International Conference on Renewable Energies and Power

Quality (ICREPQ’09)
European Association for the
Development of Renewable Energies, Valencia (Spain), 15th to 17th April, 2009
Environment and Power Quality

Design Blades of a Wind Turbine Using Flexible Multibody Modelling

Maria Augusta Neto 1, Wenbin Yu 2, Jorge A.C. Ambrósio3, Rogério Pereira Leal1
1
Departamento de Engenharia Mecânica - Faculdade de Ciência e Tecnologia da Universidade de Coimbra (Polo II),
3020 Coimbra, Portugal.
e-mail: [email protected]; rogerio.leal @dem.uc.pt1
2
Department of Mechanical and Aerospace Engineering
Utah State University, Logan, UT 84322-4130, USA
e-mail: [email protected]
3
Instituto de Engenharia Mecânica – Instituto Superior Técnico, Lisboa, Portugal.
Instituto Superior Técnico
Av. Rovisco Pais, 1041-001 Lisboa, Portugal
e-mail: [email protected]

Composite material; Rotor blade, flexible multibody systems,

mode component synthesis, automatic differentiation, design
Abstract. A methodology for the application of structural sensitivity, variational asymptotic beam section.
optimization to find the optimal layouts of composite blades,
based on a multibody model of a wind turbine, is presented.
VABS (Variational Asymptotic Beam Section) is used to com-
1. Introduction
pute the sectional constants for a generalized Timoshenko beam
model. The gravity and inertia effects are taken into account as Composite blades are the type of structures encoun-
well as the aerodynamic force, which are calculated by the tered in helicopter rotor blades and wind turbines blades
aerodynamic coefficients and transferred to the corresponding [1]. The design of composite rotor blades in modern in-
finite element nodes. The minimum lateral moment component dustrial environments requires complex and accurate
at the blade root is found by using an optimization methodology computational models, which are difficult to build, thus
that finds the optimal values for the fiber orientation of the restricting the design exploration and making it hard for
composite spar of the blade. The computational efficiency and designers to try novel lay-ups and cross-sectional con-
robustness of the optimal process relies on the accurate evalua-
figurations. Therefore, the variations of the cross-
tion of the system sensitivities. In this work a general formula-
tion for the computation of the first order analytical sensitivities sectional layouts across the wide range of modern rotor
based on the direct method is used. The direct method for sensi- blades are remarkable small compared to the wealth of
tivity calculation is obtained by differentiating the equations de- possibilities provided by the use of composite materials
fining the response of the structure with respect to the design [2].
variables. The equations of motion and sensitivities of the flexi-
ble multibody system are solved simultaneously and therefore The wind turbine modeling complexity is related with
the accelerations and velocities of the system and the sensitivi- the tools used in the construction of the models of indi-
ties of the accelerations and of the velocities are integrated in
time using a multi-step multi-order integration algorithm.
vidual and coupled components of the system. Because of
the special cross-sectional geometric features of blades,
the use of computational tools that provide fast and accu-
Keywords rate mean of assessing the blade properties opens oppor-
tunities to improve the design cycle of composite blades.
The computer program VABS (Variational Asymptotic
Beam Section analysis) allows modeling heterogeneous achieve twist-flap coupling [10,11,8]. All of these studies
anisotropic beams with arbitrary cross section geometries are conducted by performing a set of static analysis in
through a finite element discretization of the cross section which the fiber angle of the spar cap is modified in a dis-
and, therefore, permits a high-fidelity representation of crete form changing from 5 to 20 or to 25 degrees.
the blade structure in terms of an equivalent beam model
[3]. Once the blade properties are obtained, various types In this work, the optimization of the composite blade
of blade analyses, including the multibody dynamic simu- is carried out using a flexible multibody dynamics model
lation of the entire wind turbine system, become possible. of a wind turbine. The ply orientations of the laminate
The outline of the blade used in this work is a SERI-8 used to make up the spar cap of the blade are taking as
aerofoil with the chord length of 0.594 m [4]. continuous design variables. The multibody dynamic and
sensitivities analysis code is linked with general optimiza-
The wind turbine is a typical mechatronic system tion algorithms included in the package DOT/DOC [12].
consisting of mechanical, electrical and control subsys-
tems interacting with each other [5]. Regardless the im- 2. Sensitivity Design in Flexible Multibody
portance of the electrical and control subsystems on the Dynamics
performance of the wind power generator, the mechanical
subsystem is the focus of this work, in which a three- The finite element method is used to represent the
bladed wind turbine is investigated. The use of the flexi- flexible body being the body reference frame coincident
ble multibody dynamics approach to describe the large with a coordinate system rigidly attached to a point on the
motion of complex systems that have flexible components flexible component. The kinetic energy of a flexible body
is advantageous compared to standard nonlinear finite is defined by assuming that the distributed mass of the
element methodologies because it deals very efficiently flexible body is replaced by particle masses and lumped
with the very large rotations, relative kinematics of the inertias in the nodal positions of the finite elements [13].
different system components and with the deformation of The local body reference frame is aligned with the body
the structural members including the inertia coupling be- principal inertia directions and it is fixed to its center of
tween large gross motion and the structural elastodynam- mass. The sensitivity design is necessary for deterministic
ics. The application of flexible multibody dynamics for- optimization algorithms because one needs to know how
mulation allows the use of reduction techniques to obtain a change in the design variables changes the system per-
simpler models while preserving the ability to represent formance. The calculation of these sensitivities can be
the large rotations of the components with respect to each carried out analytically or numerically. In this work only
other or with respect to an inertial reference frame [6,7]. the analytical sensitivities are obtained by using auto-
For efficient modeling and accurate dynamic analysis of a matic differentiation.
wind turbine, a rigid-flexible multibody system method-
ology is recommended since it combines the rigid body The sensitivities of the equations of motion are ob-
motion and elastodynamics allowing to capture all the tained by differentiating the equations of motion of a
rigid and elastic components, couplings and important rigid-flexible multibody system with respect to the design
motions that result from the nonlinear kinematics [5]. variables b,

With the objective of reducing the transient loads of

wind turbine blades, the aeroelastic tailoring concept has
 M rr M rf ΦTqr   q
ɺɺrb   Q*b 
   
been used since the late 1980s. Using this concept the
 M fr M ff ΦTw   w
ɺɺ b  =  R *b  (1)

blade geometry and its structure are designed in such a
 Φq Φw 0   λ b   γ b* 
way that when the blade undergoes flapwise deflection  r
during operation it inherently undergoes twist as well.
Two basic approaches used either (1) sweep the blade in where the notation used is the following
the plane of rotation, allowing that the aerodynamic load-
ing to act on the outboard portions of the blade producing M ff = X T M ff X = I , M rf = M rf X , M fr = X T M fr ,
a substantial torque about the blade elastic axis, or (2) ro- Φ w = Φq f X and ΦTw = X T ΦTq f . In equation (1)
tate some of the structural laminate fibers away from the (⋅)b denotes the sensitivity of quantity (⋅) with respect to
spanwise direction inducing, in this way, a twist-flap cou- b and the sensitivities Q*b , R *b and γ b* are defined ac-
pling in the structure [8]. For a comprehensive review of cording to Ambrósio et al.[14].
prior research on aeroelastic tailoring of blades the inter-
ested reader is referred to Lobitz et al. [9] Several works
sponsored by the Sandia National Laboratories have been For the laminate optimization problem the design
focusing the structural details of various blade designs to variables used are the lay-up orientations, denoted by
vector (b = θ) . Therefore, the derivative of the stiffness composite beam elements, are attached to the hub by
matrix of the composite flexible body with respect to the revolute joints, called “pitch hinges”, allowing for the
lay-up orientations has to be accounted for. The sensitiv- relative motion of the blade with respect to the hub,
ity of the stiffness matrix of the composite beam element which is normal to the plane of rotation of the rotor.
is such that the matrix D is the only matrix that depends Flexible root retention beams, each modeled as a single
on the design variables. Thus, the sensitivity of this equa- cubic beam element, are connected to the assemblies on
tion is given by the back of the hub [18].

∂K (e)ff 
1
∂D 
(e)
Pitch hinges
= ∫  BT B J d ξ (2) Blade 3 Blade 2
∂θ −1 
∂b  Blade

BedPlate Hub
Blade 1
Tower Shaft
The partial derivative of matrix D is obtained by using Root retention
the VABS code, which is extended to consider this com- Tower beams
putation. In the generalized Timoshenko theory. The par-
tial derivative of the D matrix with respect to the design
variables, is Fig. 1. Schematic configuration of the wind turbine.

The outline of the blade is a SERI-8 aerofoil with

 ∂X ∂F 
chord length of 0.594 m being the shear web located 30%
∂D  ∂θ ∂θ 
 (3)
= of the chord length from the leading edge, as sketched in
∂θ  ∂F  ∂G 
T
Figure 2 [4]. The rotor diameter is of 20.52 m. The layer

 ∂θ 
  ∂θ  orientations used, to make the different components of
this blade, are listed in Table I. The material properties,
More details on the evaluation of the derivatives of the including density and elastic constants, are given in Table
sub matrices X, F and G are found in Neto et al. [15´,16] II. The blade cross section finite element model is
meshed with 1030 nine-node quadratic elements in order
3. Optimization of Composite Rotor Blades of a to be used in the computer code VABS.
Wind Turbine
Shear web Skin
Filled core
The methodology for the optimal design of flexible
Up spar cap
multibody systems made with composite beam compo- Low spar cap

nents is applied to the design of a three-bladed wind tur-

bine. Several models of the wind turbine are given in ref-
erences by Stol et al. [17], Bauchau and Wang [18] and
Fig. 2. Blade cross-section of the S805A/S807 aerofoil.
Hodges and Yu [1].
TABLE I. - Laminated used in the blade of the wind turbine.
The solution of the optimization problem can be ob-
tained by either deterministic or stochastic optimization
Total thick- Laminate orienta-
methods. In this work a deterministic optimization Component Material
ness (mm) tions
method, the Modified Method of Feasible Directions
(MMFD), as implemented in DOT program [12], is used. Graph- 00
Skin 2.877
ite/epoxy
A. Description of the Wind Turbine Graph- 00/00
Fill core 5.754
ite/epoxy
This application deals with the modeling of the three- Graph- 150/150/150/150
Up spar cap 2.877
ite/epoxy
bladed wind turbine depicted in Figure 1. The physical
Low spar Graph- 150/150/150/150
properties of the system are presented in Stol et al. [17] 2.877
cap ite/epoxy
and are not be repeated here. The full multibody model Graph- 450/-450/450/-450
considers the tower that is connected to a flexible bed Shear web 1.869
ite/epoxy
plate, the shaft that is connected to the bed plate, by
means of a revolute joint, and the tip of the shaft is at- TABLE II. – Material properties of the composites used on the
tached to the hub, which is modeled as a rigid body. Fi- blade.
nally, the three flexible blades, each modeled by ten cubic
 Density −(Q3 = Drag cos φ − Lift sin φ )
Material Elastic properties
(Kg/m3) Drag

Graph- E1 = 184.556 GPa υ12 = υ13 = 0.28 φ =α +β Q2 = Lift cos φ + Drag sin φ
ite/epox 1600 E2 = E3 = 10.503 GPa υ 23 = 0.33
y G12 = G13 = 7.312 GPa G23 = 3.9 GPa Apparent Wind
Lift
φ Head Wind
The induced wind forces are computed according to
the combined blade element and momentum theories, be- Real Wind

ing defined as [19]

Fig. 3. Aerodynamic forces acting on the cross section of the
R blade.
1
Lift = ∫ ρ v 2 c Cl dr (4)
x1
2 B. Multibody simulation of the wind turbine
R
1
Drag = ∫ ρ v 2 c Cd dr (5) For the purpose of demonstrating the proposed meth-
x1
2 odology the wind turbine model depicted in Figure 1 is
used. For simplicity the tower and the bed plate are not
where v is the apparent wind speed, c is the chord considered in this simulation. Modeling of these compo-
length, ρ is air density, Cl is the lift coefficient and Cd nents would be easily done, but they do not affect signifi-
is the drag coefficient. Lift and Drag forces depend on the cantly the optimization goal related with the lateral mo-
coefficients Cl and Cd , which in turn depend on the ment component at the root blade.
cross section of the blade that is being used, and on the
angle with which the wind strikes the blade. The lift force The initial angular velocity of the shaft is 5 rad/s and
acts perpendicularly to the direction of wind flow and the a constant resistant moment is applied on the shaft during
drag force acts parallel to the direction of wind flow. Be- all period of simulation. Others parameters considered are
cause the chord of the blade cross-section is constant the air density 1.2 Kg/m [3], the wind speed 10 m/s and
throughout the blade span, the finite element model com- the span of the blades 10.26 m, which corresponds to a
putes the drag and lift forces as distributed throughout the wind turbine diameter of 2×10.26 m. Because the aero-
blade span. dynamic characteristics for the S805A/S807l airfoil pro-
file are not available, the lift and the drag coefficients are
The projection of the drag and lift forces over the selected from Wolfe and Ochs [20]. For a mixed lami-
cross-section coordinate system is depend on the angle nar/turbulent calculations and an angle of attack between
between the apparent wind direction and the chord line, 0 and 9.22 degrees the experimental drag coefficient is
α , and on the twist blade angle, β . Because the blade is about 0.007, thus in these simulations the drag coefficient
is assumed constant with a value of 0.007. Nevertheless,
not twisted initially, the multibody model of the wind tur-
the value of the lift coefficient experience significant
bine accounts for the flexible twist of the cross section
modifications as the angle of attack changes. Therefore,
only. Therefore, the thrust and the driving forces that acts
in this work the least square minimization method is used
on the blade cross section are evaluated, as presented in
for fit the experimental values [20] of the lift coefficient
Figure 3. Actually, the flexible twist allows to adjust the
to a linear equation, defined as
local angle of attack which in turn leads to different coef-
ficients Cl and Cd .
Cl = 0.177249 + 0.098172 φ (6)

The flexible twist is evaluated for all nodes of the 1-

D finite element model of the blade, being time depend- where φ is defined in Figure 3. It is assumed that the an-
ent. The aerodynamic forces are transferred to the corre- gle of attack is about 5.130.
sponding finite element nodes and are used at the next
time step, thus a direct aeroelastic coupling is taking into When using the modal superposition technique, four
account. vibration modes corresponding to the first four frequen-
cies of each flexible component are selected. These
modes of vibration are obtained by performing a modal
analysis on each component of the flexible multibody sys-
tem. The structural attachment conditions used in the ei-
genproblem are the same used to fix the body coordinate
system to the flexible body, i.e., the node in the centre of 15000

mass is fixed to the body fixed frame. The simulation of 10000

the wind turbine is carried out for 10 seconds being the 5000

Blade root moments (N.m)

0
time histories of displacement, force and moment ob- -5000
Axial

tained as the results of the analysis. In Figures 4, 5, 6 and -10000 Lateral

7 the wind turbine response is presented in terms of the -15000 Transverse

rotor angular velocity, the transverse elastic displacement -20000

-25000
component at the tip of the blade and the sectional force
-30000
and moment components at the blade root, respectively. 0 1 2 3 4 5 6 7 8 9 10
Time (s)

5.04

5.02 Fig. 7. Time history of blade root moments.

Rotor angular velocity (rad/s)

5.00

4.98
C. Wind turbine blade optimization
4.96

4.94
The use of composite materials in wind turbine
blades allows taking advantage of particular effects asso-
4.92
ciated with the design of the blade material, namely the
4.90
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 twist-flap coupling that makes the blades intrinsically
Time (s) smart. This coupling effect acts as a passive mechanism
within the blade structure to enhance their static and dy-
Fig. 4. Rotor angular velocity of the wind turbine. namic behavior [21]. Besides the twist-flap coupling,
general improvements are achieved when the moment
0.005
components at the blade root are minimized.
0.000
Blade tip transverse displacement (m)

-0.005 In this work the magnitude of the minimization of the

-0.010 lateral moment, χ (t ) , at the root of blades is the objec-
-0.015
tive of the optimization procedure. The problem is formu-
-0.020
lated as the minimization of the objective function
-0.025

-0.030
te et
1
 χ 0 − e( − χ ( t ) χ1 )  dt (7)
2
Ψ 0 = ∫ f 0 (t )dt = ∫
-0.035
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
(te − ti ) ti  
Time (s) ti

Fig. 5. Time history of transverse blade tip displacement. where χ 0 and χ1 are the reference value and a scaling
factor, respectively. The simulation period is of ten sec-
10000 onds but equation (7) is only computed in the simulation
5000
period of ti = 1.2 s ≤ t ≤ te = 10 s to avoid accounting for
0
the transient period of the start of the analysis. The refer-
Blade root forces (N)

-5000
-10000
Axial
Lateral
ence value, χ 0 , is defined as being the average value of
function e( − χ ( t ) χ1 ) through the simulation period. Figure 8
-15000 Transverse
-20000
-25000 shows the time history of function f 0 for the simulation
-30000
-35000
described in the previous section.
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time (s)
The objective function defined by equation (7) is
equivalent to the area under the curve f 0 between the
Fig. 6. Time history of blade root forces. limits ti = 1.2 to te = 10 . With this objective function the
fluctuation of the lateral moment around the average
value χ 0 should be minimized, which may imply an im-
provement of the blade damping. To verify this idea, the
time history of the blade lateral moment for the initial and
for the optimum design is fitted to the function defined as
 ducing the objective function in the order of 22%. Figure
χ f (t ) = χ m + χ a e (µ t)
cos( χω t + θ ) (8) 9 shows the time history of the lateral moment, χ f at the
blade root for the initial and the optimal design.
70
TABLE III. – Summary of the optimization of the wind turbine.
60

50
Upper spar cap Lower spar
40
cap
30
Optimum Layer orien-
20 700/700/700/700 900/900/900/900
tations
10
Normalized Initial ob-
0 1.00
1.2 2.0 2.8 3.6 4.4 5.2 6.0 6.8 7.6 8.4 9.2 10.0
jective function
Time (s) Normalized Optimum
0.78
objective function
Fig. 8. Time history of function f 0 . Reduction of objective
22%
function
Due to the nonlinear nature of this function the use of Function calls 15
least square minimization methods to obtain the con-
Gradient calls 2
stants χ m , χ a , µ , χω and θ used in equation (8) leads
Number of iterations 3
to the solution of a nonlinear least square optimization
problem. Thus, in this work, the constants χ m , χ a , µ ,
χω and θ are evaluated by solving a second minimiza- 0

-5000
tion problem defined as Blade root moment (N.m)
-10000
n

∑  χ
2
minimize f (t ) − χ (t )  (objective function) -15000

k =1
-20000
Subject to ( χ m ) L ≤ χ m ≤ ( χ m )U (Side constraints)
-25000 Latera (ID)

( χ a ) L ≤ χ a ≤ ( χ a )U (9)
-30000
Lateral(OD)

( χω ) L ≤ χω ≤ ( χω )U 1.2 2.0 2.8 3.6 4.4 5.2 6.0

Time (s)
6.8 7.6 8.4 9.2 10.0

( µ ) L ≤ µ ≤ ( µ )U
(θ )L ≤ θ ≤ (θ )U Fig. 9. Lateral moment at the root blade for the initial and op-
timal designs.
where the indices U and L refer to the upper and lower
limit constraints of the design variables. The optimization The solution of the optimization problem formulated
problem expressed by equation (9) is solved using a ge- by equations (8) and (9) for the initial and optimized de-
netic algorithm, specially recommended for the solution signs are presented in Table IV. The optimum design al-
of curve fitting problems [22]. lows improving the damping coefficient of about 40% of
its initial value. The representation of function χ f (t ) for
The laminate used to model blades 1, 2 and 3 is the the initial and optimum designs is made on Figure 10.
material used in the previous section. The first design
variable used in the optimization process corresponds to TABLE IV. – Summary of the curve fitting problem for the
the lay-up orientation of the laminate used in the upper second optimization scenario.
spar cap, and the second design variable corresponds to
the lay-up orientation of the laminate used in the lower Constants Initial design Optimum design
spar cap. The initial laminate design of the blades is de-
fined in Table I. χm -1.2357 × 104 -1.1981× 104

χa -1.2934× 104 -1.31478 × 104

The optimization results are summarized in Table III. χω 8.007 7.855

The convergence to the optimum solution is very fast, re-
 θ 3.037 3.036 the cross section [3]. The incorporation of such beam
analysis into the flexible multibody code has been
µ (damping) -0.05 -0.07 showed to be a powerful approach to the modeling of re-
alistic engineering systems with slender composite com-
ponents.
0

-5000
The optimization of wind turbine blades has been
Blade root moment (N.m)

-10000 achieved using the formulation proposed. The results

-15000
show a feasible design for the wind turbine in which an
improvement on damping coefficient of the lateral mo-
-20000
ment component at the blade root is observed. Although
Lateral (ID-G)
-25000
Lateral (OD-G)
an improved design of the turbine blade was obtained
-30000 there is no assurance that it is a global optimal design.
1.2 2.0 2.8 3.6 4.4 5.2 6.0 6.8 7.6 8.4 9.2 10.0
Time (s)
Acknowledgement
Fig. 10. χ f (t ) given by equation (8), for the initial and opti-
mized designs. This study was supported by the Portuguese Foundation
for the Science and Technology and FSE under a post doc
The comparison of the initial design with the corre- scholarship. The views and the findings contained herein
sponding optimum design allows concluding that the ma- are those of the author and should not be interpreted as
jor improvement of the optimum design is obtained for necessarily representing the theoretical policies or en-
the maximum value of the lateral moment rather than for dorsement, either expressed or implied, of the Portuguese
its minimum value. Moreover, if the value of the constant Foundation for the Science and Technology.
χ m of the initial design is compared with the correspond-
References
ing optimum design it is observed that the optimum de-
sign show smaller mean lateral moments than the initial
design.
[1] D.H. Hodges and W. Yu, “A Rigorous, Engineer-
friendly Approach for Modelling Realistic Compos-
Note that the present manufacturing technology limits
ite Rotor Blades”, Wind Energy (2007), Vol. 10, pp.
the accuracy at which layer orientations can be set to in-
179-193.
crements of about 50. This fact, more than constituting a
[2] V.V. Volvoi, S. Yoon, C.-Y. Lee and D.H. Hodges,
constraint in the optimization process, enforces the use of
“Structural Optimization of Composite Rotor
optimization methods such as those based on evolution
Blades”, 45th AIAA/ASME/ASCE/AHS/ASC Struc-
strategies. However, such methods usually require several
tures, Structural Dynamics & Materials Conference,
hundreds or even thousands of expensive simulation runs
Palm Springs, California, USA, (2004), pp. 19-22.
[23], being too costly to be considered in this work at this
time. [3] W. Yu, D.H. Hodges, V. Volovoi, and C.E.S. Cesnik,
“On Timoshenko-like modeling of initially curved
4 Conclusions and twisted composite beams”, International Jour-
nal of Solids and Structures (2002), Vol. 39, pp.
A general method for the design optimization of 5101-5121.
flexible multibody systems made of composite materials [4] C.-H. Ong and S.W. Tsai, The Use of Carbon Fibers
has been presented in this work for the optimal design of in Wind Turbine Blade Design, SAND2000-0478,
wind turbine blades. The emphasis of the proposed meth- Sandia National Laboratories, Albuquerque, New
odology is on a flexible multibody formulation involving Mexico, USA, (2000).
composite beams and efficient evaluation of the design [5] X. Zhao, P. Maißer and J. Wu, “A new multibody
sensitivities. The sensitivities computation of the 6×6 modeling methodology for wind turbine structures
cross-sectional stiffness matrix of the beam element is ob- using a cardanic joint beam element”, Renewable
tained by extending the VABS code to include the Energy (2007), Vol. 32, pp. 532-546.
evaluation of the design sensitivities for anisotropic, het- [6] J. Ambrósio and J. Gonçalves, “Complex Flexible
erogeneous materials and to represent general cross- Multibody Systems with Application to Vehicle
sectional geometries. In the process the use of a costly 3- Dynamics”, Multibody Systems Dynamics (2001),
D finite element discretization is avoided without any loss vol. 6(2), pp. 163-182.
of accuracy, inherent in the simplified representations of
[7] M.A. Neto, J. Ambrósio and R. Leal, “Flexible Mul- Meeting and Exhibit, 11-14, Reno AIAA, Washing-
tibody Systems Using Composite Materials Com- ton, DC, NICH Report No. 32548, (1999), pp. 135-
ponents”, Multibody Systems Dynamics (2004), 142.
Vol. 12(4), pp. 363-405. [18] O.A. Bauchau and J. Wang, “Stability analysis of
[8] K.K. Wetzel, “Utility Scale Twist-Flap Coupled complex multibody systems”, Journal of Computa-
Blade Design”, Journal of Solar Energy Engineer- tional and Nonlinear Dynamics (2006), Vol. 1, pp.
ing (2005), Vol. 12(7), pp. 529-537. 71-80.
[9] D.W. Lobitz, P.S. Veers, G.R. Eisler, D.J. Laino, [19] W. Johnson, Helicopter Theory, Dover publications,
P.G. Migliore, and G. Bir, The Use of Twist- Inc, New Jersy, USA, (1994).
Coupled Blades to Enhance the Performance of [20] E.P. Wolfe and S.S. Ochs, CFD Calculations of
Horizontal Axis Wind Turbines, SAND2001-1003, S809 Aerodynamics Characteristics, Sandia Na-
Sandia National Laboratories, Albuquerque, New tional Laboratory Report, AIAA Aerospaces Sci-
Mexico, USA, (2001). ence Meetings, USA, (1997).
[10] K. Wetzel and J. Locke, Uncoupled and Twist-Bend [21] N.M. Karaolis, P.J. Mussgrove, and G. Jeronimidis,
Coupled Carbon-Glass Blades for the LIST Tur- Active and Passive Aeroelastic Power Control using
bine, AIAA-2004-0170, A Collection of the 2004 Asymmetric Fibre Reinforced Laminates for Wind
ASME Wind Energy Symposium Technical Papers Turbine Blades, Proc. 10th British Wind Energy
at the 42nd AIAA Aerospace Sciences Meeting and Conf., D. J. Milbrow, Ed., London, (1988), pp. 22-
Exhibit, American Institute of Aeronautics and As- 24.
tronautics and American Society of Mechanical En- [22] B. Chaparro, S. Thuillier, L. Menezes, P. Manach
gineers, New York, (2004), pp. 13–23. and J. Fernandes, “Material parameters identifica-
[11] D. Griffin, M. Zuteck, D. Berry, and T. Ashwill, tion: Gradient-based, genetic and hybrid optimiza-
Development of Prototype Carbon-Fiberglass Wind tion algorithms”, Computational Material Science
Turbine Blades: Conventional and Twist-Coupled (2008); Vol. 44(2), pp. 339-346.
Designs, AIAA-2004-0169, A Collection of the [23] P. Eberhard, F. Dignath and L. Kübler, “Parallel
2004 ASME Wind Energy Symposium Technical evolutionary optimization of multibody systems
Papers at the 42nd AIAA Aerospace Sciences Meet- with application to railway dynamics”, Multibody
ing and E hibit, American Institute of Aeronautics Systems Dynamics (2003); Vol. 9, pp. 143-164.
and Astronautics and American Society of Me-
chanical Engineers, New York, (2004), pp. 1–12.
[12] G. Vanderplaats, DOT-Design Optimization Tools,
Version 3.0 VMA Engineering , Colorado springs,
(1992).
[13] R. Cook, Concepts and Applications of Finite Ele-
ment Analysis, 2nd ed., Wiley, New York, (1987).
[14] J. Ambrósio, M.A. Neto and R. Leal, “Optimization
of a Complex Multibody System with Composite
Materials”, Multibody Systems Dynamics (2007),
Vol. 18, pp. 117-144.
[15] M.A. Neto, W. Yu and R. Leal, “Generalized Ti-
moshenko modeling of composite beam structures:
sensitivity analysis and optimal design”, Engineer-
ing Optimization (2008), (In printing).
[16] M.A. Neto, J. Ambrósio and R. Leal, “Sensitivity
analysis of flexible multibody systems using com-
posite materials components”, International Jour-
nal for Numerical Methods in Engineering (2009),
Vol. 77, pp. 386-413.
[17] K. Stol, G. Bir and M. Balas, Linearized Dynamics
and Operating Modes of a Simple Wind Turbine
Model, in Proc. Of 37th AIAA Aerospace Sciencs