2D Numerical Simulations of

Blade-Vortex Interaction in a
E. Amet
Laboratoire des Ecoulements Géophysiques
Industriels (LEGI),
Darrieus Turbine
Grenoble 38041, France
The aim of this work is to provide a detailed two-dimensional numerical analysis of the
physical phenomena occurring during dynamic stall of a Darrieus wind turbine. The flow
T. Maître is particularly complex because as the turbine rotates, the incidence angle and the blade

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021

Research Scientist
Institut National Polytechnique de Grenoble Reynolds number vary, causing unsteady effects in the flow field. At low tip speed ratio,
(INPG), a deep dynamic stall occurs on blades, leading to large hysteresis lift and drag loops
Grenoble 38031,France (primary effects). On the other hand, high tip speed ratio corresponds to attached bound-
ary layers on blades (secondary effects). The optimal efficiency occurs in the middle
range of the tip speed ratio where primary and secondary effects cohabit. To prove the
C. Pellone capacity of the modeling to handle the physics in the whole range of operating condition,
Research Scientist it is chosen to consider two tip speed ratios (␭ ⫽ 2 and ␭ ⫽ 7), the first in the primary
effect region and the second in the secondary effect region. The numerical analysis is
J.-L. Achard performed with an explicit, compressible RANS k-␻ code TURBFLOW, in a multiblock
Research Scientist structured mesh configuration. The time step and grid refinement sensitivities are exam-
ined. Results are compared qualitatively with the visualization of the vortex shedding of
Centre National de la Recherche Scientifique Brochier (1986, “Water channel experiments of dynamic stall on Darrieus wind turbine
(CNRS), blades,” J. Propul. Power, 2(5), pp. 445–449). Hysteresis lift and drag curves are com-
Grenoble 38042, France pared with the data of Laneville and Vitecoq (1986, “Dynamic stall: the case of the
vertical axis wind turbine,” Prog. Aerosp. Sci., 32, pp. 523–573).
关DOI: 10.1115/1.4000258兴

Keywords: VAWT, RANS modeling, dynamic stall, hysteresis

1 Introduction speed ratio and for solidity parameter below 0.2, because the
Glauert momentum theory is not valid outside this range 关5,6兴.
Basically, wind turbines can be divided in two groups depend-
Also, the flow is assumed quasisteady and predicted blade loads
ing on their rotational axis direction: horizontal axis wind turbine
are inaccurate. The normal component to the free-stream flow is
共HAWT, or axial flow turbine兲, and vertical axis wind turbine
neglected. The momentum models are inadequate to model the
共VAWT, or cross-flow turbine兲. VAWT has a smaller efficiency
wake developed downstream of a wind turbine farm 关7兴.
than HAWT, but did not benefit from the years of development The vortex methods applied to rotor wake problems ranges
undergone by the HAWT. A typical Darrieus turbine is a fixed from “prescribed” 共fixed兲 to “free” vortex techniques. The devel-
pitch VAWT with straight blades and its simplicity lies in the opment of a vortex wake model is based on the assumption of an
absence of the yawing mechanism. On the other hand, the rotor incompressible potential flow with all vorticity being assumed
aerodynamics is far from being simple since it involves highly concentrated within the vortex filaments 关8兴. In the vortex models,
unsteady flow fields. This unsteadiness is due to large variations in based on the vorticity equation, the blade element is replaced with
the angle of attack on the blades during their rotation. a lifting line or surfaces that depend on the azimuthal position.
Figure 1 presents a typical performance curve in term of power The advantage is that the pressure does not explicitly appear in the
coefficient C P versus the tip speed ratio ␭. This curve is divided in vorticity equation and the pressure field is not needed to compute
three regions according to the importance of the primary 共dy- for the velocity field 关5兴. The fluid velocity is obtained by adding
namic兲 and secondary 共viscous兲 effects. At low tip speed ratios, the undisturbed fluid velocity and the vortex induced velocity. On
blades undergo deep stall, leading to large dynamic loadings. the other hand, lift and drag static coefficients are needed to com-
Losses come from the large scale vortices detached from blades pute the bound element vortex strength. The vortex models are
and convected in the wake. These dynamic effects are strongly relatively higher in computational overhead, mainly because the
influenced by the solidity ␴ of the turbine. At higher tip speed Biot–Savart law is computed for every vortex shed in the flow but
ratio, the shaft, the added mass effects, and the viscosity friction they are more realistic than the momentum models.
prevail due to smaller angles of attack. Between the two zones, Ponta and Jacovkis 关9兴 combined a free vortex model with a
there is a narrow optimal transition zone, corresponding to a bal- finite element analysis 共FEVDTM兲. The free vortex model acts as
ance between primary and secondary effects. a macromodel, whose results are used as a boundary condition on
Since 1970, several aerodynamic prediction models have been the boundary of the micromodel 共the finite elements area which
formulated for Darrieus machines. The momentum models 共single include the airfoil兲. The bound vortex strength is obtained by in-
streamtube model 关1兴, multiple streamtubes model 关2兴, double- tegrating the velocity obtained with the finite element analysis.
multiple streamtubes model 关3,4兴, and other variants are all de- Once converged, the surface pressure distribution over the airfoil
rived from the actuator disk theory. They are limited to small tip is obtained by integrating the momentum equation model. Know-
ing the pressure and velocity distributions, a boundary layer
model is used to compute for the viscous shear stress over the
Contributed by the Fluids Engineering Division of ASME for publication in the
JOURNAL OF FLUIDS ENGINEERING. Manuscript received July 2, 2008; final manuscript
airfoil surface. Then, the instantaneous forces are calculated by
received August 31, 2009; published online October 21, 2009. Assoc. Editor: Joseph integrating the pressure distribution and shear stress. This method
Katz. is attractive because it is free of empirical data; rotational effects

Journal of Fluids Engineering Copyright © 2009 by ASME NOVEMBER 2009, Vol. 131 / 111103-1
 0015 VAWT with infinite aspect ratio. DES models give the most
accurate description of the generation and shedding of the vortic-
ity and an acceptable sensitivity to grid refinement 共both in time
and space兲. LES models perform worse than DES models, prob-
ably because they model the wall region less accurately. DES and
LES models present large force oscillations at frequencies higher
than the frequency of the turbine rotation, due to either a continu-
ous shedding of small vortices or either due to some numerical
convergence effects. A nonzero value of the normal force for a
zero geometric angle was found for both URANS models, while
for DES and LES models, a zero normal force was found. It is
more likely that, for a zero geometric angle, a symmetric airfoil in
circular motion develops a nonzero lift, unlike a symmetric airfoil

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021

in pure translation, which has a zero lift for a zero geometric
angle. This could be explained by the apparent camber as the
turbine rotates 共virtual camber兲 and the inertia of the fluid in the
boundary layer 关21兴.
Observe that for large scale wind turbines, compressibility ef-
Fig. 1 Typical Darrieus rotor performance CP as a function of
the tip speed ratio †7‡ fects should not be neglected when an accurate solution is needed.
Even if the upwind flow has a low Mach number, local Mach
numbers⬎ 0.3 are to be expected, due to local velocities accelera-
tion. McCroskey et al. 关22兴 observed sonic velocities on the sur-
are introduced in the finite elements analysis and static airfoil data face of an airfoil every time the Mach number exceeded 0.2.
are not needed. It does not cover the stall phenomena but provides The present work examines two operating conditions for the
instantaneous data at the border of the boundary layer that could turbine corresponding to the primary and secondary effects re-
be used to detect the dynamic stall appearance. gions. This analysis is performed with the 2D URANS compress-
All these aforementioned methods 共momentum and vortex兲 are ible solver TURBFLOW 共Ecole Centrale de Lyon, France兲. This
not able to predict accurately the physics beyond the stall angle. CFD code was chosen not for the compressibility equations, but
That is why dynamic stall models have been incorporated in the for its higher order discretization schemes, which are expected to
analysis. These methods may be divided in theoretical and 共semi兲- provide a better description of the flow unsteadiness. The chosen
empirical methods 关10兴. The theoretical models are derived from test case, corresponding to the experiment of Laneville and Vitte-
experimental air load measurements in two-dimensional flows for coq 关23兴, on a two bladed Darrieus air turbine, is incompressible.
oscillating airfoils and then the loads are described using a few This experimental work highlights the strong dependence of the
key parameters, depending on the method used. In this category lift and drag hysteresis curves of a blade with the tip speed ratio ␭.
falls the “UTRC, ␣, A, B” method 关11兴, Boeing–Vertol “Gamma” Two tip speed ratios are considered in this study: a low tip speed
function method 关12兴, Beddoes’ Time Delay method 关13兴, Gang- ratio ␭ = 2 for the primary effects analysis and a higher tip speed
wani’s method 关14兴, etc. ratio ␭ = 7 for the secondary effects analysis. The time step and
The semi-empirical methods are derived from the thin-airfoil grid refinement sensitivities for ␭ = 2 are examined. At ␭ = 2, re-
theory by using a set of linear and nonlinear equations for lift, sults are compared qualitatively with the visualization of the vor-
drag, and pitching coefficients, with several empirical coefficients, tex shedding of Brochier et al. 关24兴 共on a very close configura-
mostly deduced from unsteady airfoil measurements. In this cat- tion兲. Hysteresis lift and drag curves are also quantitatively
egory, there are the Johnson’s method 共MIT method兲 关15兴, compared with the data of Laneville and Vittecoq. The case ␭
ONERA method 关16兴, Leishman–Beddoes method 关17兴, Gormont = 7 is not covered by these experiments.
method 关12兴, etc.
The advantages of the momentum and vortex models are their
flexibility to include a wide range of empirically subcomponent
models representing various physical effects. These are hindered 2 Turbine Parameters
by two aspects: first, by the strong hypotheses discussed above,
The two-dimensional cross section of a Darrieus blade is shown
and second, by the fact that they are nondeterminist. Validation of
this approach with experimental measurements is essential. For in Fig. 2. As the turbine rotates with the angular velocity ⍀, in the
further details, see Refs. 关8,5,18兴. The CFD codes arise as a better upstream flow U⬁, the attack velocity field changes during a revo-
choice to overcome these limitations. lution, given by
The basis of all CFD codes is the numerical schemes for solu- ជ =U
ជ⬁−⍀
ជ ∧ជ
W OM 共1兲
tion of the flow equations, grid generation technique, and model-
ing the boundary layer turbulence. The most affordable is the Notice that this velocity acts upon the blade side exposed to the
RANS method. Allet et al. 关19兴 performed a two-dimensional upstream flow. Consequently, the inner suction side in the up-
CFD computation of a single-bladed VAWT turbine using a stream half disk becomes the inner pressure side in the down-
NACA 0015 airfoil. A finite element turbulent solver based on a stream half disk. At the same time, the incidence angle varies
stream function-vorticity formulation has been used with two tur- from positive to negative values.
bulence models, i.e., Cebeci–Smith and Johnson–King. The shape From geometric considerations, the velocity modulus W and the
of the instantaneous normal and tangential forces was similar with incidence angle ␣ are given by
the experimental curve but the maximum values were overpre-
dicted. Moreover, the stall vortices visualized in the corresponding W = U⬁冑1 + 2␭ cos ␪ + ␭2 共2兲
experiments are not obtained in the computed results.
More elaborated modeling combine RANS method for the
boundary layers regions and the LES method in the outer regions
共the so-called detached eddie simulation 共DES兲兲. Simão Ferreira
␣ = tan−1 冉 sin ␪
cos ␪ + ␭
冊 共3兲

et al. 关20兴 compared two URANS models 共with Spalart–Allmaras where ␪ is measured from the tip position in the counterclockwise
and k-␧ turbulence models兲, and the LES and DES models with direction 共Fig. 2兲.
experimental vorticity visualization for a single-bladed NACA The maximum incidence angle is given by

111103-2 / Vol. 131, NOVEMBER 2009 Transactions of the ASME

 Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021
Fig. 2 Forces and velocities in a Darrieus turbine †20‡
Fig. 3 Tip speed ratio effect on ␣„␪… curves; curvature param-
eter effect on Fⴱ„␣…

␣max = tan−1共1/冑␭2 − 1兲 共4兲

Laneville and Vittecoq 关23兴 defined a reduced frequency for a • An absolute reference frame 共X , Y兲 with the origin in the
Darrieus turbine, derived from that of helicopter blades for a mean center of rotation O;
zero incidence, as follows: • Lift-drag relative reference frame 共l , d兲, with the origin at
the thrust center M and the tangential and normal forces
C␣˙ max
Fⴱ = 共5兲 acting respectively in the W and perpendicular to W direc-
⍀R ⫻ 2␣max tions;
The reduced frequency can be seen as the ratio between a time • Normal-tangential relative reference frame 共n , t兲 called
scale of the fluid motion along the blade C / ⍀R on a time scale Lithental frame, with the origin at the thrust center and the
2␣max / ␣˙ max of the incidence angle variation. As stated by Mc- tangential and normal forces acting respectively in the chord
Croskey 关25兴, for reduced frequencies in excess of 0.05, the ma- and perpendicular to chord directions.
chine is prone to strong dynamic stall with the presence of vortex
shedding if the angle of attack exceeds the one for static stall. For As seen from Fig. 4, when the t-direction is tangential to the
Fⴱ values smaller than 0.05, the flow is assumed quasisteady. The circle of rotation 共blade without pitch angle兲, the thrust is given by
equivalent expression of Eq. 共5兲 for a Darrieus turbine is obtained the tangential component of the aerodynamic force in the Lithen-
tal reference frame.
by setting ␪ = ⍀t in Eq. 共3兲 as
The force coefficients are normalized 共the unitary cross-

Fⴱ = 冉 冊冉 冊
C
R
1 1
␭ − 1 2␣max
共6兲
dimension is 1 m兲 as follows:

Figure 3 is a plot on a same graph of two sets of curves: the

incidence angle ␣ versus the azimuthal angle ␪ for various tip
speed ratios ␭ and the reduced frequency Fⴱ versus the tip speed
ratio ␭ for various curvature parameters C / R. The first set shows
that for ␭ = 2 to 6, typical optimal values for water and wind
Darrieus turbines 共depending on solidity兲, the maximum angle of
incidence varies from 30 deg to 9.5 deg. These angles are large
compared with classical turbo machines applications. The static
stall angle ␣ = 12 deg is obtained for ␭ = 5. Consequently, for all ␭
values less than 5, dynamic stall potentially occurs. The second
set of curves show that for C / R ⬎ 0.1, Fⴱ is in excess of 0.05,
whatever the value of the tip speed ratio. So, if the two conditions
␭ ⬍ 5 and C / R ⬎ 0.1 are fulfilled, strong dynamic stall is to be
expected.
In practice, a distinction between air and water Darrieus tur-
bines is necessary. For water applications, because of their high
solidity 共0.5–2兲 and their low tip speed ratio, Darrieus turbines are
always associated with dynamic stall phenomena. For air, the Dar-
rieus turbine is less affected because of a smaller solidity 共0.1–
0.5兲 and higher tip speed ratios.
The forces acting on a blade, commonly applied at the quarter
chord 共thrust center兲 are depicted on the Fig. 4. The aerodynamic Fig. 4 Velocities and forces acting on a blade and their con-
force on a blade is projected on three reference frames, as follows: ventional positive directions

Journal of Fluids Engineering NOVEMBER 2009, Vol. 131 / 111103-3

 Table 1 Characteristics for the three experimental setups

Parameter Brochier et al., 1986 Fujisawa and Shibuya, 2001 Laneville and Vitecoq, 1986
Technique LDV in water channel PIV in water channel Separately measurements of N and T by strain gauge
共air兲
Blade profile NACA 0018 NACA 0018 NACA 0018
Number of blades nb 2 1 2
Blade chord C 0.02 m 0.01 m 0.061 m
Turbine radius R 0.06 m 0.03 m 0.3 m
Curvature C / R 0.33 0.33 0.20
Blade length ᐉ 0.2 m 0.135 m 0.6 m
Reynolds number ReD = 10,000 共fixed兲 ReD = 3000共fixed兲 ReC = 38,000共fixed兲
Angular velocity ⍀ variable variable 300 rpm 共fixed兲

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021

Undisturbed flow velocity U⬁ 0.15 m/s 共fixed兲 0.05 m/s 共fixed兲 variable
Tip speed ratio ␭ 2.14 3.85 1; 2; 3 2; 2.5; 3; 4; 5
Reduced frequency Fⴱ 0.3 0.22 ⬀; 0.32; 0.24 0.194; 0.164; 0.134; 0.124; 0.119
Central column diameter 0.01 m Not used 0.0381 m
Blade fixation End plates Bottom= disk Bottom= horizontal arm with force transducers
Up= free Up= wires
Aspect ratio ⬇⬁ Relatively low Relatively low

冦 冧
L adjacent cells, with 188 nodes on each blade face. The ratio for
CL = the other zones is incremented from 1.15 to 3 at the boundary. To
0.5␳C1共⍀R兲2
共7兲 properly resolve the viscous affected region, an O mesh is used
D around the airfoils and y+ at the wall adjacent cell is chosen
CD =
0.5␳C1共⍀R兲2 around 1, corresponding to a size of 25 ␮m. Grid sensitivity
study presented in Appendix A show that a finer grid provide very
similar results.
3 Numerical Study
Numerical simulations on a Darrieus turbine were carried out, 3.2 Numerical Governing Equations. The flow solver TURB-
the hysteretic behavior of which was investigated experimentally FLOW is used to perform computations on the Darrieus turbine.
by Laneville and Vittecoq 关23兴. The various parameters are shown The equations solved here are the full unsteady compressible
in Table 1. RANS equations in the absolute frame, where the laminar viscos-
The two blades have a NACA 0018 foil shape and are attached ity is assumed constant. The equations are solved for the conser-
to the rotor radius at the quarter point from the leading edge in an vative 共Favre mass-averaged兲 variable density ␳, momentum vec-
infinite flow field. The offset pitch angle is set to zero 共␣0 = 0兲. tor components ␳ui, and total energy ␳E. In the present study, the
Due to some numerical instabilities at the trailing edge, the sharp turbulence model k-␻ of Kok 关26兴 is used. To avoid unphysical
edge was transformed into a round edge with a ratio radius/chord production of eddy viscosity, the production of kinetic energy is
of 0.1%. The results are not affected by this tiny transformation. limited with a limiter relating the maximum allowable production
to the double of the dissipation of turbulent kinetic energy.
3.1 Grid Setup. The computational domain extends 30 blade To take into account the blades rotation, an arbitrary
chord lengths from the center. The mesh is a multiblock structured Lagrangian–Euler 共ALE兲 method was used 关27兴, with an adiabatic
grid composed by a mix of “O” blocks near the blades and near zero normal velocity condition on the blade surface. At the outlet,
the frontier and a “H” bloc in the center 共Fig. 5兲. The whole the atmospheric pressure was imposed, and at the inlet, the up-
domain was meshed with the GAMBIT software 共version 2.4兲 of stream velocity was used.
FLUENT and then divided in 12 blocks, roughly of the same num- The space discretization is based on a monotone upstream-
ber of grid points, to take advantage of the parallel virtual ma- centered schemes for conservation laws 共MUSCL兲 关27兴 finite vol-
chine 共PVM兲 parallel programming. The computations were per- ume formulation, with a third-order upwind scheme 关28兴 for the
formed on two Dell PowerEdge 1950 machines equipped with convective terms. To advance the solution in time, a dual-time
two Intel Xeon dual core processors. step procedure was used. Also, due to the stiffness and low con-
The reference grid contains 160,000 points and will be refer- vergence of the compressible solvers when low Mach numbers are
enced as grid N1. Two blocks contain each blade and extend half encountered 关29兴, a low Mach preconditioning is employed 关30兴.
a chord, meshed with a ratio of a maximum of 1.15 between two The solution is advanced in pseudotime with a three-step Runge–
Kutta method until a pseudosteady-state is reached. The following
numerical parameters come from a study presented in Appendix
B. The physical time step was limited to 1.38⫻ 10−4 s, which
corresponds to 0.25 deg of incremental azimuthal angle. To cover
a chord, 48 time steps are necessary. For the inner time step, a
local time step was employed with CFL= 1, which gives a
pseudotime step of 10−8 s near the blade region. To ensure stabil-
ity at the beginning of a time step, a linear slope of the CFL
number was imposed for the first 300 pseudotime steps, ranging
from 10−3 to 1. A maximum number of 3000 inner-iterations was
used 共see Appendix A兲. In the following simulations, a transient
behavior is noticed, with large differences between the first and
the second revolutions, and negligible differences between the
second and the third revolutions. Similar transient behaviors are
found in Refs. 关31,32兴. The presented results correspond to the
Fig. 5 Overview of the grid mesh N1 second revolution.

111103-4 / Vol. 131, NOVEMBER 2009 Transactions of the ASME

 Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021
Fig. 6 Computed vortices trajectories for ␭ = 2, using Q criterion, nondi-
mensionalized by „C / „⍀R……2

4 Results responding to 21% and 5% of the blade rotational speed 共9.42

m/s兲. This result confirms the strong interaction of the vortex a
4.1 Case ␭ = 2. Figure 6 shows the main vortices trajectories when its trajectory intersects the blade trajectory at the beginning
inside the turbine. The map has been obtained using the Q crite- of the second half disk of rotation. The influence of the vortex b is
rion, well adapted to identify traveling vortices 共i.e., high vorticity weaker because of a smaller circulation and a larger distance from
and low shear strain rate regions兲 关33,34兴. The numbers 0 to 7 the blade.
refer to the foil positions and the letters to the vortex. The figure In the two other cases 共Figs. 7.23 and 7.1兲 the induced veloci-
shows that only the clockwise 共c⬘兲 vortex goes through the upper ties are found equal to 2.5% and 1% of the rotational speed, re-
part of the turbine, following a straight trajectory. On the other spectively. This shows that the c⬘ vortex influence on the two
hand, the 共a-b兲 pair of vortices and the clockwise 共e兲 vortex cross blades is negligible.
also the turbine, following curvilinear trajectories. At ␪ = 0 共Fig. 7.1兲, the flow recovers an attached boundary
In Fig. 7, isocontours of vorticity and streamlines in a relative layer. The alternated vortices, seen at the rear of the wake, come
frame of reference are plotted for 24 azimuthal blade positions. from higher incidences undergone at the end of the preceding
The main coherent vortices are identified with letters as done in revolution. At ␪ = 45 deg 共Fig. 7.2兲, a counterclockwise vortex c
Refs. 关24,35兴. 共the vortex field was not analyzed in Ref. 关23兴兲. appears at the trailing edge. As this vortex grows, a new counter-
Because the present configuration is close to the Brochier one, a clockwise vortex a develops at the leading edge 共Figs. 7.3 and
similar vortex marking out is used. For convenience, the main 7.4兲. At ␪ = 71.25 deg 共Fig. 7.5兲, c vortex is shed in the wake
characteristics of the three aforementioned experimental works simultaneously with a clockwise vortex c⬘ coming from the outer
are gathered in Table 1. side boundary layer. As vortices c and c⬘ are convected in the
From Fig. 6, it appears that three regions are potentially con-
wake, the a vortex continues to develop and detaches from the
cerned with blade vortex interactions, as follows:
leading edge 共Figs. 7.6 and 7.7兲. From ␪ = 90 deg to ␪ = 105 deg
共1兲 a large region from positions 2 to 6 共␪ = 90− 270 deg兲, 共Figs. 7.8–7.12兲, the detachment of the a vortex is strongly
where the foil interacts with its own vortices, essentially a coupled with the growth of a clockwise b vortex at the trailing
and b 共Fig. 7.19兲 edge 共similar to c⬘兲. As vortices a and b are shed in the wake
共2兲 a narrow region about position 7 共␪ = 315 deg兲, where the 共Figs. 7.13–7.15兲 a counterclockwise d vortex develops at the
foil interacts with its own vortex c⬘ 共Figs. 7.22–7.24兲 leading edge. At this position, the incidence angle reaches the
共3兲 a narrow region between positions 0 and 1 共␪ maximum ␣ = 29.92 deg. In a similar way than for 共c-c⬘兲 vortices
= 0 – 45 deg兲 where the foil interacts with the c⬘ vortex and 共a-b兲 vortices, the detachment of the d vortex corresponds to
shed by the other blade 共Figs. 7.1 and 7.2兲 the growth of a clockwise e vortex at the trailing edge coming
from the outer side boundary layer 共Figs. 7.16–7.18兲. At this po-
To check the real influence of a, b and c⬘ vortices in the three sition, the foil enters in the second half disk of rotation and the
cases, their circulation have been evaluated by computing their incidence angle becomes negative. When approaching the maxi-
vorticity flux ⌫. The induced velocity on the blade is obtained mum negative angle of incidence, 共␪ = 225 deg, Fig. 7.19兲 a
using the Kelvin vortex model. Following this model, the azi- clockwise vortex f is seen at leading edge. Figures 7.19 and 7.20
muthal induced velocity at a point situated at a distance r from the depict the blade positions at two close angles of incidence ⫺28.68
vortex center, lying in the outer vortex region, is given by v␪ and ⫺28.02, before and after the maximum, respectively. Figure
= ⌫ / 共2␲r兲. The vortex influence is calculated at the quarter chord 7.19 shows that the local angle of attack exceeds by far the geo-
blade point. metrical one. This can be explained by the proximity of the a
In the first aforementioned case, the position corresponding to vortex that creates a blockage effect leading to a flow deflection in
Fig. 7.19 is chosen because, as it will be shown in the next para- the outer direction. This effect is not seen on the Fig. 7.20 due to
graph, it reveals a strong interaction. The induced velocities of the convection of the a vortex in the downstream region. As seen
vortices a and b are equal to 2m/s and 0.5m/s, respectively, cor- in the first half disk, the f vortex detachment leads to the devel-

Journal of Fluids Engineering NOVEMBER 2009, Vol. 131 / 111103-5

 Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021

Fig. 7 Vorticity vector isocontours and relative streamlines

opment and the shedding of a counter-rotative vortex g 共Figs. 7.20 vortices, detaching from leading and trailing edges, are shed in the
and 7.21兲. As the incidence angle decreases 共␪ = 293– 322 deg, wake. These pairs of vortices appear weaker than the ones ob-
Figs. 7.21–7.24兲 many other pairs of alternate counter-rotating served in the first half disk.

111103-6 / Vol. 131, NOVEMBER 2009 Transactions of the ASME

 Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021
Fig. 8 Lift coefficient for one revolution, ␭ = 2

Figures 8 and 9 show the lift and drag coefficients on one blade lations are observed between the two curves兲, some significant
versus the geometrical angle of attack ␣. The black and dotted differences exist. They can come from four factors, as follows:
lines correspond to the calculated and experimental coefficients,
respectively. The number plotted on the calculated curves corre- 共1兲 The first one is the grid sensitivity, or more generally, the
sponds to the picture number in Fig. 7. Though the global numeri- influence of the numerical parameters of the solver like the
cal and experimental curve shapes are similar 共some global trans- time step, the number of inner-iterations of the dual-time

Fig. 9 Drag coefficient for one revolution, ␭ = 2

Journal of Fluids Engineering NOVEMBER 2009, Vol. 131 / 111103-7

 stepping, and the order of the numerical schemes. A de-
tailed study reported in Appendix A shows that a more
refined grid do not produce remarkable differences on the
hysteresis lift and drag loops. It is also demonstrated that
the chosen physical and dual-time steps are small enough to
produce no influence on the results.
共2兲 The second one is the two-dimensional approximation. In
the experimental work, the blade aspect ratio, equal to 10,
seems to be too large to explain the mentioned gap. Never-
theless, this point should be further investigated.
共3兲 The third one is the physical model relevance. Several
works, as mentioned in the introduction, show that the
URANS model based on classical isotropic turbulent model

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021

is satisfying in the blade boundary layer. The main weak-
ness of this approach concerns actually the convection of
the large eddies that are detached from the blade and con-
vected inside the turbine. For this reason it is surprising that
the main gap between the calculated and experimental co-
efficients concerns the blade positions between ␪
= 332 deg and ␪ = 60 deg 共positions 24 to 3兲 corresponding
to moderate flow incidences 共␣ = −12 deg to 20 deg兲, Fig. 10 Vorticity vector for ␭ = 7 „black color-counterclock
where no or only moderate stall takes place. A theoretical rotating…
inviscid model reported in Appendix B shows that the flow
curvature effect 共due to the blade rotation兲 leads to a strong
positive lift coefficient CL = 1 for ␪ = 0° Fig. 21, close to the
numerical coefficient CL = 0.85, though the experimental curve is linear. The slope is about the half of the one given by the
one is 0. Also, data concerning the turbulence intensity up- theoretical inviscid model of Appendix B. This difference is at-
stream are not provided in the experimental work. tributed to the delay due to the strong unsteadiness of the flow.
共4兲 Finally, the last factor concerns the accuracy of the experi- The first lift drop corresponds to the detachment of the c vortex
mental data. The above considerations make these data 共positions 3 and 4兲. Then, a new raise is observed when the a
questionable in some respects. Unfortunately, no sufficient vortex develops along the inner part of the foil 共positions 5 and 6兲.
detail are found in the experimental study to quantify the The second drop corresponds to passage of the a vortex at the
uncertainties relative to the measured forces acting on the blade trailing edge 共positions 7 and 8兲. This drop is attributed to
blades. Since the blades are built with balsa wood and the streamlines that separate strongly from the chord direction at
maintained with wires at the tip, a possible cause of dis- trailing edge. From positions 9 to 14, though the incidence reaches
crepancy could be the blade torsion. If a twist is considered its maximum, the lift stabilizes 共oscillates兲 due to the strong and
only, numerical simulations 共viscid and inviscid兲 show that complex interactions of the two main vortices a and b with the
an offset pitch angle of the order of 3 deg 共trailing edge blade. From position 15 to 17, the lift decreases smoothly follow-
orientated toward the rotor兲 is sufficient to cause a zero lift ing the flow incidence decrease. For position 17 共␪ = 180 deg,
coefficient for a null incidence. ␣ = 0 deg兲 the negative value of the lift is attributed to the influ-
ence of the d vortex, which obviously interacts with the blade.
In the following the calculated lift variations presented in Fig. 8 The inviscid model, which does not take into account these vor-
are linked to the dynamic of the main vortices highlighted in Figs. tices, gives a positive lift coefficient of 0.31 coherent with the
6 and 7. The drag is not commented because its variations are in curvature effect 共Appendix B, Fig. 21兲. From positions 18 to 19,
phase with the lift ones. the blade has undergone a negative lift due to the negative flow
From positions 1 to 2 the boundary layer is attached and the lift incidence. The increase in the lift slope, in this part, is probably

Fig. 11 Computed lift coefficient for one revolution, ␭ = 7

111103-8 / Vol. 131, NOVEMBER 2009 Transactions of the ASME

 Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021
Fig. 12 Computed drag coefficient for one revolution, ␭ = 7

due to the influence of the a and b vortices as they get closer to mechanism is similar. Fujisawa and Shibuya 关35兴 mentioned the
the blade. Position 20 confirms this influence; as the incidence presence of two pairs of counter-rotating vortices, one at the small
keeps its maximum between positions 19 and 20 共⫺28 deg兲, the incidence angle and the other one at the higher incidence angle.
lift coefficient changes from ⫺1 to about 0. The corresponding The present computed results confirm the existence of two pairs of
streamlines presented on Fig. 7 show that the counterclockwise a such counter-rotating vortices, 共c-c⬘兲 and 共a-b兲, although the pair
vortex induces a flow blockage at position 19, which increases generated at low incidence angle is more quickly dissipated than
locally the negative incidence. At position 20, this flow blockage the 共a-b兲 pair of vortices.
is completely removed due to the convection of the a vortex
downstream. From position 21 to position 24 incidence decreases 4.2 Case ␭ = 7. Figure 10 presents the vorticity field for four
as the blade revolution completes. The oscillating behavior of the blades positions, actually eight positions if one blade is consid-
lift during this phase is a direct consequence of the alternate vor- ered. The flow field appears much less complicated than in the
tices developing at the leading and trailing edges and sheds in the case ␭ = 2. In the first half of the revolution alternate vortices, very
blade wake. Finally, the large hysteresis loops obtained for the lift similar to those observed in the second half of the revolution for
and the drag curves show the aforementioned strong influence of ␭ = 2, are observed. In the second half of the revolution, the alter-
the vortex structures on the blade. nate vortices are replaced by an attached boundary layer. Figures
The vortex formation and its convection downstream are found 11 and 12 present the corresponding lift and drag coefficients. The
also in visualizations by Brochier et al. 关24兴 or by Fujisawa and experimental coefficients are not available for this tip speed ratio.
Shibuya 关35兴. For low tip speed ratios 共␭ = 1 , 2 , 3兲, the counter pair It is important to notice that flow incidences are now less than 8
of vortices a-b is mentioned in the two articles and the convection deg compared with 30 deg in the case ␭ = 2. This explains the light

Fig. 13 Isovalues of axial velocity ratio UX / Uⴥ for ␪ = 0 deg, ␭ = 2

Journal of Fluids Engineering NOVEMBER 2009, Vol. 131 / 111103-9

 Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021
Fig. 14 Isovalues of axial velocity ratio UX / Uⴥ for ␪ = 0 deg, ␭ = 7

stall regime observed. The figures also show a small hysteresis for passing. Figure 15 presents the drag turbine coefficient versus the
the lift and practically no hysteresis for the drag. The hysteresis is azimuthal blade position. It confirms that the drag for ␭ = 7 is
restricted to the first half of the revolution where a light stall is larger than for ␭ = 2, the corresponding mean drags being 2.2 and
observed. This demonstrates that the importance of the hysteresis 0.64, respectively. It must also be noticed that for the ␪ = 0 deg
directly depends on the dynamic stall strength. Figure 11 also position, the turbine drag is practically zero though the flow slow-
shows a remarkable slope change between the first and second ing down, found at the same position in Figs. 13 and 14, is im-
half disks. The slope for a negative ␣ is about half of the one for portant. This explains why the behavior of the macroscopic flow
a positive ␣. This curve behavior can be explained by the varia- field around the turbine depends not only of the current flow in-
tion in the main flow velocity shown in Figs. 13 and 14 for ␭ side but also of the flow time history.
= 2 and 7, respectively. The flow velocity decreases along the
turbine, particularly for ␭ = 7. As a result, the local tip speed ratio
at the blade position, based on the local flow field, is strongly
5 Conclusion
increased for downstream blade positions. This leads to smaller The unsteady flow field in a cross-flow turbine was numerically
local flow incidences in this region and consequently to smaller computed with a 2D URANS method and a k-␻ turbulence model.
lifts values. The flow slowing down in front of the turbine comes Two extreme tip speed ratios ␭ = 2 and 7, of the straight two
from the turbine drag and is associated to a global turbine by- bladed Darrieus turbine investigated experimentally by Laneville

Fig. 15 Rotor drag coefficient

111103-10 / Vol. 131, NOVEMBER 2009 Transactions of the ASME

 Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021
Fig. 16 Influence of inner-iterations on the lift coefficient at the beginning of the dy-
namic stall

and Vittecoq 关23兴, have been considered. The grid and numerical 共a-b兲 pair, created near the maximum incidence angle in the
parameters have been chosen to avoid any sensitivity effects on upstream half disk, is the strongest. The maximum lift co-
the results. efficient corresponds to the convection of the a vortex from
共1兲 The comparison with the experiment is made at ␭ = 2 on the the leading edge to the trailing edge of the blade. The sud-
basis of the lift and drag coefficients. Though the global den lift drop is caused by the vortex passage at the trailing
numerical and experimental curve shapes are similar, some edge. It is shown that this vortex continues to influence the
significant differences exist. Several arguments lead to ask same blade at ulterior positions. It is responsible for the
oneself about the experimental data. Unfortunately, the un- strong increase in the negative lift in the downstream half
certainties of the lift and drag measurements are not pre- disk and then for its drop at the maximum negative inci-
cisely provided. dence angle. The drag varies practically in phase with the
共2兲 The dynamic stall undergone by the blades has been studied
lift during the blade rotation. The large hysteresis loops
in detail using the numerical results. For ␭ = 2, pairs of
counterclockwise vortices close to those observed in ex- obtained for the lift and the drag curves express the strong
perimental works are sequentially shed from the blade. The influence of the vortex structures on the blade.

Fig. 17 Residuals of ␳E for three time steps and various inner-iterations

Journal of Fluids Engineering NOVEMBER 2009, Vol. 131 / 111103-11

 Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021
Fig. 18 Lift coefficients for the three meshes

共3兲 The ␭ = 7 case is characterized by a weak shedding of al- Ecoulements Géophysiques Industriels 共LEGI兲 hydrodynamic tun-
ternated vortices in the upstream half disk and by an at- nel. It will allow measurement of average and instantaneous tur-
tached flow in the downstream half disk. This light stall bine torques and powers on cross-flow turbine models. Flow vi-
regime, due to smaller incidence angles, tends to flatten the sualizations and velocity measurements will also be possible.
hysteresis loops of the lift and drag curves. A strong de- Further numerical works will focus especially on the comparison
crease in the main flow is also found inside the turbine. with the experimental data obtained.
This explains the smaller slope of the lift curve in the
Once fully validated, this modeling will allow to obtain the
downstream half disk. It is shown that the flow blockage,
addressing the macroscopic flow around the turbine, is in- optimal foil section of the turbine, particularly its camber-line.
duced not only by the current turbine drag but also by its Indeed, an inverse camber-line leads to a power increase in the
time history. half upwind disk of rotation, but to a power decrease in the half
downwind disk. The choice of the optimal camber-line could be
A new experimental setup is being built in the Laboratoire des obtained only using CFD calculations.

Fig. 19 Drag coefficients for the three meshes

111103-12 / Vol. 131, NOVEMBER 2009 Transactions of the ASME

Nomenclature
D ⫽ drag force 共N兲
L ⫽ lift force 共N兲
R ⫽ turbine radius 共m兲
Fⴱ ⫽ reduced frequency
nb ⫽ number of blades
ᐉ ⫽ blade length 共m兲
C ⫽ chord of the foil 共m兲
S = 2Rᐉ ⫽ swept area 共m2兲
U⬁ ⫽ upstream fluid velocity 共m/s兲
W ⫽ relative velocity 共m/s兲
␮ ⫽ water molecular viscosity 共Kg/ m s兲

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021

␣ ⫽ incidence angle 共deg兲
␣0 ⫽ offset pitch angle 共deg兲
␣˙ = d␣ / dt ⫽ rate of change in the incidence angle 共s−1兲
␪ = ⍀t ⫽ azimuthal angle 共deg兲
⌬␪ ⫽ azimuthal increment angle 共deg兲
␭ = ⍀R / U⬁ ⫽ tip speed ratio
␴ = nbC / R ⫽ solidity
␳ ⫽ water density 共Kg/ m3兲
⍀ ⫽ angular velocity 共rad/s兲
C P = Torque⫻ ⍀ / 1 / 2␳SU3⬁ ⫽ power coefficient
ReC = ␳⍀RC / ␮ ⫽ blade Reynolds number
ReD = ␳U⬁2R / ␮ ⫽ machine Reynolds number
rpm ⫽ revolutions per minute 共rpm兲

Appendix A: Numerical Convergence and Grid Sensi-

tivity
Preliminary tests showed that a maximum of 3000 inner-loops
by time step combined with 10−7 for the residuals is sufficient
enough to ensure the overall convergence of the computation. A
specific test to check the influence of the inner-loops on the solu-
tion was carried out. The converged solution with the previous
parameters at a particular azimuthal position 共88 deg, correspond-
ing to the deep dynamic stall regime兲 was chosen using 500, 750,
1250, 1500, and 3000 inner-iterations and the solution was ad-
vanced in time for 40 time steps. The lift coefficient is plotted in
Fig. 16 and the residuals of 共␳E兲 for three time steps are plotted in
Fig. 17. An insufficient number of inner-loops is reflected in an
oscillatory behavior of lift coefficient, while from 1250 of inner-
loops the lift remains unchanged. A closer look at the convergence
rate inside a physical time step 共Fig. 17兲 shows that the oscillatory
lift coefficient comes from an insufficiently convergence rate.
From 1500 inner-iterations, no supplementary gain is obtained. Fig. 20 Vorticity isocontours and relative streamlines for the
The peaks present at the end of a time step for 1500 and 3000 three meshes and three azimuthal angles; numbers 1 to 3 refer
inner-iterations are a consequence of the filtering operator, which to positions indicated on Figs. 18 and 19
limits the turbulent variables k and sometimes ␻. It should be
noted that either a lower CFL or a smaller physical time step
mesh an azimuthal increment of 0.1 deg and a five-step Runge–
activates less often the filtering operator. In the attached flow re-
Kutta method was used instead of 0.25 deg and three-step Runge–
gime 共0 ⫾ 45 deg兲, the filtering operator is not activated and the
Kutta for the other two cases.
residuals have a steep linear descent. The choice of a CFL= 1 and The results for the lift and drag coefficients are presented in
a time step corresponding to 0.25 deg and a maximum of 3000 Figs. 18 and 19, respectively. Seen from these figures, the differ-
inner-iterations is a good compromise between convergence and ence between the lift and drag coefficients, for the finer mesh N0
stability. and the reference mesh N1 is negligible, even though the time step
To check the grid sensitivity, two other grids were chosen, sys- is smaller for the N0 mesh. It can be concluded that for the ref-
tematically dividing or multiplying in each direction by two 共mul- erence mesh, the solution is relatively time step independent.
tiplying or dividing by four the total number of cells兲 referred as The following remarks concern the three meshes:
N0 共the finer mesh兲 and N2 共the coarser grid兲. It should be noted
that for a block structured mesh with collocated grid points, this 共a兲 From ␪ = 0 deg until the beginning of the dynamic stall,
technique becomes highly prohibitive because the computational the coarser mesh N2 gives satisfactory results compared
time is multiplied by a factor of four for the finer mesh. To over- with the other two meshes, although the maximum lift
come this problem, for the finer mesh, a linear interpolation from and drag coefficient is under predicted.
the reference solution 共␭ = 2兲 was done for an azimuthal position 共b兲 A deeper decrease in both lift and drag coefficients oc-
of 55 deg and the test was stopped at 180 deg. Also, a rigorous test curs for the N0 and N1 meshes, compared with the N2
was not possible, due to the explicit CFL restrictions; for the finer mesh, which corresponds to the passage of the a vortex

Journal of Fluids Engineering NOVEMBER 2009, Vol. 131 / 111103-13

 Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021
Fig. 21 Computed lift coefficient, ␭ = 2

toward the trailing edge. From the relative stream traces VT = Vz/A0 共B4兲
in Fig. 20.1, the decrease is a consequence of a larger a
vortex on the inner-surface of the foil. The slip condition in the physical plane for one point i, located on
共c兲 The main difference between the coarse mesh and the the airfoil 共affix z P兲, has the following form in the transformed
plane:

冋 冊 册
other two meshes appears in the zone corresponding to
the convection downstream in the inner part of the rotor
of the counter-rotating pair of vortices 共a-b兲. The 共a-b兲
pair of vortices follows more closely the blade in the case
冉
Im ei␹VTP + ei␹U⬁ ei␪ + i␭
z̄ P z P
R A0
=0 共B5兲

of the finer grid 共Figs. 20.2 and 20.3兲. Their influence is where Im is the imaginary part of a complex function, and ␹ is the
particularly obvious. At ␪ = 180 deg, the coarser mesh tangent angle at the point P of the transformed foil. At instant t, in
indicates a positive lift 共as predicted by the theoretical the transformed plane, the static foil is attacked by the complex
model, Appendix B兲, while the other two meshes produce velocity VT0 = U⬁共ei␪ + i␭共z̄ P / R兲兲共z P / A0兲. The velocity is a function
a negative lift coefficient. of time and point position in the transformed plane. To simplify
共d兲 At the end of the revolution, the N2 and the N1 meshes the calculation, a steady flow is considered and the attack velocity
produce similar results. is the average velocity on the foil. The magnitude of ␾0 is calcu-
lated in order to provide a positive real value of the mean velocity
T
Vm . Thus, in the transformed plane, the velocity at infinity up-
Appendix B: Conformal Transformation stream is always directed along the positive xT axis.
A conformal mapping technique was used which transforms the In the physical plane, at any time, the flow has a period of 2␲.
complex plane z = x + iy into a complex plane zT. In the complex In the transformed plane, a space periodicity of direction ␾0
plane zT = f共z , t兲, the entrainment velocity of a point belonging to + ␲ / 2 is, thus, obtained. The potential flow is calculated in a two-
the moving complex plane z = x + iy 共Figure 4兲 becomes zero rela- dimensional cascade of transformed foils. The method is the
tive to the transformed frame of reference zT. Thus, the trans- boundary elements method 共singularities methods兲. The pressure
formed foil is at rest. It is easy to demonstrate that zT has the integrations along the foil provide the efforts that the fluid exerts
following form 共characteristics theory兲 on the foil. For the calculation of the pressure p, the Lagrange
equation is written in the physical moving plane 共generalized Ber-
zT = f共z,t兲 = U⬁t + F1关ze−i␪兴 共B1兲 nouilli equation兲 as
F1 is an arbitrary holomorphic complex function. F1 is determined 1 ⳵oxy⌽
so that the velocity at infinity vanishes. A class of functions which p + ␳兩V兩2 + ␳ + ␳⍀ Im共Vz兲 = p⬁ 共B6兲
have this property is the complex logarithm function 2 ⳵t
where p⬁ stands for the upstream pressure. The nonstationary term
zT = A0 log B0共z − z0兲 共B2兲
⳵oxy⌽ / ⳵t due to the partial derivative of the velocity potential
The complex constants are chosen so that the foil leading edge is relative to the moving plane is neglected. The results are presented
located at the origin of the plane zT. The transformation has to in Fig. 21.
respect the domain topology. In fact, z0 should not be taken inside
the foil. It is selected simply at the origin of coordinates 共z0 = 0兲.
With these restrictions the transformation becomes References
关1兴 Templin, R. J., 1974, “Aerodynamic Performance Theory for the NRC Vertical
zT = A0 log共z/zL兲 共B3兲 Axis Wind Turbine,” National Research Council of Canada, Report No. LTR-
LA-160.
where A0 = CTei␾0 / 兩log共zF / zL兲兩, and zF and zL designate the affixes 关2兴 Strickland, J. H., 1975, “The Darrieus Turbine: A Performance Prediction
of the trailing and leading edges points in the complex plane z, Model Using Multiple Streamtubes,” SANDIA Laboratory, Report No. SAND
respectively, and CT is the chord. Under the assumptions of non- 75-041.
关3兴 Loth, J. L., and McCoy, H., 1983, “Optimisation of Darrieus Turbine With an
viscous fluid and irrotational flow, the complex velocity V of the Upwind and Downwind Momentum Model,” J. Energy, 7, pp. 313–318.
physical plane is transformed into the complex velocity VT 关4兴 Paraschivoiu, I., and Delclaux, F., 1983, “Double Multiple Streamtubes. Model

111103-14 / Vol. 131, NOVEMBER 2009 Transactions of the ASME

 With Recent Improvement,” J. Energy, 7共3兲, pp. 250–255. lating Dynamic Stall in a 2D VAWT: Modeling Strategy, Verification and Vali-
关5兴 Paraschivoiu, I., 2002, Wind Turbine Design With Emphasis on Darrieus Con- dation With Particle Image Velocimetry Data,” J. Phys.: Conf. Ser., 75, p.
cept, Polytechnic, Brooklyn, NY. 012023.
关6兴 Pawsey, N. C. K., 2002, “Development and Evaluation of Passive Variable- 关21兴 Migliore, P. G., Wolfe, W. P., and Fanucci, J. B., 1980, “Flow Curvature
Pitch Vertical Axis Wind Turbine,” Ph.D. thesis, University of New South Effects on Darrieus Turbine Blade Aerodynamics,” J. Energy, 4共2兲, pp. 49–55.
Wales, Sydney, Australia. 关22兴 McCroskey, W. J., McAllistair, K. W., Carr, L., Pucci, L. W., Lambert, O., and
关7兴 Strickland, J. H., Webster, B. T., and Nguyen, T., 1979, “A Vortex Model of Indengard, R., 1981, “Dynamic Stall on Advanced Airfoils Sections,” J. Am.
the Darrieus Turbine: An Analytical and Experimental Study,” ASME J. Fluids Helicopter Soc., 26共7兲, pp. 40–50.
Eng., 101, pp. 500–505. 关23兴 Laneville, A., and Vittecoq, P., 1986, “Dynamic Stall: The Case of the Vertical
关8兴 Leishman, G. J., 2002, “Challenges in Modelling the Unsteady Aerodynamics Axis Wind Turbine,” Prog. Aerosp. Sci., 32, pp. 523–573.
of Wind Turbines,” Wind Energy, 5, pp. 85–132. 关24兴 Fraunié, P., Brochier, G., Béguier, C., and Paraschivoiu, I., 1986, “Water
关9兴 Ponta, F. L., and Jacovkis, M., 2001, “A Vortex Model for Darrieus Turbine Channel Experiments of Dynamic Stall on Darrieus Wind Turbine Blades,” J.
Using Finite Element Techniques,” Renewable Energy, 24, pp. 1–18. Propul. Power, 2共5兲, pp. 445–449.
关10兴 Leishman, G. J., 2002, Principles of Helicopter Aerodynamics, Cambridge 关25兴 McCroskey, W. J., 1972, “Dynamic Stall on Airfoils and Helicopters Rotors,”
University Press, New York. AGARD Paper No. R595.
关11兴 Bielawa, R. L., 1975, “Synthesized Unsteady Airfoil Data With Applications to

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/131/11/111103/5579491/111103_1.pdf by Universiti Teknologi Malaysia user on 15 August 2021

关26兴 Kok, J. C., 1999, “Resolving the Dependence on Free-Stream Values for the
Stall Flutter Calculations,” 31st Annual Forum of the American Helicopter
k-␻ Turbulence Model,” National Aerospace Laboratory NLR, Paper No.
Society, Washington, DC, May 13–15.
NLR-TP-99295.
关12兴 Gormont, R. E., 1973, “A Mathematical Model of Unsteady Aerodynamics and
关27兴 Donea, J., Giuliani, S., and Halleux, J. P., 1982, “An Arbitrary Lagrangian–
Radial Flow for Application to Helicopter Rotor,” USAAVLABS, Technical
Euler Finite Element Method for Transient Dynamic Fluid-Structure Interac-
Report No. 72-67.
tions,” Comput. Methods Appl. Mech. Eng., 33, pp. 689–723.
关13兴 Beddoes, T. S., 1978, “Onset of Leading Edge Separation Effects Under Dy-
namic Conditions and Low Mach Number,” 34th Annual Forum of the Ameri- 关28兴 Edwards, J. R., and Liou, M. S., 1998, “Low-Diffusion Flux-Splitting Methods
can Helicopter Society, Washington, DC, May 15–17. for Flows at All Speeds,” AIAA J., 36共9兲, pp. 1610–1617.
关14兴 Gangwani, S. T., 1981, “Prediction on Dynamic Stall and Unsteady Airloads 关29兴 Turkel, E., and Vatsa Veer, N., 2003, “Choice of Variables and Preconditioning
for Rotor Blades,” 37th Annual Forum of the American Helicopter Society, for Time Dependent Problems,” 16th AIAA Computational Fluid Dynamics
LA, May 17–20. Conference, Orlando, FL, Jun. 23–26.
关15兴 Johnson, W., 1980, Helicopter Theory, Princeton University Press, Princeton, 关30兴 Weiss, J. M., and Smith, W. A., 1995, “Preconditioning Applied to Variable
NJ. and Constant Density Flows,” AIAA J., 33共11兲, pp. 2050–2057.
关16兴 Pitot, D., 1989, “Differential Equation Modelling of Dynamic Stall,” La Re- 关31兴 Ploesteanu, C., 2004, “Etude Hydrodynamique d’un Type d’Hydrolienne à axe
cherche Aerospatiale, Paper No. 1989-6. Vertical pour les Courants Marins,” Ph.D. thesis, Institut National Polytech-
关17兴 Leishman, G. J., and Beddoes, T. S., 1989, “A Semi-Empirical Model for nique de Grenoble, France.
Dynamic Stall,” J. Am. Helicopter Soc., 34共3兲, pp. 3–17. 关32兴 Ekaterinas, J. A., and Platzer, M. F., 1998, “Computational Prediction of Air-
关18兴 Claessens, M. C., 2006, “The Design and Testing of Airfoils for Application in foil Dynamic Stall,” Prog. Aerosp. Sci., 33, pp. 759–846.
Small Vertical Axis Wind Turbines,” Ph.D. thesis, Delft University of Tech- 关33兴 Chong, M. S., and Perry, A. E., 1990, “A General Classification of Three-
nology, Delft, The Netherlands. Dimensional Flow Fields,” Phys. Fluids A, 2共5兲, pp. 765–777.
关19兴 Allet, A., Hall, É. S., and Paraschivoiu, I., 1999, “Numerical Simulation of 关34兴 Jeong, J., and Hussain, F., 1995, “On the Identification of a Vortex,” J. Fluid
Dynamic Stall Around an Airfoil in Darrieus Motion,” ASME J. Sol. Energy Mech., 285, pp. 69–94.
Eng., 121, pp. 69–76. 关35兴 Fujisawa, N., and Shibuya, S., 2001, “Observations of Dynamic Stall on Dar-
关20兴 Simão Ferreira, C. J., Bijl, L., van Bussel, G., and van Kuik, G., 2007, “Simu- rieus Wind Turbine Blades,” J. Wind Eng. Ind. Aerodyn., 89, pp. 201–214.

Journal of Fluids Engineering NOVEMBER 2009, Vol. 131 / 111103-15