Proceedings of ASME Turbo Expo 2015: Turbine Technical Conference and Exposition

GT2015
June 15 – 19, 2015, Montréal, Canada

GT2015-42284

AN EXPERIMENTAL AND NUMERICAL ASSESSMENT OF AIRFOIL POLARS FOR

USE IN DARRIEUS WIND TURBINES. PART 1 - FLOW CURVATURE EFFECTS

Alessandro Bianchini Francesco Balduzzi John M. Rainbird Joaquim Peiro

Dept. of Industrial Engineering Dept. of Industrial Engineering Dept. of Aeronautical Dept. of Aeronautical
University of Florence University of Florence Engineering, Imperial College Engineering, Imperial College
Via di Santa Marta 3 Via di Santa Marta 3 South Kensington Campus South Kensington Campus
50139, Firenze, Italy 50139, Firenze, Italy SW7 2AZ, London, UK SW7 2AZ, London, UK
Tel. +39 055 275 8773 Tel. +39 055 275 8773 Tel. +44 (0)20 7594 5119 Tel. +44 (0)20 7594 5051
Fax +39 055 275 8775 Fax +39 055 275 8775 Fax +44 (0)20 7594 1974 Fax +44 (0)20 7594 1974
[email protected] [email protected] [email protected] [email protected]

J. Michael R. Graham Giovanni Ferrara Lorenzo Ferrari *

Dept. of Aeronautical Dept. of Industrial Engineering CNR-ICCOM
Engineering, Imperial College University of Florence National Research Council of Italy
South Kensington Campus Via di Santa Marta 3 Via Madonna del Piano 10, 50019,
SW7 2AZ, London, UK 50139, Firenze, Italy Sesto Fiorentino, Italy
Tel. +44 (0)20 7594 5074 Tel. +39 055 275 8777 Tel. +39 055 5225 218
Fax +44 (0)20 7594 1974 Fax +39 055 275 8775 Fax +39 055 5225 203
[email protected] [email protected] [email protected]

ABSTRACT mounted on a Darrieus rotor with a c/R of 0.25. The blade

A better comprehension of the aerodynamic behavior of incidence and lift and drag forces were extracted from the CFD
rotating airfoils in Darrieus Vertical-axis wind turbines output using a novel incidence angle deduction technique.
(VAWTs) is crucial both for the further development of these According to virtual camber theory, the transformed airfoil
machines and for improvement of conventional design tools in this curvilinear flow should be equivalent to the NACA0018
based on zero or one-dimensional models (e.g. BEM models). in rectilinear flow, while the NACA0018 should be equivalent
When smaller rotors are designed with high chord-to- to the inverted transformed airfoil in rectilinear flow.
radius (c/R) ratios so as not to limit the blade Reynolds number, Comparisons were made between these airfoil pairings
the performance of turbine blades has been suggested to be using the CFD output and the rectilinear performance data
heavily impacted by a virtual camber effect imparted on the obtained from the wind tunnel tests and XFoil output in the
blades by the curvilinear flow they experience. form of pressure distributions and lift and drag polars.
To assess the impact of this virtual camber effect on blade Blade torque coefficients and turbine power coefficient are
and turbine performance, a standard NACA0018 airfoil and a also presented for the CFD VAWT using both blade profiles.
NACA0018 conformally transformed such that the airfoil’s
chord line follows the arc of a circle, where the ratio of the NOMENCLATURE
airfoil’s chord to the circle’s radius is 0.25 were considered. AoA Angle of Attack [deg]
For both airfoils, wind tunnel tests were carried out to assess BEM Blade Element Momentum
their aerodynamic lift and drag coefficients for Reynolds c Chord [m]
numbers of interest for Darrieus VAWTs. cD Drag Coefficient [-]
Unsteady CFD calculations have been then carried out to cL Lift Coefficient [-]
obtain curvilinear flow performance data for the same airfoils cP Power Coefficient [-]

* = Contact Author

1 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 cT Torque Coefficient [-] flow is more extreme [3,10-11]. Based on original Migliore et
CFD Computational Fluid Dynamics al.’s theory [10], the blade is virtually transformed into an
D Turbine Diameter [m] equivalent airfoil with a camber line defined by its arc of
k Turbulence Kinetic Energy [m2/s2] rotation. Migliore theorized this effect using non-dimensional
R Turbine Radius [m] theories and did not verified it on real turbines. Recent work
RANS Reynolds-Averaged Navier–Stokes has shown that the associated differences in performance
Re Reynolds Number [-] between symmetrical and cambered airfoils becoming a source
Reθ Momentum Thickness Reynolds Number [-] of error in any analysis using the original blade profile’s data
SST Shear Stress Transport [3]; moreover, it was demonstrated the using corrections to
TSR Tip-Speed Ratio [-] account for this camber can improve accuracy of BEM methods
U Wind Speed [m/s] [3,5,11]. This study sets out to quantify the impact of the effect
VAWTs Vertical Axis Wind Turbines by testing an airfoil alongside a custom made profile
y+ Dimensionless Wall Distance [-] conformally transformed to match the original airfoil’s virtually
cambered equivalent for a given turbine configuration with a
Greek letters c/R=0.25. To this purpose, unsteady CFD simulations were
α Angle of Attack (in formulas) [deg] used to evaluate the aerodynamic performance of the airfoils in
γ Intermittency [-] motion. This was processed into airfoil polars as experienced in
ϑ Azimuthal Angle of the Blade [deg] the curved flow apparent to the VAWT blades and was then
μ Dynamic Viscosity [Pa/s] paired to experimental data for the equivalent transformed
π Pressure Coefficient [-] airfoil taken in a wind tunnel at the same Reynolds numbers.
ρ Air Density [kg/Nm3]
ω Specific Turbulence Dissipation Rate [1/s] ANALYZED AIRFOILS
The investigated airfoils are displayed in Fig. 1.
Subscripts
max Maximum Value
stall Value of a Quantity at the Stall Angle
u Uncorrected

Superscripts
* Normalized Value

INTRODUCTION
Accurate numerical models of Darrieus-type VAWTs could
represent a valuable step towards a wider take up of the turbine
configuration. Most design tools used in the analysis of the
machines are adapted from horizontal-axis methods (e.g. BEM
codes), but there are several differences between the blade Figure 1 - Investigated airfoils.
operation of the two configurations that limit the applicability
of these techniques. In particular, the accuracy of the The profiles chosen are a standard NACA0018 and a
considered airfoils polars becomes crucial to ensure good NACA0018 conformally transformed so that its camber line
predictions of turbines performance [1-5]. In a parallel study follows an arc of a circle such that the ratio of the airfoil’s
[6], the authors investigated the accuracy of input airfoil polars chord to the circle’s radius (c/R ratio) is 0.25.
used to deduce blade loadings in numerical analyses at high The NACA0018 was chosen as it is commonly used in
angles of attack, providing a comparison between new VAWT studies (e.g. [1], [8] and [12]), while the cambered
experimental data and published data commonly used in airfoil was chosen to mimic the “virtual camber” effect [3,10]
VAWTs simulations [7-9]. The low torques produced by imparted on a VAWT blade during operation with a c/R (where
VAWTs at the low tip-speed ratios (TSRs) encountered during R is the radius of the turbine) of 0.25. This value was chosen
start-up mean in fact that even slight inaccuracies in these for comparability to Migliore et al’s analysis of a c/R of 0.26.
polars can have a big impact on modeled performance here. Hereon, the standard NACA0018 is referred to as the
The concern in this study is at higher TSRs, where NACA0018, while the conformally transformed NACA0018 is
modeled blade performance is hindered by differences in the referred to as the transformed airfoil to avoid confusion
flow encountered by VAWT blades compared to that of airfoils between the two. A parallel study has been completed that
under test in wind tunnels when polars are being produced due focuses in detail on the post-stall behavior of these airfoils and
to the curved flow they experience. This is a particular problem methods of extrapolating post-stall data from more common
when modeling turbines with low c/R ratios as the curvature of pre-stall polars [6].

2 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 WIND TUNNEL TESTS
Experiments on the airfoils have been conducted in the
Imperial College 3’ x 3’ low turbulence wind tunnel.
Experimental apparatus
The tunnel test section has a 915 mm x 915 mm cross
section and a length of 2390 mm. The turbulence intensity of
this tunnel is less than 0.1% when empty, making it ideal for
low Reynolds number airfoil testing [13].
A detailed description of the experimental apparatus, the
testing procedure and the data acquisition can be found in Ref.
[6] and it is not reported here for reasons of brevity.
A brief overview of the setup is, however, given below.
Both airfoils had a span of 915mm and a chord of 183mm, Figure 3 - Experimental lift coefficient of the NACA0018 airfoil
giving a chord to tunnel height ratio of 0.2. The aerofoils were between 0° and 180° [6].
mounted vertically in the wind tunnel at the half chord,
between two end plates that sit flush with the tunnel
floor/ceiling, see Fig. 2.

Figure 4 - Experimental drag coefficient of the NACA0018 airfoil

between 0° and 180° [6].

Due to the symmetry of the NACA0018, only AoAs

between 0° and 180° are reported here. For completeness, note
that the lift curve at negative incidences will be anti-symmetric
with respect to the y-axis, whereas the drag curve will be
Figure 2 - The NACA 0018 mounted in the wind tunnel.
symmetric with respect to the same axis.
Results for this study have been corrected for lift
interference, solid blockage and wake blockage using the
formulas given in ESDU 76028 [14], modified for use at
extreme incidence angles as described in Ref. [6].
Results
Results as presented in [6] are reproduced below for
convenience in Fig. 3, Fig. 4, Fig. 5 and Fig. 6.
Figure 3 and Figure 4 contain lift and drag data
respectively for the NACA0018, while Fig. 5 and Fig. 6
contain the same for the transformed airfoil.
Figure 7 finally compares the polars for the two airfoils at
Re=300k, which was the relevant regime for the present study.
Figure 5 - Experimental lift coefficient of the transformed airfoil
between -180° and 180° [6].

3 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Figure 6 - Experimental drag coefficient of the transformed airfoil Figure 8 - Lift coefficient comparison with literature data for the
between -180° and 180° [6]. NACA0018 airfoil (Re=300k).

Figure 9 - Drag coefficient comparison with literature data for the

Figure 7 - Comparison of lift and drag coefficients @ Re=300k NACA0018 airfoil (Re=300k).
between the NACA0018 airfoil and it homologous transformed to
account for virtual camber effects (c/R=0.25). There is good agreement between the results of the present
study and those of Timmer [9] in most of the lift curve, other
The differences in the post-stall region have been than in the immediate vicinity of stall, where stall angles differ.
extensively discussed by Bianchini et al. [6] The attention here Timmer’s occurs at 12°, while present experiments showed
is focused on the AoA range between -30° and +30°, which is attached flow up to 15°. Stall point is notoriously difficult to
considered of particular interest for Darrieus VAWTs, as obtain accurately, being impacted by, for example, airfoil
discrepancies in aerodynamic coefficients can have a notable surface finish, profile accuracy and 3D effects. Details on the
impact on performance prediction using, for example, a BEM manufacture of airfoils for the current study are given in [6],
code [1-5]. but Timmer does not provide this. Both data sets showed
similarly good agreement for the drag coefficient.
COMPARISON WITH PUBLISHED DATA XFoil predictions were fully compatible with both present
Results of the present test campaign were compared to experiments and Timmer data up to approximately 12°. After
similar tests from previous studies, along with results from that point, XFoil starts to make use of internal correlations to
XFoil [15], a tool based on potential flow. As the transformed extrapolate the aerodynamic coefficients and its accuracy is
airfoil was custom-made for this study, no comparable data was notably reduced both for lift and drag predictions.
available, so it has been compared only to XFoil output, As noticed by Du et al. [17] in the recent past, however,
launched by XFLR5 [16]. A Ncrit of 9 (standard wind tunnel both the aforementioned experiments and XFoil predictions
[3,15]) was conservatively used. For comparisons, a reference significantly differed from the famous data of Sheldal &
Reynolds Number of 3.0·105 was used, as this has been Klimas [8], which confirmed a general underestimation of the
commonly used in previous studies [8-9] and is comparable to lift coefficient. This data set was almost exclusively used in the
blade Reynolds numbers in a typical VAWT at rated TSRs in past in many BEM theoretical studies of wind turbines with no
moderate winds. Figure 8 and Figure 9 report the comparison direct assessment of the overall accuracy. Care must be taken,
of lift and drag coefficients for the NACA0018, respectively. however, with Sheldal & Klimas’ data for the NACA 0018 as it
is numerically extrapolated from experimental results for

4 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 thinner profiles. In the vicinity of stall, profile thickness has a As discussed earlier, the investigated geometry is
significant impact on performance, reducing the reliability of analogous to the case presented in [3], except for the single-
these extrapolations here. blade configuration, in place of the three-bladed rotor, for
which the mesh properties were determined accordingly to the
CFD CAMPAIGN results of a specific combined mesh-timestep sensitivity
In order to understand the aerodynamic behavior of the analysis. In detail, six different levels of refinement of the mesh
airfoils when rotating onboard a Darrieus turbine, and compare and three angular time-steps were considered in order to
it with static measurements, CFD simulations were carried out identify the required number of nodes in the mesh surrounding
and analyzed. In detail, the commercial code ANSYS Fluent the airfoil and the total number of mesh elements of the
[18] was used in a two-dimensional form to solve the time computational grid. The mesh settings were defined
dependent unsteady Reynolds-averaged Navier-Stokes (U- accordingly to the results of the grid-independency analysis,
RANS) equations. The fluid is air, modeled as an ideal since they were deemed to guarantee the same level of
compressible gas with standard ambient conditions, i.e. a accuracy.
pressure of 1×105 Pa and a temperature of 300 K. Some of the In particular, an unstructured mesh composed by triangular
authors have presented in recent works [3,19] the assessment elements was used for the discretization of the core flow
and validation of the main settings that have been applied to the region, except for the boundary layer region. A structured O-
CFD simulations, which have also been validated against grid was generated with a row of 50 inflated layers to include
experimental data obtaining good agreement. The results of the the entire boundary layer height. The first element height was
sensitivity analyses on the main parameters are here reported. chosen in order to guarantee that the y+ values from the flow
For additional details on the complete study see Ref. [3,19]. solutions did not exceed the limit of the SST turbulence model,
The unsteady approach requires the division of the i.e. y+ ~ 1. The expansion ratio for the growth of elements
simulation domain into two subdomains in order to allow the starting from the surface was kept below 1.1 to achieve good
rotation of the machine [20-24]: mesh quality. It was proven that a grid-independent behavior
ƒ a circular zone containing the turbine, rotating with the can be obtained by using a discretization of the airfoil surface
same angular velocity of the rotor with 530 nodes. The mesh size of the rotating region, for the
ƒ a rectangular fixed outer zone, determining the overall single-bladed configuration, results in 1.3·105 elements, while
domain extent the stationary region is discretized with 2.0·105 elements.
The two regions communicate by means of a sliding Figure 11, Figure 12 and Figure 13 show some details of
interface. For the definition of the rotor geometry, only the the spatial discretization. The rotating domain, containing the
rotating blade was taken into account, neglecting the presence rotating blade, is characterized by a progressive coarsening of
of both the supporting arms and the shaft. Figure 10 shows the the elements with the distance from the blade. The mesh is
simulation domain, where all the boundary distances are given refined in the region surrounding the blade due to the higher
as a function of the rotor diameter (D). In detail, a velocity-inlet complexity of the flow field. As suggested by [23], a control
boundary condition is imposed at the inlet section, which is circle (Fig. 12), with a diameter equal to 2∙c, was defined
placed 40 rotor diameters upwind of the rotating axis. The around the blade in order to have a better capability to control
ambient pressure condition is imposed at the outlet boundary, the elements size in the region closer to the blade itself. The
being 100 rotor diameters downwind, while a symmetrical extrusion of quadrilateral elements for the discretization of the
condition is defined for the lateral boundaries at a distance of boundary layer is clearly distinguishable in Fig. 13 for the
30 rotor diameters. The dimensions of the domain were blade leading edge.
selected on the basis of the specific sensitivity analysis
discussed in [3,19].

Figure 10 – Simulation domain. Figure 11 – Computational grid for the rotating domain.

5 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 range between 0.135° and 0.42° were used, corresponding to
the cases of lowest and highest TSR respectively.
The Coupled algorithm is employed to handle the pressure-
velocity coupling. The dedicated sensitivity analysis reported in
[19] showed that this algorithm ensured more stable results
when adopting different meshes, timesteps, or rotating speeds.
The second order upwind scheme was used for spatial
discretization of the whole set of RANS and turbulence
equations, as well as the bounded second order for time
differencing, to obtain a good resolution [20-21].
As suggested by the literature [20-24], the global
convergence of each simulation was monitored comparing the
average value of the torque coefficient (cT) over a complete
revolution between two subsequent revolutions. After a specific
sensitivity analysis, the selected threshold for convergence was
identified in a variation lower than the 0.1% of the torque
value, by far lower than the limit commonly adopted in
literature (i.e., 1%). The required number of revolutions is not a
priori known, being dependent on the rotating speed of the
turbine: in the analysis, it was comprehended in a range
between 30 and 50 revolutions.
Concerning the turbulence closure problem, Balduzzi et al.
[3,19,25] showed the effectiveness of the Menter’s shear stress
transport (SST) model in simulations involving unsteady
aerodynamics for VAWTs, as also confirmed by the wide use in
Figure 12 – Computational grid for control circle region: (A) literature [20,23-24]. The suitability of the SST model is related
Transformed airfoil; (B) NACA0018 airfoil. to its formulation [26], being the blending of the standard k-ε
model that is suitable for a shear layer problem and the Wilcox
k-ω model that is suitable for wall turbulence effects. In
particular good predictions of the global machine performance
was obtained in [19,25].
In the present study attention has been focused on a more
detailed examination of the aerodynamic behavior of a single
airfoil in motion by analyzing equivalent static pressure
coefficients on the blade profiles.
Since the prediction of the boundary layer evolution
becomes a critical issue and the blade Reynolds number for the
considered cases cannot guarantee a fully turbulent condition,
the γ-Reθ transition model was implemented. It is a four-
equation model introduced by Menter and Langtry [27-28],
derived from the SST transport equations by adding two other
transport equations, one for the intermittency (γ) and one for
the transition onset criteria, in terms of momentum-thickness
Figure 13 – Computational grid: boundary layer discretization at Reynolds number (Reθ). Lanzafame et al. [29] showed good
the leading edge. agreement between experimental data and numerical results
obtained with the transition SST turbulence model for two
The chosen mesh topology requires a grid-clustering in different types of H-Darrieus turbines.
order to have a smaller spacing between the nodes in In addition, a comparative test was carried out in order to
correspondence of both the leading and trailing edges, being evaluate the difference between the SST and the γ-Reθ
the regions experiencing the highest gradients. approaches, showing substantial consistency of results in terms
The temporal discretization was set to a constant value of both instantaneous torque curves and power coefficient
since it was shown to be almost independent on the revolution characteristics. For the present study, the authors decided to use
speed considered for the simulation. As a result, the angular only the γ-Reθ model, which allowed the evaluation of laminar
timestep becomes directly proportional to the revolution speed to turbulent transition.
of the rotor. In the present analysis, angular timesteps in the

6 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 INCIDENCE ANGLE COMPUTATION coordinates in order to be compared to those previously
In order to reconstruct VAWT blade polars, a robust obtained for an airfoil with horizontal chord. Moreover,
procedure to extrapolate the incidence angle is needed [30]. In they have to be again normalized within -1 and +1 by
conventional aerodynamics, the angle of attack (AoA) for a 2D scaling them by their maximum and minimum values.
airfoil is defined as the geometrical angle between the flow ƒ STEP 4 - For every azimuthal position, the pressure
direction and the chord. This familiar concept makes use, coefficient distribution from CFD can be compared to all
however, of the main simplification of assimilating the airfoil those calculated for the airfoils. By doing so, the
to a point, to which all vectorial quantities (i.e. relative distribution that best fits that from the CFD can be
velocity, forces, etc.) are referred. In CFD calculations of highlighted: this distribution gives an estimate of the
rotating blades, however, the definition of the AoA is less incidence angle experienced by the airfoil in motion. In
straightforward. An overview on the most reliable techniques particular, the position along the chord of the pressure peak
for determining the angle of attack on blades of horizontal-axis is used to define the incidence as the influence of flow
wind turbines has recently been provided by Guntur et al. [31]. speed has been discarded by normalizing the distributions.
In Darrieus turbine simulation the problem is more Unfortunately, as discussed in literature [31], the validity
complex still, as the angle of attack changes constantly during of this approach ceases as soon as the flow is separated. In
the rotation and the airfoils experience a wide range of AoAs, these conditions, no reliable power coefficient distribution can
including significant regions beyond the stall angles. Balduzzi be obtained with XFoil and therefore no comparison can be
et al. proposed a method for the determination of VAWT blade done to define the AoA on the airfoil. For further details on the
AoAs [3] by modifying the averaging technique of Hansen [32] proposed approach, please refer to Bianchini et al. [30].
for use in Darrieus turbines. The velocity triangles on the
blades are reconstructed by evaluating the relative wind speed RESULTS AND DISCUSSION
modulus in a properly positioned area in front of the airfoil and Seven rotating speeds were considered for the simulation
applying an inverse BEM method to estimate the induction of each blade configuration, in order to determine the power
factor. In an inverse BEM method applied to VAWTs the effect coefficient (cP) characteristic.
of velocity reduction and distortion generated by the blades- The cP curves are reported in Fig. 14 as a function of the
flow interaction is globally modeled only by a variation of the tip speed ratio (TSR). An undisturbed wind speed of 8 m/s was
induction factor [30], with no information on the distortion of imposed at the inlet boundary, leading to an investigated
the absolute wind speed. The method is affected by an intrinsic operating range from TSR=1.05 to TSR=4.71. As discussed in
uncertainty due to the selection of the averaging zone when the Balduzzi et al. [3], it was here confirmed that the blade
relative wind speed is calculated [3]. operating with the Transformed airfoil provides a higher power
On these bases, in the present study, the novel approach coefficient with a shift of the curve peak towards lower TSR.
developed by Bianchini et al. [30] was adopted and here briefly As a general remark, it is worth highlighting that the two
summarized. The method is basically made of four steps: curves are quite different, stressing the importance of a proper
ƒ STEP 1 - Based on the chord-to-radius ratio of the rotor blade design criteria in case of high c/R ratio [3].
(c/R) and the tip-speed ratio (TSR), the virtual airfoil due
to flow-curvature effects is defined based on the conformal
transformations of Migliore et al [10]. This main
assumption, first proposed by Balduzzi et al. [3] will be
demonstrated later in the present study.
ƒ STEP 2 - After the apparent Reynolds number on the
airfoils has been estimated, the pressure coefficient
distributions over the virtual airfoil must be defined for a
wide range of AoA with a relatively fine step either with a
dedicated CFD campaign (very time consuming) or with
simple codes based on panel methods (e.g. XFoil [15]).
Then, all the pressure coefficient distributions are
normalized within -1 and +1 by scaling them by their
maximum and minimum values. This solution allows
comparison of pressure distributions, which depend only
on the incidence angle with a negligible error on the exact Figure 14 – Power coefficient vs TSR for the hypothetical 1-blade
relative speed, which can be hard to define from CFD rotor using either the NACA0018 or the Transformed airfoil.
calculations [30].
ƒ STEP 3 - The pressure coefficient distributions from CFD The differences between the two airfoils can be
must be acquired from calculations at different azimuthal appreciated by analyzing the associated trends of the torque
positions and reported by a transformation of the coefficients over a revolution (Fig. 15 and Fig. 16).

7 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 distributions obtained from XFoil. For the convention of signs
and reference systems please refer to Fig. 17.

Figure 15 – Instantaneous torque coefficient over a revolution for

different TSRs: NACA0018.

Figure 17 – Signs and reference systems convention.

Based on the results shown by Balduzzi et al. [27], the

pressure coefficient distributions over the airfoils were
expected to reproduce those of the corresponding virtual airfoil
coming from conformal transformation. As mentioned, the
transformed airfoil arranged with its camber inward becomes a
virtual NACA0018 when rotated about a radius four times its
chord length, while the NACA0018 becomes a virtual
transformed airfoil with its camber outward in similar
conditions (see Fig. 17). Conversely, no correspondence could
be found with the pressure coefficient distributions of the same
airfoil in straight flow. For example, in Fig. 18 CFD data of the
Figure 16 – Instantaneous torque coefficient over a revolution for transformed airfoil in terms of normalized pressure coefficients
different TSRs: Transformed airfoil. (π*) were compared to XFoil predictions at different AoAs.

It is apparent that the energy extraction with the

transformed airfoil (which is arranged with its camber curving
towards the center of rotation, and is expected to behave like
the symmetric NACA0018 in curvilinear flow) is more
balanced between the upwind and the downwind halves of the
revolution. On the other hand, the geometrical NACA0018
(which in turn is expected to behave like the transformed airfoil
with camber curved outward away from the center of rotation)
concentrates the torque extraction in the upwind zone,
providing higher local torque coefficients [27].
To reconstruct the airfoil polars, two revolution regimes,
corresponding to TSR=2.09 and TSR=3.14, respectively, were
analyzed as they were characterized by local blade Reynolds
numbers between 250·103 and 300·103, comparable to that of
the new experimental data collected in the wind tunnel. As Figure 18 – Normalized pressure coefficients (π*) at ϑ=32.8° for
discussed, due to the limitations imposed by the approach for the transformed airfoil: CFD vs. XFoil predictions.
the determination of the incidence angle, the analysis was
restricted to an azimuthal angle range of approximately ϑ=-10° It is apparent that the CFD distribution is notably different
to ϑ=+70°. In this range, the AoA is generally sufficiently from any of those predicted by XFoil at the same Reynolds
reduced to ensure attached flow on the airfoil and so enabling number. In particular, even if for a given AoA, the negative
processing through the analysis of pressure coefficient pressure peak or the transition point are located in the same

8 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 chord position than CFD, the rest of the profile has no
coherence, being notably higher along the pressure side as
typical for a cambered profile. Conversely, the CFD trend,
although produced by a cambered airfoil, exhibits the typical
behavior of a symmetric airfoil, with a reduced pressure
gradient between the pressure and the suction sides.
Expectations were confirmed by the simulations, as shown
by Fig. 19 and Fig. 20 which report the normalized pressure
coefficients (π*) comparisons in case of the transformed and
the NACA0018 airfoils, respectively, for two azimuthal angles
of 32.8° and 58.3° degrees. Comparable agreement was found
in all the other considered azimuthal positions.

Figure 20 – Normalized pressure coefficients (π*) at two

azimuthal positions: comparison between CFD for the
NACA0018 and XFoil predictions of the equivalent virtual
transformed airfoil.

With the same approach, used in Fig. 19 and Fig. 20, the
data at all the azimuthal positions were processed to reproduce
the equivalent polars of the two airfoils during their revolution
(Fig. 21, Fig. 22, Fig. 23 and Fig. 24).

Figure 19 – Normalized pressure coefficients (π*) at two

azimuthal positions: comparison between CFD for the
transformed airfoil and XFoil predictions of the equivalent
virtual NACA0018.

Figure 19 demonstrates that, if the transformed airfoil is

used in a turbine, its aerodynamic performance will almost
perfectly reproduce the NACA0018 behavior at the same
Reynolds number, with a similar opposite relationship true in
the case of a NACA0018 blade (Fig. 20).
Moreover, it is worth noticing that the transitional
turbulence model also allowed a very accurate prediction of the Figure 21 – CFD computed lift curve of the transformed
position of the transition on the airfoil. airfoil and experimental data and XFoil predictions of the
NACA0018 @ Re=300k.

9 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 The figures demonstrate that the effects of flow curvature
first proposed by Migliore et al [10] are clearly relevant to the
CFD simulation of the VAWT. There is an excellent agreement
between the data reconstructed from the CFD turbine blade
performance and each blades equivalent virtual airfoil, whether
calculated using XFoil or obtained experimentally. The lift
coefficients are particularly well reproduced, in terms of both
the slope of the linear region and the lift coefficient at AoA=0°.
Small discrepancies can be seen in the drag coefficient
produced by the CFD computed NACA0018 airfoil, which
seems to increase more rapidly than that of the virtual
transformed airfoil, although the very low absolute value of the
drag is more sensitive to small errors produced by the analysis.
These results confirm the blade design criteria proposed by
Figure 22 – CFD computed drag curve of the Transformed Balduzzi et al. [3] and support Migliore et al’s earlier work
airfoil and experimental data and XFoil predictions of the [10]. The strong agreement between CFD, XFoil and
NACA0018 @ Re=300k. experiments greatly increases the confidence that can be placed
in the results of this study. The experimental polars are judged
to be of sufficient accuracy for future use in VAWT BEM
codes, which can be highly sensitive to inaccuracies in input
blade data.
As an example, Fig. 25 shows the comparison between the
CFD torque coefficient trend at TSR=3.14 of the transformed
airfoil and the BEM predictions using the experimental polars
either of the NACA0018 (i.e. with virtual camber effect) or of
the transformed airfoil. BEM predictions were obtained using
the VARDAR code of the University of Florence [3,5,11].
Upon examination of the figure, it is apparent that, even if
coming from a simplified modeling technique, the BEM
predictions using the polar of the virtually transformed airfoil
nicely fit the trend coming from the 2D CFD simulation, with
sound agreement on both the peak values and the location of
Figure 23 – CFD computed lift curve of the NACA0018 and the maximum torque output. On the other hand, if the turbine is
experimental data and XFoil predictions of the simulated with the BEM theory using the polars of the
Transformed airfoil @ Re=300k. “geometrical” airfoil, the agreement becomes poor as both the
maximum lift and the maximum torque location are not
accurately predicted.

Figure 24 – CFD computed drag curve of the NACA0018

and experimental data and XFoil predictions of the Figure 25 – Torque coefficient trends @ TSR=3.14: CFD
Transformed airfoil @ Re=300k. simulation of the transformed airfoil vs. BEM predictions
using different polars.

10 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Finally, Fig. 26 reports the azimuthal positions at which In order to obtain curvilinear flow performance data, CFD
the incidence angles were measured for the two revolution simulations of a VAWT with a c/R of 0.25 were completed and
regimes. In the figure, the notation for the incidence angle in blade incidence and lift and drag forces extracted from the CFD
the machine reference system is used, following the convention output. These were compared to the rectilinear data obtained
of Fig. 17, whereas previous figures were reported in the airfoil from the wind tunnel and XFoil. The airfoils were analyzed in
reference system, i.e. with a positive incidence with a flow pairings, with the transformed airfoil in a curvilinear flow
inclined towards the concave side of the airfoil. being equivalent to a “virtual” NACA0018 in rectilinear flow,
while the NACA0018 in a similarly curved flow is equivalent
to a “virtual” inverted transformed airfoil in rectilinear flow.
Data is presented comparing pressure distributions taken
about the CFD VAWT blades and through XFoil simulations of
the equivalent “virtual” airfoil. Comparisons are also presented
of lift and drag polars from the CFD VAWT blades, wind
tunnel tests and XFoil for the equivalent “virtual” airfoils.
Agreement is extremely promising across all of these,
suggesting that the virtual camber effect does significantly
impact VAWT blade performance.
This proves that by replacing turbine blade lift and drag
data with an equivalent “virtual” airfoil’s data, improvements in
the accuracy of analysis of VAWTs through, for example BEM
codes, should be possible.
Blade torque coefficients and turbine performance
Figure 26 – AoA vs. azimuthal angles for the investigated coefficients are also presented for the CFD VAWT using both
configurations. blades. These show better performance for the cambered
airfoil, suggesting that virtual camber negatively impacts
It is indeed interesting to notice that, although the two performance (note that the cambered blade corrects the effect
polars correctly matched, in the two revolution regimes the of the virtual camber, since its “virtual” pairing is symmetrical).
AoA are achieved at different azimuthal angles as obtained
with different combinations of the peripheral speed (function of ACKNOWLEDGMENTS
the TSR) and the wind speed. Thanks are due to Prof. Ennio Antonio Carnevale of the
It can be seen that the NACA0018 has an almost null University of Florence for supporting this activity.
incidence at ϑ=0°, as is to be expected. When overlaid, the
upper and lower surfaces of the transformed airfoil are REFERENCES
coincident with those of the NACA0018 at the quarter chord, [1] Paraschivoiu, I., 2002, Wind turbine design with emphasis
with the leading and trailing edges curving down below the on Darrieus concept, Polytechnic International Press,
NACA0018 equivalents. The angle between the chord lines of Montreal, Canada.
the two airfoils in this overlay can be calculated as 3.6°, [2] Kirke, B.K., 1998, “Evaluation of self-starting vertical
approximately equal to the incidence of the transformed airfoil axis wind turbines for standalone applications,” Ph.D.
at ϑ=0°, again as is to be expected. thesis, Griffith University, Gold Coast, Australia.
It can also be noticed that, by increasing the TSR (i.e. with [3] Balduzzi, F., Bianchini, A., Maleci, R., Ferrara, G. and
a narrower range of incidence angles [1]), an increased number Ferrari, L., 2015, “Blade design criteria to compensate the
of points was available for the analysis and, in general, a given flow curvature effects in H-Darrieus wind turbines,”
AoA was reached at higher azimuthal angles, as the rotational Journal of Turbomachinery, 137(1), pp. 1-10.
velocity comes to dominate the velocity triangle. [4] Marten, D., Wendler, J., Pechlivanoglou, G., Nayeri, C.N.
and Paschereit C.O., 2013, “Qblade: an open source tool
CONCLUSIONS for design and simulation of horizontal and vertical axis
Wind tunnel tests and XFoil simulations have been wind turbines,” International Journal of Emerging
completed on a standard NACA0018 airfoil and a NACA0018 Technology and Advanced Engineering (IJETAE),
airfoil conformally transformed such that the airfoil’s chord 3(special issue 3), pp. 264-269.
line follows the arc of a circle, where the ratio of the airfoil’s [5] Bianchini, A., Ferrari, L. and Magnani, S., 2011, “Start-up
chord to the circle’s radius is 0.25. The profiles were chosen to behavior of a three-bladed H-Darrieus VAWT:
test the relevance to VAWT analysis of the flow curvature experimental and numerical analysis,” Proc. of the ASME
effects imparted on a blade in curvilinear flow when compared Turbo Expo 2011, Vancouver, Canada, June 6-10, 2011.
to a blade in rectilinear flow. [6] Bianchini, A., Balduzzi, F., Rainbird, J., Peiro, J.,
Graham, J M.R., Ferrara, G. and Ferrari, L., 2015, “An

11 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Experimental and Numerical Assessment of Airfoil Polars [21] Howell, R., Qin, N., Edwards, J., and Durrani, N., 2010,
for Use in Darrieus Wind Turbines. Part 2 - Post-Stall “Wind Tunnel and Numerical Study of a Small Vertical
Data Extrapolation Methods,” Proc. of the ASME Turbo Axis Wind Turbine,” Renewable Energy, 35(2), pp. 412–
Expo 2015, Montreal, Canada, June 15-19, 2015. 422.
[7] Du, L., Berson, A. and Dominy, R.G., 2014, “NACA0018 [22] Beri, H., and Yao, Y., 2011, “Effect of Camber Airfoil on
behaviour at high angles of attack and at Reynolds Self Starting of Vertical Axis Wind Turbine,” J. Environ.
numbers appropriate for small wind turbines,” ECS Sci. Technol., 4(3), pp. 302–312.
Technical Report 2014/08. [23] Raciti Castelli, M., Englaro, A., and Benini, E., 2011,
[8] Sheldahl, R.E., and Klimas, P.C., 1981, “Aerodynamic “The Darrieus Wind Turbine: Proposal for a New
Characteristics of Seven Symmetrical Airfoil Sections Performance Prediction Model Based on CFD,” Energy,
through 180-degree Angle of Attack for Use in 36(8), pp. 4919–4934.
Aerodynamic Analysis of Vertical Axis Wind Turbines,” [24] Rossetti, A., and Pavesi, G., 2013, “Comparison of
National Renewable Energy Laboratory, Golden, Different Numerical Approaches to the Study of the H-
Colorado (USA), tech. rep. SAND80-2114. Darrieus Turbines Start-Up,” Renewable Energy,
[9] Timmer, W.A., 2008, “Two-dimensional low-Reynolds 50(February), pp. 7–19.
number wind tunnel results for airfoil NACA0018,” Wind [25] Balduzzi, F., Bianchini, A., Gigante, F.A., Ferrara, G.,
Engineering, 32(6), pp. 525-537. Campobasso, M.S., and Ferrari, L., 2015, “Parametric and
[10] Migliore, P.G., Wolfe, W.P. and Fanucci, J.B., 1980, “Flow Comparative Assessment of Navier-Stokes CFD
Curvature Effects on Darrieus Turbine Blade Methodologies for Darrieus Wind Turbine Performance
Aerodynamics,” Journal of Energy, 4(2), pp. 49-55. Analysis,” Proc. of the ASME Turbo Expo 2015,
[11] Bianchini, A., Ferrari, L. and Carnevale, E.A., 2011, “A Montreal, Canada, June 15-19, 2015.
model to account for the Virtual Camber Effect in the [26] Menter, F.R., 1994, “Two-Equation Eddy-Viscosity
Performance Prediction of an H-Darrieus VAWT Using Turbulence Models for Engineering Applications,” AIAA
the Momentum Models,” Wind Engineering, 35(4), pp. J., 32(8), pp. 1598–1605.
465-482. [27] Menter, F.R., Langtry, R.B., Likki, S.R., Suzen, Y.B.,
[12] Islam, M., Ting, D. and Fartaj, A., 2007, “Desirable Huang, P.G., and Völker, S., 2004, “A Correlation-based
Airfoil Features for Smaller-Capacity Straight-Bladed Transition Model Using Local Variables Part 1 – Model
VAWT,” Wind Engineering, 31(3), pp. 165–196. Formulation,” Proc. of the ASME Turbo Expo 2004,
[13] Selig, M.S., Deters, R.W. and Williamson, G.A., 2011, Vienna, Austria, June 14-17, 2004.
“Wind Tunnel Testing Airfoils at Low Reynolds [28] Langtry, R.B., and Menter, F. R., 2009. “Correlation-
Numbers,” Proc. of the 9th AIAA Aerospace, Sciences based transition modeling for unstructured parallelized
Meeting, AIAA, 875, pp. 4-7. computational fluid dynamics codes,” AIAA Journal,
[14] ESDU, “Lift-interference and blockage corrections for 47(12), pp. 2894–2906.
two-dimensional subsonic flow in ventilated and closed [29] Lanzafame, R., Mauro, S., Messina, M., 2014, “2D CFD
wind-tunnels,” Technical Report 76028, Engineering Modeling of H-Darrieus Wind Turbines Using a
Sciences Data Unit, 1978. Transition Turbulence Model,” Energy Procedia, 45, pp.
[15] XFoil User Guide, available online at: 131-140.
http://web.mit.edu/drela/Public/web/xfoil, last access [30] Bianchini, A., Balduzzi, F., Ferrara, G. and Ferrari, L.,
02/10/2014 2015, “A computational procedure to define the angle of
[16] www.xflr5.com/xflr5.htm, last access 13/10/2014. attack on airfoils rotating around an axis orthogonal to
[17] Du, L., Berson, A. and Dominy, R.G., 2014, “NACA0018 flow direction,” paper submitted for publication to
behaviour at high angles of attack and at Reynolds Renewable Energy.
numbers appropriate for small wind turbines,” ECS [31] Guntur, S., 2012, “Evaluation of several methods of
Technical Report 2014/08. determining the angle of attack on wind turbine blades,”
[18] Ansys, Inc., 2013, “Fluent Theory Guide,” release 14.5. 4. Proc. of The science of Making Torque from Wind 2012,
[19] Balduzzi, F., Bianchini, A., Maleci, R., Ferrara, G. and Oldenburg, Germany, 09 October 2012.
Ferrari, L., 2014, “Critical issues in the CFD simulation of [32] Hansen, M.O.L., Sørensen, N.N., Sørensen, J.N., and
Darrieus wind turbines,” Paper submitted for publication Michelsen J.A., 1997, “Extraction of lift, drag and angle
to Renewable Energy. of attack from computed 3-D viscous flow around a
[20] Maître, T., Amet, E., and Pellone, C., 2013, “Modeling of rotating blade,” Proc. of the EWEC 1997, Dublin
the Flow in a Darrieus Water Turbine: Wall Grid (Ireland), pp. 499–501.
Refinement Analysis and Comparison with Experiments,”
Renewable Energy, 51, pp. 497–512.

12 Copyright © 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 01/19/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use