Energy Conversion and Management 126 (2016) 790798

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

A computational procedure to define the incidence angle on airfoils

rotating around an axis orthogonal to flow direction
Alessandro Bianchini a, Francesco Balduzzi a, Giovanni Ferrara a, Lorenzo Ferrari b,
a
Dept. of Industrial Engineering, University of Florence, Via di Santa Marta 3, 50139 Firenze, Italy
b
CNR-ICCOM, National Research Council of Italy, c/o Dept. of Industrial Engineering, University of Florence, Via di Santa Marta 3, 50139 Firenze, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Numerical simulations provided in the last few years a significant contribution for a better understanding
Received 2 May 2016 of many phenomena connected to the flow past rotating blades. In case of airfoils rotating around an axis
Received in revised form 25 July 2016 orthogonal to flow direction, one of the most critical issues is represented by the definition of the inci-
Accepted 7 August 2016
dence angle on the airfoil from the computed flow field. Incidence indeed changes continuously as a func-
Available online 27 August 2016
tion of the azimuthal position of the blade and a distribution of peripheral speed is experienced along the
airfoils thickness due to radius variation. The possibility of reducing the flow to lumped parameters (rel-
Keywords:
ative speed modulus and direction), however, would be of capital relevance to transpose accurate CFD
Incidence angle
Angle of attack
numerical results into effective inputs to low-order models that are often exploited for preliminary
CFD design analyses. If several techniques are available for this scope in the case of blades rotating around
Computation an axis parallel to flow direction (e.g., horizontal-axis wind turbines), the definition of a robust procedure
Vertical-axis wind turbine in case the revolution axis is orthogonal to the flow is still missing.
Darrieus In the study, a novel technique has been developed using data from Darrieus-like rotating airfoils. The
method makes use of the virtual camber theory to define a virtual airfoil whose pressure coefficient dis-
tributions in straight flow are used to match those of the real airfoil in curved flow. Even if developed
originally for vertical-axis wind turbines, the method is of general validity and is thought to represent
in the near future a valuable tool for researchers to get a new insight on many complex phenomena con-
nected to flow past blades rotating around an axis orthogonal to flow direction, like for example dynamic
stall.
2016 Elsevier Ltd. All rights reserved.

1. Introduction relevance in low-order simulation methods [1], like for example

In conventional aerodynamics, the incidence angle or angle of
attack (AoA) for a 2D airfoil is defined as the geometrical angle
between the flow direction and the blade chord. This very familiar
concept makes use, however, of the main simplification of assimi-
lating the airfoil to a point, to which all vectorial quantities (e.g.,
relative velocity, forces, etc.) are referred.
The concept of incidence angle is widely used in many turbo-
machinery applications and, of course, in several analyses con-
nected to wind turbines, which are used in the present work as a
test case for developing the new method. In particular, the possi-
bility of reducing the flow characteristics to a few lumped param-
eters (i.e. relative flow velocity modulus and direction) is of capital
Corresponding author.
E-mail addresses: bianchini@vega.de.unifi.it (A. Bianchini), balduzzi@vega.de.
unifi.it (F. Balduzzi), giovanni.ferrara@unifi.it (G. Ferrara), lorenzo.ferrari@iccom.
cnr.it (L. Ferrari).
the BEM theory [2,3], and particularly in aero-elastic engineering
models, where the AoA is used as an input to enter tabulated polars
of the airfoils in the database and hence to calculate the forces act-
ing on a blade [46]. These tabulated data very often come from
wind tunnel measurements, where the main concern is to provide
good-quality rectilinear flows. This makes, however, them not
always fully representative of the airfoils behavior onboard rotat-
ing blades, where many complex phenomena interfere with the
aerodynamics of the blades themselves (e.g., 3D effects from tip
and root vortices, dynamic stall, etc.) [57].
In this view, increasingly accurate Computational Fluid
Dynamic (CFD) simulations could provide a valuable contribution
to a deeper understanding of the rotating airfoils behavior [8,9].
To transpose numerical results into more practical non-
dimensional aerodynamic parameters (e.g., lift and drag coeffi-
cients [10]), however, the lift and drag forces easily obtainable
from calculated pressure distributions over the airfoils must be
referred to the correct incidence angle [11].

http://dx.doi.org/10.1016/j.enconman.2016.08.010
0196-8904/ 2016 Elsevier Ltd. All rights reserved.
 A. Bianchini et al. / Energy Conversion and Management 126 (2016) 790798
Nomenclature
a induction factor
AoA angle of attack, deg
BEM blade element momentum
c airfoil chord, m
cL lift coefficient
cD drag coefficient
Cn normal force coefficient
Ct tangential force coefficient
CFD computational fluid dynamic
Fn normal force, N
Ft tangential force, N
HAWT horizontal axis wind turbine
y+ dimensionless wall distance
R turbine radius, m
Re Reynolds number
In conventional numerical simulations, the airfoil is generally
kept fixed while the incidence is imposed by defining an inclina-
tion angle for the freestream velocity at the inlet boundary (e.g.,
[12,13]). In the case of rotating blades, the definition of the relative
velocity on airfoils (and hence of the incidence angle) is indeed
more complex. The oncoming flow relative to blades is the vector
sum of the absolute flow and the peripheral speed and it can be
heavily affected by secondary flow structures, which make the def-
inition of a unique velocity vector extremely complex. The above
issues are even more challenging in the case of blades rotating
around an axis orthogonal to the oncoming flow (cycloidal
motion). In this case, the peripheral speed has a non-uniform dis-
tribution along the thickness of the airfoil [14], affecting both the
streamlines direction and the correct assessment of the relative
speed modulus.
With particular reference to vertical-axis wind turbines
(VAWTs), which are used in the present work as a test case to
explain the proposed method and show its potential applications,
in the recent past CFD data have been often reduced by assuming
a theoretical incidence angle coming from blade element momen-
tum (BEM) theory (e.g., [15]). This expedient, however, somehow
disabled the accuracy enhancement provided by CFD. Under these
preconditions, in the present study a novel method to define the
incidence angle from a computed CFD field past blades rotating
around an axis orthogonal to flow direction is presented. The
method, here applied to airfoils rotating in a Darrieus-wind-
turbine-like way, is thought of general validity for any blade type
experiencing the same flow conditions, e.g., helicopter blades,
Wells turbines, etc.
2. Background
The concern of defining the angle of attack from CFD simula-
tions of rotating blades has been historically addressed by wind
turbine specialists in case of horizontal-axis wind turbines
(HAWTs). In this study, attention is instead focused on blades
rotating in a plane parallel to flow direction. These applications
are primarily connected to Darrieus-type vertical-axis wind tur-
bines (VAWTs), which are experiencing a renewed interest by both
manufacturers and researchers in the wind energy world [3,16
18]. Darrieus turbines are used in the present work as a test case
for showing the method, but, as discussed, the proposed approach
is thought to be of general validity in case of any airfoil rotating
around an axis orthogonal to flow direction.
791
TSR tip-speed ratio
U wind speed, m/s
VAWT vertical axis wind turbine
W relative velocity, m/s
Greek letters
a AoA (in formulas), deg
c intermittency
# azimuthal angle, deg
p pressure coefficient
q flow density, kg/N m3
x revolution speed, rad/s
1 value at infinity
normalized value
2.1. Proposed techniques for horizontal-axis wind turbines
An interesting overview on the most reliable techniques to
define the angle of attack on HAWTs blades has been recently pro-
vided by Guntur et al. [20].
The historically most exploited technique for calculating the
AoA from a computed flow-field is represented by a sort of inverse
BEM (Blade Element Momentum) approach [1,19,20]. In this
approach, the angle of attack is calculated by means of well-
known Eqs. (1) and (2), in which the forces are a priori imposed
based either on experimental measurements or on computed
values.
1
F t # qW 2# cHcL sin a  cD cos a 1
2
1
F n # qW 2# cHcL cos a  cD sin a 2
2
It is widely acknowledged that this method gives reasonably
reliable results, which can be also again exploited in BEM codes
for later predictions [1,19]. On the other hand, the major limitation
of the approach is the use of a 1D theory. In particular, it is worth
noticing that the direction of the oncoming absolute wind is con-
sidered not to be altered by the blade-flow interaction [21], with
notable errors in the case of high deflections. Moreover, BEM mod-
els always need a tip-correction, whose accuracy becomes in turn
relevant for a correct evaluation of airfoils data.
A second technique, again developed for HAWTs, is the averag-
ing technique, applied by Hansen and Johansen [22] and Johansen
et al. [5,6]. In this approach, the local angle of attack is calculated
by reconstructing the velocity triangle on the blade: the peripheral
speed is indeed known in modulus and direction, whereas the rel-
ative speed is extrapolated from CFD calculations by analyzing the
flow in the rotor plane. As discussed by Zhong Shen et al. [1], how-
ever, since the method exploits averaged data, many points of the
computational domain must be analyzed to describe the local flow
features. Moreover, this approach is barely applicable to more gen-
eral flow conditions (e.g., a yawed flow).
On this basis, Zhong Shen et al. [1] more recently proposed a
further and more general technique, using several cross-sections
at different span positions of an HAWT, in which the velocity is
read in a specifically positioned control point and the induced
velocity is calculated by means of the associated circulations from
the estimated lift and drag forces on the blades. The same theoret-
ical approach is also reported by Guntur and Srensen [19].
792
Finally, a last technique is that reported in different ways by
both Guntur et al. [21] and Dumitrescu et al. [23], in which the
incidence angle is calculated by comparing the pressure distribu-
tion over the blade at different sections with existing data for the
selected airfoils. Guntur et al. [21], however, claimed this method
to be the less reliable for HAWTs based on wind tunnel experimen-
tal data.
2.2. Previous studies on blades rotating around an axis orthogonal to
flow direction
In the case of blades rotating around an axis orthogonal to flow
direction (e.g., in Darrieus wind turbines [18]), the problems to be
solved to correctly describe the flow field developing around the
turbine are even more complex. The incidence angle changes con-
stantly during the revolution and each airfoil experiences a wide
range of AoAs, including significant regions beyond the static stall
angle [24]. A reliable estimation of the incidence in similar applica-
tions would provide, however, great benefits for a more exhaustive
understanding of many phenomena connected to the complex
aerodynamic interactions, e.g., the description of dynamic stall
hysteresis cycles [25] or flow curvature effects (virtual camber
and virtual incidence) [2628].
A first approach to the determination of the AoA in vertical-axis
wind turbines (VAWTs) was proposed by Balduzzi et al. [11]. In
detail, the authors fitted the averaging technique of Hansen and
Johansen [22] to be used in Darrieus turbines. In their approach,
the velocity triangles on the blades (for sign convention please
refer to Fig. 1) are reconstructed by evaluating the relative wind
speed modulus in a properly positioned area in front of the airfoil
and applying an inverse BEM method to estimate the induction
factor (Eqs. (3) and (4)).
q more distorted when passing through it, hence making the approx-
U 1 1  a sin # U 1 1  a cos # xR2
2
w 3
 
U 1 1  a sin #
a arctan
U 1 1  a cos # xR
Once the AoA has been estimated, the lift and drag coefficient
can be obtained reversing Eqs. (1) and (2), as normal and tangential
forces can be straightforwardly obtained by CFD data by integrat-
ing the pressure distribution over the airfoils. The authors success-
fully applied in [11] the method to a preliminary study on flow
curvature effects in Darrieus turbines, obtaining a significant vali-
dation of the original theoretical model proposed by Migliore et al.
[14]. The validity of the approach was, however, specifically lim-
ited to the azimuthal angles in which the flow is attached to blades,
so that the oncoming relative velocity can be defined clearly by
means of a streamlines analysis. Moreover, the approach accuracy
is hindered by some intrinsic limitations, discussed in the follow-
ing section.
Fig. 1. Conventions for reference systems and AoA.
A. Bianchini et al. / Energy Conversion and Management 126 (2016) 790798
More recently, a proposal was made by Saverin and Frank [27],
in which the incidence angle is directly computed as the average
angle between the normal and tangential flow velocity compo-
nents in a rectangular region around the airfoil. This method still
presents the advantage of using only CFD data for the AoA determi-
nation but seems quite rough, as great errors can be introduced in
the analysis in case of vortices or detached flows, whose local
velocity components can dramatically affect the averaged AoA
value.
2.3. Limitations of an inverse BEM technique
An inverse BEM method applied to VAWTs or, in general, to
blades rotating around an axis orthogonal to flow direction, still
suffers from all of the limitations of a 1D model. In particular, in
a BEM-model perspective, the effect of velocity reduction and dis-
tortion generated by the blades-flow interaction is globally mod-
eled only by a variation of the induction factor (Eq. (5)).
! ! !
W x R U 1 1  a 5
Even if this simplification has historically led to sufficiently reli-
able results in terms of global turbine performance estimation, the
description of the local flow field is very rough as the strong inter-
action between the upwind and the downwind halves of the revo-
lution makes the flow deflection even more pronounced [28].
For example, Fig. 2 displays the dimensionless velocity contours
around the three-bladed turbine studied by Balduzzi et al. [11]
rotating at TSR = 3.3. The velocity was normalized by the free-
stream velocity U1.
Upon examination of the figure, it is apparent that the absolute
velocity is strongly deflected when approaching the rotor and even
imation of Eq. (5) very relevant for a correct prediction of the AoA.
Moreover, beyond physical approximations, this method is
affected by an intrinsic uncertainty due to the selection of the aver-
4 aging zone where the relative wind speed is calculated. For exam-
ple, in Fig. 3 two azimuthal positions are compared for the
aforementioned rotor, corresponding to # = 10 and # = 54, respec-
tively. In the figures, the averaging zone used by Balduzzi et al. [11]
is also reported.
Fig. 2. Normalized velocity contours and streamlines around the turbine of [11] at
TSR = 3.3.
 A. Bianchini et al. / Energy Conversion and Management 126 (2016) 790798 793

Fig. 3. Normalized relative speed on an airfoil of the turbine of [11] at # = 10 and # = 54, respectively.

In the figure, the relative speed was normalized by the mean
value in the zone itself; by doing so, the contours actually repre-
sent the relative error in the estimation of the modulus of W. The
figures clearly show that in some azimuthal positions the identifi-
cation of the streamline corresponding to the hypothetical relative
speed is extremely difficult, making the selection of the zone loca-
tion and dimensions pivotal for a correct prediction. In particular,
assuming that the considered zone was able to correctly define
the relative speed (it was indeed verified in the present study), it
is apparent that there are several azimuthal positions (e.g.,
# = 54) where the sensitivity of the results on the averaging zone
is reduced. On the other hand, many other positions (e.g., # = 10)
are characterized by velocity fields in which a shift of the averaging
zone is connected to large variations of the relative speed estima-
tions, making then the impact of the operators choices relevant to
the final result. Fig. 3 finally suggests that, even if the operator was
able to properly select the averaging zone, the method must be
limited to those azimuthal angles in which the flow is clearly
attached to the blades and no disturbances are present in the flow
approaching the blades.
a circle having a radius equal to that of the revolution trajectory
(Fig. 4). This behavior has been recently assessed by Rainbird
et al. [26]. Even if, of course, the airfoils geometry is not modified
during the revolution, the aforementioned effect on their aerody-
namic performance is of particular reference for the calculation
of the incidence angle, as it will be shown later on in the study.
Stated the above, the virtual airfoil is obtained by a conformal
transformation procedure based on the chord-to-radius ratio of the
rotor (c/R) and the ratio between the peripheral speed and the
mainstream velocity (tip-speed ratio, or TSR, for wind turbines).
Equations for the conformal transformation are not reported here
for brevity reasons, but can be easily found in [31]. As one may
argue, the difference between the real airfoil and its virtual homol-
ogous is progressively reduced when the c/R ratio and/or the tip-
speed ratio are reduced [26]. For example, according to Migliore
and Wolfe [31], the virtual camber effect in symmetrical NACA
4-digits airfoils is almost negligible (i.e. induced camber lower
than 0.2%) below c/R = 0.025. The analysis of the virtual airfoil,
however, constantly increases the accuracy of the analyses in
blades rotating around an axis orthogonal to flow direction [11,26].

3. Proposed method

Within these preconditions, a novel method to estimate the

incidence from CFD simulations was developed. The model, specif-
ically conceived for airfoils rotating around an axis orthogonal to
flow direction (e.g., Darrieus VAWTs), combines the use of pressure
coefficient profiles [29] with the use of flow curvature corrections
that have been highlighted as a major point for this type of appli-
cations [11,14,26,30]. The theoretical approach withstanding the
model has been successfully applied by some of the authors to
the analysis of virtual camber [26] and virtual incidence effects
[28], which both needed an estimation of the incidence angle on
the airfoil. From those studies, the model has been further
improved and made more robust and systematic. The whole proce-
dure is made of four, easily automatable, main steps.

3.1. Step 1

Based on the original Migliores theory [14], an airfoil in motion

in a cycloidal way behaves like a virtual one in straight flow having
the same thickness distribution but the camber line fitting an arc of Fig. 4. Virtual camber effect.
794 A. Bianchini et al. / Energy Conversion and Management 126 (2016) 790798
3.2. Step 2
As a second step, the attended flow conditions on the airfoils are
estimated. In further detail, since the conventional aerodynamic
coefficients (lift, drag and moment) are strongly affected by the
Reynolds number (Re), a proper estimation of the blade Re (based
on the relative speed and the airfoils chord). A first approximation
of the relative speed can be straightforwardly obtained from CFD
data, e.g., by defining an averaging zone like that proposed in
[11]. Otherwise, a relatively simple BEM theory can be used with
negligible errors on the final accuracy of the method.
Once the Reynolds number has been estimated, the aerody-
namic performance of the virtual airfoil defined at Step 1 can be
calculated (as a function of the AoA) as if it is working in rectilinear
flow. According to Migliores theory, these data can be used in fact
as a benchmark for CFD data. In particular, according to the
HAWTs method of [21,23], the pressure coefficient distributions
are here considered as the most relevant data for comparison.
To obtain these distributions, either a dedicated CFD campaign
(very time consuming, but often accurate [12,13]) or simpler codes
based on panel methods (e.g., XFoil [32]) can be used. The results of
these latter codes are quite accurate in the first part of the polar
curve (low-medium AoAs), while poor results are generally pro-
vided after the static stall angle [11]. In the present application,
however, due to the flow distortions that are generated around
the turbine for higher incidences, the range of angles provided by
panel method codes is generally suitable. In case a software like
XFoil is used, an Ncrit = 9 (standard wind tunnel [32]) is suggested
together with the free transition over the airfoil.
Finally, all the pressure coefficient distributions in the present
method are normalized within 1 and +1 by simply scaling them
by their maximum and minimum values. This expedient does not
alter the physical shape of the distribution but allows one to com-
pare pressure distributions, independent of the magnitude of the
relative velocity.
Fig. 5. Conceptual scheme of the proposed method.
3.3. Step 3
Moving to the analysis of CFD results, the pressure coefficient
distributions around the airfoils are acquired from calculations at
different azimuthal positions over the revolution. Then, they are
again normalized within 1 and +1, being divided by their maxi-
mum and minimum values to purge the effect of the relative speed
modulus.
3.4. Step 4
For every azimuthal position acquired in Step 3, the pressure
coefficient distributions from CFD are finally compared to all those
calculated for the virtual airfoils, either manually or with an auto-
matic procedure. By doing so, the distribution that best fits the one
coming from CFD can be highlighted: this distribution gives an
estimation of the incidence angle experienced by the airfoil in
motion. In particular, the position along the chord of the pressure
peak is used to define the incidence as the influence of flow speed
has been discarded by normalizing the distributions.
3.5. Benefits and limitations of the proposed model
The proposed approach (summarized in the chart of Fig. 5) is
able to overcome two of the main limitations of the inverse BEM
approach, i.e.: (a) the uncertainty in selecting the zone in which
the magnitude of the relative velocity is calculated and (b) the
inaccuracy in the evaluation of the magnitude itself.
These two points are indeed crucial in case of blades rotating
around an axis orthogonal to the main flow direction, where a
mutual interaction between the upwind and the downwind halves
of the revolution is present and cyclic changes in incidence are
experienced. Based on the dimensionless pressure coefficients over
the airfoil, the incidence is indeed defined independently from the
absolute value of the pressure around the airfoil and then of the
 A. Bianchini et al. / Energy Conversion and Management 126 (2016) 790798
relative speed. Unfortunately, in case the performance of the vir-
tual airfoil are obtained with XFoil, the validity of this approach
ceases as soon as the flow is separated, as no reliable power coef-
ficient distribution can be obtained in these conditions [11]. How-
ever, a recent work [33] showed that sufficiently accurate
estimations of the AoA can be also obtained for high AoAs by focus-
ing the attention only on the pressure distribution over the leading
edge of the airfoils; in this sense, if the pressure distributions of the
virtual profile are obtained by CFD, the proposed method could be
extended to all the expected AoAs, allowing, for example, the
reconstruction of dynamic stall hysteresis cycles.
Moreover, it is worth pointing out that the main novelty of the
approach lies on the method to account for the virtual camber in
the aerodynamic performance evaluation of the airfoils. The
matching between the CFD pressure profiles for the geometric air-
foil and those coming from straight-flow simulations of the trans-
formed airfoil is indeed a key point for the whole method, which is
thought to be of capital relevance for accurate predictions based on
the last studies on these problems [26].
4. Examples of model application to Darrieus vawts
To assess the model robustness and accuracy, it is here applied
to two study cases, already exploited by the authors for an investi-
gation on the flow curvature effects in Darrieus wind turbine air-
foils [26]. In details, a conventional NACA0018 airfoil of 0.2 m
chord (null pitch angle with respect to the tangential direction)
was studied while rotating in two curvilinear paths having a
chord-to-radius ratio of 0.114 and 0.25, respectively.
4.1. Numerical model
The motion of the airfoil was simulated with fully unsteady CFD
simulations. A detailed description of CFD settings is reported in
[26] and not completely repeated here for originality issues. In
[26], an extensive experimental validation is also provided. The
main features of the simulation model are, however, briefly sum-
marized in this sections and finally resumed in Table 1.
The commercial code ANSYS Fluent [34] was used in a two-
dimensional form to solve the time-dependent unsteady
Reynolds-averaged Navier-Stokes equations in the pressure based
formulation. The fluid was air, modeled as an ideal compressible
gas with standard ambient conditions, i.e. a pressure of
1.1  105 Pa and a temperature of 300 K. In this condition, a refer-
ence kinematic viscosity of 1.568  105 m2/s was considered.
The Coupled algorithm was employed to handle the pressure-
velocity coupling [35]. The global convergence of each simulation
was monitored by considering the difference between the mean
values of the torque coefficient over two subsequent revolutions
normalized by the mean value over the second period of the pair.
The periodicity error threshold was set to 0.1% [35].
Table 1
Main simulation settings.
Geometry and operating conditions
Airfoil NACA0018
Wind speed (U) 8 [m/s]
Chord (c) 0.2 [m]
Radii (R) 0.8 and 1.75 [m]
TSR 3.1
Main CFD parameters
Simulation code ANSYS FLUENT
Turbulence model c-Reh transition model
Algorithm Transient - coupled
y+ on the airfoil Always < 1
Mesh size [# elements] 400,000
795
Exploiting the sliding-mesh model of the solver, the simulation
domain was divided into two subdomains in order to allow the
rotation of the turbine, as proposed by [15]. The final dimensions
of both domains (reported in [26]) were defined according to the
sensitivity analysis reported in [35] in order to allow a full devel-
opment of the turbine wake.
The mesh settings were defined accordingly to the results of the
grid-independency analysis reported in [34], where six different
levels of refinement of the mesh were tested. An unstructured
mesh composed by triangular elements was used for the dis-
cretization of the core flow region, whereas the entire boundary
layer region has been described with a structured O-grid of 50 lay-
ers. The airfoils surface was discretized with approximately 600
nodes. Based on the requirements of the selected turbulence
model, the first cell height was selected as to provide a dimension-
less wall distance (y+) during the rotor revolution constantly lower
than 1. The expansion ratio for the growth of elements starting
from the surface was kept below 1.1 to achieve good mesh quality
in proximity of the airfoil.
Concerning the turbulence closure problem, Balduzzi et al. [35],
showed the effectiveness of Menters shear stress transport (SST)
[36] model in performance simulations involving unsteady aerody-
namics for VAWTs, as also confirmed by wide use in literature. In
the present study, however, attention has been focused on a more
detailed examination of the aerodynamic behavior of each single
airfoil in motion by analyzing equivalent static pressure coeffi-
cients 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 c-Reh transition model (derived by Menter and Lang-
try from the SST model [37]) was implemented, despite its
increased computational cost. Lanzafame et al. [38] recently
showed good agreement between experimental data and numeri-
cal results obtained with the transition turbulence model for two
different types of H-Darrieus turbines [26,28].
For the purposes of the present analyses, a relevant TSR of
approximately 3.1 was selected in both the study cases. Due to
the geometric differences between the two rotors, this condition
corresponds to a stable functioning point in case of the rotor with
c/R = 0.25 [26] and an unstable one in case of the c/R = 0.114 [33].
This selection is of particular interest for the following analyses
since, it reinforces the assessment of model stability for different
attended operating conditions.
In both cases, however, the attended Reynolds numbers experi-
enced by the blades are within the range covered by accurate
experimental data of lift and drag coefficients available from wind
tunnel measurements [26] and used here for comparison.
4.2. Assessment of model capabilities
Due to the lack of experimental results able to provide a reliable
estimation of the incidence angle during the functioning of these
turbines, a real validation of the proposed model is not possible
yet. However, it is worth evaluating the capabilities of the pro-
posed methodology in comparison to previous models and to the
state-of-the-art in the performance analysis of these machines.
A first, qualitative assessment of the robustness of the method
is in fact presented in Fig. 6, where the typical lift curves of the
NACA0018 airfoil with both tested chord-to-radius ratios were
reconstructed reversing Eqs. (1) and (2), based on the incidence
angles calculated using the new method. In particular, in Fig. 6
the calculated values from CFD are compared to the predictions
of XFoil and the experimental data collected in the wind tunnel
for two airfoils exactly shaped as the virtual transformation of
the NACA0018 based on c/R = 0.114 and c/R = 0.25, respectively
[26].
796 A. Bianchini et al. / Energy Conversion and Management 126 (2016) 790798
Fig. 6. CFD computed lift curve of the NACA0018 and experimental data and XFoil
predictions of the Transformed airfoils (c/R = 0.114 and c/R = 0.25) @ Re = 300 K.
Upon examination of the figure, good agreement can be noticed
between the results at both chord-to-radius ratios, confirming the
suitability of the proposed method for the determination of the
AoA from CFD computations. Similar agreement would not have
been reproduced if either the assumption of the virtual camber
effect or the incidence selection procedure by comparing the pres-
sure profiles were not fulfilled.
The benefits of the proposed approach are even more visible by
analyzing some relevant azimuthal position of the blades, in which
the present method is able to properly correct the common (incor-
rect) estimation of the AoA. For example, it is well known that sim-
plified models, like the BEM theory, predict a null incidence at the
Fig. 7. Pressure patterns over the airfoil at: (a) # = 0 (theoretical AoA = 0); (b) # = 180 (theoretical AoA = 0).
two relevant positions of # = 0 and # = 180, where the peripheral
speed and the wind speed are thought to be aligned [18]. The com-
puted flow field in these positions in the considered case study
with c/R = 0.114 are reported in Fig. 7. Upon examination of the fig-
ure, if this assumption is generally true in case of # = 0 (where the
pressure pattern is almost symmetrical in the pressure and suction
sides on the airfoil), it is apparent that at # = 180 the airfoil exper-
iments a non-null incidence, due to a distorted direction of the
absolute wind speed. As discussed, this represents a main limita-
tion of a low-order method like the BEM analysis, since no correc-
tion model is presently available to account for flow distortions of
the wind speed induced by the flow-turbine interaction [28]; bias
errors are then unavoidably induced by the use of tabulated aero-
dynamic polars.
In the present case, based on the computed tangential force (Eq.
(1)) and the preliminary estimation of AoA = 0 at # = 180, one
would indeed predict a lift coefficient of the airfoil of approxi-
mately 0.48 at AoA = 0, i.e. almost twice the correct value reported
in Fig. 6. By applying the present method for AoA estimation
(which, of course, does not need any preliminary flow visualization
during the automatized procedure), an AoA = 2 is predicted at
# = 180. By doing so, the lift coefficient of the airfoil (Fig. 6) is
properly related to its functioning polar curve.
As a final remark, however, it is worth pointing out that a key
point of the presented method is then the higher stability in com-
parison to the former proposal of [11]. In the new procedure, no
user-selected variable is needed, whereas the inverse BEM tech-
nique actually needs an arbitrary definition of the fluid zone in
which the relative speed is calculated. The impact of a wrong posi-
tioning of the averaging zone could indeed heavily impact on the
results, especially for high angles of attack near the stall of the air-
foils. It is worth remarking anyhow that the new model still needs
a first estimation of the relative speed (see Step 1 of the proce-
dure), but it is now used only to estimate the Re number on the air-
foil: small errors in the velocity estimation would then have a
notably lower impact on the overall reconstruction of incidence.
5. Prospects and discussion
Due to novelty of accurate CFD simulations for the aerodynamic
problems connected to presently analyzed applications and the
lack of experimental data able to provide an estimation of the inci-
dence on these blades, it is presently difficult to quantify the
impact of the proposed method on the performance analysis of real
machines.
 A. Bianchini et al. / Energy Conversion and Management 126 (2016) 790798
Fig. 8. Comparison between computed incidence angles and theoretical predictions
using the BEM code (BEM data limited between 50 to improve the readability vs.
CFD results).
In the near future, however, the applications of the proposed
model could be very many, allowing the use of computed CFD data
to develop new theoretical models to be transferred into low-order
simulation methods, like for example new dynamic stall models,
since the possible extension of the model to high AoAs discussed
in Section 3 could allow an accurate reconstruction of lift and drag
stall hysteresis cycles.
In this view, a more reliable estimation of the incidence angle
during the revolution would provide great benefits also to struc-
tural analyses, inducing a more accurate definition of the stresses
Fig. 9. Comparison between computed tangential and normal force coefficients and
theoretical predictions using the BEM code (BEM data limited between 50 to
improve the readability vs. CFD results).
797
acting on the rotors in motion. For example, in Fig. 8 the predicted
incidence angles calculated from CFD for the NACA0018 used as a
test case are compared to those predicted by a BEM code, recently
validated with experiments and to be considered as the state-of-
the-art for this kind of simulation models [3,57,24].
It can be noticed that the predictions of an even good BEM
model are quite different from CFD-based results. A different slope
of the variation trend can be noticed, especially towards higher
incidence angles. Moreover, as discussed in the previous section,
in a real flow, where the streamlines are not perfectly parallel to
flow direction (as assumed by the BEM model), errors in the correct
evaluation of incidence when the airfoil is aligned with the flow
direction (# = 0 and # = 180) can be also attended.
If one then compare the performance predictions of the BEM
code, using the wrong AoAs, with the same quantities calculated
by CFD accounting for the correct AoAs (obtained with the present
method), the impact of the discrepancies of Fig. 8 can be readily
appreciated. In particular, Fig. 9 compares the tangential (Ct) and
normal (Cn) coefficients (i.e. the tangential and normal forces
divided by the kinetic energy of the relative flow and the turbine
area) coming from the computed incidences or those predicted
by the BEM code.
Upon examination of the figure, it is apparent that, assuming
that CFD reconstructed data can be considered representative of
the real phenomena, the BEM model is not always able to provide
accurate estimations of the attended force during the revolution. In
particular, as far as the incidence angles increase, even small
changes in the considered AoA induce notable variation in the pre-
dicted aerodynamic forces, with a direct impact on both the
extracted torque and the stresses on the rotor. In case of BEM
codes, in addition, due to their intrinsic sensitivity to the tabulated
polars [24], more accurate estimations of the AoA could provide
notable accuracy improvements. In other words, future studies
able to correctly describe the AoA trends of airfoils in cycloidal
motion, could lead to better low-order models, able to more accu-
rately describe the functioning of these machines and then to pro-
vide more useful indications to designers in the development of
new machines.
6. Conclusions
A novel approach for the determination from a CFD computed
flow field of the incidence angle on an airfoil rotating around an
axis orthogonal to flow direction was proposed.
The method is based on the application of the virtual camber
effect to define the virtual airfoil to which compare the aerody-
namic performance in straight flow. Once the transformed airfoil
is defined, the pressure coefficient distributions over it are defined
based on panel methods, CFD and/or experiments and normalized
between 1 and +1 in order to discard the influence of the magni-
tude of relative speed.
Then, the CFD distributions over the real airfoil in motion are
extrapolated from the computed curvilinear flow field for all the
azimuthal positions in which an attached flow is ensured. Each
of them is normalized and then compared to those obtained in
straight flow for the transformed airfoil at different AoAs until
the correct match is found. By definition, the resulting AoA corre-
sponds to the incidence angle globally experienced by the rotating
airfoil.
The method has been tested here on a NACA0018 airfoil rotat-
ing with two chord-to-radius ratios of 0.114 and 0.25, respectively,
obtaining sound agreement between the results of processed data
and theoretical expectations. Moreover, a superior robustness of
the present method in comparison to previous proposal has been
798 A. Bianchini et al. / Energy Conversion and Management 126 (2016) 790798

assessed, since no user-selected variable is anymore needed in [16] Tjiu W, Marnoto T, Mat S, Ruslan MH, Sopian K. Darrieus vertical axis wind
turbine for power generation I: assessment of Darrieus VAWT configurations.
the method.
Renew Energy 2015;75(March 2015):5067. http://dx.doi.org/10.1016/j.
The proposed procedure can be generally applied to any airfoil renene.2014.09.038.
rotating around an axis orthogonal to a straight flow and is thought [17] Aslam Bhutta MM, Hayat N, Farooq AU, Ali Z, Jamil ShR, Hussain Z. Vertical axis
to provide in the near future a powerful tool for a further under- wind turbine a review of various configurations and design techniques.
Renew Sustain Energy Rev 2012;16(4):192639. http://dx.doi.org/10.1016/j.
standing of the complex flow phenomena involved in similar cases. rser.2011.12.004.
[18] Paraschivoiu I. Wind turbine design with emphasis on Darrieus
Acknowledgment concept. Montreal (Canada): Polytechnic International Press; 2002.
[19] Guntur S, Srensen NN. Evaluation of several methods of determining the
angle of attack on wind turbine blades. In: Proc of the science of making torque
The authors would like to acknowledge Prof. Ennio Antonio Car- from wind 2012, Oldenburg (Germany), 0911 October.
nevale of the University of Florence (Italy) for supporting this [20] Lindenburg C. Investigation into rotor blade aerodynamics analysis of the
stationary measurements on the UAE Phase-VI rotor in the NASA-Ames wind
research activity. tunnel. Tech rep no ECN C-03-025; 2003.
[21] Guntur S, Bak C, Srensen NN. Analysis of 3D stall models for wind turbine
References blades using data from the MEXICO experiment. In: Proc of the 13th
international conference on wind engineering, Amsterdam, The Netherlands,
July 1015.
[1] Zhong Shen W, Hansen MOL, Srensen JN. Determination of Angle of Attack
[22] Hansen MOL, Johansen J. Tip studies using CFD and comparison with tip loss
(AOA) for rotating blades. In: Peinke J et al., editors. Wind energy, proceedings
models. Wind Energy 2004;7(4):34356. http://dx.doi.org/10.1002/we.12.
of the euromech colloquium. New York (USA): Springer; 2007. p. 2059. http://
Special Issue: The science of making torque from wind. Part I: Aerodynamics
dx.doi.org/10.1007/978-3-540-33866-6_37.
and aeroelasticity.
[2] Camporeale SM, Magi V. Streamtube model for analysis of vertical axis variable
[23] Dumitrescu H, Cardos V, Frunzulica F, Dumitrache A. Determination of angle of
pitch turbine for marine currents energy conversion. Energy Convers Manage
attack for rotating blades. INCAS Bull 2012;4(12):3742. http://dx.doi.org/
2000;41(16):181127. http://dx.doi.org/10.1016/S0196-8904(99)00183-1.
10.13111/2066-8201.
[3] Bianchini A, Ferrara G, Ferrari L. Design guidelines for H-Darrieus wind
[24] Marten D, Bianchini A, Pechlivanoglou G, Balduzzi F, Nayeri CN, Ferrara G, et al.
turbines: optimization of the annual energy yield. Energy Convers Manage
Effects of airfoils polar data in the stall region on the estimation of Darrieus
2015;89(1 January 2015):690707. http://dx.doi.org/10.1016/j.
wind turbines performance. In: Proc of the ASME turbo expo 2016, Seoul,
enconman.2014.10.03.
South Korea, June 1317. p. 137.
[4] Bianchini A, Cangioli F, Papini S, Rindi A, Carnevale EA, Ferrari L. Structural
[25] Simao-Ferreira C, van Zuijlen A, Bijl H, van Bussel G, van Kuik G. Simulating
analysis of a small H-Darrieus wind turbine using beam models: development
dynamic stall on a two-dimensional vertical-axis wind turbine: verification
and assessment. J Turbomach 2015;137(1):111. http://dx.doi.org/10.1115/
and validation with particle image velocimetry data. Wind Energy
1.4028212.
2010;13:117. http://dx.doi.org/10.1002/we.330.
[5] Johansen J, Srensen NN. Aerofoil characteristics from 3D CFD rotor
[26] Rainbird J, Bianchini A, Balduzzi F, Peiro J, Graham JMR, Ferrara G, et al. On the
computations. Wind Energy 2004;7(4):28394. http://dx.doi.org/10.1002/
influence of virtual camber effect on airfoil polars for use in simulations of
we.12. Special Issue: The science of making torque from wind. Part I:
Darrieus wind turbines. Energy Convers Manage December 2015;106:37384.
Aerodynamics and aeroelasticity.
http://dx.doi.org/10.1016/j.enconman.2015.09.053.
[6] Bianchini A, Balduzzi F, Rainbird J, Peiro J, Graham JMR, Ferrara G, et al. An
[27] Saverin J, Frank S. The loading cycle of a H-blade type vertical axis wind
experimental and numerical assessment of airfoil polars for use in Darrieus
turbine. In: Proc of the ASME turbo expo 2015, Montreal (Canada), 1519 June.
wind turbines. Part 1 Flow curvature effects. J Eng Gas Turb Power 138(3): 1
http://dx.doi.org/10.1115/GT2015-42185.
10. http://dx.doi.org/10.1115/1.4031269.
[28] Bianchini A, Balduzzi F, Ferrara G, Ferrari L. Virtual incidence effect on rotating
[7] Bianchini A, Balduzzi F, Rainbird J, Peiro J, Graham JMR, Ferrara G, et al. An
airfoils in Darrieus wind turbines. Energy Convers Manage 2016;111(1 March
experimental and numerical assessment of airfoil polars for use in darrieus
2016):32938. http://dx.doi.org/10.1016/j.enconman.2015.12.05.
wind turbines. Part 2 Post-stall data extrapolation methods. J Eng Gas
[29] Bianchini A, Ferrara G, Ferrari L. Pitch optimization in small-size darrieus wind
Turbines Power 2016;138(3):110. http://dx.doi.org/10.1115/1.4031270.
turbines. Energy Proc 2015;81(December 2015):12232. http://dx.doi.org/
[8] Wang Y, Sun X, Dong X, Zhu B, Huang D, Zheng Z. Numerical investigation on
10.1016/j.egypro.2015.12.067.
aerodynamic performance of a novel vertical axis wind turbine with adaptive
[30] Bianchini A, Carnevale EA, Ferrari L. A model to account for the virtual camber
blades. Energy Convers Manage 2016;108(15 January 2016):27586. http://dx.
effect in the performance prediction of an H-Darrieus VAWT using the
doi.org/10.1016/j.enconman.2015.11.00.
momentum models. Wind Eng 2011;35(4):46582. http://dx.doi.org/10.1260/
[9] Raciti Castelli M, Englaro A, Benini E. The Darrieus wind turbine: proposal for a
0309-524X.35.4.465.
new performance prediction model based on CFD. Energy 2011;36
[31] Migliore PG, Wolfe WP. The effects of flow curvature on the aerodynamics of
(8):491934. http://dx.doi.org/10.1016/j.energy.2011.05.036.
Darrieus wind turbines. US Dept Of Energy (DOE), Tech rep. ORO-5135-77/7;
[10] Yang H, Shen W, Xu H, Hong Z, Liu C. Prediction of the wind turbine
1980.
performance by using BEM with airfoil data extracted from CFD. Renew Energy
[32] XFoil User Guide. <http://web.mit.edu/drela/Public/web/xfoil> [last access 15/
2014;70(October 2014):10715. http://dx.doi.org/10.1016/j.renene.2014.
04/2016].
05.002.
[33] Bianchini A, Balduzzi F, Ferrara G, Ferrari L. Critical analysis of dynamic stall
[11] Balduzzi F, Bianchini A, Maleci R, Ferrara G, Ferrari L. Blade design criteria to
models in low-order simulation models for vertical-axis wind turbines. Energy
compensate the flow curvature effects in H-Darrieus wind turbines. J
Proc; 2016 [in press].
Turbomach 2014;137(1):110. http://dx.doi.org/10.1115/1.4028245.
[34] ANSYS Inc. FLUENT theory guide, release 16.0; 2015.
[12] Petrilli JL, Paul RC, Gopalarathnam A, Frink NT. A CFD database for airfoils and
[35] Balduzzi F, Bianchini A, Maleci R, Ferrara G, Ferrari L. Critical issues in the CFD
wings at post-stall angles of attack. In: Proc of the 31st AIAA applied
simulation of Darrieus wind turbines. Renew Energy 2016;85(Jan
aerodynamics conference, San Diego (CA), USA, 2427 June. http://dx.doi.org/
2016):41935. http://dx.doi.org/10.1016/j.renene.2015.06.048.
10.2514/6.2013-2916.
[36] Menter FR. Two-equation Eddy-viscosity turbulence models for engineering
[13] Ribeiro AFP, Awruch AM, Gomes HM. An airfoil optimization technique for
applications. AIAA J 1994;32(8):1598605.
wind turbines. Appl Math Model 2012;36(October):4898907. http://dx.doi.
[37] Langtry RB, Menter FR. Correlation-based transition modeling for unstructured
org/10.1016/j.apm.2011.12.026.
parallelized computational fluid dynamics codes. AIAA J 2009;47
[14] Migliore PG, Wolfe WP, Fanucci JB. Flow curvature effects on Darrieus turbine
(12):2894906.
blade aerodynamics. J Energy 1980;4(2):4955. http://dx.doi.org/10.2514/
[38] Lanzafame R, Mauro S, Messina M. 2D CFD modeling of H-Darrieus wind
3.62459.
turbines using a transition turbulence model. Energy Proc 2014;45:13140.
[15] Amet E, Maitre T, Pellone C, Achard JL. 2D numerical simulations of blade-
http://dx.doi.org/10.1016/j.egypro.2014.01.015.
vortex interaction in a darrieus turbine. J Fluids Eng
2009;131:111103.1111103.15. http://dx.doi.org/10.1115/1.400025.