Journal of Materials Processing Technology 167 (2005) 463–471

Optimisation of wind turbine blades

M. Jureczko, M. Pawlak, A. M˛eżyk ∗
Faculty of Mechanical Engineering, Silesian University of Technology, ul. Konarskiego 18a, 44-100 Gliwice, Poland

Abstract

The manufacturing cost of WT blade is about 15–20% of wind turbine production cost. The expenses of innovations in design of blades
represent the small amount of overall cost of wind turbine production. Profits coming from better structural model, use of suitable composite
materials and better techniques of manufacturing, both the blades and composite materials, causes necessity of application of numerical
modeling and optimization techniques.
When designing a wind turbine, the goal is to attain the highest possible power output under specified atmospheric conditions. From the
technical point of view, this depends on the shape of the blade. The change of the shape of blade is one of the methods to modify stiffness
and stability, but it may influence aerodynamic efficiency of wind turbine. Other method to change dynamic and mechanical properties of
wind turbine is modifying the composite material, which the blade is made of. The problem of determining the optimal shape of blade and
determining the optimal composite material is a complex one, as the mathematical description of aerodynamic load is complex and a number
of constraints and objectives have to be satisfied.
These considerations have prompted the authors to take up the problem of the multi-criteria optimum design of wind turbine blades. The
aim of this study was to develop a computer program package that would enable optimisation of wind turbine blades with regard to a number
of criteria.
© 2005 Published by Elsevier B.V.

Keywords: Wind turbine blade; Composite materials; Finite element analysis; Optimisation

1. Properties of the blade
The aerodynamic profiles of wind turbine blades have cru-
cial influence on aerodynamic efficiency of wind turbine.
However, when blades are longer than 45 m the dynamic
behaviour of the blade must be also taken into account. Then,
the position and shape of spars have to be considered and
analysed. In the article [9] is mentioned that the location of
the main spar together with the location of the stiffening ribs
will have the biggest influence on the bending modes of the
blade.
The model of blade (see Fig. 1) made of shell elements
was used in multi-criteria optimisation procedure. Accord-
ing to Ref. [8], the blade is to be twisted around the elastic
axis. The position of elastic center can be changed by modi-
fying the position of spars and its shape. The solid model of
the blade is created in order to obtain required properties of
∗ Corresponding author.
E-mail address: Arkadiusz.Mezyk@polsl.pl (A. M˛eżyk).
the blade and position of spars. We are interested about the
properties of the cross-sections. Using commercial software
MSC Patran, there is possible to get information about area,
moments of inertia, shear centers and centroids. The output
data, for example, cross-section of the blade is presented in
Fig. 2.
The blade was divided to 26 cross-sections, for each of
them were received the similar data.
Now the cross-sections of the blade must be twisted along
the shear axis. But the question is what with the position of
spars, if they are also twisted. The conclusion is that they
are not twisted in similar way as aerofoils. Leaving the spars
straight, the blade would have the form shown in Fig. 3.
The commercial blades does not have the spars positioned
in this kind. The reason is the aerodynamic damping phenom-
ena. Twist of the blade decide about value of aerodynamic
loads, but also the direction in which the blade will vibrate.
The blade with twisted spars is presented in Fig. 4.
The twist of spars decides about pitch of principal bend-
ing axes. Aerodynamic damping is a very important dynamic

0924-0136/$ – see front matter © 2005 Published by Elsevier B.V.

doi:10.1016/j.jmatprotec.2005.06.055
464 M. Jureczko et al. / Journal of Materials Processing Technology 167 (2005) 463–471
Fig. 1. The structural model of the blade with spars, before twist.
Fig. 2. The calculated position of shear center and centroid.
aspect. The negative value of aerodynamic damping means
that some additional energy is added to the blade during vibra-
tion and the amplitude of vibration is increased. Aerodynamic
damping has in plane and out of plane components. Damp-
ing in-plane direction will have the negative value if the blade
section produce the power. If damping in out of plane direc-
Fig. 3. The twisted blade with straight spar.
Fig. 4. The twisted blade with twisted spars.
tion is positive, by twisting the blade the value of out of plane
damping will decrease (but must be still positive) and the in-
plane damping will receive positives values. Due to the twist
of spars, the blade will vibrate not clearly edgewise or flap-
wise [5].
The second mode shape of the blade is presented in
Figs. 5 and 6.
2. Aerodynamic loads
The analysis of aerodynamic loads in our case is based
on the Blade Element Momentum method (BEM [2,10]). As
the input date, we have the free stream wind velocity V0 ,
geometry of aerodynamic profiles used in blade with the CL
and CD as the aerodynamic lift and drag coefficients unique
for each of them. In literature, they are presented as CL (α)
and CD (α), which means that they depend on the angle of
attack α (Fig. 7).
 M. Jureczko et al. / Journal of Materials Processing Technology 167 (2005) 463–471 465

The pitch angle is the angle between the axial wind veloc-
ity and the tangential wind velocity:
V0 (1 − a)
tan φ = (1)
(1 + a ) · ω · r
Relative velocity of the wind is calculated from the equa-
tion:
ω · r(1 + a )
Vrel = (2)
cos φ
The aerodynamic loads are expressed in the following
forms:
Lift:
1
L= · ρ · Vrel
2
· c · CL (3)
2
Drag:
Fig. 5. The second mode of the blade, first edgewise mode, frequency 1
1.5975 Hz. D= · ρ · Vrel
2
· c · CD (4)
2
Thrust:
FN = L cos φ + D sin φ (5)
Fig. 6. The second mode of the blade, flapwise.
Torque:

BEM method is an iterative method, at the beginning the FT = L sin φ − D cos φ (6)
value of axial retardation coefficient a is assumed to be zero,
where ρ is the density of the air and c is the chord of aero-
the results of simulation (calculated a) are compared with the
dynamic profile.
initial value. If their values differ the calculation is repeated
with the received axial retardation coefficient as initial value. Blades used in aeroplanes have the order of the gravity,
If they agree the calculation is finished. Below there are aerodynamic and shear centers different than blades used in
shortly presented the equations used in BEM method [2]. wind turbines.

Fig. 7. The diagram of wind velocity components.

466 M. Jureczko et al. / Journal of Materials Processing Technology 167 (2005) 463–471
Fig. 8. The location of the shear, gravity and aerodynamic centers.
According to Ref. [6] on the left there are elastic centers,
gravity centers and aerodynamic centers. In modern wind
turbines the elastic center is between gravity and aerodynamic
center. The position of the spars decide about position of shear
axis and the order of centers.
As can be seen in Fig. 8 in aerodynamic center there are
working aerodynamic loads:
• tangential aerodynamic loads FT ;
• normal aerodynamic loads FN .
Also, the aerodynamic damping according to Ref. [6] is
working directly in aerodynamic center. In the gravity center
the gravity forces are working but also inertial loads, which
are the results of rotation of the blade has to be mentioned. The
elastic axis is the axis about which the aerodynamic profiles
are twisted. The elastic axis is created from elastic centers of
cross-sections.
To analyse the dynamics of the blade by the finite ele-
ment method analysis, the blade model as the Timoshenko
beam elements was created. Before it, the assumption was
made that the elastic axis is created by shear centers from
each cross-section. The grid points are created along the
aerodynamic centers. The aerodynamic forces are applied
directly to the grid point, the forces act at the aerody-
namic centers. Gravity forces are applied to the gravity
centers. The shear center and aerodynamic center is off-
set from the neutral, the loads should cause the element to
twist.
On the model of the blade presented below there are
marked three lines, red are aerodynamic centers, green are
elastic centers and black are gravity centers.
In our research, the aerodynamic analysis was done, based
on modified BEM theory described in Refs. [2,10].
We do not have pressure forces along the blade to fully
analyse the shell model. To speed up calculation, three-
dimensional model of the blade is reduced to beam elements
with data presented in Fig. 9 as input. Beam elements are
modeled along the aerodynamic centers, including shear
and gravity centers as vector offsets. The limitation of this
method is the ability to use this model only in linear analysis,
such as modal, static and linear transient analysis.
3. State of load on the blade
The analysis of the state of load on the wind turbine blade
is intended to verify whether the turbine will withstand the
action of load within appropriate safety range. Various cases
of load on the blade, resulting from the action of various
external factors on the turbine, have to be considered. The
following types of states of load on a wind turbine blade can
be distinguished:
• Aerodynamic loads of a wind turbine blade are shown in
Fig. 10.
• Mass loads, as the wind turbine blade is slender, the loads
associated with its inertia are limited to the loads generated
 M. Jureczko et al. / Journal of Materials Processing Technology 167 (2005) 463–471
Fig. 9. The model of the wind turbine with marked aerodynamic, shear and
gravity centers on the blades.
by its weight, which causes sinusoidal loads the frequency
of which corresponds to the rotor Fig. 11 shows the deter-
mined state of load on wind turbine blade. Both mass and
aerodynamic loads were investigated.
4. Material of the blade
The considered blade is made of composite materials con-
taining more than one bonded material, each with different
structural properties. One of the materials, called the rein-
forcing phase is embedded in the other material of the matrix
Fig. 10. The local forces on the blade.
467
Fig. 11. State of load on wind turbine blade.
phase. If the composite is designed and fabricated correctly,
it combines the strength of the reinforcement with the tough-
ness of the matrix to achieve a combination of desirable
properties not available in any single conventional material.
The main advantage of composite materials is the potential
for a high ratio of stiffness to weight. Composites used for
typical engineering applications are advanced fiber or lam-
inated composites, such as fiberglass, glass epoxy, graphite
epoxy and boron epoxy. Composites are somewhat more dif-
ficult to model than an isotropic material such as iron or steel.
The special care must be taken in defining the properties and
orientations of the various layers since each layer may have
different orthotropic material properties.
Majority of wind turbine blades is made of fiberglass rein-
forced with polyester or epoxy resin. Construction using
wood–epoxy or other materials also can be found. Small
turbine blades are made of steel or aluminum, but they are
heavier.
Lighter and more effective blades decrease material
requirements for other wind turbines component making
468 M. Jureczko et al. / Journal of Materials Processing Technology 167 (2005) 463–471

overall costs to be lower. Longer blades require another mate- Table 1

rials to be applied, usually carbon-based composites. Carbon Investigated composite materials
fiber composites allow to lower blade’s mass (from 20 to 18 T Material Elastic modulus [GPa] Density [kg/m3 ]
at 61.5 m long blade). Carbon-based composites allows also Kevlar 149 179 1470
to reconstruct older blades made of fiberglass reducing mass Technora 70 1390
and increasing its stiffness. However, use of carbon materials Glass E 76 2540
Glass S 88 2540
requires increased accuracy and makes manufacturing costs
to be higher.

Table 2
Natural frequency of blade, which was made of different materials
5. FEM model of wind turbine blade
No mode shape Frequency [Hz]

The FEM model of the wind turbine blade with a NACA Glass E Glass E Technora Kevlar 149
63–212 airfoil was created using APDL language in ANSYS. 1 0.72853 0.79534 0.98482 1.4996
It is a parametric model, as the thickness of the shell, com- 2 1.6909 1.8459 2.2857 3.4805
3 2.0845 2.2756 2.8178 4.2907
posite material, which blade is made, number of stiffening
4 4.1426 4.5224 5.5999 8.527
ribs and their arrangement were the model parameters that 5 6.0169 6.5686 8.1336 12.385
were input from the authors’ program that implemented a 6 7.1806 7.839 9.7067 14.78
modified genetic algorithm. A given parametric file can be 7 10.975 11.982 14.836 22.591
used to create various blade models, modify their thickness 8 14.548 15.882 19.666 29.946
9 15.035 16.414 20.324 30.948
and basic dimensions.
10 16.44 17.947 22.223 33.84
The created FEM model of the blade consists of 124,042
elements, 55,044 nodes and 327 areas meshed. The 8-nodal
shell of the SHELL 63 type with 6 degrees of freedom was
chosen as finite elements, what enabling specification of any 6. Selection of composite materials of wind turbine
thickness at each node of the chosen element. Selection of blade
elements in a numerical model of a blade enables specifi-
cation of various thicknesses and material data and defining The aim of this study was estimation the influence of
various types of elements. composite materials, which the blade is made, on dynamical
The following simplifying assumption was made when properties of wind turbine blades. The composite materials,
creating the numerical model of the blade: which were investigated are presented in Table 1.
Modal analysis was made for blade with constant elements
• the manner of connecting the shell with supporting webs thickness. Table 2 shows natural frequency of blade, which
and stiffening ribs was neglected. was made of different materials.
Materials with lower density such as fiber aramid (Tech-
The blade was treated as a clamped beam with its geometry nora) have higher natural frequencies and bigger deflection.
determined in the manner described in previous section. The comparison of fibreglass S-type and fibreglass E-type
The FEM blade model is shown in Fig. 12. shows increase of natural frequencies without eigenvalue
change. Another material—Kevlar, increases both natural
frequencies and eigenvalue.

7. Formulation of the problem of multi-criteria

discrete–continuous optimisation

It is not possible to formulate the problem of optimum

design of wind turbine blades as a single-criterion optimi-
sation task because this process requires many criteria to be
taken into account. In many cases, these criteria are mutually
incomparable, uncountable and sometimes even contradic-
tory, which precludes their simultaneous optimisation.
The authors have taken into account the following criteria
in the process of wind turbine blades design:
• minimisation of generated blade vibrations;
Fig. 12. FEM blade model. • maximisation of output generated;
 M. Jureczko et al. / Journal of Materials Processing Technology 167 (2005) 463–471 469

• minimisation of blade material cost; Table 3

• local and global stability of blade structure; Genetic algorithm—no copying
• fulfilment of appropriate strength requirements by the Variable
blade structure. x1 (m) 0.0201
x2 (m) 0.0118
x3 10
Minimisation of vibrations is a good way to successful x4 (m) 1.25, 33.75
design of blade structure and at the same time it contributes
Found in population 22
to other benefits, such as lower cost or high stability. How-
Value of objective function (kg) 1645.1
ever, when minimising vibrations of the blade, the natural Duration of computation (s) 125352
frequency of the blade must be separated from the harmonic
vibration associated with rotor rotation. Such an approach
prevents the occurrence of resonance, which under high optimisation task
amplitude of vibration could lead to destruction of the struc- min
f (x) (7)
x∈Ω
ture. Frequency spacing is one of the methods of isolating
frequencies. where Ω is the domain of possible solutions within the space
The amplitude of generated vibrations of the wind turbine of objects, x the column matrix of design variables and f(x)
blade depends on its stiffness, which is a function of material is the objective function, i.e. mass of blade.
density, thickness of shell, number of stiffening ribs and their The column matrix of design variables can be represented
arrangement along blade span. Therefore, when the vibration in the following form:
minimisation criterion be taken into account, the wind turbine xT = [x1, x2, x3, x4] (8)
blade should be provided with the highest possible stiffness.
Such a formulation of the optimisation problem also sat- where x1 is the shell thickness, x2 the web thickness, x3 the
isfies the criterion of generated output maximisation, as the number of stiffening ribs and x4 is the arrangement of stiff-
output of a wind turbine depends also on the optimum shape ening ribs.
of blades, i.e. on their optimum geometrical features. The other criteria were expressed in the form of inequality
The mass and fabrication cost of a blade depend on the limitations:
same parameters as the amplitude of blade vibrations. If
• stresses generated in the blade cannot exceed permissible
the cost minimisation criterion were considered, then the
stresses—compliance with appropriate strength require-
optimisation task would have to be formulated as a weight
ments of the structure
minimisation task. However, in order to ensure stability of
the structure, the weight should be maximised. σ(x) ≤ σdop (9)
Furthermore, to meet the strength requirements of the
structure, optimisation of maximum displacements of the • displacement of individual nodes in the numerical model
blade in transverse direction would have to be carried out with of the blade cannot exceed the set value—global stability
a limiting condition that permissible stresses be not exceeded. must be ensured
In view of the above considerations, the authors formu- u(x) ≤ 0.1 m (10)
lated the problem of optimum design of wind turbine blades
as a multi-criteria optimisation task, which enables simul- • displacement of the nose of the numerical model of the
taneous investigation of several criteria. The values of indi- blade cannot exceed the set value—local stability must be
vidual criteria depend on parameters of continuous nature ensured
(thickness of shell, thickness of webs) and of discrete nature
uTIP (x) ≤ 0.15 m (11)
(number of stiffening ribs, their arrangement along blade
span). • separation of natural frequency of the blade from harmonic
The optimisation task was formulated as a discrete– vibration associated with rotor rotation
continuous multi-objective problem.
freqBLADE
= freqROTOR (12)

Table 4
8. Formulation of the optimisation problem Genetic algorithm—copying permitted
Variable
After having carried out the study of formulating an opti- x1 (m) 0.0182
misation criterion when minimising vibrations, consisting in x2 (m) 0.0144
the investigation of the effect of the various criteria discussed x3 2
x4 (m) 1.75, 2.25
above on the amplitude of vibrations of the wind turbine
blade, the authors formulated the problem of multi-criteria Found in population 43
discrete–continuous optimisation using the ␧ limitations Value of objective function (kg) 1557.4
Duration of computation (s) 120916
method (described, for instance, in Ref. [4]) a single criterion
470 M. Jureczko et al. / Journal of Materials Processing Technology 167 (2005) 463–471

Fig. 13. History of optimisation by means of a modified genetic algorithm with no copying of the best individual.

Further limitations apply to the values of design variables.
These can be expressed by means of the following matrix
formula:
xdown ≤ x ≤ xup
where xdown is the column matrix of lower bound variables
and xup is the column matrix of upper bound variables.
Optimisation calculations were done with the use of the
authors’ program that implemented a modified genetic algo-
rithm, for which the following assumptions were made:
• number of individuals: 20;
• number of populations (STOP criterion): 50;
(13) • probability of crossing: 0.7;
• probability of mutation: 0.03.
The applied probabilities of genetic operations were
adopted on the basis of studies performed, inter alia, by the
authors and presented, for instance, in Ref. [1].

Fig. 14. History of optimisation by means of a modified genetic algorithm with copying of the best individual.
 M. Jureczko et al. / Journal of Materials Processing Technology 167 (2005) 463–471
9. Numerical calculation results
Selected results obtained for the optimisation problem are
given in Table 3 (optimisation with the use of the modified
genetic algorithm and with no copies of the best individual)
and Table 4 (optimisation with the use of the modified genetic
algorithm and with copies of the best individual).
Fig. 13 shows chosen courses of optimisation with the
use of the modified genetic algorithm including the copy-
ing of the best individual, whereas Fig. 14 shows chosen
courses of optimisation with the use of the modified genetic
algorithm where the copying of the best individual was
permitted.
Because the genetic algorithms this the stochastic methods
of search of optimum, which belong to Pareto-optimal set,
individual curves solutions (appointed with different colours)
represent the course of individual optimisation process.
10. Conclusion
The developed numerical model of the wind turbine blade
and the computer program package for performing multi-
criteria discrete–continuous optimisation of wind turbine
blades are of general nature. Various blade models can be
created by means of an ANSYS parametric file; thicknesses
and main dimensions of the model blade can be varied. The
authors’ program that implements a modified genetic algo-
rithm enables optimisation of various objective functions
subjective to various constraints.
Subsequent to determination of the optimised profile of
the wind turbine blade could be machined using Pre-Pregs
471
or Vacuum-Assisted Resin Transfer Moulding (VARTM)
method [8,11].
Acknowledgement
This investigation was realized within a framework of
Project No. 4T07C06828 funded by Scientific Committee
(KBN).
References
[1] P. Bachorz, M. Marcinkowska, Relation between efficiency of a
genetic algorithm and assumed optimization parameters, in: Pro-
ceedings of the Scientific Conference Applied Mechanics, 2002, pp.
11–16.
[2] O.L. Hansen Martin, Aerodynamics of Wind Turbines, James &
James, London, 2002.
[4] R.T. Marler, J.S. Arora, Survey of multi-objective optimization meth-
ods for engineering, Struct. Multidisciplinary Optimisation 26 (2004)
369–395.
[5] J.T. Petersen, H.A. Madsen, et al., Prediction of Dynamic Loads
and Induced Vibrations in Stall, Risø-R-1045 (EN), Risø National
Laboratory, Roskilde, Denmark, 1998.
[6] R.L. Bisplinghoff, H. Ashley, R.L. Halfman, Aeroelasticity, Dover
Publications Inc., New York.
[8] Guidelines for Design of Wind Turbines, second ed., Det Norske
Veritas and Riso National Laboratory, Jydsk Centraltrykkeri, Den-
mark, 2002.
[9] One-Dimensional Variations: Blades, Dutch Offshore Wind Energy
Converter Project, LM Glasfiber Holland BV, 2003.
[10] A. Spera David, Wind Turbine Technology, Fundamental Concepts
in Wind Turbine Engineering, ASME Press, New York, 1998.
[11] J. Śleziona, Podstawy Technologii Kompozytów, Silesian University
of Technology, Gliwice, 1998.