Engineering Structures: Sciencedirect

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Engineering Structures 207 (2020) 110209

Contents lists available at ScienceDirect

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

Optimal tuning of the tuned mass damper (TMD) for rotating wind turbine T
blades
Zili Zhang
Department of Engineering, Aarhus University, 8000 Aarhus, Denmark

ARTICLE INFO ABSTRACT

Keywords: This paper presents the explicit formulas for optimal tuning of the tuned mass damper (TMD) for damping
Tuned mass damper edgewise vibrations in rotating wind turbine blades. A 2-DOF model is first established for the rotating blade-
Wind turbine blade TMD system, which acts as the basis for analytically deriving the optimal frequency-tuning and the optimal
Edgewise vibration damping-tuning formulas. Two optimality criteria are considered, one based on equal dynamic amplification at
Optimal tuning
two neutral frequencies and the other one based on equal modal damping ratio of the free vibration modes. As a
Vibration control
Complex natural frequencies
result, two slightly different optimal frequency-tuning formulas are obtained. The formula for optimal TMD
damping ratio is obtained by means of dynamic amplification analysis, and the optimal damping ratio turns out
to be dependent on the rotor rotational speed. Comparison of the two optimal frequency-tuning formulas are
performed in terms of the dynamic amplification curves and the root loci, using the NREL 5 MW wind turbine as
the numerical example. For realistic rotor rotational speeds, the two optimal frequency-tuning formulas lead to
almost identical TMD parameters.

1. Introduction with misaligned wind inflow [4]. During onset of aeroelastic instability,
the sum of structural damping and aerodynamic damping becomes
Modern multi-megawatt wind turbines are designed and manu- negative, and energy is pumped from the wind flow to the structure,
factured with increasingly larger rotor blades and taller towers, in order resulting in exponentially increasing vibrational amplitude. Therefore,
to capture more wind energy during its lifetime and to lower the le- adding damping into edgewise mode becomes an important design
velized cost of energy (LCoE). Recently, GE Renewable Energy has consideration for protecting the wind turbine blades from damage
developed the HALIADE-X 12 MW wind turbine, with a blade length of during the design period.
107 m and a tower height of 260 m. The increased size of wind turbines There has been increasing research interest in vibration control of
leads to the fact that the blades and the tower are becoming more wind turbine blades in the past few years. Different types of vibration
flexible, thus more vulnerable to dynamic excitations from wind and control devices have been proposed, either by directly changing the
sea waves. Structural vibrations of the wind turbine components not aerodynamic loads experienced by the blades, i.e. aerodynamic control
only contribute to the fatigue damage but also adversely affect the devices, or by introducing extra damping to the structure through in-
power production [1]. stalled mechanical dampers, i.e. structural control devices [5]. Aero-
The modes of vibrations in wind turbine blades are classified as dynamic control devices are normally small devices distributed along
flapwise and edgewise modes, representing vibrations out of the rotor the span of the blade, and through a combination of sensing, control
plane and in the rotor plane, respectively. Under normal operational and actuation, these devises dynamically affect the flow and thus the
conditions (the flow is attached to the blade), the flapwise vibration is aerodynamic loads on the blades. Typical examples are the trailing edge
associated with very strong aerodynamic damping [2], implying the flap (TEF) [6], the microtabs [7], the synthetic jets [8] and the vortex
response to be mainly quasi-static. On the contrary, the edgewise vi- generators [9]. All these devices are functioning by changing lift coef-
bration is related to light or almost no aerodynamic damping [2,3], and ficients of the airfoil, and thereby mainly affect the out-of-plane (flap-
is dominated by dynamic response rather than quasi-static. There is also wise) blade loads. The lightly damped edgewise vibration is almost not
risk of aeroelastic instability in the edgewise mode for some combina- affected by them. Furthermore, controller conflicts may take place due
tions of blade properties and operational conditions, e.g. wind turbine to multiple objectives such as the load reduction and the power
with high performance rotor operating close to stall, or parked turbine smoothing. On the other hand, structural control technologies, which

E-mail address: zili_zhang@eng.au.dk.

https://doi.org/10.1016/j.engstruct.2020.110209
Received 11 July 2019; Received in revised form 7 December 2019; Accepted 9 January 2020
Available online 08 February 2020
0141-0296/ © 2020 Elsevier Ltd. All rights reserved.
Z. Zhang Engineering Structures 207 (2020) 110209

Fig. 1. Definition of coordinate systems, geometry and degrees of freedom.

have achieved significant success in civil engineering structures, turn theoretical framework has been developed for the blade-mounted TMD.
out to be a promising alternative and in recent years are being in- For the present system, two extra parameters enter into the system
creasingly investigated for application in wind turbine blades. equations, i.e. the rotational speed of the rotor and the installation lo-
Passive, semi-active and passive control solutions have all been cation of the TMD (since it's not possible to install the TMD at the blade
proposed. Several types of pendulum-analogous passive devices have tip), making it different from all the previous studies. A 2-DOF model is
been introduced to be installed inside the blade for damping edgewise first established for edgewise vibration of a rotating blade equipped
vibrations, including roller dampers [14], tuned liquid column dampers with a TMD, by means of analytical dynamics. The blade is modelled as
(TLCD) [15], circular liquid column dampers (CLCD) [16] and tuned a rotating cantilever Euler-Bernoulli beam with varying cross-sectional
liquid dampers [17,18]. Mathematical models have been developed for properties. Both the geometric stiffness due to the centrifugal force and
these novel devices, enabling time-domain simulations of the coupled the centripetal softening effect are included in blade edgewise vibra-
wind turbine-damper system. However, explicit formulas for optimal tion, and both effects are dependent on the rotor rotational speed. Next,
damper parameters have not been given, and numerical optimization explicit formulas for both the optimal frequency tuning and optimal
needs to be performed for damper design. Semi-active tuned mass damping tuning of the TMD are developed by means of the dynamic
damper (TMD) was investigated for controlling blade flapwise vibra- amplification analysis, based on the criterion that the dynamic ampli-
tions [10], although the modal damping in this direction is already very fication at three specific excitation frequencies are identical. Further-
high due to the aerodynamic damping. Active TMDs have also been more, an alternative optimal frequency-tuning formula (slightly dif-
studied for mitigating blade edgewise vibrations, and the active TMD ferent from the previous one) is also derived by means of complex
achieves greater response reductions than the passive counterpart [11]. natural frequency analysis, based on the criterion that the two modal
Further, Krenk et al. [12] proposed an active strut mounted near the damping ratios of the free vibration modes of the 2-DOF system are
root of each blade for suppressing blade vibrations, the concept of always equal. Comparison of the two formulas are performed in terms
which is based on resonant interaction between the rotor and the of the dynamic amplification curves and the root loci, using a 5 MW
controller. Staino et al. [13] presented the use of active tendons wind turbine as the numerical example. Parametric study shows that
mounted inside each blade for active control of edgewise vibrations, within realistic rotor rotational speed range, the two frequency-tuning
which allows a variable control force to be applied in the edgewise formulas lead to almost identical TMD parameters.
direction.
Comparing with active devices, passive and semi-active dampers are
more simple, reliable, cost-effective, and require little external power. 2. Equations of motion
This means passive or semi-active dampers might be more applicable by
the wind industry. The classical TMD has been extensively used in civil 2.1. Definition of the problem
engineering applications. Analytical formulas have been developed for
optimal design of TMDs attached to an undamped or a damped SDOF The model of a rotating blade equipped with a TMD is illustrated in
system (which can represent one modal coordinate of the structure), Fig. 1. The edgewise vibration of the blade is described in the moving
subjected to harmonic excitations [19,20,21] or white noise excitations (x1, x2 , x3) -coordinate system. x1 axis is unidirectional to the mean
[22,23]. The classical optimal design procedure for SDOF-TMD system wind velocity, and x3 axis is placed along the undeformed blade axis
can be extended to cases where the TMD is mounted to a flexible orientating from the hub towards blade tip. Each blade (assumed to be
structure, by accounting for the additional background flexibility from identical) is modelled as a Euler–Bernoulli beam, and its mass per unit
non-resonant modes [24]. Further, various numerical optimization length and bending stiffness in the edgewise direction are denoted µ (x3)
techniques [25,26,27] have also been proposed for optimal design of and EI (x3) , respectively. The TMD is installed inside the blade, and it
TMDs, which enable taking the structural damping and other types of consists a block mass md connected to the blade through a spring with
excitations into account. However, there is presently a gap in the lit- spring stiffness kd and a viscous damper with damping coefficient cd .
erature and studies reported on optimal design of the TMD for vibration For describing the velocity vector of the TMD, a fixed global
control of rotating wind turbine blades are sparse. (X1 , X2 , X3 ) -coordinate system is also introduced as will be shown in the
This paper aims to address the knowledge gap through establishing following.
analytical formulas for optimal tuning of the TMD installed inside a The rotational speed of the wind turbine rotor is denoted . Then,
rotating blade. Inspired by Krenk's work [20], a comprehensive the azimuthal angle j (t ) for blade j is given by:

2
Z. Zhang Engineering Structures 207 (2020) 110209

2 centrifugal force. From Euler-Bernoulli beam theory, k 0 can be ex-


j (t ) = t+ (j 1), j = 1, 2, 3
3 (1) pressed as:
In the following, only blade j = 1 is considered since all three blades 2 2
L d 2 (x3) d (x3)
have identical geometrical and structural parameters. Hence, k0 ( ) = EI (x3) + N (x3 , ) dx3
0 dx32 dx3 (9)
j (t ) = (t ) = t .
The local edgewise displacement field u2 (x3, t ) of the rotating blade
where N (x3, ) is the centrifugal axial force per unit length along the
is defined positive in the positive x2 direction. u2 (x3, t ) can be dis-
blade, which is written as:
cretized by either finite element method or modal expansion method.
Since the TMD is only targeted to damp a single vibrational mode and N (x3, )= 2
L
µ( ) d
the fundamental mode (1st mode) is the most critical mode for wind- x3 (10)
excited vibrations, u2 (x3, t ) is described by a single degree of freedom Therefore, the modal stiffness k 0 ( ) is dependent on the rotor ro-
q (t ) as: tational speed , and is reduced to pure elastic stiffness for non-rotating
2
u2 (x3, t ) = (x3) q (t ) (2) blade, i.e. k 0 (0) =
L
EI (x3)
d2 (x3)
dx3 .
0 dx32
where (x3) is the fundamental edgewise mode shape of the blade, Then, the fundamental angular eigenfrequency of the blade is ob-
which is normalized to 1.0 at the blade tip. q (t ) is defined positive in tained as:
the negative x2 direction, due to the fact that the blade will deform in
k0 ( )
the negative x2 direction in reality. 0( )=
Assuming the TMD to be mounted at the coordinate x3 = x 0 , the
m0 (11)
local displacement of the blade at this position is written as: which is dependent on as well.
u2 (x 0 , t ) = (x 0) q (t ) = aq (t ) (3) Using Eqs. (7) and (8), the equations of motion of this 2-DOF system
can be obtained from the stationarity conditions of the Euler-Lagrange
where a = (x 0) . Due to the normalization of the mode shape (x3) , a equations [28]:
takes values between 0 and 1 depending on the mounting location x 0 .
Furthermore, the displacement of the TMD relative to the deformed d T T
+
U
= f (t ) c0 q
blade is denoted u (t ) , as shown in Fig. 1. Therefore, q (t ) and u (t ) make dt q q q
up the degrees of freedom of the 2-DOF blade-TMD system.
(m 0 + a2md ) q¨ + amd u¨ + c0 q + [k 0 (m 0 + a2md ) 2] q amd 2u

2.2. Analytical dynamics = f (t ) + amd g sin (12)

The x2 and x3 components of the velocity field of the blade can be d T T U


+ = cd u
respectively written as: dt u u u
v2 (x3 , t ) = x3 (x3 ) q (t ) amd q¨ + md u¨ + cd u amd 2q + (k d md 2) u = md g sin (13)
v3 (x3 , t ) = (x3) q (t ) (4)
where f (t ) denotes the modal load on the blade due to turbulence and
The components of the position vector and velocity vector of the the gravity. c0 q (t ) indicates the linear viscous damping force. f (t ) can
TMD in the fixed global (X2 , X3) -coordinate system (as shown in Fig. 1) be calculated using a more sophisticated 13-DOF aeroelastic wind tur-
are given by: bine model [5]. c0 = 2 0 m 0 0 indicates the modal damping coefficient
of the primary structure, with 0 being the related modal damping ratio.
X2, d (t ) = x 0 sin aqcos ucos It should be noted that the term (m 0 + a2md ) 2 in Eq. (12) signifies
X3, d (t ) = x 0 cos aqsin usin (5) the centripetal softening effect in the edgewise vibration.
Eqs. (12) and (13) can be combined into the matrix form:
V2, d (t ) = (x 0 + aq + u)cos + (aq + u) sin
m 0 + a2md amd q¨ c0 0 q
V3, d (t ) = (x 0 + aq)sin (aq + u) cos (6) + +
amd md u¨ 0 cd u
Hence, the total kinetic energy of the system (i.e. one blade and one k0 (m 0 + a2md ) 2 amd 2 q f (t ) + amd g sin
TMD) becomes: =
amd 2 kd md 2 u md g sin
1 L 1
T (t ) = µ (x3)(v22 (x3, t ) + v32 (x3, t )) dx3 + md (V2,2 d (t ) + V3,2 d (t )) (14)
2 0 2
1 2 2 2 1 2
= m 0 (q + q ) + m1 q + m2
2 2 3. Optimal tuning of the blade-mounted TMD using the dynamic
1 amplification
+ md [(x 0 + aq + u)2 + (a q + u)2]
2 (7)
L 3.1. Dynamic amplification analysis of the 2-DOF system
where m 0 = 0
µ (x3) 2 (x ) dx
3 3 denotes the modal mass of the blade.
L
Further, m1 = 0
µ (x3) x3 (x3) dx3 is a mass parameter coupling the For deriving analytical expressions of the optimal tuning conditions
L
modal response with the rotor rotation. m2 = 0 µ (x3) x32 dx3 is the mass of the TMD, the following approximations are made: (1) The primary
moment of inertia of the blade with respect to the hub. It should be structure has negligible damping, i.e. c0 = 0 . This is reasonable because
noted that m1 and m2 will disappear when inserting Eq. (7) into the of the inherently low structural damping of the blade and the negligible
Euler-Lagrange equation in the following. aerodynamic damping of the edgewise mode. (2) The two terms
The total potential energy of the system is written as: amd gsin and md gsin in the load vector are negligible comparing
1 1 with the term f , due to the insignificant TMD mass md .
U (t ) = md g (x 0 cos aqsin usin ) + k 0 q2 + kd u2 Represent the responses q (t ) , u (t ) and load f (t ) in terms of har-
2 2 (8)
monic components with angular frequency :
where g is the gravitational acceleration. k 0 denotes the modal stiffness
of the blade including the geometric stiffness resulting from the q (t ) = q0 ei t , u (t ) = u0 ei t , f (t ) = f0 e i t
(15)

3
Z. Zhang Engineering Structures 207 (2020) 110209

where q0 and u 0 are the complex amplitudes of the forced responses of From Eq. (16), the dynamic amplification can be expressed in the
the blade (fundamental edgewise modal response) and of the TMD, following format [20]:
respectively. f0 is the real amplitude of the load. i is the complex unit. 2
Inserting Eq. (15) into Eq. (14), the normalized amplitude of the q0 A + 2i d B q0 A2 + (2 d )2B2
= =
structural response can be written as: f0 / k 0 C + 2i d D f0 / k 0 C 2 + (2 d ) 2D 2 (21)

2 2 2 2
q0 0 ( d + 2i d d )
=
f0 / k 0 4 [ 2
0 + (1 + a2µ) 2
d 2 2] 2 +[ 2
d
2
0
2
d (1 + a2µ) 2 2
0
2 + 4] + 2i d d [ 2
0 (1 + a2µ) 2 (1 + a2µ) 2]
(16)

with the following normalized parameters of the TMD introduced: The condition of the dynamic amplification being independent of d
leads to:
md kd cd
µ= , d = , d =
m0 md 2 k d md (17) A2 B2
= 2 AD = ± BC
C2 D (22)
where µ , d and d are the mass ratio, natural frequency (tuning fre-
quency) and damping ratio of the TMD, respectively. Inserting expressions of A , B , C and D and realizing that the plus
Similarly, the normalized amplitude of the relative TMD motion can sign in Eq. (22) only results in trivial solution = 0 , the following
be written as: equation fulfilled by = M and = N is obtained:

u0 a 02 2 + a 2
0
2
=
f0 / k 0 4 [ 2
0 + (1 + a2µ) 2
d 2 2] 2 +[ 2
d
2
0
2
d (1 + a2µ) 2 2
0
2 + 4] + 2i d d [ 2
0 (1 + a2µ) 2 (1 + a2µ) 2]
(18)

The normalized form of the so-called frequency equation (or char- 2 2


(2 + a2µ) 4 2[ (2 + a2µ ) 2 + (1 + a2µ) d]
2
acteristic polynomial) is obtained by setting the denominator in Eq. 0

(16) (or in Eq. (18)) to zero: + [2 2


d
2
0 2 2
d (1 + a2µ ) 2 2 2
0
2 + (2 + a2µ ) 4] =0 (23)
4 [ 2
0 + (1 + a2µ ) 2
d 2 2] 2 +[ 2
d
2
0
2
d (1 + a2µ ) 2 from which the sum of 2
M and 2
N is directly given by:
2 2 4] 2 2 2
0 + + 2i d d [ 0 (1 + a2µ) 2 (1 + a2µ ) 2] =0 2 2 2[ 0 (2 + a2µ) 2 + (1 + a2µ ) d]
M + N =
(19) 2+ a2µ (24)
which determines the complex natural frequencies of the free vi- Since at these two frequencies the dynamic amplifications are in-
bration modes of the blade-TMD system, as will be elaborated in dependent of d , the simplest formulation is the one with d , which
Section 4. from Eq. (16) becomes:
As a special case, in the limit of infinity TMD damping ratio d , i.e.
, the viscous damper constrains the relative motion u (t ) of the q0 ±1
d =
TMD mass. In this case, the solution to Eq. (19) becomes real, which f0 / k 0 M ,N
1 (1 + a2µ )( / 0)
2 (1 + a2µ )( / 0)
2
(25)
represents the natural frequency of the blade with a locked damper:
Equal dynamic amplification at M and N leads to:
2
0() 2 1
=
1 + a2µ (20) 1 (1 + a2µ)( / 0)
2
(1 + a2µ)( M / 0 )2
This natural frequency will act as a reference frequency and play an 1
=
important role in deriving the analytical formulas in the present paper. 1 (1 + a2µ)( / 0 ) 2 (1 + a2µ)( N / 0)
2
(26)
For optimal tuning of the TMD, the optimal values of d and d need
which gives the second equation for 2
M + N:
2
to be determined, once µ is preselected. This can be conventionally
done based on the dynamic amplification of the structural response, 2 2 2 2
0 2(1 + a2µ) 2
+ =
which is the absolute value of Eq. (16). M N
1 + a2µ (27)

Comparing Eqs. (24) and (27), the explicit formula for the optimal
3.2. Optimal frequency tuning by equalized dynamic amplification
frequency of the TMD becomes:

From the classical TMD theory [19,20], it is known that around 0 0( ) d 1


( d )opt = =
there are two excitation frequencies M and N where the dynamic 1 + a2µ 1 + a2µ
0 opt (28)
amplifications are independent of the TMD damping ratio d . M and N
are named the neutral frequencies. As will be theoretically proved in where the rotational speed only enters implicity through 0 ( ). It is
the following, it turns out that this is also true for the present case, i.e. a interesting to note that the optimal frequency ratio ( d / 0 )opt given by
rotating blade installed with a TMD. The classical TMD theory [19,20] Eq. (28) is independent of , which is an extremely nice property. From
gives the optimal frequency-tuning criterion, i.e. the optimal frequency the optimal value ( d )opt , the optimal spring stiffness kd of the TMD is
d is determined by setting the dynamic amplification at these two determined as kd = ( d )2opt md .
neutral frequencies equal. This will also be the first criterion to be used Inserting the optimal frequency ( d )opt given by Eq. (28) into Eq.
in the present paper. (23), the quadratic equation with the roots being M 2
and N2 can be

4
Z. Zhang Engineering Structures 207 (2020) 110209

written as: structural motion, the explicit formulas for optimal frequency tuning
4 2 ( d / 0 )opt and optimal damping tuning ( d )opt of the blade-mounted
2 2a2µ 2 a2µ 4
4 2 2 2 + =0 TMD are obtained, as given by Eqs. (28) and (34), respectively.
2 + a2µ (29)
where (the natural frequency in the case of a locked damper) has 4. Complex natural frequencies and equal modal damping
been given by Eq. (20). The solutions to Eq. (29) can be written ex- analysis
plicitly as:
Once the values of µ , a , , 0 , d and d are given, the quartic
4
2 2 4 2 2a2µ 2 2 a2µ 4
frequency equation Eq. (19) will result in four complex roots, among
M, N = ±
2 + a2µ (30) which 1 and 2 are located in the first quadrant, and the symmetrically
located roots ¯ 1 and ¯ 2 are in the second quadrant, with the bar
which indicate that the locations of the two neutral frequencies are
symbol denoting the complex conjugate. 1 and 2 (or ¯ 1 and ¯ 2)
dependent on the rotor rotational speed . Moreover, Eq. (30) also
represent the complex natural frequencies of the free vibration modes
implies that is located in the interval between M and N . Inserting
of the blade-TMD system and thus reveal certain dynamic character-
Eq. (30) into Eq. (25), the dynamic amplification at M and N be-
istics of this 2-DOF system. By varying the TMD damping ratio d from
comes:
zero to infinity (using a large value in practice), the so-called root locus
q0 2 + a2µ can be obtained [20,24].
= For the classical SDOF-TMD system, the optimal frequency tuning
f0 / k 0 a2µ (31)
M ,N
criterion (equal dynamic amplification at the two neutral frequencies)
which turns out to be independent of . This implies that although the as presented in Section 3.2 is equivalent to the existence of a local bi-
locations of the neutral frequencies vary as changes, the dynamic furcation frequency (corresponding to the TMD damping ratio ),
amplification at the neutral frequencies is unchanged as long as a and µ which is the intersection of two branches in the root locus [20]. This
are fixed. This - independence property is only obtained when ( d )opt will further ensure that for all cases with d , the modal damping
given by Eq. (28) has been applied. Eq. (31) will be used for de- ratios of the two free vibration modes are identical [20,30]. However,
termining the optimal damping ratio of the TMD in Section 3.3. for the present case of rotating blade with TMD mounted, the above
indicated equivalence is lost, i.e. equal dynamic amplification at the
3.3. Optimal damping tuning two neutral frequencies does not guarantee the existence of the fre-
quency bifurcation point as well as the equal modal damping, as will be
Den Hartog [19] and Krenk [20] proposed different criteria for the shown in the numerical example in Section 5. On the other hand, this
optimal damping ratio of the TMD. It turns out that these two criteria gives an alternative way of deriving the optimal frequency-tuning for-
lead to similar response reduction effect [20,29], but the Krenk cri- mula (slightly different from that in Eq. (28)), obtained by the criterion
terion provides a fairly level behavior of the dynamic amplification of that the existence of the bifurcation frequency is ensured.
the relative TMD motion [20]. Therefore, the same principle as in Krenk
criterion is employed in the present study, i.e. the dynamic amplifica- 4.1. Optimal frequency tuning by ensuring the existence of the bifurcation
tion is equal at three excitation frequencies M , N and . This means frequency
that the TMD damping ratio should be determined so that the dynamic
amplification at equals that in Eq. (31). The existence of the local bifurcation frequency implies that the
Inserting the expressions of given by Eq. (20) and ( d )opt given quartic frequency equation can be expressed as:
by Eq. (28) into Eq. (16), the normalized amplitude (complex) of the ( )2 ( + ¯ ) 2 = 0 (35)
structural response at the excitation frequency = becomes:
This equation can be rewritten in a form similar to Eq. (19):
a2µ 04 3
0
2
0 2
q0 + 2i d 1 + a2µ 4 4i Im[ ] 3 (2| |2 + 4Im[ ]2 ) 2 + 4i Im[ ]| |2 +| |4 = 0
(1 + a2µ)2 1 + a2µ
=
f0 / k 0 a2µ 04 (36)
(1 + a2µ)2
Next, comparison of the coefficients in Eq. (36) and Eq. (19) is to be
2i d (1 + a2µ) 1 2
=1 performed. The magnitude of the bifurcation frequency, | |, follows
2
a2µ 1 + a2µ 0 (32) directly by comparing the ratio of the coefficients to the linear and
which indicates that the real part of the normalized amplitude at cubic terms:
= is always 1, while its imaginary part is dependent on d , , a , µ 2
0
and 0 . The dynamic amplification at = then follows from Eq. | |2 = 2 | |=
1 + a2µ (37)
(32) as:
where Eq. (20) has been used. Eq. (37) indicates that the bifurcation
2
q0 4 d (1 + a2µ) 2 1 2
frequency is located on a circle around the origin of the complex
= 1+
f0 / k 0 a4µ2 1 + a2µ 2
0 (33) plane with a radius equal to .
Comparison of the constant term in Eqs. (36) and (19) provides:
By equating the dynamic amplification at to that at M and N
2 a4µ2 ) 2
given by Eq. (31), the optimal TMD damping ratio is determined as: | |4 = 02 d2 2 2
d (1 + a µ)
2 2
0
2+ 4 d
=
1 0 (1
0 opt 1 + a2µ 2
0 (1 + a2µ) 2
2
1 a2µ 0 (38)
( d )opt = 2
2 1 + a2µ 0 (1 + a2µ) 2
(34) which gives the alternative optimal frequency tuning formula.
Comparing with the optimal frequency-tuning formula in Eq. (28), Eq.
which for prescribed values of µ and a reaches its minimum value
1 a2µ
(38) results in a slightly larger frequency ratio, due to the term inside
( d )opt = 2 1 + a2µ
when = 0 (non-rotating blade). Eq. (34) denotes the square root operator. Moreover, unlike Eq. (28), ( d / 0 )opt given by
that the optimal TMD damping ratio is dependent on and 0 , which is Eq. (38) is dependent on 0 and . On the other hand, Eq. (38) becomes
on the contrary to the optimal frequency ratio Eq. (28). identical to Eq. (28) when = 0 (non-rotating blade).
Hence, based on the analysis of the dynamic amplification of the Eq. (38) might also be Taylor truncated into the following

5
Z. Zhang Engineering Structures 207 (2020) 110209

expression: ensures that the modal damping ratios of the two free vibration
2
modes are equal as long as the TMD damping ratio d . This means
1 (1 a4µ2) 2 1 a2µ 2
that the two roots in the first quadrant (also the two roots in the second
d 0
= +
0 opt 1 + a2µ 2
0 (1 + a2µ) 2 1 + a2µ 2 2
0 (1 + a2µ) 2 (39)
quadrant) are inverse points with respected to a circle centered at origin
which more clearly indicates the deviation from Eq. (28). Since a2µ with radius [20]. Therefore, the four roots in the complex plane can
has a very small value (<0.1), the difference between Eq. (39) and Eq. be written as j , 2 / ¯ j , ¯ j and 2
/ j with j = 1 or 2. The quartic
(28) turns out to be insignificant (for realistic rotor rotational speed ). frequency equation can thus be written as:
This implies that even though enters into the optimal frequency- 2 2
tuning formula in Eq. (38), it only slightly influences the frequency- ( j) ( + ¯j) + =0
tuning of the TMD. ¯j j (44)
Hence, two slightly different optimal frequency-tuning formulas
Eq. (44) can be rewritten in a symmetric format [20]:
have been developed, i.e. Eq. (28) and Eq. (38), based on different
optimality criteria. This also implies that for the present case of a ro- 4 4i r 3 (4r 2 2 + 4 2) 2 2 + 4i r 3
+ 4
=0 (45)
tating blade equipped with a TMD, the optimal frequency tuning is not
with the two parameters and r defined as:
unique, but with two options. The optimal TMD damping ratio, Eq.
(34), is obtained based on the optimal frequency-tuning formula Eq. =
Im[ j ]
, j = 1, 2
| j|
(28). On the other hand, it's difficult to derive the optimal TMD
damping ratio when Eq. (38) is chosen for the optimal frequency
tuning, because the nice property of equal dynamic amplification is
r=
1
2 ( | j|
+ | j| )= | 1 |+| 2 |
2 (46)
lost. Nevertheless, in the present paper, Eq. (34) will still be used as the The parameter = 1 = 2 is the modal damping ratio, which is
optimal TMD damping ratio when Eq. (38) is used for optimal fre- identical for the two modes. As for the parameter r , it can be observed
quency tuning, partly because Eq. (28) and Eq. (38) actually give si- that r 1, and r = 1 is reached at the bifurcation frequency.
milar results for realistic values of rotor rotational speed , and partly By comparison of the coefficient to the linear term of Eq. (45) and of
because the choice of optimality criterion for the TMD damping ratio is Eq. (19) with the optimal frequency tuning Eq. (38) inserted, the modal
quite arbitrary even for the classical SDOF-TMD system [19,20]. damping ratio can be obtained as:
Next, comparison of the coefficient to the cubic term in Eqs. (36)
and (19) leads to: d 0 (1 + a2µ )( 2
0 + (a4µ2 1) 2)
= 2
2 2r 0 (1 + a2µ ) 2
(47)
0 0 (1 a4µ2) 2
4Im[ ] = 2 d (1 + a2µ ) Im[ ] = 2
2 0 (1 + a2µ ) 2
Comparison of the coefficient to the quadratic term of Eqs. (45) and
(40) (19) will provide anther equation for and r . The two unknowns and
r can then be solved numerically using this equation together with Eq.
where is the TMD damping ratio at the bifurcation frequency, and
(47). However, an approximate value of can be obtained directly from
is determined by further comparison of the coefficient to the quadratic
term: Eq. (47) without numerical calculation, i.e. d
2
, since r 1 and the
term inside the parenthesis is also approximately 1. This indicates that
2 2
2| |2 + 4Im[ ]2 = 0 + (1 + a2µ ) d 2 2
the modal damping ratio is approximately half of the TMD damping
a2µ 2 ratio d , for all cases with d .
0
= 2 For d > , bifurcation takes place and the modal damping ratios of
1 + a2µ (1 a4µ2 ) 2
(41)
0
the 2-DOF system are no longer equal.
where Im [ ] has been eliminated using Eq. (40). Eq. (41) is reduced to
a2µ 5. Numerical example
= when = 0 . It should be emphasized that when Eq. (28) is
1 + a2µ
used, does not exist so the expression in Eq. (41) merely provides a Data from the NREL 5 MW reference turbine [31] were employed to
value of the TMD damping ratio without any specific meaning. calibrate the structural model of the wind turbine blade. Each blade has
a length of 61.5 m and an overall mass of 17740 kg, with the edgewise
4.2. Equal modal damping ratio of the 2-DOF system bending stiffness, the mass per unit length and the fundamental edge-
wise mode shape given by [31]. The modal mass m 0 is thus calculated
The modal damping ratio j ( j = 1, 2 ) is defined as the relative as m 0 = 1.354 × 103 kg.
magnitude of the imaginary part of the complex frequency j (the two The rated rotational speed of the rotor is 0 = 1.27 rad/s, corre-
roots of Eq. (19) located in the first quadrant): sponding to the rated wind speed of 11.4 m/s. For normal operational
Im[ j ] conditions, the rotor rotational speed increases from 0 to 0 when the
j = , j = 1, 2 wind speed increases from the cut-in speed (3 m/s) to the rated speed
| j| (42)
(11.4 m/s), and keeps constant at 0 for wind speed larger than 11.4 m/
The common modal damping ratio at the bifurcation point s (up to the cut-out wind speed 25 m/s) due to the functioning of the
( d = ) is then determined by inserting Eqs. (37), (40) and (41) into pitch controller. Fig. 2 shows the fundamental edgewise eigenfrequency
Eq. (42):

0
2
0 ((1 a4µ2 )) 2 a2µ 2
0
= 2
= 2
,
2 0 ((1 a4µ2 )) 2 2 0 (1 + a2µ ) 2

for d = (43)

From the first equality in Eq. (43), it is observed that 2


since the
term inside the parenthesis is approximately 1. This implies that the
common modal damping is about half the TMD damping ratio at the
bifurcation frequency . Fig. 2. Fundamental edgewise eigenfrequency 0 of the blade as a function of
As indicated in [20], the existence of the local bifurcation frequency the rotational speed .

6
Z. Zhang Engineering Structures 207 (2020) 110209

Fig. 3. Dynamic amplification of the structural response for µ = 0.05 (md = 67.7 kg) and x 0 = 45 m. Optimal frequency-tuning Eq. (28) has been used. (a) = 0 (non-
rotating blade). (b) = 0 (1.27 rad/s). (c) = 3 0 (3.81 rad/s).

of the blade as a function of the rotor rotational speed (from 0 to Next, Fig. 4 shows the corresponding results when the alternative
3 0 ). It should be noted that in reality the rotor speed will not exceed tuning formula Eq. (38) is used. For the case of = 0 , Fig. 4(a) is ex-
0 (unless the brake system breaks down), and the chosen range of 0 to actly the same as Fig. 3(a) because Eq. (38) becomes identical to Eq.
3 0 is just to illustrate how the centrifugal stiffening effect influences (28) when = 0 . For cases of 0 , the optimal frequency-tuning
the fundamental edgewise eigenfrequency 0 . It is seen that within formula Eq. (38) leads to asymmetric configuration of the curves, i.e.
normal conditions (0 to 0 ), 0 only changes slightly. the dynamic amplifications at the two neutral frequencies are not equal
In the following, the results obtained from the two optimal fre- anymore. The unbalance is insignificant for = 0 , and becomes more
quency-tuning formulas Eqs. (28) and (38) are compared in terms of the significant for = 3 0 . Even though Eq. (34) is still defined as the
dynamic amplification (absolute value of Eq. (16)) and the root loci. optimal TMD damping ratio when Eq. (38) is used for frequency tuning,
the nice property of equal dynamic amplification at M , N and is
lost. Nevertheless, the “optimal” damping formula Eq. (34) results in a
5.1. Dynamic amplification fairly level behaviour of the curve between M and N (although
slightly inclined), which is superior than the under-optimal and over-
Fig. 3 shows the dynamic amplification of the structural response optimal cases.
(as a function of the normalized frequency / 0 ) with the optimal Fig. 5 shows the dynamic amplification of the relative TMD motion
frequency-tuning formula Eq. (28) used, for the mass ratio µ = 0.05 and (with µ = 0.05 and x 0 = 45 m) obtained as the absolute value of Eq.
the TMD mounting location x 0 = 45 m. Three different rotational (18), when the optimal frequency-tuning formula Eq. (28) is used. For
speeds have been considered, i.e. = 0 , 0 , 3 0 , where = 3 0 is each case of , it is seen that the dynamic amplification is independent
regarded as an extreme case. For each , three different TMD damping of the TMD damping ratio d at the particular frequency = .
ratios were considered, including ( d )opt given by Eq. (34), 0.8 × ( d )opt Moreover, ( d )opt given by Eq. (34) leads to a quite flat behavior of the
(representing under-optimal damping), and the damping ratio given dynamic amplification of the relative TMD motion. As indicated in
by Eq. (40) (representing over-optimal damping). It should be reminded [24], the first three derivatives with respect to frequency actually
that the optimal frequency tuning Eq. (28) does not guarantee the ex- vanish at when ( d )opt is used, which demonstrates some kind of
istence of the bifurcation frequency , so the use of Eq. (40) here flatness. On the other hand, any value of d smaller than ( d )opt (such as
merely aims to provide a damping ratio that is larger than the optimal 0.8 × ( d )opt ) results in peaks exceeding the dynamic amplification at
one. , and any value of d larger than ( d )opt (such as ) leads to a single
We can observe that regardless of and d , the dynamic amplifi- narrow peak in the dynamic amplification curve. The above indicated
cation at the two neutral frequencies M and N are always identical, in characteristics hold regardless of the rotor rotational speed , as long as
accordance with Eq. (31). When ( d )opt is used, the dynamic amplifi- the optimal frequency-tuning formula Eq. (28) is used. Similar to the
cation at becomes equal to that at M and N , as shown by the observations in Fig. 3, as increases, the curves become broader and
dashed horizontal line in the figure. ( d )opt also results in a fairly level move towards left. It should be noted that if the alternative tuning
behavior of the curve between the neutral frequencies. On the other formula Eq. (38) is used, the curves for 0 (not shown here for
hand, the curve with under-optimal damping shows a trough between briefness) will not be flat but inclined, similar to those in Fig. 4(b) and
the neutral frequencies, and the curve with over-optimal damping ( ) (c).
has a narrow peak at the frequency rather than two peaks. Com-
paring Fig. 3(a), (b) and (c), it is seen that when increases, the curves 5.2. Root loci of the 2-DOF blade-TMD system
become broader and move towards left, but the symmetric configura-
tion of the curves are maintained in all cases. Comparison of Figs. 3 and 4 demonstrates that formula Eq. (28) is

Fig. 4. Dynamic amplification of the structural response for µ = 0.05 (md = 67.7 kg) and x 0 = 45 m. Optimal frequency-tuning Eq. (38) has been used. (a) = 0 (non-
rotating blade). (b) = 0 (1.27 rad/s). (c) = 3 0 (3.81 rad/s).

7
Z. Zhang Engineering Structures 207 (2020) 110209

Fig. 5. Dynamic amplification of the relative TMD motion for µ = 0.05 (md = 67.7 kg) and x 0 = 45 m. Optimal frequency-tuning Eq. (38) has been used. (a) =0
(non-rotating blade). (b) = 0 (1.27 rad/s). (c) = 3 0 (3.81 rad/s).

superior than Eq. (38) in terms of the balanced dynamic amplification. the optimal TMD damping ratio Eq. (34), and two dashed straight lines
However, Eq. (38) is superior in terms of the resulting equal modal connecting each of these two poionts through the origin of the complex
damping, as will be shown in this subsection. plane (not shown here) are also illustrated. For the case of = 0, these
Fig. 6 shows the root loci diagram (by increasing the TMD damping two lines overlap with each other, implying that the two natural fre-
ratio d from zero to a large value) of the 2-DOF model in the complex quencies are inverse points with respect to a circle centered at origin
plane, when the optimal frequency-tuning formula Eq. (28) is used. with radius . This further confirms that the two modal damping ra-
Two different TMD mounting positions (x 0 = 25 m and x 0 = 45 m) and tios are identical. Actually, the two modal damping ratios are identical
four different rotational speeds ( = 0, 0 , 2 0 and 3 0 ) have been as long as 0 d . The above observations apply to both x 0 = 25 m
considered, where = 2 0 and = 3 0 are regarded as extreme cases. (blue curves) and x 0 = 45 m (red curves), with the difference that the
At zero TMD damping d = 0 , the two roots are located on the real axis. area encircled by the two local branches is smaller for x 0 = 25 m than
As the damping increases, the two roots move into the positive ima- for x 0 = 45 m.
ginary half-plane, and the arrows in the figure indicate how the two The results for cases of 0 are shown in Fig. 6(b)–(d). On the
roots move with increasing d . Fig. 6(a) shows the results when = 0, contrary to Fig. 6(a), the bifurcation point does not exist anymore in
and the two local branches intersect at the bifurcation point (as in- Fig. 6(b)–(d). In each figure, it is seen that the right-side branch forms a
dicated by the square) when d reaches . As d increases further, the local curve ending at , while the left-side branch has non-local
branches split into an increasing branch of non-local character and a character with increasing imaginary part. The two branches don't in-
decreasing branch ending at (indicated by the cross). The two circles tersect with each other. This property becomes more and more sig-
indicate the two natural frequencies ( 1 )opt and ( 2 )opt corresponding to nificant when increases from 0 to 3 0 . Moreover, the two dashed

Fig. 6. Root loci for µ = 0.05 (md = 67.7 kg), x 0 = 25 m and x 0 = 45 m considered. Optimal frequency-tuning Eq. (28) has been used. Cross: . Circle: complex
natural frequencies ( 1 )opt and ( 2 )opt corresponding to ( d )opt . Square: bifurcation frequency . (a) = 0 (non-rotating blade). (b) = 0 (1.27 rad/s), no bi-
furcation point. (c) = 2 0 (2.54 rad/s), no bifurcation point. (d) = 3 0 (3.81 rad/s), no bifurcation point.

8
Z. Zhang Engineering Structures 207 (2020) 110209

Fig. 7. Root loci for µ = 0.05 (md = 67.7 kg), x 0 = 25 m and x 0 = 45 m considered. Optimal frequency-tuning Eq. (38) has been used. Cross: . Circle: complex
natural frequencies ( 1 )opt and ( 2 )opt corresponding to ( d )opt . Square: bifurcation frequency . (a) = 0 (non-rotating blade). (b) = 0 (1.27 rad/s). (c) = 2 0
(2.54 rad/s). (d) = 3 0 (3.81 rad/s).

lines connecting each of the two circles to the origin do not coincide, (38) is superior than Eq. (28) in terms of the equal modal damping
implying that the modal damping ratios of the 2-DOF system are no ratio. However, comparing Fig. 6(b) with Fig. 7(b), it is seen that the
longer equal. Again, this difference is negligible for = 0 , but is more difference is minor. This means that for realistic rotor rotational speeds
significant for = 3 0 . (0 0 ), the difference resulted from Eq. (38) and Eq. (28) is
Next, Fig. 7 shows the corresponding results when the alternative negligible. This can also be observed from the comparison of Fig. 3(b)
tuning formula Eq. (38) is employed. For the case of = 0 , Fig. 7(a) is and Fig. 4(b). Considering that the optimal TMD damping ratio Eq. (34)
exactly the same as Fig. 6(a). For the cases of 0 , the tuning formula has been theoretically derived based on Eq. (28), it is suggested to use
Eq. (38) guarantees the existence of the bifurcation point, regardless of Eqs. (28) and Eq. (34) for the optimal tuning of the TMD for rotating
the rotational speed . Therefore, similar characteristics of the curves blades in practice.
are obtained for Fig. 7(a)–(d). Moreover, equal modal damping ratio is Fig. 8 further illustrates how the modal damping ratios 1 and 2 of
also guaranteed by Eq. (38), as long as the TMD damping ratio is in the the 2-DOF system (obtained by Eq. (42)) change with the TMD damping
range of 0 d . ratio d , when the optimal frequency-tuning formula Eq. (28) is used.
Hence, comparison of Figs. 6 and 7 demonstrates that formula Eq. Fig. 8(a) shows the results when = 0 . In this case, the two modal

Fig. 8. Modal damping ratios j of the 2-DOF system as a function of the TMD damping ratio d. µ = 0.05 (md = 67.7 kg), x 0 = 45 m. Optimal frequency-tuning Eq.
1
(28) has been used. Blue line: 1st modal damping ratio 1. Red line: 2nd modal damping ratio 2. Black dashed straight line: = 2 d
. (a) = 0 (non-rotating blade). (b)
= 0 (1.27 rad/s). (c) =3 0 (3.81 rad/s). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)

9
Z. Zhang Engineering Structures 207 (2020) 110209

damping ratios are equal ( 1 = 2 = ) when 0 d , and bifurcate (ii) For = 0 , Eq. (28) and Eq. (38) lead to the same optimal TMD
into two branches (one increases almost linearly and one decreases parameters as well as the same modal damping ratios, as already re-
asymptotically to zero) when d > . The intersection point is the bi- vealed in Sections 5.1 and 5.2. For 0 , Eq. (28) and Eq. (38) result in
furcation point, corresponding to the square in Fig. 6(a). Besides, a slightly different values of ( d / 0 )opt , and the difference becomes more
black dashed straight line showing the relationship = 2 d is also
1
significant as x 0 increases and as increases. ( 1 )(1) opt and ( 2 )opt are
(1)

plotted. It is seen that actually = 2 d predicts the relationship between different for 0 , while ( 1 )(2) and are always equal. The
1 (2)
opt ( )
2 opt
and d quite well before the bifurcation point, as already indicated by common value of ( 1 )(2) opt and ( ) (2)
2 opt acts almost as the mean value of
Eq. (47). For the cases of 0 shown in Fig. 8(b) and (c), the equal ( 1 )(1) and ( ) (1)
. The difference between ( ) (1)
and ( ) (1)
is insignif-
opt 2 opt 1 opt 2 opt
modal damping property is lost (there is no bifurcation point any more), icant for = 0 and becomes more distinct for = 3 0 . Therefore, for
and 1 2 whenever d > 0 , although the difference is insignificant for realistic rotor rotational speeds (0 0 ), the optimal TMD para-
the case of = 0 in Fig. 8(b). The relationship = 2 d still predicts the
1
meters and the corresponding system dynamic properties resulted from
initial part of the curve quite well. Eq. (28) and from Eq. (38) can be considered unchanged.
It is important to emphasize that larger modal damping ratios of the (iii) As increases with given x 0 , ( d )opt slightly increases.
2-DOF system do not necessarily lead to better performance of the TMD. Moreover, ( d / 0 )opt calculated by Eq. (28) is unchanged, while it
Take the case of = 0 (Fig. 8(a)) as an example, the common modal slightly increases when calculated by Eq. (28). Again, the change of
damping ratio reaches its maximum at the bifurcation point when optimal TMD parameters is insignificant with changing for realistic
d = . However as shown in Fig. 3(a), leads to a single narrow peak speed range (0 0 ). This makes it possible to employ a passive
for the dynamic amplification, which clearly indicates the non-optim- TMD in damping blade edgewise vibrations for the present example
ality of . Since ( d )opt < , the two modal damping ratios ( 1 )opt and investigated (NREL 5 MW wind turbine).
( 2 )opt obtained from ( d )opt is smaller than the modal damping ratio Further, the corresponding results for a larger TMD mass ratio
obtained from (Eq. (42)). µ = 0.1 is shown in Table 2. Similar observations can be made as from
Essentially, the two modal damping ratios are obtained from the Table 1. Comparing the results in Table 2 and Table 1, it is observed
two complex natural frequencies of the free vibration modes of the 2- that for given and x 0 , the increase of µ leads to the increase of ( d )opt
DOF system (which are obtained by solving the frequency equation in and the decrease of ( d / 0 )opt . Actually, the increase of µ is equivalent
Eq. (19)). The free vibration modes imply that the resonance effect with to the increase of x 0 , which is consistent with the fact that a2µ acts as
respect to the external loads is not taken into account, and this is the effective mass ratio entering in all the optimal tuning formulas.
consistent with the fact that the frequency equation is only related to Moreover, the corresponding modal damping ratios (of the 2-DOF
the denominator of Eq. (16). On the other hand, the dynamic amplifi- system) are also increased comparing with those in Table 1. Again, it
cation takes into account both the denominator and the numerator of should be emphasized that the modal damping ratios do not represent
Eq. (16), and is thus more representative of the TMD performance. This the performance of the TMD. The increase of the modal damping ratios
further justifies the preference of using Eq. (28) as the optimal fre- is merely the result of the increased ( d )opt , so that the optimal perfor-
quency-tuning formula, even though it can not guarantee the equal mance of the TMD (with increased µ ) is achieved. Essentially, the
modal damping ratio. performance of the optimized blade-mounted TMD is completely de-
termined by the effective mass ratio a2µ , as clearly revealed by the
expression of the dynamic amplification in Eq. (31).
5.3. Parametric study
5.4. Comparison with the optimal tuning formulas using = 0 (the classic
Table 1 gives the optimal parameters of the TMD (mass ratio
tuning)
µ = 0.05) and the resulting modal damping ratios of the 2-DOF system,
considering three different values of and three different values of x 0 .
The frequency-tuning formula Eq. (28) essentially represents the
The two optimal frequency-tuning formulas Eqs. (28) and (38) have
classic TMD frequency-tuning without rotational effect [20] using the
both been considered for comparison. The two modal damping ratios
effective mass ratio a2µ . Fig. 9 shows the optimal frequency-tuning ratio
obtained by using Eqs. (28) are denoted ( 1 )(1) opt and ( 2 )opt , while they are
(1)

denoted ( 1 )opt and ( 2 )opt when Eqs. (38) is used. From Table 1, the
(2) (2) ( )d
0 opt
as a function of a2µ , using Eq. (28), Eq. (38) and Eq. (39) (Taylor
following observations are to be made: truncated Eq. (38)). It is seen that Eq. (38) and Eq. (39) basically result
(i) For a given , as x 0 increases, ( d )opt increases and ( d / 0 )opt in almost same result even for a very large (unrealistic) value
decreases. Correspondingly, the two modal damping ratios also increase ofa2µ = 0.4. The deviation of Eq. (38) from Eq. (28) increases as a2µ
with x 0 , since they follow the same changing tendency of the TMD increases. As shown in Fig. 9(a), for the case of = 0 , this deviation is
damping ratio d as long as 0 d ast , as already shown in Fig. 8. It is not very significant for realistic values of a2µ ( < 0.1) . This is the reason
seen that the values of the modal damping ratios are approximately half why the difference resulted from the two optimal frequency-tuning
of ( d )opt in all cases (especially for small values of and x 0 ). formulas (Eqs. (28) and (38)) is insignificant for = 0 . For the case of

Table 1
Optimal parameters of the TMD and the resulting modal damping ratios of the 2-DOF system (with the optimal TMD parameters), for the case of µ = 0.05 (md = 67.7
kg).
x 0 [m] a ( d )opt
( )
d
0 opt
Eq. (28) ( )
d
0 opt
Eq. (38) ( 1 )(1)
opt ( 2 )(1)
opt ( 1 )(2)
opt ( 2 )(2)
opt

0 25 0.135 0.0214 0.9991 0.9991 0.0107 0.0107 0.0107 0.0107


35 0.298 0.0470 0.9956 0.9956 0.0236 0.0236 0.0236 0.0236
45 0.516 0.0811 0.9868 0.9868 0.0408 0.0408 0.0408 0.0408
0 25 0.135 0.0217 0.9991 0.9991 0.0110 0.0110 0.0110 0.0110
35 0.298 0.0478 0.9956 0.9957 0.0243 0.0244 0.0243 0.0243
45 0.516 0.0824 0.9868 0.9871 0.0420 0.0422 0.0421 0.0421
3 0 25 0.135 0.0240 0.9991 0.9992 0.0134 0.0136 0.0135 0.0135
35 0.298 0.0528 0.9956 0.9962 0.0294 0.0301 0.0297 0.0297
45 0.516 0.0912 0.9868 0.9886 0.0506 0.0528 0.0517 0.0517

10
Z. Zhang Engineering Structures 207 (2020) 110209

Table 2
Optimal parameters of the TMD and the resulting modal damping ratios of the 2-DOF system (with the optimal TMD parameters), for the case of µ = 0.1 (md = 135.4
kg).
x 0 [m] a ( d )opt
( ) d
0 opt
Eq. (28) ( )d
0 opt
Eq. (38) ( 1 )(1)
opt ( 2 )(1)
opt ( 1 )(2)
opt ( 2 )(2)
opt

0 25 0.135 0.0302 0.9982 0.9982 0.0151 0.0151 0.0151 0.0151


35 0.298 0.0663 0.9912 0.9912 0.0333 0.0333 0.0333 0.0333
45 0.516 0.1139 0.9740 0.9740 0.0576 0.0576 0.0576 0.0576
0 25 0.135 0.0307 0.9982 0.9982 0.0156 0.0156 0.0156 0.0156
35 0.298 0.0674 0.9912 0.9913 0.0343 0.0345 0.0344 0.0344
45 0.516 0.1158 0.9740 0.9745 0.0593 0.0598 0.0596 0.0596
3 0 25 0.135 0.0339 0.9982 0.9984 0.0189 0.0192 0.0191 0.0191
35 0.298 0.0746 0.9912 0.9924 0.0414 0.0429 0.0421 0.0421
45 0.516 0.1284 0.9740 0.9775 0.0711 0.0757 0.0733 0.0733

Fig. 9. Optimal frequency-tuning ratio as a function of the non-dimensional parameter a2µ . (a) = 0 (1.27 rad/s). (b) =3 0 (3.81 rad/s).

Fig. 10. Performance comparison when using the optimal damping ratio Eq. (34) and the classic tuning formula. µ = 0.05 (md = 67.7 kg) and x 0 = 45 m. Optimal
frequency-tuning Eq. (28) has been used. (a) = 0 (1.27 rad/s). (b) = 3 0 (3.81 rad/s).

= 3 0 as shown in Fig. 9(b), this deviation is much more significant. 6. Conclusions


Furthermore, it is interesting to investigate how much difference it
will cause by using the classic damping-tuning formula [20] with the This paper develops explicit formulas for optimal tuning of the TMD
effective mass ratio a2µ inserted. In the present paper, the classic for damping edgewise vibrations of rotating wind turbine blades. A 2-
damping-tuning formula is recovered by assigning = 0 in Eq. (34). DOF model has been established for the blade-TMD system, which acts
Fig. 10 compares the dynamic amplification of the structural response as the basis for analytically deriving the optimal frequency-tuning and
when using the optimal damping ratio Eq. (34) and the classic formula the optimal damping-tuning formulas. Two different optimality criteria
1 a2µ
, for cases and have been used for deriving the optimal frequency-tuning formula, the
( d )classic = 2 1 + a2µ
= 0 =3 0. As seen in
1st one based on the equal dynamic amplification and the 2nd one based
Fig. 10(a), the difference caused by using ( d )classic and ( d )opt is not very on the equal modal damping ratio (equivalently, the existence of the
significant for the case = 0 . The use of ( d )classic leads to a slightly
bifurcation frequency). Unlike the classical SDOF-TMD system where
under-optimal damping ratio, and the curve with ( d )classic shows a
the two criteria result in the same tuning formula, in the present study,
slightly deeper trough between the two neutral frequencies. On the
the rotational effect leads to the fact that two slightly different optimal
other hand, as increases from 0 to 3 0 (Fig. 10(b)), the deviation
frequency-tuning formulas, Eq. (28) and Eq. (38), are obtained from the
becomes much more significant, and the use of ( d )classic leads to a quite
two criteria. The optimal frequency ratio obtained from the 1st criterion
large trough indicating non-optimal behavior of the blade-mounted
is independent on the rotor rotational speed , while that from the 2nd
TMD.
criterion is dependent on . Based on the optimal frequency ratio Eq.

11
Z. Zhang Engineering Structures 207 (2020) 110209

(28), the optimal TMD damping ratio Eq. (34) has been derived, which [5] Zhang Z. Passive and Active Vibration Control of Renewable Energy Structures.
is dependent on . For theoretical completeness as well as the fact that Denmark: Aalborg University; 2015. Ph.D thesis.
[6] Barlas TK, van Kuik GA. Review of state of the art in smart rotor control research for
dynamic amplification represents the performance of TMD better, it is wind turbines. Prog Aerosp Sci 2010;46(1):1–27.
suggested to use Eqs. (28) and (34) for the optimal tuning of the blade- [7] Chow R, Van Dam C. Computational investigations of deploying load control mi-
mounted TMD, although the equal modal damping ratio is not guar- crotabs on a wind turbine airfoil. 45th AIAA Aerospace Sciences Meeting and
Exhibit Reno, NV. 2007.
anteed. [8] Maldonado V, Farnsworth J, Gressick W, Amitay M. Active control of flow se-
For the considered NREL 5 MW wind turbine, the difference resulted paration and structural vibrations of wind turbine blades. Wind Energy 2010;13(2-
from the two optimal frequency-tuning formulas turns out to be insig- 3):221–37.
[9] Prince SA, Badalamenti C, Regas C. The application of passive air jet vortex-gen-
nificant for realistic rotor rotational speeds. Furthermore, the use of erators to stall suppression on wind turbine blades. Wind Energy
classic damping tuning (disregarding ) only leads to slight deviation 2017;20(1):109–23.
from the optimal damping tuning (Eq. (34)) for the considered wind [10] Arrigan J, Pakrashi V, Basu B, Nagarajaiah S. The application of passive air jet
vortex-generators to stall suppression on wind turbine blades. Wind Energy
turbine. The difference becomes more significant as increases (such
2017;20(1):109–23.
as the extreme case of = 3 0 ). Larger rotor rotational speed might be [11] Fitzgerald B, Basu B, Nielsen SRK. Active tuned mass dampers for control of in-
encountered for some other types of wind turbines. In such cases, the plane vibrations of wind turbine blades. Struct Control Health Monitor
distinction between Eq. (28) and Eq. (38) becomes more important, and 2013;20(12):1377–96.
[12] Krenk S, Svendsen MN, Høgsberg J. Resonant vibration control of three-bladed
Eq. (38) provides the alternative optimal frequency tuning if equal wind turbine rotors. AIAA J 2012;50(1):148–1146.
modal damping ratio is preferred. [13] Staino A, Basu B, Nielsen SRK. Actuator control of edgewise vibrations in wind
turbine blades. J Sound Vib 2012;331(6):1233–56.
[14] Zhang Z, Li J, Nielsen SRK, Basu B. Mitigation of edgewise vibrations in wind
CRediT authorship contribution statement turbine blades by means of roller dampers. J Sound Vib 2014;333(21):5283–98.
[15] Zhang Z, Basu B, Nielsen SRK. Tuned liquid column dampers for mitigation of
Zili Zhang: Conceptualization, Data curation, Formal analysis, edgewise vibrations in rotating wind turbine blades. Struct Control Health Monitor
2015;22(3):500–17.
Funding acquisition, Investigation, Methodology, Project administra- [16] Basu B, Zhang Z, Nielsen SRK. Damping of edgewise vibration in wind turbine
tion, Resources, Software, Supervision, Validation, Visualization, blades by means of circular liquid dampers. Wind Energy 2016;19(2):213–26.
Writing - original draft, Writing - review & editing. [17] Zhang Z, Nielsen SRK, Basu B, Li J. Nonlinear modeling of tuned liquid dampers
(TLDs) in rotating wind turbine blades for damping edgewise vibrations. J Fluids
Struct 2015;59:252–69.
Declaration of Competing Interest [18] Zhang Z, Staino A, Basu B, Nielsen SRK. Performance evaluation of full-scale tuned
liquid dampers (TLDs) for vibration control of large wind turbines using real-time
hybrid testing. Eng Struct 2016;126:417–31.
The authors declare that they have no known competing financial
[19] Den Hartog JP. AMechanical vibrations. Courier Corporation 1985.
interests or personal relationships that could have appeared to influ- [20] Krenk S. Frequency analysis of the tuned mass damper. J Appl Mech
ence the work reported in this paper. 2005;72(6):936–42.
[21] Ghosh A, Basu B. A closed-form optimal tuning criterion for TMD in damped
structures. Struct Control Health Monitor 2007;14(4):681–92.
Acknowledgments [22] Warburton GB. Optimum absorber parameters for various combinations of response
and excitation parameters. Earthquake Eng Struct Dyn 1982;10(3):381–401.
The supports of Aarhus University Research Foundation under the [23] Chang CC. Mass dampers and their optimal designs for building vibration control.
Eng Struct 1999;21(5):454–63.
AUFF Assistant Professor Starting Grant (AUFF-E-2017-7-20) and the [24] Krenk S, Høgsberg J. Tuned mass absorber on a flexible structure. J Sound Vib
Committee of Science and Technology of Shanghai China (Grant No. 2014;333(6):1577–95.
18160712800) are highly appreciated. [25] Hoang N, Fujino Y, Warnitchai P. Optimal tuned mass damper for seismic appli-
cations and practical design formulas. Eng Struct 2008;30(3):707–15.
[26] Leung AY, Zhang H. Particle swarm optimization of tuned mass dampers. Eng Struct
References 2009;31(3):715–28.
[27] Marano GC, Greco R, Chiaia B. A comparison between different optimization cri-
teria for tuned mass dampers design. J Sound Vib 2010;329(23):4880–90.
[1] Ahlström A. Influence of wind turbine flexibility on loads and power production.
[28] Pars LA. A Treatise on Analytical Dynamics. Ox Bow Press Woodbridge; 1979.
Wind Energy 2006;9(3):237–49.
[29] Tubino F, Piccardo G. Tuned mass damper optimization for the mitigation of
[2] Hansen MH. Aeroelastic instability problems for wind turbines. Wind Energy
human-induced vibrations of pedestrian bridges. Meccanica 2015;50(3):809–24.
2007;10(6):551–77.
[30] Krenk S, Høgsberg J. Equal modal damping design for a family of resonant vi-
[3] Thomsen K, Petersen JT, Nim E, øye S, Petersen B. A method for determination of
bration control formats. J Vib Control 2013;19(9):1294–315.
damping for edgewise blade vibrations. Wind Energy 2000;3(4):233–46.
[31] Jonkman J, Butterfield S, Musial W, Scott G. Definition of 5-MW reference wind
[4] Bir G, Jonkman J. Aeroelastic instabilities of large offshore and onshore wind tur-
turbine for offshore system development. National Renewable Energy Laboratory,
bines. EAWE 2007 Torque from Wind Conference. 2007, August 28-31, Lyngby,
Technical Report; 2009. NREL/TP-500-38060, Golden, Colorado.
Denmark.

12

You might also like