Time domain flutter analysis
Time domain flutter analysis
An International Journal
To cite this article: Praveen Shakya , Mohammed Rabius Sunny & Dipak Kumar Maiti
(2020): Time domain flutter analysis of bend-twist coupled large composite wind turbine
blades: a parametric study, Mechanics Based Design of Structures and Machines, DOI:
10.1080/15397734.2020.1824796
Article views: 17
1. Introduction
The long slender structure of modern composite wind turbine blades adds to the vulnerability to
aeroelastic instabilities. Therefore aeroelastic based criteria are finding growing importance in the
design of these blades. Lobitz and Veers (1998) showed that the flutter speed of a 20 kW horizon-
tal axis wind turbine (HAWT) rotor stiff blade of 5 m radius is several times the operating speed.
For 1.5 MW HAWT rotor blade of 35 m radius, the flutter speed is about two times of its operat-
ing speed (Lobitz 2004). With increase in size of blade, the difference between flutter and operat-
ing speeds reduces. This motivated several studies on the flutter behavior of MW size wind
turbine blades (Hansen 2004; Hafeez and El-Badawy 2018; Howison, Thomas, and Ekici 2018).
Stiffness coupling has been observed to have significant effect on aeroelastic stability (Hong
and Chopra 1985; Natori and Nemat-Nasser 1986; Stablein, Hansen, and Verelst 2017). Hence,
use of bend-twist stiffness coupling for increasing the flutter speed has been proposed by several
researchers. Nowadays, modern wind turbine blades are made up of composite laminates. In such
blades stiffness coupling can be achieved by avoiding anti-symmetric layup of composite lami-
nates. Ha, Hayat, and Xu (2014) studied the effect of change in ply angle of NREL 5 MW wind
turbine blade. They reported the reduction in the blade failure index and mass by changing the
skin angles. Hayat and Ha (2015) predicted the flutter of composite wind turbine blade having
unbalanced laminates. Quasi-steady aerodynamic model was used in their analysis. Flutter ana-
lysis of a single blade of 5 MW wind turbine blade was done using the eigenvalue-based approach.
Hayat et al. (2016) did a transient analysis to predict the flutter instability. Bend-twist stiffness
coupling was achieved by considering the unbalance layup (i.e. change in the ply angle, material
properties and thickness of plies) in composite laminates. They showed that the introduction of
CONTACT Mohammed Rabius Sunny sunny@aero.iitkgp.ac.in Department of Aerospace Engineering, IIT Kharagpur,
Kharagpur, West Bengal, 721302, India.
Communicated by Francesco Tornabene.
ß 2020 Taylor & Francis Group, LLC
2 P. SHAKYA ET AL.
carbon fiber in composite blade increases the flapwise and torsion stiffness and reduces the flutter
instability. Zhou, Huang, and Li (2018) analyzed flutter performance of a bend-twist coupled
wind turbine blade. BEM method and unsteady dynamic stall model were combined with nonlin-
ear beam model. Unbalance in the skin laminates showed no significant difference in flutter per-
formance for nonlinear beam model. However, flutter limit obtained from nonlinear model was
observed to be less than to that obtained from linear model.
Effect of the aerodynamic model on the predicted aeroelastic behavior can be observed
through several studies reported in open literature. Steady aerodynamic models have been
observed to show less flutter speed as compared to unsteady aerodynamic models (Lobitz 2004;
Hafeez and El-Badawy 2018). High fidelity computational fluid dynamic (CFD) methods have
been used to model the unsteady aerodynamics by few researchers (Carri on et al. 2014; Leble and
Barakos 2016). However, the CFD models suffer from limitations due to significant computa-
tional complexity. Generally, three-dimensional structural geometry is needed for such aeroelastic
analysis. It is observed from open literature that for high rotor speed aeroelasticity results
obtained using Theodorsen’s theory is significantly different from the results obtained using time
domain models like BEM model, vortex model, CFD models (Hauptmann et al. 2014; Hayat and
Ha 2015; Farsadi and Kayran 2016). Non-inclusion of wake vorticity and return shed wake are
some of limitation of Theodorsen’s theory based aerodynamic model (Silva and Donadon 2013).
However, most of the studies on the effect of blade characteristics such as bend twist coupling on
flutter speed have been performed through eigenvalue analysis based approach using
Theodorsen’s theory based aerodynamic model due to its computational simplicity. Lobitz (2004)
used the Theodorsen’s theory based unsteady aerodynamic model to predict the flutter of MW-
size wind turbine blade. Pourazarm, Sadeghi, and Lackner (2016) used Theodorsen’s theory based
unsteady aerodynamic model to calculate the flutter speed of different MW-size wind turbine
blades. In their study, it was observed that flutter frequency can be reduced by increasing the
flapwise natural frequency. Hansen (2004) showed that flutter speed is decreased by reducing the
torsional stiffness.
BEM method yields relatively more accurate results than Theodorsen’s model with much lower
computational requirement than CFD based models and it can be used for 1D beam model.
Unsteady BEM method which has the provision for including dynamic wake and stall model is a
further more efficient alternative. In this article, the authors study flutter behavior of bend-twist
coupled composite wind turbine blade in time-domain through transient analysis using unsteady
BEM based aerodynamic model. The authors in their previous article (Shakya, Sunny, and Maiti
2019) showed a study of flutter behavior of bend-twist coupled composite wind turbine blade in
frequency domain. Eigenvalue-based approach using Theodorsen’s theory based aerodynamic
model was used to predict the flutter of unbalanced laminates. Effects of parameters like unbal-
ance in the spar cap, asymmetry in the skin etc. on flutter limit was showed. This article shows a
similar study conducted in time domain. A detailed comparison of the results obtained through
this time domain approach using unsteady BEM based aerodynamic model and the frequency
domain approach using Theodorsen’s theory based aerodynamic model mentioned in Shakya,
Sunny, and Maiti (2019) for different parameters has been presented.
2. Mathematical model
2.1. Aerodynamic model
Glauert (1930) presented BEM method to calculate thrust, power and steady loads on rotating
rotor blades. Wind turbine blades are affected by unsteadiness in wind due to wind shear, wind
turbulence and presence of tower. Unsteady BEM method which is an improved version of the
BEM method takes account of all these parameters. Dynamic inflow model and yaw model are
MECHANICS BASED DESIGN OF STRUCTURES AND MACHINES 3
Figure 1. A typical cross-sectional view of a wind turbine blade with velocity and force components.
such correction models used to incorporate unsteadiness in BEM model. Dynamic wake/inflow
model accounts for change in time delay between two rotor loading configuration in which wake
goes from one equilibrium to another equilibrium state, thus changing the induced velocities
(Øye 1991; Snel and Schepers 1995). Yaw model is used to correctly predict the restoring yaw
moment. Blade in upstream experiences higher wind speed producing higher load than down-
stream blade. Yaw model enhances the yaw stability by turning the rotor more into the wind
(Tangler 1982; Snel and Schepers 1995).
Wind velocity is calculated along the blade span by transforming the wind velocity from iner-
tial coordinates to the body coordinates using a transformation matrix (Hansen 2013). Relative
wind velocity is calculated as the vector sum of wind velocity V 0 , induced velocity W, rotational
speed V rot and negative of structural velocity V s : It can be written as follows:
V rel ¼ V 0 þ W þ V rot V s
! ! ! ! !
Vrel, y Vy Wy Xrcos hcone Vs, y (1)
¼ þ þ
Vrel, z Vz Wz 0 Vs, z
where Vrel, y , Vrel, z are relative wind velocities in edgewise and flapwise directions, respectively;
Vy , Vz are wind velocities in edgewise and flapwise directions, respectively; Wy , Wz are induced
velocities in edgewise and flapwise directions, respectively; r is local radius of the blade; X is con-
stant rotor speed of the blade; hcone is cone angle; Vs, z , Vs, y are structural velocities in flapwise
and edgewise directions, respectively.
Figure 1 shows the cross-sectional view of the wind turbine blade. The flow angle ðwÞ and
angle of attack ðaÞ are determined as follows:
tan w ¼ Vrel, z =Vrel, y (2)
a ¼ w ð/ þ hÞ (3)
where h, / are the pre-twist angle and twist angle of the wind turbine blade.
The aerodynamic load acting on the blade are computed as follows:
pz ¼ Lcos w þ Dsin w (4)
py ¼ Lsin w Dcos w (5)
The expressions for lift L, drag D and moment M acting on wind turbine blade are given
below.
4 P. SHAKYA ET AL.
1 1 1
L ¼ q1 Vrel
2
cCl ; D ¼ q1 Vrel
2
cCd and M ¼ q1 Vrel
2 2
c Cm (6)
2 2 2
Here pz , py are the normal and tangential loads; c is the chord length of the airfoil; q1 is air
density; Cl , Cd , Cm are the coefficient of lift, the coefficient of drag, and the coefficient of
moment, respectively.
The normal and tangential component of induced velocities for each blade are as follows
(Hansen 2013):
BLcos w
Wn ¼ Wz ¼ (7)
4pq1 rFtiploss V 0 þ fg nðn :W Þ
BLsin w
Wt ¼ Wy ¼ (8)
4pq1 rFtiploss V 0 þ fg nðn :WÞ
where B is the number of blades; n is the normal vector to the rotor plane; Ftiploss is Prandtl’s
tip loss correction factor and fg is Glauert correction.
Equations (7) and (8) have been solved by assuming the old value of induced velocities (initially
induced velocities are assumed to be zero). These equations are not solved iteratively because the
value of induced velocities changes slowly in time due to the phenomena of dynamic wake as
mentioned in Hansen (2013). The induced velocities calculated in Eqs. (7) and (8) are quasi-
steady, i.e. they give correct value if the wake is in equilibrium with aerodynamic loads.
A dynamic wake model proposed by Øye (1991) is shown below.
dW int dW qs
W int þ s1 ¼ W qs þ 0:6s1 (9)
dt dt
dW f
W f þ s2 ¼ W int (10)
dt
W qs , W int , W f are the quasi-steady value, intermediate value and the final filtered induced value,
respectively. Time constants s1 , s2 and axial induction factor a can be calculated as shown
below.
1:1 R
s1 ¼ (11)
ð1 1:3aÞ V 0
2 !
r
s2 ¼ 0:39 0:26 s1 (12)
R
jV 0 j jV 0 þ nðn :W Þj
a¼ (13)
jV 0 j
Thrust is changed due to change in pitch angle. Therefore, a dynamic wake model for induced
velocity is incorporated to capture the time behavior of the loads. This model is very useful for
pitch regulated wind turbine blades.
Yaw/Tilt model is also incorporated in unsteady BEM model for azimuthal variation of the
induced velocity. Yaw model proposed by Glauert (1930) is presented as follows:
r v
W ¼ W 0 1 þ tan cos hwing h0 (14)
R 2
where v is the wake skew angle; W 0 is the mean induced velocity found by Eqs. (9) and (10); h0
is the angle in which the blade is deepest into the wake. The skew angle is calculated as shown
below.
MECHANICS BASED DESIGN OF STRUCTURES AND MACHINES 5
Figure 2. Coordinates of the blade at undeformed (dashed black) and deformed states (black).
n : ðV 0 þ nðn :W ÞÞ
cos v ¼ (15)
jnjjV 0 þ nðn :W Þj
The skew angle is calculated at a radial position r/R ¼ 0.7, and W is taken as old value of
induced velocities.
Introducing the yaw model, the induced velocity can be known for new azimuthal position.
Angle of attack is calculated from Eq. (3). Aerodynamic loads are determined using the
Eqs. (4)–(6).
where dU, dT, dW are variation in strain energy, variation in kinetic energy and virtual work
done respectively between any two time points t1 , t2 :
6 P. SHAKYA ET AL.
Figure 3. Finite element discretization of the blade: (a) having N number of elements with the (b) eth element of beam having
elemental length le :
We have considered flapwise, edgewise bending and twist of the blade in this study.
Considering linear strain–displacement relationship and neglecting the nonlinear terms in the
derived equations (Hodges and Dowell 1974; Hong and Chopra 1985), the linear differential
equations of motion for rotating composite wind turbine blade are given below.
" ðL ! #00 ðL !0
0
EIy w00 em / X2 qAxdx þ B1 /0 w0 X2 qAxdx X2 mem / þ m w € ¼ pz
€ þ em /
x x
(17)
ðL !0
00
0 00 0
EIz v B2 / v X qAxdx
2
mX2 v þ m€v ¼ py (18)
x
" ðL ! #0
2
GJ þ Km2 2 0
X qAxdx / þ B1 w00 B2 v00 þ X2 qA Km2 Km1
2 €
/ þ qAKm2 /
x
ðL ! (19)
X2 qAxdx em w00 þ X2 mxem w0 þ mem w
€ ¼M
x
at x ¼ 0 w ¼ 0, w0 ¼ 0, v ¼ 0, v0 ¼ 0 and / ¼ 0
000 000 (20)
x ¼ L w00 ¼ 0, w ¼ 0, v00 ¼ 0, v ¼ 0 and /0 ¼ 0
where w, v and / are flapwise, edgewise and torsional displacements, respectively; GJ, EIy , EIz
are torsional, flapwise and edgewise stiffness, respectively; B1 and B2 are the flap-torsional, edge-
torsional stiffness, respectively; Km , Km1 and Km2 are polar radius of gyration, mass radius of
gyration and axis perpendicular to the chord through the elastic axis, respectively; q and m are
the blade density and mass per unit length, respectively; em , A are mass offset and cross-sectional
area of blade, respectively.
MECHANICS BASED DESIGN OF STRUCTURES AND MACHINES 7
Within element e, the flapwise, edgewise and torsional displacements (w, v and /) are
approximated as follows:
X
4 X
4 X
2
wðxÞ ¼ Nie die , vðxÞ ¼ Nie gie and /ðxÞ ¼ Pie /ei (21)
i¼1 i¼1 i¼1
where Nie and Pie are Hermite cubic and Lagrangian interpolation functions (Appendix).
Structural properties are calculated at each node of the wind turbine blade. For an element e,
the structural properties are obtained through averaging of nodal structural properties. The elem-
ental equations are obtained as follows:
€ g þ ½K e fX g ¼ fF e g
½Me fX (22)
where ½Me , ½K e are the elemental mass and stiffness matrices; fF e g is the elemental load vector
which is computed using the unsteady BEM aerodynamic model; fX g represents the displacement
vector as fX g ¼ de g e /e ; fX
T
€ g and fX_ g represent the acceleration and velocity vector.
After assembling the elemental matrices, applying the essential boundary conditions defined in
Eq. (20) and adding a Rayleigh damping matrix ½C defined in Eq. (24), a set of coupled ordinary
differential equations shown in Eq. (23) are obtained.
€ g þ ½CfX_ g þ ½K fX g ¼ fF g
½MfX (23)
Rayleigh damping matrix ½C is calculated as follows:
½C ¼ a1 ½M þ b1 ½K
a1 ¼ 2f2 x2 b1 x22
(24)
2f x2 2f1 x1
b1 ¼ 2 2
x2 x21
where a1 is mass proportional Rayleigh damping coefficient; b1 is stiffness proportional Rayleigh
damping coefficient; f1 and f2 are damping ratio based on the structural damping which are con-
sidered as 2–5%, x1 and x2 are the natural frequencies.
Free vibration analysis has been performed by neglecting the Rayleigh damping matrix and n
number of natural frequencies and mode shapes are calculated. The displacement fXg is repre-
sented as a linear combination of n modes as follows:
fX g ¼ u1 Y1 þ u2 Y2 þ ::: þ uN YN ¼ ½un fY gn (25)
where u1 , u2 , uN are the mode shape vectors; Y1 , Y2 , YN are the generalize modal amplitudes; ½u
represents the modal matrix of first n coupled modes and fYg is the vector of n generalized
modal amplitudes.
8 P. SHAKYA ET AL.
Substituting Eq. (25) into Eq. (23) and pre-multiplying by the transpose of the modal matrix
½uT , the rearranged equation is expressed as:
T € g þ ½uT ½C½ufY_ g þ ½uT ½K ½ufY g ¼ ½uT fF g
½u ½M½ufY (26)
Total twenty mode shapes have been considered in ½u: Beta-Newmark’s method with step size
of 0.0001 sec has been implemented to obtain the modal amplitude. Displacement, velocity in
flapwise and edgewise directions, and twist are calculated from the modal amplitude obtained
from Eq. (26). Angle of attack depends upon the twist angle as shown in Eq. (3). At every inter-
val of time, structural velocities and twist from structural model change the aerodynamic loads.
From new aerodynamic loads, we get new twist and structural velocities. Thus, aerodynamic and
structural models involve two way coupling through the twist and structural velocities. A flow
chart of this iterative solution process is shown in Figure 4.
cross-sectional view and materials of SNL 61.5 blade. SNL 61.5 blade has been used for paramet-
ric study in the present work. SNL 61.5 blade has laminates with no flap-torsional stiffness (B1 )
and edge-torsional stiffness (B2 ). These couplings can be achieved by making the laminates unbal-
anced by changing the lamination sequence. To study the effect of these couplings due to unbal-
ance in the lamination sequence on the flutter speed different cases involving various lamination
sequence at different parts of the blade has been considered. Table 1 presents these cases. SNL
Triax properties for different values of h having laminate layup ½6h2 =½02 s are obtained from
Shakya, Sunny, and Maiti (2019).
5. Validation
Wind turbine blade models namely, WindPACT 1.5 MW, SNL 61.5 and NREL 5 MW, are used to
validate the natural frequencies corresponding to flapwise, edgewise bending and twist of the
blade. Table 2 shows a comparison of the natural frequency of WindPACT 1.5 MW wind turbine
blade obtained in our study and previous studies. Natural frequencies are in close agreement with
previous studies. Table 3 presents the comparison of the natural frequencies of SNL 61.5,
obtained by us and Resor (2013). They are in close agreement.
Natural frequencies of NREL 5 MW baseline blade are observed to be in close agreement with
other studies as shown in Table 4. Figure 6 presents the static aeroelastic response of wind tur-
bine blade which is compared with other studies. Steady-state deflection is calculated at a rated
wind speed with a constant inflow and no gravity. It can be noticed that the flapwise deflection is
in close agreement with Xudong et al. (2009) and edgewise deflection is in close agreement with
Jeong et al. (2013).
Figure 6. Comparison of steady-state deflections of NREL 5 MW wind turbine blade in flapwise and edgewise direction with
other studies.
Table 5. General characteristic of WindPACT 1.5 MW, SNL 61.5 and NREL 5 MW wind turbine blades.
WindPACT 1.5 MW NREL 5 MW/SNL 61.5 SNL 100-00
Wind speed (m/s) 11.4 11.2 11.3
Rotor diameter (m) 70 123 205
Tower height (m) 84 90 146.4
Rotational speed (rpm) 20.5 12.1 7.44
Tilt angle ( ) 5 5 5
Cone angle ( ) 0 2.5 2.5
Cut-in and cutout wind speed (m/s) 3, 25 3, 25 3, 25
X ¼15.8 rpm, the flapwise deflection, edgewise deflection and twist show higher amplitude which indicates
the onset of flutter. Through eigenvalue-based analysis using Theodorsen’s theory as aerodynamic model,
the critical flutter speed of this blade has been found to be 21.8 rpm as reported in previous study (Shakya,
Sunny, and Maiti 2019). In Theodorsen’s theory, it assumes that blade is rotating in still air, i.e. inflow com-
ponent of velocity is neglected and only rotor plane velocity is included, and wake vorticity due to other
rotor blades is neglected. Unsteady BEM method includes a dynamic wake model which includes the wake
vorticity due to other rotor blades. This difference in modeling may be considered as the reason behind the
difference between the flutter speeds obtained through these methods.
Figure 9 shows the phase diagram of SNL 61.5 blade at different rotational speeds. Point ‘o’
represents initial point of the phase portrait which begins at origin. It is observed that at rated
rotor speed X ¼ 12:1 rpm phase diagram shows the stability. Flutter instability occurs at rotor
speed of 15.8 rpm when the third flapwise and first torsion mode are coupled.
Figure 7. Modal amplitude of flapwise and torsion mode of SNL 61.5 wind turbine blade at different rotational speeds (a)
X ¼12.1 rpm (b) X ¼15.2 rpm (c) X ¼15.8 rpm.
MECHANICS BASED DESIGN OF STRUCTURES AND MACHINES 13
Figure 7. Continued.
phase diagram. At rated rotor speed X ¼20.5 rpm, the transient response reaches steady-state
after some time, and similar observation can be made from the phase diagram. When the rotor
speed is increased to 27.7 rpm the phase plot shows stability in flapwise bending response, how-
ever in edgewise bending and twist response shows instability indicating flutter. At rotor speed of
28 rpm, instability occurs in all three responses. Critical flutter speed of this blade reported in
other studies through eigenvalue analysis using Theodorsen’s model (Lobitz 2004; Owens et al.
2013; Pourazarm, Sadeghi, and Lackner 2016) is around 42–46 rpm. Same as previous cases, the
flutter speed predicted by the time domain approach using unsteady BEM model is less as com-
pared to the Eigen solution based approach using Theodorsen’s model. Reason for this low speed
is stated earlier.
Figure 8. Transient response of SNL 61.5 blade for increased rotor speeds in (a) flapwise (b) edgewise (c) twist directions.
Figure 9. Phase diagram of transient response of SNL 61.5 blade in flapwise, edgewise and twist directions for increased
rotor speeds.
Figure 10. Transient response of WindPACT 1.5 MW blade for increased rotor speeds in (a) flapwise (b) edgewise (c)
twist directions.
Figure 11. Phase diagram of transient response of WindPACT 1.5 MW blade in flapwise, edgewise and twist directions for
increased rotor speeds.
16 P. SHAKYA ET AL.
Figure 12. Transient response of SNL 100-00 wind turbine blade for increased rotor speeds in (a) flapwise (b) edgewise (c)
twist directions.
phase diagram. Bending and twist response of the blade are observed to achieve the steady-state
after some time for the rated rotational speed, i.e. 12.1 rpm. When the rotor speed is increased to
17.5 rpm phase diagram shows the stability in the edgewise bending and twist response and fur-
ther increase in the rotor speed to 17.6 rpm shows instability in all three responses. Meng (2011)
has presented the flutter speed in time-domain of NREL 5 MW wind turbine blade which is in
close agreement with present study. Flutter speed of this blade obtained through Eigen solution
using Theodorsen’s theory based aerodynamic model reported in the previous studies (Hansen
2007; Hayat and Ha 2015; Shakya, Sunny, and Maiti 2019) is in the range of 22–27 rpm.
6.2. Comparative parametric study of SNL 61.5 wind turbine blade with various composite
structural properties
A comparative parametric study is conducted by changing fiber angle orientation in skin of SNL
61.5. Critical flutter speeds for these subcases obtained through time-domain analysis using
unsteady BEM model have been compared with those obtained through eigenvalue analysis using
Theodorsen’s model reported in Shakya, Sunny, and Maiti (2019). Table 1 presents these cases as
mentioned in wind turbine blade models section.
Figure 13. Phase diagram of transient response of SNL 100-00 blade in flapwise, edgewise and twist directions for increased
rotor speeds.
Figure 14. Transient response of NREL 5 MW blade for increased rotor speeds in (a) flapwise (b) edgewise (c) twist directions.
18 P. SHAKYA ET AL.
Figure 15. Phase diagram of transient response of NREL 5 MW blade in flapwise, edgewise and twist directions for increased
rotor speeds.
Figure 16. Comparison between the flutter speeds obtained through eigenvalue (Shakya et al. 2019) and time-domain
approaches considering symmetric skin with different ply angles of SNL 61.5 blade.
25 , h2 ¼ 45 and h1 ¼ 35 , h2 ¼ 45 is less as compared to the difference obtained through
eigenvalue analysis. Both the approaches show that layup with off-axis fiber angles h1 ¼
60 , h2 ¼ 45 leads to less flutter speed as compared to that of the baseline model (i.e.
h1 ¼ 45 , h2 ¼ 45 ). It can be observed that subcase I and II show higher critical flutter speed.
MECHANICS BASED DESIGN OF STRUCTURES AND MACHINES 19
Figure 17. Comparison between the flutter speeds obtained through eigenvalue (Shakya et al. 2019) and time-domain
approaches considering symmetric skin with different ply angles in spar cap of SNL 61.5 blade.
Figure 18. Comparison between the flutter speeds obtained through eigenvalue (Shakya et al. 2019) and time-domain
approaches considering asymmetric skin with different ply angles in spar cap of SNL 61.5 blade.
Therefore, the off-axis fiber angles h1 ¼ 25 , h2 ¼ 45 and h1 ¼ 35 , h2 ¼ 45 are used in the
next case set of analyses.
In case 2, fiber angle orientation of SNL triax has been varied in spar cap only considering
symmetric skin. Figure 17 presents a comparison of critical flutter speed of the blade using both
approaches. Flutter speed obtained through time-domain based approach is lower than eigenvalue
based approach. However, it is observed that the difference between the critical flutter speeds cor-
responding to both off-axis fiber angles (i.e. h1 ¼ 25 , h2 ¼ 45 and h1 ¼ 35 , h2 ¼ 45 ) is found
to be almost negligible using both approaches. However, subcase II shows slightly higher critical
flutter speed as compared to subcase I of case 2.
20 P. SHAKYA ET AL.
Figure 19. Comparison between the flutter speeds obtained through eigenvalue (Shakya et al. 2019) and time-domain
approaches considering asymmetric skin with different ply angles of SNL 61.5 blade.
6.2.2. Asymmetric laminate with varying fiber angle sequence in suction side
Case 3 involves studying the effect of changing the off-axis fiber angles only in the suction side
of spar cap of the blade keeping the skin asymmetric. Figure 18 shows the comparison of the crit-
ical flutter speed of the blade at different off-axis fiber angle using both approaches. Both the
approaches show that critical flutter speed of both off-axis fiber angles (i.e. h1 ¼ 25 , h2 ¼ 45
and h1 ¼ 35 , h2 ¼ 45 ) are almost negligible. It is observed that critical flutter speed of a blade
with asymmetric skin in spar cap (case 3) is higher than that of a blade with symmetric skin
(case 2) in spar cap of the blade.
Case 4 involves studying the effect of changing the off-axis fiber angles in the suction side of
entire blade keeping the skin asymmetric. Figure 19 shows the comparison of the critical flutter
speed of the blade at different off-axis fiber angle using both approaches. It can be noticed that
for asymmetric skin layup the off-axis fiber angle h1 ¼ 25 , h2 ¼ 45 shows higher critical flutter
speed as compared to off-axis fiber angle h1 ¼ 35 , h2 ¼ 45 which gives the highest flutter speed
for symmetric skin layup. Through eigenvalue analysis, it shows 100% increment from the base-
line model while through time domain analysis, the observed increment is 81%. In case 2, it can
be noticed that time domain analysis shows lower flutter speed of the blade as compared to
eigenvalue analysis. However, the maximum flutter speed is achieved at the same layup sequence.
7. Conclusion
In this article, aeroelastic stability analysis conducted through time-domain analysis using
unsteady BEM model is presented. Four MW-sized wind turbine blades namely, WindPACT
1.5 MW, SNL 61.5, NREL 5 MW, and SNL 100-00 are considered for modal and flutter instability
analyses. Time-domain analysis has been conducted using unsteady BEM theory for the aero-
dynamic load calculation and a structural beam model accounting for flapwise bending, edgewise
bending and twisting with linear strain–displacement relationship. The discretization of the gov-
erning partial differential equation of the blade structure has been done using FEM. Interaction
between aerodynamic and structural variables has been taken into consideration at every interval
of time. Effect of various parameters like composite lamination sequence, symmetry/asymmetry
in only the spar cap and full blade etc. on the flutter speeds has been studied through a paramet-
ric study. Important observations from this study are summarized below.
MECHANICS BASED DESIGN OF STRUCTURES AND MACHINES 21
Flutter speed of NREL 5 MW blades obtained using our time domain analysis model shows
close agreement with time domain analysis based estimation of Meng (2011).
Asymmetric skin in the blade shows higher critical flutter speed as compared to symmetric
skin in the blade. Through time-domain analysis, it has been observed that up to 81% incre-
ment of flutter speed from the baseline model can be achieved by choosing proper set of off-
axis fiber angle orientation in the suction side of the entire blade.
Both time and frequency domain approaches show the maximum flutter speed for the same
set of off-axis h1 ¼ 25 , h2 ¼ 45 fiber angle orientation.
Time domain analysis using unsteady BEM model shows lower flutter speed as compared to
Eigen value analysis using Theodorsen’s theory possibly due to the inclusion of return wake
velocity in unsteady BEM method.
Hence, because of its lower computational requirement the eigenvalue-based approach using
the Theodorsen’s theory can be used to find the optimum fiber angle orientation of constituent
plies and time-domain analysis using unsteady BEM method can be used to get a better estimate
of flutter speed. Use of high fidelity CFD methods and consideration of geometric nonlinearity
for even more accurate estimate of flutter speed can be a future extension of this work.
Conflict of interest
The authors declared no potential conflicts of interest with respect to the research, authorship
and publication of this article.
References
Bouzidi, I., A. Hadjoui, and A. Fellah. 2020. Dynamic analysis of functionally graded rotor-blade system using clas-
sical version of the finite element method. Mechanics Based Design of Structures and Machines. doi:10.1080/
15397734.2019.1706558.
Carrion, M., R. Steijl, M. Woodgate, G. N. Barakos, X. Munduate, and S. Gomez-Iradi. 2014. Aeroelastic analysis
of wind turbines using a tightly coupled CFD-CSD method. Journal of Fluids and Structures 50:392–415. doi:10.
1016/j.jfluidstructs.2014.06.029.
Cheng, J., H. Xu, and A. Yan. 2006. Frequency analysis of a rotating cantilever beam using assumed mode method with
coupling effect. Mechanics Based Design of Structures and Machines 34 (1):25–47. doi:10.1080/15367730500501587.
Farsadi, T., and A. Kayran. 2016. Aeroelastic instability analysis of composite rotating blades based on Loewy’s and
Theodorsen’s unsteady aerodynamics. The 2016 World Congress on Advances in Civil Environmental, and
Materials Research.
Glauert, H. 1930. The force and moment on an oscillating aerofoil. Berlin: Springer.
Griffin, D. A. 2001. WindPACT turbine design scaling studies technical area 1—Composite blades for 80-to 120-
meter rotor. National Renewable Energy Laboratory NREL/SR-500.
Griffith, D. T., and T. D. Ashwill. 2011. The Sandia 100-meter all-glass baseline wind turbine blade: SNL 100-00.
Sandia National Laboratories Technical Report SAND2011-3779.
Ha, S. K., K. Hayat, and L. Xu. 2014. Effect of shallow-angled skins on the structural performance of the large-scale
wind turbine blade. Renewable Energy. 71:100–12. doi:10.1016/j.renene.2014.05.023.
Hafeez, M. M. A., and A. A. El-Badawy. 2018. Flutter limit investigation for a horizontal axis wind turbine blade.
Journal of Vibration and Acoustics 140 (4):041014. doi:10.1115/1.4039402.
Hansen, M. H. 2004. Stability analysis of three-bladed turbines using an eigenvalue approach. In: 42nd AIAA
Aerospace sciences meeting and exhibit, Proceedings of the ASME Wind Energy Symposium, Reno, Nevada.
doi:10.2514/6.2004-505.
Hansen, M. H. 2007. Aeroelastic instability problems for wind turbine blade. Wind Energy 10 (6):551–77. doi:10.
1002/we.242.
Hansen, M. O. L. 2013. Aerodynamics of wind turbines, 2nd ed. Abingdon, Oxon: Routledge. doi:10.4324/
9781849770408.
Hauptmann, S., M. B€ ulk, L. Sch€on, S. Erbsl€oh, K. Boorsma, F. Grasso, M. K€ uhn, and P. W. Cheng. 2014.
Comparison of the lifting-line free vortex wake method and the blade-element-momentum theory regarding the
simulated loads of multi-MW wind turbines. Journal of Physics: Conference Series 555:012050. doi:10.1088/1742-
6596/555/1/012050.
22 P. SHAKYA ET AL.
Hayat, K., and S. K. Ha. 2015. Flutter performance of large-scale wind turbine blade with shallow-angled skin.
Composite Structures 132:575–83. doi:10.1016/j.compstruct.2015.05.073.
Hayat, K., A. G. M. de Lecea, C. D. Moriones, and S. K. Ha. 2016. Flutter performance of bend-twist coupled
large-scale wind turbine blades. Journal of Sound and Vibration 370:149–62. doi:10.1016/j.jsv.2016.01.032.
Hodges, E. H., and E. H. Dowell. 1974. Nonlinear equation of motion for elastic bending and torsion of twisted
non-uniform blades. NASA TND 7818.
Hong, C. H., and I. Chopra. 1985. Aeroelastic stability analysis of a composite rotor blade. Journal of the American
Helicopter Society 30 (2):57–67. doi:10.1109/ICMSAO.2013.6552604.
Howison, J., J. Thomas, and K. Ekici. 2018. Aeroelastic analysis of a wind turbine blade using the harmonic bal-
ance method. Wind Energy 21 (4):226–41. doi:10.1002/we.2157.
Jeong, M. S., I. Lee, S. J. Yoo, and K. C. Park. 2013. Torsional stiffness effects on the dynamics stability of a hori-
zontal axis wind turbine blade. Energies 6 (4):2242–61. doi:10.3390/en6042242.
Jonkman, J. M., S. Butterfield, W. Musial, and G. Scott. 2009. Definition of a 5-MW reference wind turbine for off-
shore system development. Technical Report NREL/TP.
Kar, R. C., and T. Sujata. 1991. Dynamic stability of a rotating beam with various boundary conditions. Computers
& Structures 40 (3):753–73. doi:10.1016/0045-7949(91)90243-F.
Leble, V., and G. Barakos. 2016. Demonstration of a coupled floating offshore wind turbine analysis with high-
fidelity methods. Journal of Fluids and Structures 62:272–93. doi:10.1016/j.jfluidstructs.2016.02.001.
Lobitz, D. W. 2004. Aeroelastic stability predictions for MW-sized wind turbine blades. Wind Energy 7 (3):211–24.
doi:10.1002/we.120.
Lobitz, D. W., and P. S. Veers. 1998. Aeroelastic behavior of twist-coupled HAWT blades. Proceedings of the
ASME/AIAA Wind Energy Symposium, Reno, Nevada. doi:10.2514/6.1998-29.
Meng, F. 2011. Aeroelastic stability analysis for large-scale wind turbines. PhD thesis dissertation, TU Delft.
Natori, M., and S. Nemat-Nasser. 1986. Application of a mixed variational approach to aeroelastic stability analysis
of a nonuniform blade. Journal of Structural Mechanics 14 (1):5–31. doi:10.1080/03601218608907508.
Owens, B. C., D. T. Griffith, B. R. Resor, and J. E. Hurtado. 2013. Impact of modeling approach on flutter predictions
for very large blade designs. Proceedings of the American Helicopter Society 69th Annual Forum, Phoenix, AZ.
Øye, S. 1991. Dynamic stall, simulated as a time lag of separation. In Proceedings of the 4th IEA Symposium on the
Aerodynamics of Wind Turbines, ed. K. F. McAnulty.
Pourazarm, P., Y. M. Sadeghi, and M. A. Lackner. 2016. A parametric study of coupled-mode flutter for MW-size
wind turbine blades. Wind Energy 19 (3):497–514. doi:10.1002/we.1847.
Resor, B. R. 2013. Definition of a 5 MW/61.5 m wind turbine blade reference model. Sandia National Laboratories
Technical Report SAND2013-2569.
Rinker, J., and K. Dykes. 2018. WindPACT Reference Wind Turbines. Nrel/Tp 5000–67667.
Shakya, P., M. R. Sunny, and D. K. Maiti. 2019. A parametric study of flutter behavior of a composite wind turbine
blade with bend-twist coupling. Composite Structures 207:764–75. doi:10.1016/j.compstruct.2018.09.064.
Silva, C. T., and M. V. Donadon. 2013. Unsteady blade element-momentum method including returning wake
effects. Journal of Aerospace Technology and Management 5 (1):27–42. doi:10.5028/jatm.v5i1.163.
Snel, H., and J. G. Schepers. 1995. Joint investigation of dynamic inflow effects and implementation of an engineer-
ing method. ECN-C. Netherlands Energy Research Foundation.
Stablein, A. R., M. H. Hansen, and D. R. Verelst. 2017. Modal properties and stability of bend-twist coupled wind
turbine blade. Wind Energy Science 2 (1):343–60. doi:10.5194/wes-2-343-2017.
Tangler, J. L. 1982. Comparison of wind turbine performance prediction and measurement. Journal of Solar Energy
Engineering 104 (2):84–8. doi:10.1115/1.3266290.
Xiao, S., B. Chen, and Q. Du. 2005. On dynamic behavior of a cantilever beam with tip mass in a centrifugal field.
Mechanics Based Design of Structures and Machines 33 (1):79–98. doi:10.1081/SME-200048325.
Xudong, W., W. Z. Shen, W. J. Zhu, J. N. Sorensen, and C. Jin. 2009. Shape optimization of wind turbine blades.
Wind Energy 12 (8):781–803. doi:10.1002/we.335.
Zhou, X., K. Huang, and Z. Li. 2018. Effects of bend-twist coupling on flutter limits of composite wind turbine
blades. Composite Structures 192:317–26. doi:10.1016/j.compstruct.2018.02.071.
Appendix
The elemental mass and stiffness matrix in time-domain are as follows
2 Ð Ð 3
m fN g41 fN gT14 dx 0 me fN g41 fPgT12 dx
6 Ð
m 7
6 7
½Me ¼ 6 0
T
mfN g41 fN g14 dx 0 7 (A.1)
4Ð Ð 5
T T
mem fPg21 fN g14 dx 0 qAkm fPg21 fPg12 dx
2
MECHANICS BASED DESIGN OF STRUCTURES AND MACHINES 23
2 3
K11 K12 K13
½K ¼ 4 K21
e
K22 K23 5 (A.2)
K31 K32 K33
where
ð le T ð le T
d2 N d2 N dN dN
e
K11 ¼ EIy dx þ A 1 dx
0 dx2 41 dx2 14 0 dx 41 dx 14
e
K12 ¼ ½0
ð le T ð le ðle T
dN dP dN d2 N dP
e
K13 ¼ A1 em dx þ mX2 xem fPgT12 dx þ B1 dx
0 dx 41 dx 12 0 dx 41 dx2 41 dx 12
0
e
K21 ¼ ½0
ð le T ð le T ðle
d2 N d2 N dN dN T
e
K22 ¼ EIy dx þ A1 dx mX2 fN g41 fN g14 dx
0 dx2 41 dx2 14 0 dx 41 dx 14
0
ðle T
d2 N dP
e
K23 ¼ B2 dx
dx2 41 dx 12
0
ð le T ð le T ðle T
dP dN dN dP d2 N
e
K31 ¼ A 1 em dx þ mX2 xem fPg21 dx þ B1 dx
0 dx 21 dx 14 0 dx 14 dx 21 dx2 14
0
ðle T
dP d2 N
e
K32 ¼ B2 dx
dx 21 dx2 14
0
ð le T ð le
dP dP 2
2 f g T
e
K33 ¼ GJ þ A1 Km
2
dx þ X2 qA Km2 Km1 P 21 fPg12 dx
0 dx 21 dx 12 0
ðL
A1 ¼ X2 qAxdx,
x
T T
fNg ¼ N1 ðxÞ N2 ðxÞ N3 ðxÞ N4 ðxÞ andfPg ¼ P1 ðxÞ P2 ðxÞ
2 3
x x x 2
N 1 ð xÞ ¼ 1 3 þ2 , N2 ðxÞ ¼ x 1
le le le
2 3
x x x2 x
N3 ðxÞ ¼ 3 2 , N 4 ð xÞ ¼ 1
le le le le
x x
P1 ðxÞ ¼ 1 , P2 ðxÞ ¼
le le
le is the elemental length.
The elemental force vector is as follows
8Ð 9
>
> le
f N g p dx >
>
> 0
< 41 z >
=
Ð
fF e g ¼ le
f N g p dx (A.3)
>
> Ðl
0 41 y
>
>
>
: e fPg M dx > ;
0 21