You are on page 1of 20

Fluid Structure Interaction of Wind Turbine Airfoils

Ouahiba Guerri
1*
, Aziz Hamdouni
2
and Anas Sakout
2
1.
Centre de Dveloppement des Energies Renouvelables BP 62, Bouzarah, CP 16 340, Alger, Algeria.
Email: o_guerri@yahoo.com
2.
LEPTIAB, Dpart. Gnie Civil et Mcanique, Ple Sciences et Technologies, Univ. de La Rochelle,
Av. Michel Crpeau, 17042, La Rochelle, Cedex 01, France. Email: ahamdoun@univ-lr.fr and
asakout@univ-lr.fr
*
Corresponding author
WIND ENGINEERING VOLUME 32, NO. 6, 2008 PP 539557 539
ABSTRACT
Flow induced vibrations of two airfoils used in wind turbine blades are investigated by a
strong coupled fluid structure interaction approach. The method is based on a general
Computational Fluid Dynamics (CFD) code that solves the Navier-Stokes equations defined
in Arbitrary Lagrangian Eulerian (ALE) coordinates by a finite volume method. A
straightforward technique is implemented in a user subroutine for the coupling of the CFD
code to a structural dynamics program to determine the airfoil displacements due to the
aerodynamics forces and for updating the grid at each time step. Simulations are carried out
for a free pitch oscillating airfoil and for a combined pitch and vertical oscillating airfoil.
Beforehand, the problem of the flow around a forced pitch oscillating airfoil is considered to
check the reliability of the moving mesh technique and the CFD computations.
All computations are performed in 2D, incompressible and low Reynolds number flows.
Keywords: Aeroelasticity. Unsteady aerodynamics. Incompressible Navier-Stokes
equations. ALE formulation. Moving mesh.
1. INTRODUCTION
Wind turbines are subjected to a hard environment such as (i) the atmospheric turbulence,
(ii) the ground boundary layer, (iii) the rapid variations in wind speed and direction and (iv)
the tower shadow for downwind turbine. This stochastic inflow, associated to the architecture
of the rotor leads to a 3D unsteady aerodynamics and dynamic stall [1, 2]. The dynamic stall
results in fluctuating blade force and blade oscillations, known as aeroelastic phenomena.
Problems of aeroelastic stability can be encountered on the new large wind turbine blades as
well as on the rotating wind turbine blades and on parked wind turbine blades at high wind
speeds [3].
Most of the wind turbine aeroelastic analyses were performed using engineering methods
where the blade forces were often computed by the BEM theory and the dynamic stall was
modeled with empirical models such as the ONERA or the Beddoes-Leishman models. The
dynamic response of the wind turbine was thereafter determined using structural
computational tools [3-5]. Displacements of the structure were then computed after
convergence (or partial convergence) of the aerodynamic part of the aeroelastic code. The
problems of viscous fluid flow and elastic body deformation were then studied separately.
These weak coupled approaches are limited to small deformations and low non-linearity. In
most aeroelastic problems, the interaction between these two media and the modeling of the
unsteady aerodynamics and dynamic stall are of great importance for the aeroelastic
stability study. We need then to perform aeroelastic computations by means of a strong
coupled approach with Fluid Structure Interaction (FSI) techniques where the aerodynamic
forces are computed from the solution of the time-accurate Navier-Stokes equations and used
for the solution of the rotor dynamic equations to determine the response of the structure at
each time step.
FSI techniques have been widely used in many industrial problems [6, 7], for aerospace
applications where most papers are applied to compressible flows [8]. It has also been shown
that compressibility effect plays a significant role in aeroelastic stability limits. As wind
turbines operate in an incompressible environment, specific studies have to be performed for
wind turbine blades. To the authors knowledge, the main contribution with an FSI approach
applied to wind turbine blades has been carried out in the frame of the European project
KnowBlade where Ellipsys3D, an in house Computational Fluid Dynamics (CFD) code, and a
structural code have been used to simulate flap-lead/lag vibrations of a wind turbine blade
[9]. Recently, Svacek et al. [10] presented the study of the classical flutter by a FSI approach
where the incompressible fluid equations were solved by the finite element method using an
Arbitrary Lagrangian Eulerian (ALE) formulation of the Navier Stokes equations.
In this paper, the problem of classical flutter phenomenon deals with a strong coupled
method where the dynamic response of the wind turbine blade is determined in time accurate
sequences. The aim of this study is to propose a straightforward technique to be used with
general computer tools. The numerical simulations are performed using StarCD, an industrial
CFD code which is coupled to a computational structural program for the solution of the
dynamic equations of the blade. A technique is implemented in a user subroutine for the
coupling of both codes and for updating the grid. The viscous flow induced vibrations on two
airfoils are then simulated. First of all, the moving mesh technique is applied to a forced
oscillating airfoil to check the reliability of the CFD computations. The hysteresis loops of the
dynamic stall phenomenon are then highlighted. All computations are carried in a 2D,
incompressible and laminar flow. In addition, for one case study, comparison is made between
laminar and turbulent computations carried out with the solution of the Reynolds Averaged
Navier-Stokes equations (RANS).
2. NUMERICAL APPROACH
The fluid flow equations, the dynamic equations of the oscillating airfoil, the coupling
approach and the equations describing the moving mesh technique are summarized in the
following.
2.1. The fluid equations
The fluid governing equations are described in ALE coordinates where the grid is considered
as a referential frame moving with an arbitrary velocity. A detailed description of the
derivation of the ALE formulation can be found in the literature (see e.g. Refs. 11 or 12). There,
the expression of the Navier-Stokes equations defined in ALE coordinates is given.
Let
f
R
2
a 2D spatial domain occupied by the fluid. The incompressible fluid equations
in ALE form are written as, in tensor notation:
(1)

=
u
x
i
i
0
540 FLUID STRUCTURE INTERACTION OF WIND TURBINE AIRFOILS
(2)
where t is the time, x
i
the Cartesian coordinate of a point of
f
, u
i
the absolute velocity
component in the direction i, u
cj
the velocity of the moving grid, p the pressure, the fluid
density, g the determinant of the metric tensor and
ij
the stress tensor components defined
as in the case of laminar flows,
(3)
with , the molecular dynamic fluid viscosity. For turbulent flows, u
i
and p denote their
ensemble averaged values and the stress tensor
ij
is expressed as:
(4)
where u
i
denotes the fluctuating part of the instantaneous velocity and the over bar
denotes the ensemble averaging process. The term u
i
u
j
is the additional Reynolds stresses
due to turbulence that are described here by the SST k-turbulence model of Menter [13]. This
model is a combination of the k- and k- models which uses the k- model near the wall but
switches through a function to a k- model when away from the wall, closer to the upper limit
of the boundary layer. Consequently, the k- equation is transformed into a k- formulation.
The SST k- model of Menter has shown to give results for flow with strong adverse pressure
gradients that are far higher than those obtained with either the original k- model and its
variants or the k- model [14, 15].
When the RANS equations are written in ALE formulation, the terms u
i
k/x
i
and u
i
k/x
i
of
the transport equations for the turbulence energy k and the turbulent specific dissipation
are expressed as:
and (5)
Geometric conservation law. An additional equation called the Space - Conservation Law
(SCL) is enforced to avoid errors that could be induced by the moving grid cells:
(6)
where V is the cell volume, S the surface enclosing the volume V and n is the surface vector.
This equation describes the conservation of space when the control volume changes its shape
and/or position with time [16].
Boundary conditions. An overview of the computational domain is shown in Fig. 1. The
boundary conditions are defined as follows. The inflow condition is applied to the West
boundary having defined the free-stream velocity. The outflow condition is defined at the East
d
dt
dV u ndS
V
cj
S

=0

( )

u u
x
i ci
i

( )

u u k
x
i ci
i

ij
i
j
j
i
i j
u
x
u
x
u u =

\
)


' '

ij
i
j
j
i
u
x
u
x
=

\
)



1
g
g u
t
u u u
x x
p
x
i
i j cj
j
ij
j

( )



( ) ( )



( )



ii
WIND ENGINEERING VOLUME 32, NO. 6, 2008 541
boundary. A symmetry condition is applied to the South and the North boundaries. At the
airfoil surface, the fluid velocity is equal to the airfoil speed:
(7)
where u is the fluid velocity adjacent to the airfoil face and v
b
is the velocity of the airfoil.
Figure 1: Block topology
Computation of the aerodynamic forces. The resultant of the aerodynamic force is
obtained from the solution of the Navier-Stokes equations by integrating the pressure p and
the shear stress
w
over the blade surface S :
(8)
where:
n

is the outward-pointing vector normal to the airfoil surface


N is the number of the airfoil faces
p
i
,
wi
and S
i
are the wall cell pressure, shear stress and elementary area
respectively
n

i
is the outward-pointing vector normal to S
i
The drag force F
x
and the lift force F
y
are computed as:
(9)
where X

and Y

are the unit vectors in the parallel and vertical directions of the flow,
respectively. The lift and drag coefficients C
L
and C
D
are then derived with:
F F Y
y i
i
N
=
=


1
F F X
X i
i
N
=
=


F p n n dS p n n S
w
S
i i wi i i
i
N
=
( )
=
( )

=

1

u v
b
=
542 FLUID STRUCTURE INTERACTION OF WIND TURBINE AIRFOILS
and (10)
where A
ref
is the plan form area of the blade.
Algorithm and discretisation schemes. The Navier Stokes equations are solved by a finite
volume method. The PISO algorithm is applied for the solution of the coupled pressure
velocity equations. The first order differencing UPWIND scheme is used for the discretisation
of the convection-diffusion terms. Higher order schemes are more accurate; however
instability problems can be encountered when the Peclet number is too high. As it is important
to distinguish the numerical instabilities of the solution from the unsteady flow associated
with the airfoil motions, the computations are performed with the first order UPWIND
scheme, less accurate but more stable. The temporal discretisation is performed with the
implicit - scheme with = 0.8.
2.2. The dynamic equations
Two ways are applied for the simulation of the flow around the vibrating airfoils: (i) forced
oscillations and (ii) free oscillations.
Forced oscillating airfoil: The instantaneous angle of attack of the airfoil in forced pitch
oscillations is given by the relation:
(11)
where
0
is the mean angle of attack of the oscillating airfoil,
m
is the oscillation amplitude
and = 2f with f the frequency of oscillation.
As the airfoil displacements are imposed, the coupling procedure is reduced to the
updating of the computational grid in accordance with the airfoil motion at each time step.
Free oscillating airfoil: It is assumed that the airfoil is an elastic body and the airfoil
elasticity is described by a simple two degrees of freedom model. The dynamic equations of
the airfoil are defined from the Lagrange equations. In the case of an airfoil in pitch oscillations
(or torsion around the elastic axis EO) and in flap-wise oscillations (or vibrations in the
vertical direction) as drawn in Fig. 2, the non linear equations that describe the airfoil
oscillations are written as:
(12a)
(12b)
where , y,
.
, y
.
,
..
and y
..
are the airfoil rotational and vertical displacement, velocity and
acceleration respectively; F
y
is the vertical component of the aerodynamic force; M
0
the
torsion moment; m is the airfoil mass; S

mr
G
with r
G
the distance between the centre of
gravity and the elastic axis EO; J

the inertia moment around the elastic axis EO, k

and k
y
are
the torsion and bending stiffness; C

and C
y
are defined as:
(13a)
(13b) C m
y y y
=2
C J

=2
J C S y k M

=
0
m y C y S k y F
y y y
=


( ) sin( ) t t
m
=
0

C
F
A V
D
X
ref
=

1 2
2

C
F
A V
L
Y
ref
=

1 2
2

WIND ENGINEERING VOLUME 32, NO. 6, 2008 543


Figure 2: The airfoil scheme
with

and
y
the damping coefficients, and are the airfoil
natural frequencies.
Time integration of the dynamic equations. The algorithm used for the solution of the
dynamic equations is based on a trapezoidal time advancement scheme. Equations (12a) and
(12b) that describe the airfoil oscillations are first re-written in a matricial form:
(14)
where q is a generalized coordinate, q
.
and q
..
the first and second derivatives with respect
to time. M, C and K are the mass, damping and stiffness matrices respectively and R the
resultant force matrices. Then, given t the time step, the terms q
.
and q
..
are discretised using
a trapezoidal time advancement scheme:
(15)
(16)
The substitution of these discretised forms of q
.
and q
..
in the equation (14) leads to the
applied algorithm written as, at each time step:
(17)
(18)
(19)
All computations are started with an initial displacement and zero vibration velocity.
R R M
t
q
t
q q C
t
q q
e
n n n n n n n
=

l
l
l
l

1
2
4 4 2

\
)

K
t
M
t
C K
e
=
4 2
2

K q R
e
n
e
n
=
1 1
q
t
q q q
n n n n
=
( )

1 1
2

q
t
q q q
n n n n
=
( )

1 1
2

M q C q K q R

y y
k m =

= k J
544 FLUID STRUCTURE INTERACTION OF WIND TURBINE AIRFOILS
2.3. The coupling scheme
At each time step, the angular and vertical displacements of the airfoil surface are derived
from the solution of the dynamic equations with:
(20)
(21)
where the superscripts nand n+1 refer to times t
n
and t
n+1
respectively. and y are then
used to move the mesh nodes.
This explicit coupling procedure is first order accurate scheme and it does not conserve
energy at the moving fluid - solid interface. It is well known that implicit algorithms are more
suitable [11]. However, implicit schemes are more computationally expensive and heavy to
implement and to integrate in a commercial CFD code. Therefore, they are not used in this
study.
2.4. The moving mesh technique
The airfoil is located at the centre of an O-H block structured grid (Fig. 1) that allows a meshing
technique easy to implement. The moving mesh method is based on algebraic interpolations.
Given:
i. R
n
i
and
n
i
the cylindrical coordinates of the vertices V
i
at time t
n
;
ii. x
n
i
and y
n
i
the related Cartesian coordinates;
iii. and y the airfoil displacements that are calculated with the solution of the
dynamic equations;
The displacement of the vertices is performed as follow:
The vertices of the circular sub-domain of radius R
1
move at the airfoil velocity:
(22)
(23)
(24)
For the vertices of the annular sub-domain delimited by the circles of radius R
1
and
R
2
, the applied relations are:
(25)
(26)
where:
(27)
(28) y R y
n n n
1
1
1 1
1
=
( )
sin

i
n
i
n i
R R
R R

\
)

1 1
2 1
1
y R y y
N
i
n
i
n
i
n n
R



=
( )

( )
1
1 1
1
2 1
1
1
sin
x R
i
n
i
n
i
n
=
( )
1 1
cos
y R y
i
n
i
n
i
n
=
( )

1 1
sin
x R
i
n
i
n
i
n
=
( )
1 1
cos

i
n
i
n
=
1

y y y
n n
=
1
=
n n

1
WIND ENGINEERING VOLUME 32, NO. 6, 2008 545
(29)
N
R
is the number of radial cells in the annular sub domain delimited by the radius R
2
and R
1
(see in Fig. 3 the illustration of the vertices notation).
The vertices of the outer domain with radius R
i
R
2
are stationary.
With this moving mesh algorithm, the mesh distortions are small and the original mesh
quality is preserved. Given an appropriate choice of the radius R
1
and R
2
, this technique can
also be applied even for large displacements of the airfoil. This meshing technique is
implemented in a user subroutine called by the CFD code at the beginning of each time step.
Figure 3: Sketch of the vertices notation
3. APPLICATION
Two airfoil models have been selected for this study: a symmetrical NACA 0012 airfoil used on
vertical axis wind turbine blades and a cambered NACA 63
2
415 used on large horizontal axis
wind turbine blades.
In all computations, the properties of the fluid are those of air with a density = 1.2 kg/m
3
and a dynamic viscosity coefficient = 1.8 10
-5
Pas. The stability of the solution for the flow
field is checked with the maximal value of the CUNO parameter, a number identical to the CFL
number but whose calculation is based on the face fluxes. It is recommended to ensure that the
peak values of the CUNO do not exceed 300 (http://www.adapco-online.com/). However, this
criterion proved to be insufficient and the time step has been fixed so as to get CUNO values
lower than 20.
Computational mesh. The computational domain extends to a distance equal to 8.1 C at
upstream and the outlet boundary is located at a distance equal to 15 C . The North and South
boundaries are located at 10 C respectively. The grid is an O-H block structured mesh and
consist of 111000 cells with 720 cells at the periphery of the sub-domain of radius R
1
, 500 cells
around the airfoil and approximately 10 grid points vertically in the airfoil boundary layer
(Fig. 4).
V
i-1
(R
i-1
,
i-1
)
V
1
(R
1
,
1
)
V
2
(R
2
,
2
)
V
i
(R
i
,
i
)
y R
2 2 2
=
( )
sin
546 FLUID STRUCTURE INTERACTION OF WIND TURBINE AIRFOILS
Figure 4: The mesh near the airfoil
3.1. NACA 0012 in forced pitch oscillations
These computations are performed for an airfoil with a chord length C = 1 m. The axis of
rotation of the airfoil is located at 25 % chord length from the leading edge. The computations
are performed with the chord Reynolds number Re = U

C/ = 10
4
and a constant
integration time step t = 10
-2
s. The instantaneous value of the angle of attack is given by
equation (10) with
0
= 0 and
m
= 10 . Two cases are selected for the oscillation frequency
corresponding to k
*
= 0.19 and 0.45 where k
*
= C/2U

is the dimensionless reduced frequency


based on half chord and the free-stream velocity. At time t = 0 s, the airfoil incidence is zero.
First, computations are carried out for an airfoil with a fixed angle of attack to get the initial
flow field. The simulations for the airfoil in forced oscillations are initiated when the solution
becomes stable.
All results are depicted as a function of the dimensionless time t
*
= t/T where T = 2/ is
the period of oscillation. The time functions of the force and moment coefficients are shown in
Fig. 5, compared to the oscillations of the airfoil:
The lift and moment coefficients have the same frequency as that of the airfoil
oscillation. However, the force and moment coefficient time histories are not pure
sinusoidal. Higher harmonics from the fundamental pitching airfoil frequency can
be observed
For the reduced frequency k
*
= 0.19, a slight lag has been observed in the phase
difference between the lift force and the angle of attack. The variation of the lift and
moment are in phase with the airfoil oscillations when k
*
= 0.45. Consequently, an
increase is expected in the phase lead as k
*
increases. Similar behavior has been
found with the linear theory of Theodorsen.
The drag frequency is one half the lift frequency. Similarly, the behavior for the drag
history was found by Yang et al. [17] with Euler computations of the flow over an
oscillating airfoil. In Ref. 18 (p. 47), it was reported that this has been shown in
experiments and that it is a consequence of the geometry of the vortex street. These
results remind us that the flow around a circular cylinder when a regular Karman
street is observed (see. eg. [19]).
WIND ENGINEERING VOLUME 32, NO. 6, 2008 547
The hysteresis loops C
L
() and C
D
() are shown in Fig. 6. When the airfoil is oscillating,
both lift and drag coefficients increase compared to the 2D steady state. The mean values of
the drag coefficient corresponding to = 0 are C
D
= 0.0775 and 0.0684 with k
*
= 0.19 and 0.45
respectively. In previous computations of the flow around a fixed NACA 0012 airfoil at zero
incidence, it was found that C
D
= 0.0374. The lift coefficients vary roughly between 0.37 and
0.12 with k
*
= 0.19 and 0.45 respectively. For these two case studies, the lift hysteresis loops are
then widen when the reduced frequency is lower. It is found a small decrease of C
L max
when
the reduced frequency k
*
increases. Similar results have been found with the linear theory of
Theodorsen.
Figure 5: Time histories of the lift, drag and moment coefficients for the forced oscillating NACA 0012
airfoil
548 FLUID STRUCTURE INTERACTION OF WIND TURBINE AIRFOILS
Figure 6: Hysteresis loops of the lift, drag and moment coefficients for the forced oscillating NACA
0012 airfoil
3.2. NACA 0012 in free pitch oscillations
The model is a NACA 0012 airfoil with a chord length C = 0.12 m and a span of 0.50 m that
oscillates around an axis located at 25% of C from the leading edge. The airfoil natural
frequency is

= 57 and the damping ratio is = 0.05. These computations have been


performed with the chord Reynolds number Re = 76 000 and the initial angle of attack

init
= 10. The initial flow field is determined by simulations that have been carried out for an
airfoil at the fixed 10 incidence until time t = 0.60 s (corresponding to the dimensionless time
t
*
= U

t/C 48), when the time wise variations of the force coefficients become periodic. The
airfoil is then released in the fluid and the computations are performed in a strong coupled
Fluid Structure Interaction.
WIND ENGINEERING VOLUME 32, NO. 6, 2008 549
The time-wise variation of the airfoil position is shown on Fig. 7 (a). The curve shows that
the airfoil oscillations are damped and that the pitch angle tends to a zero incidence. This can
be seen on Fig. 7 (b) that shows the variation of the torsion moment as a function of the angle
of attack where counter-clockwise rotation indicates positive aerodynamic damping. The
time history of the lift coefficient, not shown here, shows that the lift coefficient decreases
quickly as soon as the profile is free oscillating, until reaching a zero lift coefficient. This
positive aerodynamic damping was expected as the elastic axis is located at the aerodynamic
centre: the pitching moment always lags the airfoil motion and the aerodynamic damping in
pitch tends to suppress torsional oscillations at all frequencies [20].
Figure 7: Time histories of the pitch oscillations and moment versus the pitch angle for the free
oscillating NACA 0012 airfoil
550 FLUID STRUCTURE INTERACTION OF WIND TURBINE AIRFOILS
3.3. NACA 632 415 airfoil in pitch and vertical oscillations
The model is similar to that of Ref. [10] and has a chord length C = 0.30 m, a span of 0.50 m and
the following elastic parameters:
m= 8.6622 10
-2
kg , S

= 7.79673 10
-4
kgm, J

= 4.87291 10
-4
kgm
2
k
y
= 105.109 N/m, k

= 3.695582 Nm/rad
C
y
= 105.109 and C

= 3.695582 10
-3
The elastic axis EO is localized at 40 % of Cfrom the leading edge and the centre of gravity
G is at 37 % from the leading edge. Computations have been started with the airfoil initial
position y
init
= +0.050 m and
init
= 6, where y
init
and
init
are the initial vertical and angular
positions, respectively. The simulations have been performed with the free stream velocities
U

= 2, 26 and 45 m/s. With U

= 45 m/s, the simulations have been carried out for laminar and
turbulent flows for comparison purposes.
As previously, the initial flow-field is determined with computations carried out for a fixed
airfoil at the above initial position. The airfoil is then released in the fluid and the computations
have been performed in FSI. The obtained results have been depicted according the speed
index, a dimensionless parameter defined as where and
is the mass ratio. The aim of this was to verify if the flutter point will be
attained or not. With the applied values of the airfoil elastic parameters, the velocities U

= 2,
26 and 45 m/s correspond to the speed indices V* = 0.15, 1.96 and 3.40, respectively.
The contours of velocity magnitudes obtained with the different values of V* are shown in
Figs. 8 and 9. U
max
is the maximal velocity magnitude obtained over the airfoil surface. At time
t = 0 s, when the airfoil is at rest, the ratio U
max
/U

increases with the free stream velocity (Fig.


8). When U

= 2 m/s, it is found that the vortex streets are larger in the near wake of the airfoil
than with the higher free stream velocities. On Fig. 9, the results are also referenced to the
dimensionless time t
*
= U

t/C. When V* = 0.15 (U



= 2 m/s), the ratio U
max
/U

increases at the
beginning of the oscillations and remains important at time t
*
= 7.0 with a value U
max
/U

= 2.10.
However, with V* = 1.96 and 3.40 (U

= 26 and 45 m/s, respectively) the ratio U
max
/U

decreases quickly as soon as the airfoil is released in the fluid. At V* = 3.40 , U


max
/U

= 1.30 at
time t
*
= 7.5 and with turbulent computations, lower values of the velocity magnitude have
been found (Figs. 9 (e) and 9 (f)).
The time-wise variations of the force components are shown in Fig. 10. Higher values of the
force coefficients are found whenU

= 2 m/s. In addition, laminar and turbulent computations
resulted in different values of the aerodynamic forces with lower lift and drag coefficients for
turbulent computations.
As expected, the CFD results are sensitive to the free stream velocity value. However, it
has been found that the value of the inlet fluid velocity does not influence the airfoil responses.
The phenomenon of divergence has not been observed. On Fig. 11, it can be observed that after
some time, the pitch motion becomes periodic while the vertical displacement shows a
positive damping. In addition, due to the frequency ratio
y
/

= 0.4, the vertical motion has a


smaller frequency than the pitch motion. The non-influence of the free stream velocity on the
dynamic response of the airfoil is due to the airfoil elasticity parameters and damping and to
the low values of the speed indices. In addition, due to different initial airfoil positions, our
computed airfoil displacements are different from those of Ref. [10]. Indeed, it was shown in [21]
that the motion of oscillating airfoils is stable or unstable depending on both values of the
speed index V
*
and the initial position.

s
m C =
( )
4
2
U U C *=
( )

V U
s
* * =
WIND ENGINEERING VOLUME 32, NO. 6, 2008 551
Other simulations whose results are presented in another paper have been performed for
an airfoil without dampening and with different value of the airfoil natural frequency. The
influence of the free-stream velocity on the airfoil oscillations is then more significant.
Figure 8: Contours of the velocity magnitude around the free oscillating NACA 63
2
415 airfoil at time t = 0 s
552 FLUID STRUCTURE INTERACTION OF WIND TURBINE AIRFOILS
Figure 9: Contours of the velocity magnitude around the free oscillating NACA 63
2
415 airfoil at
selected time
WIND ENGINEERING VOLUME 32, NO. 6, 2008 553
Figure 10: Time histories of the lift coefficient for the free oscillating NACA 63
2
415 airfoil
554 FLUID STRUCTURE INTERACTION OF WIND TURBINE AIRFOILS
Figure 11: Time histories of the pitch and vertical oscillations for the free oscillating NACA 63
2
415 airfoil
4. SUMMARY AND CONCLUSION
Flow induced vibrations have been investigated for two airfoils used on wind turbine blades.
The method is based on a strong coupled Fluid Structure Interaction approach where the
dynamic blade response due to the fluid forces is determined in time accurate sequence. CFD
computations have been performed for incompressible and low Reynolds number flows, using
WIND ENGINEERING VOLUME 32, NO. 6, 2008 555
a commercial CFD code that solves the Navier Stokes equations defined in ALE coordinates.
A straightforward coupling and meshing technique was implemented in a user subroutine
called by the CFD code at each time step.
The coupling method was successfully applied for the computation of the viscous laminar
flows around (i) a forced pitch oscillating airfoil where the hysteresis loops of the dynamic
stall were highlighted, (ii) a free pitch oscillating airfoil and (iii) an airfoil in combined pitch
and vertical oscillations (classic flutter). As wind turbines operate in turbulent environment,
computations were also performed with the resolution of the incompressible Reynolds
averaged Navier-Stokes equations for one case of study. Laminar and turbulent flow
computations resulted in different values of the aerodynamic force and similar airfoil
response.
ACKNOWLEDGMENTS
The first author is grateful for the support provided by the La Rochelle University and the
Rgion Poitou - Charentes (France) in form of grants.
REFERENCES
1. Huyer, S.A., Simms, D. and Robinson, M.C., 1996, Unsteady aerodynamics associated
with a horizontal-axis wind turbine, AIAA journal, vol. 34, N7, 1410 1419
2. Leishman J.G., 2002, Challenges in modeling the unsteady aerodynamics of wind
turbines, AIAA paper N 2002-0037
3. Hansen, M.O.L., Sorensen, J.N., Voutsinas, S., Sorensen, N. and Madsen, H.Aa., State of the
art in wind turbines aerodynamics and aeroelasticity, 2006, Progress in Aerospace
Sciences, vol. 42, pp. 285-330
4. Lindenburg, C. and Snel, H., 2003, Aeroelastic stability analysis tools for large wind
turbine rotor blades, ECN report (http://www.ecn.nl)
5. Rasmussen F., Hansen, M.H., Thomsen K., Larsen T.J., Bertagnolio F., Johansen J., Madsen
H.A., Bak C. and Hansen A.M., 2003, Present Status of Aeroelasticity of Wind Turbines,
Wind Energy, vol. 6, 213-228
6. Souli M. and Zolesio J.P., 2001, Arbitrary Lagrangian - Eulerian and free surface methods
in fluid mechanics, Comput. Methods Appl. Mech. Engrg., vol. 191, pp. 451 - 466
7. Longatte, E., Bendjeddou, Z. and Souli, M., Methods for numerical study of tube bundle
vibrations in cross-flows, 2003, Journal of Fluids and Structures, vol. 18, 513-528
8. Chen X.Y. and Zha G.C., 2005, Fully coupled fluid - structural interactions using an
efficient high resolution upwind scheme, J. Fluids and structures, 1105 - 1125
9. Bertagnolio F., Sorensen N.N., Johansen J., Politis E.S., Nikolaou I.G. and Chaviaropoulos
P.K., 2004, Wind turbine blade aerodynamics and aeroelasticity (KNOW-BLADE),
RISOE report N RIS-R-1492 (EN)
10. Svacek P., Feistauer M. and Horacek J., 2007, Numerical simulation of flow induced
airfoil vibrations with large amplitudes, J. Fluids and Structures, vol. 23, pp. 391- 411
11. Abouri D., Parry A., Hamdouni A. and Longatte E., 2006, A stable fluid - structure
interaction algorithm: application to industrial problems, ASME J. Pressure Vessel
Technology, vol. 128, pp. 516-524
12. Souli, M., Ouahsine, A. and Lewin L., 2000, ALE formulation for Fluid Structure
Interaction problems, Comput. Methods Appl. Mech. Engrg., vol. 190, 659 - 675
556 FLUID STRUCTURE INTERACTION OF WIND TURBINE AIRFOILS
13. Menter F.R., 1994, Two-equation eddy-viscosity turbulence models for engineering
applications, AIAA Journal, vol. 32, N 8, pp. 1598 1605
14. Guerri O., Bouhadef K. and Harhad A., 2006, Turbulence flow simulation of the NREL
S809 airfoil, Wind Engineering, pp. 287- 301
15. Catalano, P. and Amato, M., 2003, An evaluation of RANS turbulence modelling for
aerodynamic applications, Aerospace Science and Technology, vol. 7, 493 - 509
16. Ferziger J.H. and Peric M, 2002, Computational methods for fluid dynamics, Springer-
Verlag (3rd Ed.)
17. Yang, S., Luo, S., Liu, F. and Tsai, H.M., Computations of the flows over flapping airfoil by
the Euler equations, 2005, AIAA 43 rd Aerospace Sciences Meeting, Reno, NV
18. Brevins, R.D., 1993 Flow - induced vibrations, 2nd ed., Van Nostrand Reinhold, New York
(ISBN 1 - 57524 - 183 - 8)
19. Placzek, A., Sigrist, J.F. and Hamdouni, A., 2009, Numerical simulation of an oscillating
cylinder in a cross-flow at low Reynolds number: Forced and free oscillations,
Computers and Fluids, vol. 38, 80 - 100
20. McCroskey, W.J., 1982, Unsteady airfoils, Ann. Rev. Fluid Mech., vol. 14, 285 - 311
21. Mahajan, A.J., Kaza, K.R. and Dowell, E.H., 1993, Semi-empirical model for prediction of
unsteady forces on an airfoil with application to flutter, Journal of Fluids and
Structures, vol. 7, 87 - 103
WIND ENGINEERING VOLUME 32, NO. 6, 2008 557

You might also like