect and the damage. In Fig. 14(b), one can find that the response spectrum magnitude in the
case of the TMD is mitigated dramatically to a minimal level when the STMD is used. Fig. 15(a) and (b) show the comparison
of time history and response spectrum of the nacelle and tower top response under LC4. Similar conclusions can be obtained
 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305 299

0.285
Identified dominant frequency

Identified dominant frequency/STMD frequency(Hz)

0.28 STMD retuned frequency

0.275
Foundation damage

0.27

0.265

0.26 Tower damage

0.255

0.25

0.245

0.24
0 50 100 150 200 250 300
Time (s)

Fig. 12. Identified dominant frequency and STMD retuned frequency.

(a)
3.5
Passive TMD
Nacelle/tower top fore−aft disp. (m)

3 STMD
2.5

1.5

0.5

−0.5

−1
0 50 100 150 200 250 300
Time(s)
x 10
7 (b)
5
Nacelle Disp. PSD (m2/Hz)

4 Passive TMD
X: 0.2675 STMD
Y: 4.164e+07
3

2
X: 0.2675
1 Y: 1.677e+07

0
0.2 0.25 0.3
Frequency(Hz)

Fig. 13. Nacelle/tower top displacement mitigation comparison under LC5. (a) Time-history, (b) response spectrum.

as under LC5. Moreover, multiple peaks appear in Fig. 15 (b) in the case of TMD while these peaks are mitigated to a minimal
level by the STMD.
In addition to control the response of the nacelle and the tower top, the STMD can provide effective mitigation to the
foundation. Fig. 16 shows the foundation in-plane displacement under LC5 and DC3. It is found that the response maximum
and the RMS is further mitigated by 30% and 25% in the case of the STMD. Fig. 17 shows the foundation in-plane rotation. It
is found that the response maximum and the RMS is further mitigated by 25% and 50% in the case of the STMD. Therefore,
the STMD can better mitigate the cyclic fatigue load applied to the foundation so as to prolong its fatigue life and reduce the
potential maintenance cost.
300 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305

(a)
5
Passive TMD

Nacelle/tower top fore−aft disp. (m)

STMD
4

−1

−2
0 50 100 150 200 250 300
Time(s)

x 10
7 (b)
15
Nacelle Disp. PSD (m /Hz)

X: 0.2375 Passive TMD

2

Y: 1.349e+08
10 STMD

0
0.2 0.21 0.22 0.23 0.24 0.25 0.26 0.27 0.28 0.29 0.3
Frequency(Hz)

Fig. 14. Response spectrum of the nacelle displacement under LC5 with damage case DC3. (a) Time-history, (b) response spectrum.

(a)

3 Passive TMD
Nacelle/tower top fore−aft disp. (m)

STMD
2.5

1.5

0.5

−0.5

−1

0 50 100 150 200 250 300

Time(s)

x 10
7 (b)
8
Nacelle Disp. PSD (m /Hz)

Passive TMD
2

6
X: 0.24 STMD
Y: 6.188e+07
X: 0.245
4 Y: 3.161e+07
X: 0.25
Y: 1.557e+07
2

0
0.2 0.21 0.22 0.23 0.24 0.25 0.26 0.27 0.28 0.29 0.3
Frequency(Hz)

Fig. 15. Nacelle/tower top displacement mitigation comparison under LC4 with damage case DC3. (a) Time-history, (b) response spectrum.
 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305 301

−3
x 10

Foundation in−plane disp. (m)

Passive TMD
4 STMD

−2

0 50 100 150 200 250 300

Time(s)

Fig. 16. Foundation in-plane displacement under LC5 with damage case DC3.

−3
Foundation in−plane rotation (rad)

x 10
Passive TMD
10 STMD

−5
0 50 100 150 200 250 300 350 400
Time(s)

Fig. 17. Foundation in-plane rotation under LC5 with damage case DC3.

(a) (b)
10 4
TMD−DC1 STMD−DC1 TMD−DC1 STMD−DC1
9 TMD−DC2 STMD−DC2 TMD−DC2 STMD−DC2
3.5
TMD−DC3 STMD−DC3 TMD−DC3 STMD−DC3
8
Maximum stroke(m)

3
RMS of stroke(m)

7
2.5
6

2
5

1.5
4

3 1
0.01 0.02 0.03 0.04 0.05 0.01 0.02 0.03 0.04 0.05
Mass ratio µ Mass ratio µ

Fig. 18. Passive TMD and STMD stroke comparison under LC5 with different damage cases.

Stroke is a critical index when evaluating the performance of TMD because of the limited space in the nacelle. Fig. 18(a)
and (b) compare the stroke maximum and the RMS between the STMD and the TMD with various mass ratios under LC5 and
with the three damage cases DC1  DC3. It is found that both the stroke maximum and the RMS of the STMD are approx-
imately 70% of that of the TMD with different mass ratios under the three damage cases.
Fig. 19(a) and (b) illustrate the comparison of reduction of the response maximum and RMS between the passive TMD and
the STMD under LC5 with the three damage cases DC1  DC3. It can be found in Fig. 19(a) and (b) that with given damage
case and mass ratio, the STMD can significantly increase the mitigation by more than 50% when compared to the TMD. One
can see that both the STMD and the TMD provide the best mitigation in damage case DC3. However, negative reduction can
302 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305

(a) (b)
0.8
0.8
0.6
0.6
0.4
0.4
Reduction of Maximum

Reduction of RMS
0.2
0.2
0
0

−0.2
−0.2

−0.4 −0.4
TMD−DC1 STMD−DC1 TMD−DC1 STMD−DC1
−0.6 TMD−DC2 STMD−DC2 −0.6 TMD−DC2 STMD−DC2
TMD−DC3 STMD−DC3 TMD−DC3 STMD−DC3
−0.8 −0.8
0.01 0.02 0.03 0.04 0.05 0.01 0.02 0.03 0.04 0.05
Mass ratio µ Mass ratio µ

Fig. 19. Reduction of nacelle displacement under LC5 with different damage cases.

be observed in the case of TMD with damage case DC1 and DC2, signaling that the severely off-tuned TMD will amplify rather
than mitigate the structural response when tower damage happens before or during the earthquake. In comparison, the
STMD can mitigate the structural response consistently regardless of the damage. In addition, it can be observed that the
marginal reduction effect decreases as the mass ratio increases in the case of the STMD, which agrees with the findings pre-
sented in Ref. [29]. Based on the results in Fig. 19(a) and (b), a mass ratio of 3% might be an appropriate option for the STMD
of the monopile offshore wind turbines.

6. Conclusions

An analytical model of the monopile fixed-bottom offshore wind turbine controlled by an STMD and subjected to multi-
hazards consisting of wind, wave and earthquake is established and presented. The seismic loading model, which is rarely
reported in existing papers, is explicitly presented and integrated with the aerodynamic and hydrodynamic loading models.
Complete Matlab code has been developed to compute the aerodynamic, hydrodynamic and seismic loading and to simulate
the dynamic response of the offshore wind turbine system controlled by the STMD. Based on the results and discussions pre-
sented, the following four key conclusions can be obtained:

1. Soil effects generally reduce the monopile offshore wind turbine natural frequency. When damage occurs to the founda-
tion and the tower, natural frequency of the nacelle and the tower will be decreased further. Time-frequency spectrum of
the tower top response can reflect the damage development with good accuracy.
2. Dynamic response of the nacelle and the tower is amplified under the combined effects of damage and multi-hazards. The
time-frequency spectrum peak has an average increase of around 50% and the response time history maximum can be
increased by up to 300%.
3. Under soil effects, damage and multi-hazards, the STMD is consistently effective in controlling the response of the tower
and the foundation, thereby protecting the monopile offshore wind turbines and prolonging its fatigue life. In addition,
the STMD experiences a smaller stroke than the passive TMD, which is beneficial to wind turbines from the perspective of
practical application.
4. The passive TMD which is tuned to the original structural natural frequency without considering soil effects and potential
damage becomes off-tuned when damage occurs. As damage accumulates in the system, the detuning effect is escalated
and the passive TMD eventually becomes completely ineffective. In some scenarios, the off-tuned TMD will amplify rather
than mitigate the response, which is a potential threat to the system.

Acknowledgment

This work was supported by Louisiana State University Start-up Fund (fund number is 127150013), the Faculty Research
Grant (fund number is 127159132) and the Innovation in Engineering Research (FIRE) Grant provided by the College of Engi-
neering at Louisiana State University. Portions of this research were conducted with high performance computing resources
provided by Louisiana State University. The author is grateful for all the support.
 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305 303

Appendix A

System matrices involved in Eq. (15) for the uncontrolled 10-DOF system are:
2 3
m1 0 0 0 0 0 0 m18 0 0
6 0 m1 0 0 0 0 0 m28 0 0 7
6 7
6 7
6 0 0 m1 0 0 0 0 m38 0 0 7
6 7
6 0 0 0 m2 0 0 m47 0 m47 hm47 7
6 7
6 0 0 0 0 m2 0 m47 0 m47 hm47 7
e 6 7
M¼6 7 ð53Þ
6 0 0 0 0 0 m2 m47 0 m47 hm47 7
6 7
6 0 0 0 m47 m47 m47 m7 0 m7 hm7 7
6 7
6m m28 m38 0 0 0 0 m8 0 0 7
6 18 7
6 7
4 0 0 0 m47 m47 m47 m7 0 m9 hm7 5
2
0 0 0 hm47 hm47 hm47 hm7 0 hm7 If þ h m7
where
Z R Z R Z R
m1 ¼ 
mðrÞ/2
1e dr; m2 ¼

mðrÞ/2
1f dr; mj8 ¼

mðrÞ/1e drcoswj ; j ¼ 1; 2; 3
0 0 0

Z R Z h
m47 ¼ 
mðrÞ/1f dr; M 1t ¼ MðzÞ/21t dz; m8 ¼ m7 ¼ 3m0 þ M nac þ Mhub þ M 1t
0 0

Z h
m9 ¼ 3m0 þ M nac þ Mhub þ M t þ M f ; M t ¼ MðzÞdz
0

Parameter m0 denotes the mass of a single blade, M nac and M hub denote the mass of the nacelle and the hub, M denote the
mass density (kg/m) of the tower, Mt denotes the mass of the tower, h0 denotes the tower height, M f and If denote the mass
and moment of inertia of the foundation.
2 3
kb1;eg 0 0 0 0 0 0 0 0 0
6 0 kb2;eg 0 0 0 0 0 0 0 07
6 7
6 7
6 0 0 kb3;eg 0 0 0 0 0 0 07
6 7
6 0 0 0 kb1;fp 0 0 0 0 0 07
6 7
6 7
6 0 0 0 0 kb2;fp 0 0 0 0 07
e ¼6
K 7 ð54Þ
6 0 0 0 0 0 kb3;fp 0 0 0 07
6 7
6 7
6 0 0 0 0 0 0 k7 0 0 07
6 7
6 X2 m X2 m28 X2 m38 07
6 18 0 0 0 0 k8 0 7
6 7
4 0 0 0 0 0 0 0 0 kx 05
0 0 0 0 0 0 0 0 0 k/
where

kbj;eg ¼ keg þ kge;eg  kgr;eg coswj  X2 m1 ; kbj;fp ¼ kfp þ kge;fp  kgr;fp coswj ;

Z h
2
k7 ¼ k8 ¼ EIt ð/001t Þ dz
0

2 3
cb;eg 0 0 0 0 0 0 0 0 0
6 0 cb;eg 0 0 0 0 0 0 0 07
6 7
6 7
6 0 0 cb;eg 0 0 0 0 0 0 07
6 7
6 0 0 0 cb;fp 0 0 0 0 0 07
6 7
6 7
6 0 0 0 0 cb;fp 0 0 0 0 07
e ¼6
C 7 ð55Þ
6 0 0 0 0 0 cb;fp 0 0 0 07
6 7
6 7
6 0 0 0 0 0 0 c7 0 0 07
6 7
6 2Xm 81  82
2Xm  83
2Xm 0 0 0 0 c8 0 07
6 7
6 7
4 0 0 0 0 0 0 0 0 cx 05
0 0 0 0 0 0 0 0 0 c/
304 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305

where
Z R
 8j ¼
m 
mðrÞ/1e drsinwj ; j ¼ 1; 2; 3
0

cbj;eg and cbj;fp denote the edgewise and flapwise structural and aerodynamic damping of the blade, c7 and c8 denote the
damping in the fore-aft and side-side directions, including the structural and aerodynamic damping property of the nacelle
and the tower, cx and c/ denote the in-plane translational and rotational damping of the foundation.

Appendix B

System matrices involved in Eq. (15) for the controlled 11-DOF system are:
2 3
m1 0 0 0 0 0 0 m18 0 0 0
6 0 0 7
6 m1 0 0 0 0 0 m28 0 0 7
6 7
6 0 0 m1 0 0 0 0 m38 0 0 0 7
6 7
6 0 0 7
6 0 0 m2 0 0 m47 0 m47 hm47 7
6 7
6 0 0 0 0 m2 0 m47 0 m47 hm47 0 7
6 7
e 6 7
M¼6 0 0 0 0 0 m2 m47 0 m47 hm47 0 7 ð56Þ
6 7
6 0 0 0 m47 m47 m47 m7 þ ms 0 m7 hm7 ms 7
6 7
6 7
6 m18 m28 m38 0 0 0 0 m8 0 0 0 7
6 7
6 0 0 0 m47 m47 m47 m7 0 m9 þ ms hm7 ms 7
6 7
6 2 7
4 0 0 0 hm47 hm47 hm47 hm7 0 hm7 I1 þ h ms hms 5
0 0 0 0 0 0 ms 0 ms hms ms
2
where I1 ¼ If þ h m7 ; ms denotes the mass of the STMD
2 3
kb1;eg 0 0 0 0 0 0 0 0 0 0
6 07
6 0 kb2;eg 0 0 0 0 0 0 0 0 7
6 7
6 0 0 kb3;eg 0 0 0 0 0 0 0 07
6 7
6 7
6 0 0 0 kb1;fp 0 0 0 0 0 0 07
6 7
6 0 0 0 0 kb2;fp 0 0 0 0 0 07
6 7
e ¼6
K 6 0 0 0 0 0 kb3;fp 0 0 0 0
7
07 ð57Þ
6 7
6 0 0 0 0 0 0 k7 0 0 0 07
6 7
6 7
6 X2 m18 X2 m28 X2 m38 0 0 0 0 k8 0 0 07
6 7
6 0 0 0 0 0 0 0 0 kx 0 07
6 7
6 7
4 0 0 0 0 0 0 0 0 0 k/ 05
0 0 0 0 0 0 0 0 0 0 ks
2 3
cb;eg 0 0 0 0 0 0 0 0 0 0
6 0 7
6 0 cb;eg 0 0 0 0 0 0 0 0 7
6 7
6 0 0 cb;eg 0 0 0 0 0 0 0 0 7
6 7
6 0 7
6 0 0 0 cb;fp 0 0 0 0 0 0 7
6 7
6 0 0 0 0 cb;fp 0 0 0 0 0 0 7
6 7
e ¼6
C 6 0 0 0 0 0 cb;fp 0 0 0 0 0 7
7
ð58Þ
6 7
6 0 0 0 0 0 0 c7 0 0 0 cs 7
6 7
6  81  82  83 7
6 2Xm 2Xm 2Xm 0 0 0 0 c8 0 0 0 7
6 7
6 0 0 0 0 0 0 0 0 cx 0 0 7
6 7
6 7
4 0 0 0 0 0 0 0 0 0 c/ 0 5
0 0 0 0 0 0 0 0 0 0 cs

References

[1] G. Housner, L.A. Bergman, T.K. Caughey, A.G. Chassiakos, et al, Structural control: past, present, and future, J. Eng. Mech. 123 (9) (1997) 897–971.
[2] B. Spencer, S. Nagarajaiah, State of the art of structural control, J. Struct. Eng. 129 (7) (2003) 845–856.
[3] P. Murtagh, A. Ghosh, B. Basu, B. Broderick, Passive control of wind turbine vibrations including blade/tower interaction and rotationally sampled
turbulence, Wind Energy 11 (2007) 305–317.
 C. Sun / Mechanical Systems and Signal Processing 99 (2018) 285–305 305

[4] M. Lackner, M. Rotea, Passive structural control of offshore wind turbines, Wind Energy 14 (2011) 373–388.
[5] S. Colwell, B. Basu, Tuned liquid column dampers in offshore wind turbines for structural control, Eng. Struct. 31 (2009) 358–368.
[6] A. Douglas, A. Jennifer, M. James, V. Schaun, W. Tae, Evaluation of Aluminum Highway Sign Truss Designs and Standards for Wind and Truck Gust
Loadings, Illinois Department of Transportation, Physical Research Report No. 153, FHWA/IL/PRR 153, Springfield, IL, 2006.
[7] S. Nagarajaiah, E. Sonmez, Structures with semiactive variable stiffness single/multiple tuned mass dampers, J. Struct. Eng. 133 (1) (2007) 67–77.
[8] S. Nagarajaiah, Adaptive passive, semiactive, smart tuned mass dampers: identification and control using empirical mode decomposition, hilbert
transform, and short-term fourier transform, Struct. Control Health Monit. 16 (7–8) (2009) 800–841.
[9] C. Sun, S. Nagarajaiah, A. Dick, Experimental investigation of vibration attenuation using nonlinear tuned mass damper and pendulum tuned mass
damper in parallel, Nonlinear Dynam. 78 (4) (2014) 2699–2715.
[10] F. Weber, Semi-active vibration absorber based on real-time controlled MR damper, Mech. Syst. Signal Process. 46 (2014) 272–288.
[11] C. Huang, J. Arrigan, S. Nagarajaiah, B. Basu, Semi-active algorithm for edgewise vibration control in floating wind turbine blades, in: Conference: 12th
Biennial International Conference on Engineering, Construction, and Operations in Challenging Environments, 2010, pp. 2097–2110.
[12] J. Arrigan, V. Pakrashi, B. Basu, S. Nagarajaiah, Control of flapwise vibrations in wind turbine blades using semi-active tuned mass dampers, Struct.
Control Health Monit. 18 (2011) 840–851.
[13] C. Sun, S. Nagarajaiah, Study on semi-active tuned mass damper with variable damping and stiffness under seismic excitations, Struct. Control Health
Monit. 21 (6) (2014) 890–906.
[14] V. Gupta, M. Sharma, N. Thakur, Active structural vibration control: robust to temperature variations, Mech. Syst. Signal Process. 33 (2012) 167–180.
[15] C. Dogruer, A. Pirsoltan, Active vibration control of a single-stage spur gearbox, Mech. Syst. Signal Process. 85 (2017) 429–444.
[16] A. Staino, B. Basu, S. Nielsen, Actuator control of edgewise vibrations in wind turbine blades, J. Sound Vib. 331 (2012) 1233–1256.
[17] B. Fitzgerald, B. Basu, Cable connected active tuned mass dampers for control of in-plane, J. Sound Vib. 333 (2014) 5980–6004.
[18] A. Veletsos, B. Verbic, Vibration of viscoelastic foundations, Earthq. Eng. Struct. Dynam. 2 (1) (1973) 87–102.
[19] M. Achmus, Y. Kuo, K. Rahman, Behavior of monopile foundations under cyclic lateral load, Comput. Geotech. 36 (5) (2009) 725–735.
[20] B. Fitzgerald, B. Basu, Structural control of wind turbines with soil structure interaction included, Eng. Struct. 11 (2016) 131–151.
[21] B. Jonkman, L. Kilcher, TurbSim User’s Guide: Version 1.06.00, National Renewable Energy Laboratory, Technical Report, 2012.
[22] M. Hansen, Aerodynamics of Wind Turbines, James & James (Science Publishers) Ltd, 2000.
[23] O. Faltinsen, Sea Loads on Ships and Offshore Structures, Cambridge University Press, Cambridge, UK, 1990.
[24] IEC, Wind turbines, Part 3: design requirements for offshore wind turbines, IEC 61400-3 (ed. 1), International Electrotechnical Commission, Geneva,
Switzerland, 2009.
[25] C. Sun, Structural Vibration Control of Nonlinear Systems Using the Smart Tuned Mass Damper (STMD) and the Nonlinear Tuned Mass Damper (NTMD)
in Parallel Thesis, Rice University, Houston, 2013.
[26] J. Jonkman, S. Butterfield, W. Musial, G. Scott, Definition of a 5-MW Reference Wind Turbine for Offshore Systems Development, Technical Report,
NREL, 2009.
[27] W. Carswell, J. Johansson, F. Lovholt, S. Arwade, C. Madshus, D. DeGroot, A. Myers, Foundation damping and the dynamics of offshore wind turbine
monopiles, Renew. Energy 80 (2015) 724–736.
[28] G. Warburton, Optimum absorber parameters for minimizing vibration response, Earthq. Eng. Struct. Dynam. 9 (3) (1981) 251–262.
[29] C. Sun, R. Eason, S. Nagarajaiah, A. Dick, Hardening Düffing oscillator attenuation using a nonlinear TMD, a semi-active TMD and a multiple TMD, J.
Sound Vib. 332 (4) (2013) 674–686.