Sie sind auf Seite 1von 15

ARTICLE IN PRESS

Renewable Energy 32 (2007) 916930


www.elsevier.com/locate/renene

Combined analytical/FEA-based coupled aero


structure simulation of a wind turbine with
bendtwist adaptive blades
Alireza Maheri, Siamak Noroozi, John Vinney
Faculty of CEMS, University of the West of England-Bristol, Coldharbour Lane, Bristol, BS16 1QY, UK
Received 5 January 2006; accepted 10 April 2006
Available online 12 June 2006

Abstract
The simulation of wind turbines with bendtwist adaptive blades is a coupled aero-structure (CAS)
procedure. The blade twist due to elastic coupling is a required parameter for wind turbine
performance evaluation and can be predicted through a nite element (FE) structural analyser. FEAbased codes are far too slow to be useful in the aerodynamic design/optimisation of a blade. This
paper presents a combined analytical/FEA-based method for CAS simulation of wind turbines
utilising bendtwist adaptive blades. This method of simulation employs the induced twist
distribution and the ap bending at the hub of the blade predicted through a FEA-based CAS
simulation at a reference wind turbine run condition to determine the wind turbine performance at
other wind turbine run conditions. This reduces the computational time signicantly and makes the
aerodynamic design/optimisation of bendtwist adaptive blades practical. Comparison of the results
of a case study which applies both combined analytical/FEA-based and FEA-based CAS simulation
shows that when using the combined method the required computational time for generating a power
curve reduces to less than 5%, while the relative difference between the predicted powers by two
methods is only about 1%.
r 2006 Elsevier Ltd. All rights reserved.
Keywords: Wind turbine; Adaptive blade; Elastic coupling; Coupled-aero-structure simulation

Corresponding author. Tel.: +44 0117 3282643; fax: +44 0117 3283800.

E-mail address: Ali2.Maheri@uwe.ac.uk (A. Maheri).


0960-1481/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.renene.2006.04.007

ARTICLE IN PRESS
A. Maheri et al. / Renewable Energy 32 (2007) 916930

917

1. Introduction
Simulation of a wind turbine with adaptive blades is different from the simulation of
ordinary wind turbines. In simulation of wind turbines with conventional blades, the twist
angle of the blade is a combination of known parameters of pre-twist and pitch angle
whilst the elastic torsional deformation is assumed to be negligibly small (rigid blade
assumption). Knowing topology and aerodynamic characteristics of the blades together
with the wind turbine run condition, an aerodynamic code can be employed as a wind
turbine simulator to predict the wind turbine performance. Fig. 1 illustrates the simulation
of an ordinary wind turbine schematically.
In the case of wind turbines with adaptive blades, even by neglecting the torsion due to
off-axis aerodynamic loading, the angle of attack is still under the inuence of elastic
torsional deformation generated due to elastic coupling, called induced twist. Induced twist
is a variable parameter that depends on the wind turbine run condition as well as
aerodynamic, material and structural characteristics of the blade. In the other words an
adaptive blade has dynamic topology that varies with the wind speed and/or rotor angular
velocity. The same approach as shown in Fig. 1 is still functional for simulation of a wind
turbine with adaptive blades but the induced twist b, must be given to have the blade
topology known.
1.1. Coupled aero-structure (CAS) simulation
In adaptive blades made of bend-twist-coupled materials, the ap bending in the blade
due to the aerodynamic force bends the blade and consequently the elastic coupling
produces an induced twist. In these blades the source load for the induced twist is the
aerodynamic force, which depends on the rotor angular velocity, wind velocity and blade
aerodynamic characteristics. The source load affects the induced twist and the induced
twist affects the source load. In the other words there is an interaction between the induced
twist and the source load. This interaction makes the simulation of these types of wind
turbines a coupled aero-structure process that needs an internal correction loop. In a
coupled aero-structure simulation, the effect of the induced twist on the initial loading
situation is taken into account. Correcting the load, induced twist will be re-calculated.
This sequence repeats until a converged solution has been reached. A schematic description
of a CAS simulation is shown in Fig. 2. The lack of a reliable analytical model for the
analysis of anisotropic composite shell structures makes the nite element (FE) method the

Blade geometry
and topology

Blade
aerodynamic
load predictor

Aerodynamic
loading

Wind turbine
performance
calculator

Aerofoils
aerodynamic
characteristics
Wind turbine
run-condition

Wind turbine
performance

Fig. 1. Simulation of a wind turbine with ordinary blades.

ARTICLE IN PRESS
A. Maheri et al. / Renewable Energy 32 (2007) 916930

918

Blade materialand
structural characteristics
Wind turbine
run-condition

2i
Blade
aerodynamic
load predictor

Aerofoils
aerodynamic
characteristics

3i
Source
load

Blade
FEA

Topology of
unloaded blade

Aerodynamic
Loading

1i
Blade
geometry and
topology

Blade
topology
corrector

Wind turbine
performance

Order of operation

n
ni

Initiator

Order of operation
in correction loop
Operation

Induced twist predictor

Data/Result

Induced
twist

4
Wind turbine
performance
calculator
Correction Loop

Fig. 2. Coupled aero-structure (CAS) simulation of a wind turbine with bendtwist adaptive blades.

rst choice for the analysis of these types of structure. Therefore, in a coupled aerostructure analysis approach a FE code is necessary to predict the induced twist while an
aerodynamic code predicts the blade loading and wind turbine performance.
There are many practical limitations involved in attempting to use commercial codes for
a CAS simulation. Using a commercial FE code, running in a loop, could for example be
very cumbersome and inefcient. An interface would also be necessary in order to convert
the blade aerodynamic loading, calculated by the aerodynamic code, to a proper load le
for the FE solver. A fundamental requirement of a coupled aero-structure code is that the
mesh must be efcient. Generating a mesh based on an automatic algorithm, with its sizing
function only based on the geometry of the blade, does not necessarily yield an efcient
mesh suitable for an iterative code. In the FE part of the simulation the emphasis is only on
the blade deformations, which inuence the aerodynamics of the turbine, rather than the
stress analysis and dynamic response of the blade. Therefore, the FE part only needs to
predict the blade induced twist and no stress recovery is required. All of these factors make
utilising commercial codes inefcient and problematic. Maheri et al. [1] developed an aerostructure code, called WTAB, especially designed for this purpose. Many unique features
of this code (i.e. using a force-adaptive semi-automatic mesh generator, applying a
convergence accelerator in the BEMT-based aerodynamic load predictor and employing a
sophisticated element based on natural mode FE formulation) have made it fast, accurate

ARTICLE IN PRESS
A. Maheri et al. / Renewable Energy 32 (2007) 916930

919

and therefore efcient approach for the CAS simulation of wind turbines with bendtwist
adaptive blades.
1.2. Objective evaluation in blade design
The aerodynamic design of a blade involves the determination of the geometry and
topology of the blade, which satises certain objective(s) subjected to constraints. A
complete aerodynamic design normally has two steps of primary design based on one of
the available theories and optimisation. Preliminary design is a simple and straightforward
process but since theories behind the preliminary design are simplied, the output of the
rst step is a long way removed from an optimum design. As a matter of fact, the
optimisation takes up a signicant proportion of the total time and effort in blade
aerodynamic design. In the design/optimisation process after selecting each set of design
parameters, either manually based on the designers experience or through an automatic
search in the design space, objective evaluation must be carried out. In optimisation
algorithms based on either maximising the annual energy capture or minimising the energy
cost, the objective evaluation needs the average power of the wind turbine. This means that
for a single objective evaluation a power curve is required. To generate an accurate power
curve, the simulator must run for different wind velocities between the cut-in and cut-out
velocities with small increments. Although employing an aero-structure code like WTAB
makes the simulation practical the speed of a FE-based code falls short of what is required
in the aerodynamic design/optimisation of an adaptive blade.
1.3. Analytical analysis of deformation of adaptive blades
In general, rotor blades have different structural parts, each of them for a particular
purpose. Namely, skin shell and spars are the load-carrying parts, the internal ller foam
maintains the shape of the aerofoil and the leading-edge cap protects the blade skin from
erosion. Since the effect of internal lling parts in load carrying is negligible, a rotor blade
can be analysed as a single-or multi-cell closed thin/thick-walled beam. In composite rotor
blades the wall undergoes different deforming behaviours, some of them known as nonclassical, such as large torsional warping and coupled in-and out-of-plane cross-sectional
warping. Transverse shear, cross-sectional warping, the effects of restrained warping and
elastic coupling should be consistently accounted for in a beam model to obtain accurate
results. To capture the non-classical behaviours of composite blades, many analytical
studies have been carried out. These studies can be categorised based on their different
approaches that they have followed in model development (stiffness/ exibility/ mixed
methods) and different theories that they have used in beam modelling (EulerBernoulli/
Timoshenko/ higher-order shear deformation theories in bending and transverse shear
modelling and St Venant/ Vlasov theories in torsion modelling).
The displacement or stiffness method has been used by many investigators, among them,
Reheld et al. [2], Smith and Chopra [3], Chandra and Chopra [4,5], Kim and White [68],
Bhaskar and Librescu [9], Song and Liberscu [10] and more recently Qin and Librescu [11].
This method is based on a proposed displacement eld over the blade wall. Warping
functions are involved in the displacement eld, and except for simple geometries, there is
no method to propose a correct warping distribution. The force or exibility method is
based on a proposed stress eld in the shell, while the warping distribution can be obtained

ARTICLE IN PRESS
920

A. Maheri et al. / Renewable Energy 32 (2007) 916930

through the equilibrium equations. This method provides a systematic method in warpingfunction determination. Libove [12] and Johnson et al. [13] are among the researchers who
applied this method for analysis of thin-walled beams. Recently, Jung et al. [14] used a
mixed displacementforce method to analyse open and closed thin/thick walled beams with
uniform cross-section.
The EulerBernoulli theory of bending is widely used for thin-walled beams. Early
investigations on composite blades were based on this theory. Works of Manseld and
Sobey [15], Bauld and Tzeng [16], Bauchau [17], Berdichevsky et al. [18] and Badir et al.
[19] are examples of composite blade modellings based on this theory. This theory does not
include the effect of transverse shear (through the assumption of planes normal to the axis
of the beam remain plane and normal to the axis after loading), and hence it is not
applicable to the short or thick beams or those made of high-strength composite materials
with high level of anisotropy. Timoshenkos beam theory also referred to as rst order
shear deformation theory, assumes that the planes normal to the axis remain plane but not
normal to the axis after loading. Many investigators, among them, Reheld [20], Smith
and Chopra [3], Chandra and Chopra [4,5] and Song and Librescu [10] used this theory.
Among the investigators using the higher-order shear deformation theories, Kim and
White [68] and Qin and Librescu [11] used a secondary warping function to rene the shell
deformation.
St Venants torsion analysis is based on the assumption of unconstraint warping, which
means the effect of end constraints are limited only to a small region near the ends and
vanishes rapidly inward. For composite beams, the end effects can persist over a longer
length of span and can be more important in the accuracy of obtained results. Manseld
and Sobey [15], Bauchau [17], Smith and Chopra [3], Berdichevsky et al. [18] and Cesnik et
al. [21] used St Venant theory in torsion modelling. Models proposed by Reheld [20],
Chandra and Chopra [4,5], Badir et al. [19], Song and Librescu [10], McCarthy et al. [22],
Kim and White [68] and Qin and Librescu [11] take the restrained or warping torsion into
account by involving the coupling between extension, bending, torsion, out-of-plane
warping and transverse shear deformations.
All deformations of the cross-section of a beam due to a combined loading situation
(bending, shear and torsion) can be expressed by either in or out-of-plane displacements of
the cross section. The in-plane warping accounts for the lateral distortion of the crosssection due to Poissons effect, which is especially important for the case of open crosssections and thin-walled closed cross-sections. In the case of closed thick-walled crosssections the in-plane warping can be neglected reasonably as the thickness increases. In
beam theories incorporating the rened shear deformation models in the shell, the out-ofplane warping can be expressed in terms of a combination of two primary and secondary
functions. Primary warping represents the displacement of the points located on the
contour (or mid-line in the case of open sections), while warping across the beam wall
thickness is referred to as secondary warping. The inuence of the secondary warping on
the accuracy of the obtained deformation becomes more obvious as the thickness of the
shell increases.
In modelling of bendtwist elastic coupled beams the torsional warping used in the
theory has a great inuence on the predicted results [23]. Despite the importance of torsion
warping, the proposed torsional warping functions are limited to very simple geometries
like one- and multi-cell rectangular box beams. This is due to the lack of a general analytical
method to nd such functions for different geometries. All the displacement-based

ARTICLE IN PRESS
A. Maheri et al. / Renewable Energy 32 (2007) 916930

921

methods, mentioned above, in the validation position, have been compared with the
experimental and/or FEA data only for simple geometries (mostly rectangular box beam).
As a matter of fact, despite of excellent proposed models (for example, Kim and White [7]);
it is the lack of knowledge about the warping functions that has put the FEM-based
techniques at the rst position in analysis of composite blades.

2. Combined analytical/FEA-based CAS simulation


In the pursuit of developing a closed-form formulation for predicting the induced twist
the effects of the in- and out-of-plane warping is so evident that the analytical methods
with simplications on warping functions cannot be trusted as the structural analyser.
In a combined analytical/FEA-based approach, in CAS simulation of wind turbines
with bendtwist adaptive blades, the simplifying assumptions are applied on the blade
deformation and loading rather than warping functions. This makes it possible to express
the induced twist in the blade in terms of ap bending (source load of induced twist in
bendtwist adaptive blades) and blade effective-stiffness distributions. The effectivestiffness distribution can be determined by running a FEA-based CAS simulation at a
reference wind turbine run condition.
This method of simulation is very effective when a power curve, or the average power, is
desired. It applies the FE analysis of the blade only once and employs the obtained results

FE-based CAS simulation


of wind turbine at a
reference wind velocity 1
Wind turbine
run-condition

Aerofoils
aerodynamic
characteristics

4i

5
Correction Loop

Topology
of
unloaded
blade

Induced twist
calculator
based on
analytical
model 2i

Blade
topology
corrector

Induced twist

3i
Blade
geometry and
topology

Wind turbine
performance
calculator

Source
load

Blade
aerodynamic
load predictor

Aerodynamic
loading

Reference
induced twist

Wind turbine
performance

Order of operation

Order of operation
in correctionloop

ni

Operation

Initiator
Induced twist predictor

Data

Fig. 3. Combined analytical/FEA-based CAS simulation of a wind turbine with bendtwist adaptive blades.

ARTICLE IN PRESS
922

A. Maheri et al. / Renewable Energy 32 (2007) 916930

for other wind turbine run conditions (wind velocities, rotor speed and blade pitch angle).
It can be also very useful in rened design or optimisation, where the variations of
structural parameters are limited. Fig. 3 shows a schematic description of simulation of a
wind turbine using this method.
3. Forcedisplacement relations in composite closed thin-/thick-walled beams
Following the assumptions of Kim and White [7] in establishing the kinematics relations,
which are:







The contours of the original beam cross-sections do not deform in their own planes,
the out-of-plane displacement of the beam cross-section due to bending and shear is
assumed to be described by a cubic function of the cross-sectional coordinates (higherorder shear deformable theory),
beam shell behaves as a thick shell. This implies that the transverse shear effects of the
shell are also taken into account,
primary and secondary torsional warping effects are taken into account,
All deformations are small and within linear elasticity,

the forcedisplacement relations for the case of circumferentially asymmetric stiffness


(caused by mirror lay-up and generating bendtwist elastic coupling), is given by two
uncoupled sets of forcedisplacement equations as follows:
9 2
9
8
38
u00 >
K 11 K 12 K 13 >
>
=
=
< F >
<
7 v0  y
Vy 6
K
K
0
z
,
(1)
21
22
4
5
0
>
>
>
;
;
:V >
:
0
w0  yy
K 31
0
K 33
z
9 2
8
K 44
>
=
< T >
6
M y 4 K 54
>
;
:M >
K 64
z

K 45
K 55
0

38 0 9
b
K 46 >
=
< >
0
7
y
0 5
.
y
> >
K 66 : y0z ;

(2)

Stiffnesses Kij are given in Kim and White [7].


Theoretically, having the mechanical properties of the material of the blade and the
blade cross-section, together with warping function c cp cs , the stiffnesses Kij can be
determined. But, as mentioned earlier, warping functions are only available for very simple
geometries and in the case of a blade cross-section, the warping function is unknown.
Therefore, Eq. (2) cannot be used analytically to determine the blade deformation (here, of
particular interest, the blade twist).
4. Simplied forcedisplacement relations in bendtwist-coupled composite blades
FE analysis of the blade using shell or 3D solid elements gives nodal displacements,
which can be used to nd the blade deformation. But since there are only 3 equations (Eqs.
(2)) and 7 independent unknowns (the stiffnesses Kijs), these results cannot be used to nd
the unknown.

ARTICLE IN PRESS
A. Maheri et al. / Renewable Energy 32 (2007) 916930

923

The objective of this section is to use some simplifying assumptions in order to reduce
the number of independent unknowns (Kijs) to the number of equations. By this, one can
apply the results of FEA to nd the new (combined) unknowns. Having these unknowns
and using the simplied analytical model, makes the CAS simulation of wind turbine at
different run conditions possible while FEA-based simulation takes place only once. Two
simplifying assumptions have been made as follows.
Assumption 1. It is assumed that comparing with the ap-wise slope yy and twist of the
blade b, the edge wise slope of the blade is negligible (yzE0). Fig. 4 shows the angular
deections of the blade of an approximate model of a 2-blade AWT-27 wind turbine with
the assumption of having bendtwist blades. For turbine specications refer to the
appendix and Ref. [24].
This assumption reduces Eq. (2) to the following form.
)
#(
  "
qb=qx
K 44 K 45
T


,
qyy qx
K 54 K 55
M

(3)

where, respectively, M and yy are ap bending moment and ap-wise slope of the blade
and T and b are twisting moment and the blade twist.
Assumption 2. It is assumed that the internal torque due to the off-axis aerodynamic
loading of the blade, in comparison with the torque produced due to elastic coupling, is
negligible. This assumption can be justied by recalling that the torsion due to elastic
coupling caused by ap bending moment depends on the magnitude of lift and drag forces,
while the torsion due to the off-axis loading depends on the magnitude of pitching
moment. And also, normally, order of magnitudes of pitching moment coefcient is much
less than of that of lift and drag coefcients.
Remembering that in the case of circumferentially asymmetric stiffness the bending
moment generates the induced torque, by the second assumption one can express the
torque T as a fraction of the ap bending moment M, as follows:
T M.

Angular deflections (deg)

(4)
5
4
3
2
1
0
0

0.2

0.4

0.6

0.8

r
Twist

Flap-wise slope

Edge-wise slope

Fig. 4. Angular deections of a typical bendtwist wind turbine blade.

ARTICLE IN PRESS
A. Maheri et al. / Renewable Energy 32 (2007) 916930

924

Omitting qyy =qx between Eqs. (3) and combining the result with Eq. (4) yields to the
following equation.
M K qb=qx,

(5)

in which K is bend-twist effective stiffness and is dened by


K

K 44 K 55  K 45 K 54
.
K 55  K 45

(6)

By reducing the system of Eqs. (2) to Eq. (5), once a FEA carries out, the unknown K
will be determined and can be used in the simple closed-form analysis of the blade (Eq. (5)),
to nd the induced twist at other loading condition.
Eq. (5) is the base of combined analytical/FEA-based CAS approach in simulation of
the wind turbines utilising bendtwist adaptive blades. To be more convenient, the x-axis is
replaced by the radial r-axis, with its origin located at the centre of rotor rather than the
blade root. By this, Eq. (5) can be used to nd the induced twist distribution as follows
Z 
M hub r M  r 
br
dr .
(7)
K max 0 K  r
The normalised radial location r, normalised ap bending M and normalised effective
stiffness K appearing in the above equation are dened by
r  Rhub
,
R  Rhub

(8)

M

M
,
M hub

(9)

K

K
,
K max

r

and
(10)

where Mhub is the ap bending at the hub (blade root), Kmax is the maximum effective
stiffness of the blade and R and Rhub are rotor and hub radius, respectively.
Figs. 57 show the variation of normalised ap bending distributions with respect to
wind velocity, rotor speed and pitch angle, respectively. These data are the results of FEAbased CAS simulation of wind turbine of Appendix A.
These gures represent that normalised ap bending moment is a very weak function of
wind turbine run condition and lead one to this conclusion that the effects of wind velocity,
rotor speed and pitch angle variations on the normalised ap bending moment M, can be
neglected.
Since the effective stiffness K KKmax does not depend on the wind turbine runcondition and aerodynamic characteristics of the blade and given that the normalised ap
bending M is a weak function of wind turbine run-condition, it is therefore logical to
assume that the term
Z r  
1
M r 
dr
K max 0 K  r
in Eq. (7) is only a function of material and cross-sectional properties of the blade. As a
result of this the induced twist produced in the blade of a wind turbine varies only due to

ARTICLE IN PRESS
Normalised Flap Bending, M

A. Maheri et al. / Renewable Energy 32 (2007) 916930

925

V=5m/s
V=10m/s
V=15m/s
V=20m/s
V=25m/s

0.8
0.6
0.4
0.2
0
0

0.2

0.4

0.6

0.8

Radial Location, r

1
Tip

Normalised Flap Bending, M

Fig. 5. Variation of normalised ap bending distributions against wind velocity. Pitch 1.21; O 53.33 rpm.

1
RPM=40
RPM=53.3
RPM=60

0.8
0.6
0.4
0.2
0
0

0.2

0.4

0.6

0.8

Radial Location, r

1
Tip

Normalised Flap Bending,M

Fig. 6. Variation of normalised ap bending distributions against rotor speed. Pitch 1.21; V 10 m/s.

Pitch= -2deg
Pitch= -1deg
Pitch= 0 deg
Pitch= 1 deg
Pitch= 2 deg

0.8
0.6
0.4
0.2
0
0

0.2

0.4

0.6

RadialLocation, r

0.8

1
Tip

Fig. 7. Variation of normalised ap bending distributions against blade pitch angle: V 10 m/s; O 53.33 rpm.

the variation of the ap bending moment at the hub, Mhub, (as a result of the variations of
wind velocity, rotor speed and blade pitch angle). This allows us to conclude that the above
term is invariant with respect to the wind turbine run condition and once the induced twist

ARTICLE IN PRESS
926

A. Maheri et al. / Renewable Energy 32 (2007) 916930

b(r) is calculated at a reference wind velocity, pitch angle and rotor speed, it can be
used for prediction of b(r) at other wind velocities, pitch angles and rotor speeds
as follows
br ; V ; O; pitch M hub V ; O; pitch

br ref
.
M hub;ref

(11)

In which, b(r)ref b(r,Vref, Oref, pitchref) is the reference induced twist and
Mhub,ref Mhub(Vref, Oref, pitchref) is the reference ap bending at the hub, both obtained
from the FEA-based CAS simulation of the wind turbine, at reference point.
5. Simulation algorithm
The following algorithm shows how this method can be used to generate a power curve.
To run this algorithm, a BEMT-based aerodynamic code has been used.
1. At reference run condition (Vref, Oref, pitchref) do:
1.1 Run a FEA-based CAS simulation to nd the induced twist distribution and hub
ap bending moment at that run-condition, b(r)ref and Mhub,ref.
2. For wind velocities from the cut-in velocity of VI to the cut-out velocity of Vo do:
2.1 Assume a ap bending moment at hub Mhub. This stage has illustrated by the
double dashed box in Fig. 3.
2.2 Calculate induced twist distribution through Eq. (11).
2.3 Apply the required corrections to the blade topology (here, blade twist) and using
an ordinary wind turbine simulator, nd the performance of the wind turbine and
also
 re-calculate the ap bending moment at the hub, Mhub, new.
2.4 If M hub  M hub;new 4, then replace Mhub by Mhub,new and go back to Step 2.2;
otherwise use the results of Step (2.3) as the nal results for the present wind
velocity.

6. Evaluation of the methodDiscussion


To evaluate this method of simulation, the wind turbine of Appendix A has been
simulated over a range of wind velocities of 5 to 25 m/s, by increments of 1 m/s, in order to
obtain the power curve. Running WTAB at reference wind velocity of 10 m/s, rotor speed
of 53.33 rpm and pitch angle of 1.2 1 toward stall gives the reference data of b(r)ref and
Mhub,ref (see Tables A.1, A.2).
Fig. 8 shows the predicted tip induced twists by combined method at different wind
velocities and compares them with the results of FEA-based CAS simulation. Fig. 9
compares the power curves obtained by two different methods of simulation.
Fig. 10 represents the relative difference (RD) between the predictions by combined
Analytical/FEA-based and FEA-based CAS simulations. The maximum RD in the
predicted tip-induced twist is about 7% corresponding to the wind velocity of V 24 m/s.
According to this gure the combined analytical/FEA-based CAS simulation yields
excellent results in power prediction with a maximum RD of about 1.5% with respect to a
complete FEA-based CAS simulation. The advantage of this method becomes more

ARTICLE IN PRESS
Tip Induced Twist (deg)

A. Maheri et al. / Renewable Energy 32 (2007) 916930

927

10
9
8
7
6
5
4
3
2
5

10

15

20

25

Wind Velocity (m/s)


FEA-based CAS Simulation (WTAB)
Combined Analytical/FEA-based CAS Simulation
Fig. 8. Tip-induced twist. Combined analytical/FEA-based and FEA-based CAS simulation.

600
Power (KW)

500
400
300
200
100
0
5

15

10

25

20

Wind Velocity (m/s)


FEA-based CAS Simulation (WTAB)
Combined Analytical/FEA -based CAS Simulation

Relativedifference,
RD (%)

Fig. 9. Power curve. Combined analytical/FEA-based and FEA-based CAS simulations.

8
6
4
2
0
5

10

15

20

25

Wind velocity (m/s)


Relative differencein the predicted tip induced twist
Relative differencein the predicted power
Fig. 10. Relative difference in the results of combined analytical/FEA-based and FEA-based CAS simulations
Vref 10 m/s.

ARTICLE IN PRESS
A. Maheri et al. / Renewable Energy 32 (2007) 916930

Relative difference,
RD (%)

928

10
8
6
4
2
0
5

10

15

20

25

Wind velocity (m/s)


Relative difference in the predicted tip induced twist
Relative difference in the predicted power
Fig. 11. Relative difference in the results of combined analytical/FEA-based and FEA-based CAS simulations
Vref 14 m/s.

noticeable when remembering that to generate a power curve the combined analytical/
FEA-based simulation runs a FEA-based CAS simulation only once. For this particular
case study, the computation time is less than 5% of a FEA-based CAS simulation. Fig. 10
also indicates that
1. the variation of the RD in the predicted power follows the variation of RD in the
predicted tip induced twist but with a smoother trend and smaller magnitudes,
2. RD in the predicted power in low and high wind velocities is less sensitive to
the magnitude of the RD in the predicted tip-induced twist. Therefore, to improve
the overall accuracy, one may chose a moderate wind speed as the reference point.
Fig. 11 shows the RDs obtained, based on a moderate reference velocity of Vref 14 m/
s. As expected, an improvement in the overall accuracy of the power prediction,
particularly in moderate wind velocities, can be observed. The maximum RD in predicted
power is limited to 1% in the wind velocity range of 6 to 21 m/s.
In conclusion the combined method is very successful in reducing the computational
time with negligible reduction in the overall accuracy. As a matter of fact the application of
this method can be expanded to a complete aerodynamic optimisation procedure by
involving structural and material congurations and chord and pre-twist distributions, as
far Ras the
variations are limited to narrow intervals such that the variation of
r M  r
1

K max 0 K  r dr can be neglected.
7. Summary
The simulation of wind turbines with bendtwist adaptive blades is a coupled aerostructure procedure. The blade twist due to elastic coupling is a required parameter for
wind turbine performance evaluation and can be predicted through a FEA-based
structural analyser. FEA-based codes are far too slow to be useful in the aerodynamic
design/optimisation of a blade. In pursuit of developing an analytical model that gives the
induced twist, due to elastic coupling as a function of wind turbine run condition, rstly it

ARTICLE IN PRESS
A. Maheri et al. / Renewable Energy 32 (2007) 916930

929

has been shown that the effects of wind velocity, rotor speed and pitch angle variations on
the induced twist can all be represented by the magnitude of ap bending at the hub of the
blade. Furthermore, an analytical model that relates the induced twist to the ap bending
at the hub has been developed. The simulation method developed takes account of the
effect of dependency of the induced twist to the material and structural properties of the
blade by using the results of a FEA-based coupled aero-structure simulation at a reference
wind turbine run condition. A combined analytical/FEA-based CAS simulation employs
the induced twist distribution and the ap bending at the hub of the blade predicted
through a FEA-based CAS simulation at a reference wind turbine run condition to
determine the wind turbine performance at other wind turbine run conditions. This
reduces the computational time signicantly and makes the aerodynamic design/
optimisation of bend-twist adaptive blades practical.

Appendix A
Blade and wind turbine specications and material properties of the blade
Tables A.1, A.2

Table A.1
Blade and wind turbine specications
Rotor radius, R
Hub radius, Rhub
No of blades, B
Blades conning angle, d
Rotor angular velocity, O
Pitch angle
Rated power, Prated
Yaw
Tilt

13.757 m
1.184 m
2
71
53.333 rpm
1.21 TS
300 KW
0
0

Table A.2
Material properties of the blade
Shell thickness, t
Lay-up
Fibre orientation, y
E1
E2 E3
G12 G13
G23
v12 v13
v23

10 mm
mirror
201
141.96 GPa
9.79 GPa
6.14 GPa
4.83 GPa
0.50
0.42

ARTICLE IN PRESS
930

A. Maheri et al. / Renewable Energy 32 (2007) 916930

References
[1] Maheri A, Noroozi S, Toomer C, Vinney J. WTAB, a computer program for predicting the performance of
horizontal axis wind turbines with adaptive blades. Renewable Energy, in press, accepted, 2005, doi:10.1016/
j.renene.2005.09.023.
[2] Reheld LW, Atilgan AR, Hodges DH. Nonclassical behavior of thin-walled composite beams with closed
cross-sections. J Am Helicopter Soc 1990;35(2):4250.
[3] Smith EC, Chopra I. Formulation and evaluation of an analytical model for composite box-beams. J Am
Helicopter Society 1991;36(3):2335.
[4] Chandra R, Chopra I. Experimental and Theoretical Analysis of Composite I-Beams with Elastic Coupling.
AIAA J 1991;29(12):2197206.
[5] Chandra R, Chopra I. Structural behavior of two-cell composite rotor blades with elastic couplings. AIAA J
1992;30(12):291421.
[6] Kim C, White SR. Laminated composite thin- and thick-walled beam theory and coupled deformation.
Recent Adv Compos Mater 1995; ASME MD56.
[7] Kim C, White SR. Analysis of thick hollow composite beams under general loadings. Compos Struct
1996;34:26377.
[8] Kim C, White SR. Thick-walled composite beam theory including 3-d elastic effects and torsional warping.
Int J Solids Struct 1997;34(3132):423759.
[9] Bhaskar K, Librescu L. A geometrically non-linear theory for laminated anisotropic thin-walled beams. Int J
Eng Sci 1995;33(9):133144.
[10] Song O, Liberscu L. Structural Modelling and Free Vibration Analysis of Rotating Composite Beams with
Closed Cross Sections. J Am Helicopter Soc 1997;42(4):35869.
[11] Qin Z, Librescu L. On a shear-deformable theory of anisotropic thin-walled beams: Further contribution and
validation. J Compos Struct 2002;56:34558.
[12] Libove C. Stresses and rate of twist in single-cell thin-walled beams with anisotropic walls. AlAA J
1988;26(9):110718.
[13] Johnson ER, Vasilieev VV, Vasiliev DV. Anisotropic Thin Walled Beams with Closed Cross-sectional
Contours. In: Proceedings of the AIAA/ASME/ASCE/AHS/ASC 39th structures, structural dynamics and
materials conference, AIAA, Reston, VA, 1998. p. 500.
[14] Jung SN, Nagaraj VT, Chopra I. Rened structural model for thin- and thick-walled composite rotor blades.
AIAA J 2002;40(1):10516.
[15] Manseld EH, Sobey AJ. The bre composite helicopter blade, Part 1: Stiffness properties, Part 2: prospects
for aeroelastic tailoring. Aeronau Q 1979:41349.
[16] Bauld NR, Tzeng LS. A Vlasov Theory of Fiber Reinforced Beams with Thin-. Walled Open Cross Sections.
Int J Solids Struct 1984;20(3):27797.
[17] Bauchau OA. A beam theory for anisotropic materials. J Appl Mech 1985;52:41622.
[18] Berdichevsky VL, Armanios E, Badir A. theory of anisotropic thin-walled closed-cross-section beams.
Compos Eng 1992;2(57):41132.
[19] Badir AM, Berdichevsky VL, Armanios EA. Theory of composite thin-walled opened-cross-section beams.
In: Proceedings of the AIAA/ASME/ScCE/ASC 34th structures, structural dynamics and material
conference, AIAA, Washington DC, 1993; 276170.
[20] Reheld, LW. Design analysis methodology for composite rotor blades. In: Proceedings of the seventh DoD/
NASA conference on brous composites in structural design, Denver, CO, Jun 1720, 1985.
[21] Cesnik CES, Sutyrin VG, Hodges DH. A rened composite beam theory based on the variational
asymptotical method. In: Proceedings of the 34th AIAA/ASME/ASCE/AHS/ASC structures, structural
dynamics and materials Conference, La Jolla, CA, 1922 April, 1993. p. 271020.
[22] McCarthy TR, Thomas R, Chattopadhyay A. A rened higher order composite BoxBeam Theory.
In: Proceedings of the 37th AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics and materials
conference, Salt Lake City, UT, Apr 1517, 1996.
[23] Jung SN, Nagaraj VT, Chopra I. Assessment of composite rotor blade modelling techniques. J Am
Helicopter Soc 1999;44(3):188205.
[24] Buhl ML, NWTC Design Codes, WT_Perf Version 3.1 http://wind.nrel.gov/designcodes/simulators/wtperf/.
Last modied 17December2004; accessed 10 January 2005.

Das könnte Ihnen auch gefallen