Simulation of Dynamic Stall For A NACA 0012 Airfoil Using A Vortex Method

Journal of Fluids and Structures 17 (2003) 855874
www.elsevier.nl/locate/jnlabr/yjfls
Simulation of dynamic stall for a NACA 0012 airfoil

using a vortex method
M.H. Akbari, S.J. Price*
Department of Mechanical Engineering, McGill University, 817 Sherbrooke St. West, Montr!eal, Qu!ebec, Canada H3A 2K6
Received 26 September 2001; accepted 6 January 2003
Abstract
The unsteady, incompressible, viscous laminar ow over a NACA 0012 airfoil is simulated, and the effects of several
parameters investigated. A vortex method is used to solve the two-dimensional NavierStokes equations in the
vorticity/stream-function form. By applying an operator-splitting method, the convection and diffusion equations
are solved sequentially at each time step. The convection equation is solved using the vortex-in-cell method, and the
diffusion equation using a second-order ADI nite difference scheme. The airfoil prole is obtained by mapping a circle
in the computational domain into the physical domain through a Joukowski transformation. The effects of several
parameters are investigated, such as the reduced frequency, mean angle of attack, location of pitch axis, and the
Reynolds number. It is observed that the reduced frequency has the most inuence on the ow eld.
r 2003 Elsevier Science Ltd. All rights reserved.
1. Introduction
Stall of an airfoil is characterized by a sudden drop in its normal force following an increase in the angle of attack
that causes ow separation. Dynamic stall, on the other hand, is an unsteady phenomenon, where the airfoil pitches
through the static stall angle, while the normal force continues to increase beyond its maximum value for unstalled
conditions. This phenomenon is often associated with the formation of a leading-edge vortex, called the dynamic stall
vortex, that travels along the airfoil surface as it grows, and nally separates from the airfoil at its trailing edge.
However, some researchers (e.g., Jumper and Huge, 1991) attribute this lift overshoot to a delay in separation at high
incidence, rather than the formation and subsequent convection of the dynamic stall vortex.
The dynamic stall phenomena was rst encountered in the helicopter industry, where large torsional oscillations of
the blades was observed and attributed to the periodic stalling and unstalling of each blade on the retreating side of the
rotor disk. This was considered to be a serious problem, limiting the forward speed and gross weight of the helicopter
(Crimi, 1973). This phenomenon may also occur, for example, on axial-ow compressor blades, large wind-mill rotors,
or the wings of modern ghters during fast-pitching maneuvers.
Examples of early analytical studies of unsteady airfoil stall can be found in Carta (1967a, b), Ericsson (1967), and
Ericsson and Reding (1971). Ericsson and Reding (1972) analyzed the dynamic stall of a helicopter blade section using a
quasi-steady theory. Static experimental data were used as input to predict the dynamic stall characteristics of the
airfoil; hence, the applicability of this theory was limited by the availability of the required static experimental data.
They found the technique to be successful so long as the reduced frequency, k pfo c=UN ; was not high ko0:5; where
fo ; c; and UN are the oscillation frequency, chord length, and free-stream velocity, respectively. Further improvements
and modications to their method can be found, for example, in Ericsson and Reding (1980, 1988).
*Corresponding author.
E-mail address: [email protected] (S.J. Price).
0889-9746/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0889-9746(03)00018-5
M.H. Akbari, S.J. Price / Journal of Fluids and Structures 17 (2003) 855874
856
McCroskey et al. (1976) experimentally studied dynamic stall, and the associated unsteady boundary layer
separation, in incompressible ow at high Reynolds numbers on leading-edge-modied NACA 0012 airfoils. They
found the Reynolds number to have a small effect on the normal force and pitching moment coefcients, Cn and Cm : On
the other hand, the effects of reduced frequency and leading-edge prole of the airfoil were found to be major. An
experimental investigation was performed by Lorber and Carta (1988) to study dynamic stall aerodynamics of an airfoil
at constant pitch rate and high Reynolds number (up to 4 106 ). It was observed that the strength of the dynamic stall
vortex increased with pitch rate, and decreased with increasing Mach number. The starting flow over a NACA 0012
airfoil oscillating at large incidences was investigated by Ohmi et al. (1991), using ow visualization experiments and
numerical calculations. Reynolds numbers from 1500 to 104 ; and reduced frequencies (in this case, f n fc=2UN ) of
0.11.0 were considered. The reduced frequency was identied as a key parameter in determining the vortex wake
pattern of the oscillating airfoil. Reynolds number effects were found to be small compared to other parameters. Raffel
et al. (1995) investigated the ow eld over a NACA 0012 airfoil oscillating in pitch in a low-speed wind tunnel by
means of particle image velocimetry (PIV). During the upstroke a dynamic stall vortex formed at the leading edge,
which at a 24 ; for example, extended over 75% of the upper surface of the airfoil. They observed a strong
aperiodicity of the ow structure during the downstroke motion from cycle to cycle.
In the simulations presented here, the dynamic stall characteristics of a pitching NACA 0012 airfoil at Reynolds
numbers of 3000 and 104 are studied using a vortex method for laminar ows, and the effects of the reduced frequency,
mean angle of attack, and location of the pitch axis are investigated.
2. Solution method
The governing equations for the incompressible ow of a Newtonian uid in two dimensions are the NavierStokes
and continuity equations, which can be expressed in terms of the vorticity and stream-function as the following
@o
u ro nr2 o;
@t
r2 c
o:
In these equations u u; v is the velocity vector, n is the kinematic viscosity, and o and c are the vorticity and streamfunction, respectively. Eq. (2) is the Poisson equation for stream-function which replaces the continuity equation.
Following Chorin (1973), an operator-splitting method is applied to the governing equations, and the following
two equations are obtained, which are solved sequentially in each time step
@o

u ro;
@t
@o
nr2 o:
@t
In this study, the vorticity eld is discretized into a number of point vortices, and the time domain is also discretized
into small time steps. Eq. (3) states that the total derivative of vorticity is zero during convection, or that the vorticity
of a particle is constant as it is convected with the local velocity. Hence, in order to satisfy the convection part of the
governing equations, the velocity eld is obtained, and each vortex particle is convected with the local velocity for the
period of the time step. To do that, Eq. (2) is solved for the stream-function (given the vorticity eld), and the velocity
eld is then computed as the curl of the stream function. Next, the diffusion process is simulated by solving Eq. (4)
using a second-order ADI nite difference scheme. The no-ow boundary condition is satised during the solution of
the convection problem, and the no-slip boundary condition is satised by the creation of new vorticity on the body
surface during each time step.
In the present vortex method, a computational grid is used for the following purposes: (a) solving the Poisson
equation for the stream-function given the vorticity eld, (b) calculating the velocity eld given the stream-function, (c)
solving the diffusion equation for the vorticity eld. Also, the new vorticity is created at the location of the rst row of
grid nodes, on the surface of the body. The computational grid is xed to the body, and hence, moves with the airfoil as
it oscillates.
Further specic details of the numerical methods used in this study (including derivation of equations, accuracy of
the numerical schemes, the computational grid, and calculation of the force coefcients) are given in Akbari and Price
(2000) and Akbari (1999), and hence are not repeated here for the sake of brevity. It is noted here, however, that the
present vortex method is applicable only to two-dimensional, incompressible ows, the numerical schemes are
857
unconditionally stable, and the nite-difference formulations (for the solution of the Poisson and diffusion equations)
are nominally second-order accurate.
In this study, a pitching airfoil is considered, with the instantaneous angle of attack given by
a a% A cosOt;
where a is the angle of attack at time t; a% is the mean angle of attack, A is the amplitude of the pitching oscillation (in
rad), and O 2pf ; with f being the frequency of oscillation.
3. Airfoil prole
In this study, conformal transformations are used to map a circle onto an airfoil, and vice versa. The general form of
the transformation F : zx; y-wx; Z that transforms a circle onto our desired prole is given by

C
B2
w Fz
z
P e
ia P;
6
z
where z x iy is a point on or outside the circle in the z-plane, w x iZ is the corresponding point on or outside the
transformed prole in the w-plane, and B; C; a; and P are control parameters that are described below with the help of
Fig. 1.
As can be seen in Fig. 1, C is the interception point of the circle and the positive side of the x-axis. The center of the
circle is located at point
e; d; P is the location of the rotation axis of the prole in the w-plane, and a is the angle of
rotation, positive in the clockwise direction. B in Eq. (6) is a control parameter that, along with the eccentricity of the
circle, determines the shape of the prole in the transformed plane (w-plane).
In this article, an airfoil prole is considered that closely resembles the NACA 0012 airfoil, by setting e 0:068;
d 0; and B 0:04 in Eq. (6). The proles of the airfoil obtained by our transformation and that of a NACA 0012
airfoil are compared with each other in Fig. 2, and, as can be seen, they are indistinguishable from one another. A closeup view of the two proles in the vicinity of the leading- and trailing-edges is given in Fig. 3. As seen, in these areas the
differences between the two proles are noticeable. However, one should consider the scale of the plots in order to
realize the insignicance of these differences between the two proles. Therefore, we will consider our prole to be a
very close resemblance of a NACA 0012 airfoil.
4. Results and discussion

The simulation results from our parameteric study of the dynamic stall ow of the NACA 0012 airfoil are presented
here. A summary of the case-studies in this investigation is shown in Table 1. In this work emphasis is put on
investigating the effects of reduced frequency of oscillation of the airfoil. In all case-studies the ow is started
(, )
(C, 0)
z-plane
w-plane
Fig. 1. The Joukowski transformation of the z-plane (circle) into the w-plane (airfoil).
858
0.2
-0.2
0
0.25
0.5
0.75
x
Fig. 2. Comparison of a NACA 0012 airfoil prole and that obtained from our transformation: , NACA 0012 prole; - - -, prole
from transformation.
0.02
0.02
-0.02
-0.02
0
(a)
0.04
0.96
(b)
Fig. 3. Close-up view of the two proles in the vicinity of: (a) the leading-edge, (b) the trailing-edge: , NACA 0012 prole; - - -,
prole from transformation.
Table 1
Input parameters for simulations of cross-ow over a pitching NACA 0012 airfoil
Re
a%
Da
Axis
104
104
104
3000
104
104
0.15
0.25
0.50
0.25
0.25
0.25
15
15
15
15
20
15
10
10
10
10
10
10
c=4
c=4
c=4
c=4
c=4
c=2
impulsively at a steady angle of attack amin : After the ow achieves steady state at this steady angle, the pitching
oscillation of the airfoil is started at the given frequency, and a few oscillation cycles of the airfoil are simulated.
In the following, the force and pitching moment coefcients for all cases are presented and analyzed, and the ow
elds are studied. In all numerical simulations a time-step size of Dt 0:01 and 300 360 grids have been used. It
should be noted that the numerical method used in this study is unconditionally stable, and the selected numerical
parameters produced grid-independent results.
4.1. Static airfoil

A number of ow simulations are rst presented for steady angles of attack of the airfoil. In this study the static stall
angle is predicted at aE15 for a Reynolds number of Re 104 : The steady values of the force and pitching moment
coefcients versus angle of attack are plotted in Fig. 4. In Fig. 4(a), the normal force coefcient predictions for
Re 104 are compared with the experimental results taken from Abbott and von Doenhoff (1959) for Re 6 106 ;
0.4
1.5
0.2
+
Cm
Cn
0.5
0
+
+
-0.2
+
-0.4
-0.5
0
(a)
859
10
15
20
(b)
10
15
20
0.4
0.3
Cd
0.2
0.1
0
-0.1
0
(c)
10
15
20
Fig. 4. The force and pitching moment coefcients at steady angles of attack: (a) Cn ; (b) Cm ; (c) Cd : present numerical results for
Re 104 ; W experimental results due to Johari and Durgin (1998) for Re 1:02 105 ; experimental results from Abbott and von
Doenhoff (1959) for Re 6 106 :
and from Johari and Durgin (1998) for Re 1:02 105 (all data are for a NACA 0012 airfoil). As seen in this gure,
the ow Reynolds number has some effect on the normal force coefcient. A consistent decrease with decreasing
Reynolds number (over a range of two orders of magnitude) is observed in the magnitude of the normal force
coefcient. This effect has been reported previously for a large change in the Reynolds number. For example, Massey
(1979) presented the static lift and drag coefcients versus angle of attack for a particular airfoil at two Reynolds
numbers of 4:3 105 and 2:1 104 ; where for any given angle Cl is considerably lower at the lower Reynolds number
(e.g., at a 10 ; 0.49 compared to 1.4), and Cd is higher (e.g., at a 10 ; 0.19 compared to 0.07). By comparison,
Fig. 4(a) shows a reasonable agreement of our numerical results with the experimental data, considering the differences
in the Reynolds number. The static stall occurs at aE16 in the experiments reported by Abbott and von Doenhoff, and
it is predicted at aE15 in this study.
As seen in Fig. 4(b), the pitching moment coefcient (around the quarter-chord) in the experimental data is
approximately zero for ap12 ; and then it drops to negative values at higher angles of attack. However, in our
simulations the pitching moment coefcient is predicted at positive values for some angles of attack of less than 12 ;
nonetheless, the trend in our data follows that in the experimental data for aX12 : The differences in the magnitude of
the moment coefcient for a > 12 are attributed to the large difference in the Reynolds number. Note that the normal
force coefcient is lower at a lower Reynolds number, hence, with similar locations for the aerodynamic center, a lower
pitching moment coefcient is expected at a lower Reynolds number.
Shown in Fig. 4(c) are the calculated drag coefcients in this study, along with some experimental data from Abbott
and von Doenhoff (1959). No experimental data for Cd is available for a > 8 : The scale of this plot is too large for the
variation of experimental Cd with a to be clearly seen. Our numerical results also predict small values for Cd at ap8 ;
followed by a considerable increase at higher angles of attack. The drag coefcient predicted in the present study has a
higher magnitude than that reported from experiments for ap8 ; where experimental data are available. However, note
should be made that these data are for Reynolds numbers of two orders of magnitude difference. A higher drag
coefcient is expected at a lower Reynolds number for the same airfoil at the same angle of attack (e.g., Massey, 1979).
Overall, the simulation results from this study for steady angles of attack are in reasonable agreement with available
experimental data (considering the Reynolds number differences), particularly in terms of the lift coefcient and the
static stall angle.
860
4.2. A flow visualization comparison

Before presenting the ow eld results for the main case-studies, we shall rst study a case for which some
experimental ow visualization results are available.
Ohmi et al. (1991) studied the starting flow over a NACA 0012 airfoil for Reynolds numbers of between 1500 and 104 ;
and at different reduced frequencies. Shown in Fig. 5 are their experimental ow visualization results for the streamline
patterns of the unsteady ow past the pitching airfoil with1 k 0:2p; Re 3000; a% 30 ; Da 15 ; a0 15 ; and
x=c 0:5: The pitching oscillation of the airfoil started from the minimum incidence with the impulsive start of the
ow, and one oscillation cycle was completed during the experiment. As can be seen in the gure, the ow remains
attached to the airfoil up to dimensionless time tn tUN =c 1 after the impulsive start, when aE25 : Then, the ow
starts to separate from the leading-edge, and a leading-edge vortex forms, grows, and is eventually shed from the upper
surface some time later during the downstroke. During the remainder of the oscillation cycle smaller vortex structures
are shed from both the upper surface and the trailing-edge of the airfoil.
The streamline plots from our simulation of the same ow as that studied by Ohmi et al. are presented in Fig. 6. Each
part of this gure corresponds to the similar part of Fig. 5. The two sets of plots are in a very good agreement, especially
for times up to tn 1:5: At this time the ow starts to separate from the airfoil, with the formation of a leading-edge
vortex. The size of this vortex at tn 2 is larger in the experimental results compared to that predicted in our
simulation, as seen in parts (d) of the two gures. The overall structure of the ow eld is still very similar at tn 3;
including the wavy streamlines originating from the trailing-edge. As seen in parts (g) of the gures, the size and
position of the trailing-edge vortex at tn 3:5 are very similar in the two sets of results. The formation of a trailing-edge
vortex from tn 4 to 4.5 can be seen in both set of plots.
Overall, a high degree of similarity is observed between our simulation results for the ow eld and those from the
experiment. This serves as a validation of the accuracy of the results obtained from our numerical simulations.
4.3. Dynamic stall flow
4.3.1. Effect of reduced frequency
Shown in Fig. 7 are the normal force Cn ; pitching moment Cm ; and drag Cd coefcients during ve oscillation
cycles of the airfoil at a% 15 ; with Da 10 and a reduced frequency of 0:5; at Re 104 : As seen in the gure, after
completion of the second cycle no transient effects are present in the hysteresis loops; the differences that exist from one
cycle to the next after the second cycle of oscillation are due to the unsteady nature of the ow. Hence, it is believed that
simulating even fewer number of oscillation cycles would still capture the overall characteristics of the ow eld.
Shown in Fig. 8 are the normal force, pitching moment, and drag coefcients, averaged over the last two oscillation
cycles as a function of reduced frequency, along with the variation of the same coefcients for steady angles of attack. It
is observed that the normal force coefcient in the pitching cases increase well beyond that at the static stall. Note that
stall occurs at aE15 in the steady case, when the normal force coefcient is Cn E0:77:
Hysteresis effects are observed in the force and pitching moment coefcients at all reduced frequencies. At the lowest
reduced frequency, k 0:15; stall occurs at aE21:5 : In this case an overall positive damping in the pitching moment
loop is observed. This is concluded by comparing the areas of clockwise and counter-clockwise loops in the plot of the
pitching moment versus angle of attack. When the frequency is increased to k 0:25 stall is delayed to amax 25 ; and
the positive-damping area in the pitching moment loop decreases compared to that obtained with k 0:15: When the
reduced frequency is increased further to k 0:5 the net damping in the pitching moment loop becomes negative. In
this case the drag coefcient loop is narrower, and the normal force coefcient takes negative values during the
downstrokes.
By comparing the results of these three case-studies, it is concluded that increasing the reduced frequency can slightly
increase the normal force overshoot up to a limit of 1:6; after which any further increase in the reduced frequency has
no apparent effect on the overshoot. However, an increase in the reduced frequency also modies the general prole
of the hysteresis loops. The minimum normal force coefcient decreases, and the drag coefcient loop becomes
narrower with increasing reduced frequency. But, perhaps the most important effect of increasing the reduced frequency
is the signicant increase of the negative damping in the pitching moment cycle. Reducing negative damping in the
pitching moment cycle of the airfoil is important in the stability of its torsional oscillations.
In order to validate our numerical results for the force coefcients of a pitching airfoil a comparison with some
experimental results, or other reliable numerical data, is desirable. However, all the available experimental and
numerical results for the dynamic stall ow that also include the measurement or calculation of force coefcients, cover
1
Note that k 0:2p is higher than that of the other results presented in this paper.
861
Fig. 5. Streamline plots for the NACA 0012 airfoil at Re 3000; f n 1:0; a% 30 ; Da 15 ; and a0 15 ; taken from Ohmi et al.
(1991). (a) tn 0:5; a 17:9 ; (b) tn 1:0; a 25:4 ; (c) tn 1:5; a 34:6 ; (d) tn 2:0; a 42:1 ; (e) tn 2:5; a 45:0 ; (f) tn 3:0;
a 42:1 ; (g) tn 3:5; a 34:6 ; (h) tn 4:0; a 25:4 ; (i) tn 4:5; a 17:9 ; (j) tn 5:0; a 15:0 :
the turbulent ow regime, at high Reynolds numbers of at least 106 : Our numerical method, being for a laminar regime,
is not expected to produce accurate results for Reynolds numbers of higher than about 104 : This large difference in the
Reynolds number (104 versus 106 ) has considerable effects on the behavior of the ow eld and the force coefcients.
862
Fig. 6. Streamline plots for the NACA 0012 airfoil at Re 3000; f n 1:0; a% 30 ; Da 15 ; and a0 15 ; from our simulation. (a)
tn 0:5; a 17:9 ; (b) tn 1:0; a 25:4 ; (c) tn 1:5; a 34:6 ; (d) tn 2:0; a 42:1 ; (e) tn 2:5; a 45:0 ; (f) tn 3:0; a 42:1 ;
(g) tn 3:5; a 34:6 ; (h) tn 4:0; a 25:4 ; (i) tn 4:5; a 17:9 ; (j) tn 5:0; a 15:0 :
0.2
1.5
0.1
Cn
Cm
0.5
-0.1
-0.2
863
-0.3
-0.5
0
(a)
10
15
20
25
30
10
15
(b)
20
25
30
0.7
0.6
Cd
0.5
0.4
0.3
0.2
0.1
0
0
10
(c)
15
20
25
30
Cn
0.2
0.2
0.2
Cm
Cm
Cm
-0.2
-0.2
0.5
0.5
0.5
0
0
(a)
Cd
-0.2
Cd
Cd
Cn
Cn
Fig. 7. The force and pitching moment coefcients versus angle of attack over ve oscillation cycles for the airfoil at a% 15 ;
Da 10 ; k 0:5; Re 104 ; (a) Cn ; (b) Cm ; (c) Cd :
10 15 20 25 30
0
0
(b)
10 15 20 25 30
(c)
10 15 20 25 30
Fig. 8. The hysteresis loops for the normal force Cn ; pitching moment Cm ; and drag Cd coefcients for the airfoil at a% 15 and
Re 104 oscillating with Da 10 ; x=c 0:25; and reduced frequency: (a) 0.15, (b) 0.25, (c) 0.5: , pitching case; - - -, steady case.
864
Hence, a qualitative comparison is made between our calculated force coefcients and a set of reliable data at the lowest
Reynolds number available to us.
Presented in Fig. 9 is a qualitative comparison of our results for k 0:15 and 0:25; a% 15 ; and Da 10 ; with the
numerical results due to Tuncer et al. (1990) for the dynamic stall of the same airfoil at Re 106 oscillating at the same
frequencies and angles of attack. Note that no data for Cd are presented in Tuncer et al. (1990) for the case with
k 0:25:
Some differences are observed between the two sets of results. Higher values of Cn and lower values of Cd are
reported for Re 106 compared to our calculated results for Re 104 : This behavior was expected due to the Reynolds
number difference. Stall is predicted at aE21:5 in our simulation for k 0:15; while no stall is observed for this case at
Re 106 : Because of this, a minimum value for Cm is obtained at aE21:5 in our calculations, while at Re 106 Cm
continues decreasing with decreasing a: However, for k 0:25 no stall effects are observed in the Cn loop in either
study. This is evident from the shape of the loops where Cn continues to increase with a up to the maximum angle of
attack. It is interesting to see the negative-damping region in the Cm loop reported in Tuncers results for k 0:25: The
same behavior is observed at a higher reduced frequency of k 0:5 in our simulations for Re 104 :
It seems that our results for k 0:25 and 0.5 are, respectively, more similar to Tuncers numerical results for k 0:15
and 0.25. This observation suggests that a Reynolds number increase can also intensify the effects of the reduced
0.2
0.2
0
Cm
(a)
Cm
Cn
Cn
-0.2
-0.2
-0.4
-0.4
(b)
10
15
20
25
30
Cd
0.5
0
(c)
10
15
20
25
30
Fig. 9. Comparison of the hysteresis loops for the force and pitching moment coefcients of the NACA 0012 airfoil oscillating about
c=4 between a 5 and 25 : , our numerical results for Re 104 ;
3

3
; numerical results due to Tuncer et al. (1990) for
Re 106 : (a) Cn ; (b) Cm ; (c) Cd : on the left k 0:15; on the right k 0:25:
865
frequency. Overall, despite such a large difference in the Reynolds number, some quantitative agreement is observed
between the two sets of results in terms of the hysteresis loops.
The ow eld development for one cycle of oscillation of the airfoil is now presented for each case-study considered
so far. In the vorticity contour plots that follow, the white and black areas indicate positive and negative vorticity,
respectively.
Shown in Fig. 10 are a sequence of vorticity contour plots for the ow around the airfoil at several angles of incidence
during the third cycle of oscillation of the airfoil at a% 15 ; with Da 10 and a reduced frequency of 0:15; at
Re 104 : No apparent indication of ow separation is observed on the airfoil up to aE13:6 (Fig. 10d). At aE16:7 a
leading-edge separation bubble starts forming on the airfoil. Because of the relatively small scale of the vorticity
contour shown in Fig. 10(e), this separation bubble is not easily seen; however, the separation bubble is apparent when
larger scale versions of this gure are examined. Note that static stall occurs at aE15 : The separation bubble is seen to
cover most of the upper surface of the airfoil when aE19 (Fig. 10f). At aE22 the leading-edge vortex is shed from the
airfoil. At the maximum incidence a 25 a vortex forms at the trailing-edge, that promotes the separation of the
leading-edge vortex.
Fig. 10. Vorticity contour plots at several angles of attack for a pitching NACA 0012 airfoil, at k 0:15; Da 10 ; a% 15 ; rotation
axis at c=4; and Re 104 :
866
During the downstroke the trailing-edge vortex is shed, and a new leading-edge vortex grows and is about to be shed
when aE22:2 : In parts (j) to (m) of the gure two other smaller trailing-edge vortices can be seen shedding during the
downstroke. In the mean time, two vortical structures (with opposite rotation to the trailing-edge vortices) are also shed
from the upper surface of the airfoil. When a reaches E7:8 during the downstroke the ow is already attached at the
leading-edge, and by the time the cycle is complete the ow is reattached over the entire surface of the airfoil.
Fig. 8(a) shows that the normal force coefcient in this case continues to increase during the upstroke up to aE21 :
This is the point at which ow separates from the airfoil, as seen in Fig. 10(e)(g). Near the maximum incidence, the
normal force suddenly increases again; this is due to the trailing-edge vortex being formed completely and giving a
suction effect over large part of the upper surface. During the downstroke, till aE7:8 ; the normal force coefcient is
smaller than during the upstroke; this is attributed to the ow separation during this part of the cycle. Near the end of
the downstroke the normal force increases again, this is coincident with ow reattachment on the upper surface.
The vorticity contour plots for the airfoil at a% 15 ; Da 10 ; and k 0:25 are shown for the third oscillation cycle
in Fig. 11. At the beginning of the cycle the ow is not completely attached to the surface, and there are vortices in the
867
near wake which are due to the separation and vortex shedding from the previous cycle. As time increases the ow
becomes fully attached to the surface and the small vortices in the wake are convected downstream. The ow remains
fully attached to the surface up to aE16:5 : At aE19:5 the ow starts to separate from the leading-edge, and at aE22
the ow is fully separated with the leading-edge vortex being fully formed and covering two-thirds of the upper surface.
Note that in the previous case, k 0:15; this situation was observed at aE19:3 : At amax the leading-edge vortex is shed
from the airfoil at about mid-chord position on the upper surface. Later during the downstroke at aE22:3 a trailingedge vortex is formed while the leading-edge vortex is being convected in the near wake. The ow over the upper surface
and in the wake remain vortical during the downstroke, as two smaller vortices from the trailing-edge and other vortical
structures from the upper surface are shed. At the end of the cycle the ow reattaches to most of the upper surface. Note
the similarities in the ow eld between the beginning and the end of the cycle.
It is interesting to note that the ow eld remains quiet during most of the upstroke, and much of the separation
and vortex-shedding activity occurs between aE22 in the upstroke and aE8 in the downstroke. Also interesting to
note is the speed at which the ow eld changes near amax in the upstroke. During a change in incidence of only about
3 ; from aE22 to 25 ; the ow eld bursts into a vortical ow and remains as such in the downstroke down to
aE7 :
As seen in the normal force coefcient plot for this case, Fig. 8(b), Cn continues to increase up to amax during the
upstroke; that is when the ow completely separates from the surface. In this case the ow separates at a higher
incidence compared to the previous case with a lower reduced frequency, k 0:15: Also, similar to the previous case,
both the separation and reattachment of the ow start from the leading-edge.
Shown in Fig. 12 are the vorticity contour plots for the airfoil at a% 15 ; Da 10 ; and k 0:5 during the third
oscillation cycle. At the beginning of the cycle the wake is still vortical and the ow remains separated over the rear half
of the upper surface of the airfoil as a result of separation in the previous cycle. In this case the frequency of oscillation
is so high that the ow cannot recover from the separation which occurred in the previous cycle. As time increases the
vortices on the airfoil are convected downstream, and the ow becomes attached to the entire upper surface, even
though the angle of attack is increasing. This is seen in parts (b)(e) of the gure for incidences up to aE16 : At aE19
the rst indication of a leading-edge separation bubble can be observed, but the ow is still attached over most of the
surface. At aE22:4 this separation bubble increases in size. Note that the narrow separation region on the rear half of
the upper surface is the result of the previous cycle of oscillation, and is not due to the newly formed separation bubble.
At a 25 the leading-edge vortex has completely formed and has travelled half-way along the surface. This should be
compared with the previous two cases, k 0:15 and 0:25; in which the leading-edge vortex had already travelled further
downstream at this angle of attack. The leading-edge vortex continues to evolve and travel downstream during the
downstroke, down to aE13:3 : At aE10:4 ; as the leading-edge vortex is being shed, a trailing-edge vortex starts to
form. During the remainder of the downstroke the trailing-edge vortex continues to grow, and is later shed from the
airfoil. This is while the remains of the rst leading-edge vortex, and the smaller, secondary vortices formed on the
upper surface during the downstroke, prevent the ow from reattaching.
The separated ow at such small angles of attack results in a high-pressure area on the upper surface of the airfoil,
such that the lift coefcient takes negative values, as seen in Fig. 8(c). For most of the downstroke the leading-edge
vortex covers all of the upper surface aft of the quarter-chord, and hence, the low-pressure region on that part of the
airfoil results in a negative pitching moment between aE25 and 12 ; see Fig. 8(c). Note the considerable differences in
the ow eld between this case, with the highest reduced frequency, k 0:5; and the previous two cases, with k 0:15
and 0:25: In the present case the ow remains attached to most of the upper surface during most of the upstroke
(Fig. 12be), the leading-edge vortex is formed and released entirely during the downstroke, and only one trailing-edge
vortex is shed per cycle of oscillation. Also both the separation and reattachment of the ow started from the leadingedge, similar to the previous two cases with lower reduced frequencies.
4.3.2. Effect of Reynolds number

Results for a Reynolds number of Re 3000; a value about three times lower than in other case-studies, are shown in
Fig. 13(a). If these results are compared with those of Fig. 8(b) the effect of Reynolds number can be seen. In general,
the hysteresis loops in the two cases are very similar. However, the minimum value of Cm is around
0:14 for
Re 3000; and around
0:19 for Re 104 : Also, the area of the negative-damping region near amax is larger for the
lower Reynolds number. Other than these differences in detail, the two sets of hysteresis loops are very similar for these
two cases. These results suggest that, compared to other parameters, a Reynolds number difference of this order has
only minor effects on the behavior of the force and pitching moment coefcients in the dynamic stall characteristics of
this airfoil.
868
4.3.3. Effect of mean incidence

If the mean incidence is increased to a% 20 the resulting force and moment coefcients are as plotted in Fig. 13(b).
The force coefcient hysteresis loops suggest that stall occurs at aE26:5 : Comparing the results for this case with those
for a mean incidence of 15 ; Fig. 8(b), shows that an increase in the mean incidence has considerable effect on the force
and moment coefcients of a pitching airfoil. In particular, for the results at the higher value of incidence the hysteresis
loops are wider (higher hysteresis effects); the maximum value and the average of the force coefcients are greater; and
the negative damping in the pitching moment loop is smaller.
The vorticity contour plots during the third oscillation cycle for the airfoil at a% 20 ; Da 10 ; and k 0:25 are
shown in Fig. 14. At the beginning of the cycle the ow is still separated over parts of the upper surface as a result of the
ow separation which occurred during the previous cycle. However, later at aE12:6 the ow reattaches to parts of the
surface. At aE15:5 a new separation bubble starts to form on the upper surface of the airfoil. The development of this
separation bubble into a vortex is observed during aE18:5 to 24:5 : At aE27 this vortex is about to shed from the
surface.
At the maximum incidence of the airfoil a trailing-edge vortex forms, producing a large region of ow separation on
the surface. During the downstroke two other trailing-edge vortices form and are shed, while two large-scale structures
of positive vorticity form and are released from the upper surface. As the end of the oscillation cycle is approached the
869
formation of any further vortices ceases, but the ow remains separated over most of the upper surface. Compared to
the simulation with a lower mean incidence a% 15 ; in this case the ow remains separated during a longer period of
the oscillation cycle and the wake is more vortical and violent.
As observed in Fig. 14(f) and (g), the ow becomes massively separated at aE25 during the upstroke. This is
associated with a steep drop in the normal force coefcient, as shown in Fig. 13(b). Also, referring to Fig. 14(e) and (k),
for example, one can explain the behavior of the pitching moment coefcient in the upstroke compared to the
downstroke, at aE21 : During the upstroke at aE21 ; shown in Fig. 14(e), there is a large vortex over most of the
upper surface aft of the rotation axis, resulting in a low-pressure area and producing a nose-down pitching moment
(a negative Cm ). The surface pressure coefcient Cp distribution for this particular time is shown in Fig. 15; as seen,
the positive area of the pressure distribution aft of the quarter-chord is larger compared to that ahead of it, and hence
a nose-down pitching moment.
On the other hand, during the downstroke at around the same angle, shown in Fig. 14(k), the ow over most of the
upper surface is separated, resulting in a high-pressure area and producing a nose-up pitching moment, thus a
positive Cm : The surface pressure coefcient distribution at this particular time is plotted in Fig. 16, which shows that
the pressure on the upper surface is even higher than that on the lower surface aft of the quarter-chord. The net
positive area of the pressure distribution aft of the quarter-chord is less than that ahead of it, hence, a nose-up
pitching moment.
0.2
0.2
Cn
Cn
Cn
4.3.4. Effect of pitch axis location

If the pitch axis is located at the mid-chord x=c 0:5 rather than the quarter-chord x=c 0:25; the force and
moment coefcients, averaged over the last two oscillation cycles, are as plotted in Fig. 13(c).
A comparison of the force coefcient loops for this case with those shown in Fig. 8(b), when the pitch axis is at the
quarter-chord, suggests that moving the pitching axis location backwards does not have much effect on the force
coefcients, but modies signicantly the pitching moment hysteresis loop. In this case a considerable increase in
negative damping even at low reduced frequencies is observed.
0.4
Cm
Cm
Cm
0.2
0
0
-0.2
-0.2
0.5
0.5
0.5
(a)
Cd
Cd
Cd
-0.2
0
0
10 15 20 25 30
(b)
0
5
10 15 20 25 30 35
(c)
10 15 20 25 30
Fig. 13. The hysteresis loops for the normal force Cn ; pitching moment Cm ; and drag Cd coefcients for the airfoil oscillating with
Da 10 and k 0:25; and: (a) a% 15 ; Re 3000; x=c 0:25; (b) a% 20 ; Re 104 ; x=c 0:25; (c) a% 15 ; Re 104 ; x=c 0:5:
, pitching case; - - -, steady case.
870
- Cp
2
1
0
-1
0
0.2
0.4
0.6
0.8
x/c
Fig. 15. The surface pressure coefcient distribution at aC21:5 during the upstroke over the airfoil oscillating with k 0:25; Da
10 ; a% 20 ; x=c 0:25; and Re 104 :
871
- Cp
2
1
0
-1
0
0.2
0.4
0.6
0.8
x/c
Fig. 16. The surface pressure coefcient distribution at aC21:5 during the downstroke over the airfoil oscillating with k 0:25;
Da 10 ; a% 20 ; x=c 0:25; and Re 104 :
872
- Cp
2
1
0
-1
0
0.2
0.4
0.6
0.8
x/c
Fig. 18. The surface pressure coefcient distribution at aC19:5 during the upstroke over the airfoil oscillating with k 0:25; Da
10 ; a% 15 ; x=c 0:25; and Re 104 :
The vorticity contour plots during the third oscillation cycle for the airfoil in this case, with x=c 0:5; a% 15 ;
Da 10 ; and k 0:25; are shown in Fig. 17. The ow is attached over most of the upper surface of the airfoil at the
beginning of the oscillation cycle, while the wake is still vortical due to the previous cycle. The ow remains attached
during the upstroke up to aE16:5 ; as was the situation with the pitch axis located at the quarter-chord. At aE19:5 a
separation bubble forms at the leading-edge which then develops into a vortex. At amax this leading-edge vortex has
already travelled two-thirds of the airfoil upper surface.
In the downstroke part of the cycle a trailing-edge vortex forms, while the leading-edge vortex is being shed from the
upper surface. As the downstroke continues the rst trailing-edge vortex sheds, as seen in Fig. 17(k). Meanwhile, a
second vortex forms on the upper surface, and is later shed from the surface. The ow starts to reattach at the leadingedge at aE7:9 : Even at a 5 ; at the end of the cycle, the ow is not completely attached to the surface.
Comparing Fig. 17 with Fig. 11, with pitch axis located at the mid-chord and quarter-chord, respectively, shows that
the ow elds in these two cases are very similar during most of the upstroke. Even at amax the two ow elds have
considerable similarities. The most signicant difference is in the timing of the formation and shedding of the trailingedge vortices, which for x=c 0:5 is slightly behind that for x=c 0:25; as can be observed by comparing parts (k) with
(m) of Figs. 17 and 11. The resemblance between the ow elds can also be detected from the similarities between the
force coefcient plots of the two cases, shown in Figs. 8(b) and 13(c). However, the pitching moment coefcient loops
for the two cases are quite different, which was predictable given the difference in location of the pitching axis. For
x=c 0:5 the pitching moment is always positive, which means that the aerodynamic center is always ahead of the pitch
axis.2
The behavior of the pitching moment, as shown in Fig. 13(c), can be explained as follows. During most of the
upstroke the ow is either attached near the leading-edge, or a vortex is present in this area. In either case a lowpressure region is maintained in this area. For example, at aE19:5 the leading-edge vortex covers the front half of the
upper surface, as seen Fig. 17(f). The surface pressure coefcient distribution for this instant of time is shown in Fig. 18.
It can be seen that the pressure ahead of the mid-chord is much lower than that aft of the mid-chord, and thus a noseup pitching moment about the mid-chord is produced. Also, during the downstroke the intensity of the separated ow
is higher aft of the mid-chord, as seen in Fig. 17(j)(n), keeping a lower pressure ahead of the mid-chord compared to
that aft of the mid-chord. For example, Fig. 17(j) shows a large separated ow region aft of the mid-chord on the upper
surface at aE19:4 during the downstroke. The surface pressure coefcient distribution for this instant of time is shown
in Fig. 19. It is clear that the pressure on the upper surface aft of the mid-chord is even higher than that on the lower
surface. The pressure ahead of the mid-chord, on the other hand, is lower on the upper surface compared to that on the
lower surface. Hence, a nose-up pitching moment is produced about the mid-chord. However, Figs. 18 and 19 show
that the pressure on the upper surface ahead of the mid-chord is much lower (or,
Cp is higher) during the upstroke
compared to that during the downstroke, and hence, the positive pitching moment is higher during the upstroke.
2
This, in turn, was predictable, since the aerodynamic center is usually located around the quarter-chord for most conventional
airfoils.
873
- Cp
2
1
0
-1
0
0.2
0.4
0.6
0.8
x/c
Fig. 19. The surface pressure coefcient distribution at aC19:4 during the downstroke over the airfoil oscillating with k 0:25;
Da 10 ; a% 15 ; x=c 0:25; and Re 104 :
5. Conclusions
In this paper the pitching oscillations of a NACA 0012 airfoil were studied, and the effects of some parameters on the
dynamic stall characteristics of the airfoil were investigated.
It was observed that the pitching oscillations of the airfoil delayed the ow separation to higher incidences compared
to the static stall case. The normal force was observed to increase well beyond that at the static stall angle. Separation
was observed to start from the leading-edge, followed by the formation and convection of a vortex along the surface of
the airfoil. Trailing-edge vortices were also observed to form and shed from the airfoil. During the downstroke of the
oscillation one or more secondary vortices were observed to form and release from the upper surface, following the
release of the primary dynamic stall vortex.
The reduced frequency of oscillation was found to have a major effect on the ow eld and the force coefcients of
the airfoil. Flow separation was delayed to higher incidences with increasing reduced frequency. Hence, the peak in the
normal force coefcient was observed at a higher angle of attack with increasing the reduced frequency. It was found
that the maximum normal force slightly increased, and the minimum pitching moment decreased, with increasing
reduced frequency. However, perhaps the most important effect due to an increase in the reduced frequency was a
considerable increase in the negative damping in the pitching moment coefcient hysteresis loop observed at the
highest reduced frequency.
The Reynolds number of the ow, on the other hand, was found to have little effect on the dynamic stall
characteristics of the airfoil, at least for the range of Re 3000104 considered in our simulations.
Increasing the mean incidence caused ow separation to occur earlier in an oscillation cycle, even at lower angles of
attack compared to a case with a lower mean incidence. The ow eld in this case was found to be more vortical, and
the wake to be wider and more violent. Also, the hysteresis effects in the force and pitching moment coefcient loops
intensied, and the maximum normal force and drag coefcients increased. On the other hand, the minimum pitching
moment increased (its absolute value decreased), and the negative damping in the pitching moment loop was
suppressed.
The simulations showed that moving the location of the pitch axis backwards, from the quarter-chord to the midchord, had minor effects on the ow eld in terms of the ow separation and the timing of the formation and release of
the leading- and trailing-edge vortices, as well as on the hysteresis loops of the force coefcients. For example, it was
found that this change of location caused the trailing-edge vortex to be shed slightly later during the downstroke, the
normal force coefcient loop to become slightly thinner, and the drag coefcient loop slightly wider. However, this
change of the pitch axis location caused a dramatic change in the pitching moment loop: the pitching moment became
positive during the entire cycle, and a large negative damping region appeared in the pitching moment loop. This
behavior of the pitching moment loop was explained using the ow eld information and the surface pressure
coefcient distribution plots.
Acknowledgements
The authors gratefully acknowledge the nancial support of the Natural Sciences and Engineering Research Council
of Canada (NSERC) and Les Fonds FCAR du Qu!ebec.
874
References
Abbott, I.H., von Doenhoff, A.E., 1959. Theory of Wing Sections. Dover, New York.
Akbari, M.H., 1999. Bluff-body ow simulations using vortex methods. Ph.D. Thesis, McGill University, Montr!eal, Canada.
Akbari, M.H., Price, S.J., 2000. Simulation of the ow over elliptic airfoils oscillating at large angles of attack. Journal of Fluids and
Structures 14, 757777.
Carta, F.O., 1967a. An analysis of the stall utter instability of helicopter rotor blades. Journal of the American Helicopter Society 12,
118.
Carta, F.O., 1967b. Unsteady normal force on an airfoil in a periodically stalled inlet ow. Journal of Aircraft 4, 416421.
Chorin, A.J., 1973. Numerical study of slightly viscous ow. Journal of Fluid Mechanics 57, 785796.
Crimi, P., 1973. Dynamic stall. AGARD-AG-172.
Ericsson, L.E., 1967. Comment on unsteady airfoil stall. Journal of Aircraft 4, 478480.
Ericsson, L.E., Reding, J.P., 1971. Unsteady airfoil stall, review and extension. Journal of Aircraft 8, 609616.
Ericsson, L.E., Reding, J.P., 1972. Dynamic stall of helicopter blades. Journal of the American Helicopter Society 17, 1119.
Ericsson, L.E., Reding, J.P., 1980. Dynamic stall at high frequency and large amplitude. Journal of Aircraft 17, 136142.
Ericsson, L.E., Reding, J.P., 1988. Fluid mechanics of dynamic stall. Part I: unsteady ow concepts. Journal of Fluids and Structures 2,
133.
Johari, H., Durgin, W.W., 1998. Direct measurement of circulation using ultrasound. Experiments in Fluids 25, 445454.
Jumper, E.J., Huge, R.J., 1991. Simple theories of dynamic stall that are helpful in interpreting computational results. Computer
Physics Communications 65, 158163.
Lorber, P.F., Carta, F.O., 1988. Airfoil dynamic stall at constant pitch rate and high Reynolds number. Journal of Aircraft 25, 548
556.
Massey, B.S., 1979. Mechanics of Fluids, 4th Edition. Van Nostrand Reinhold, New York.
McCroskey, W.J., Carr, L.W., McAlister, K.W., 1976. Dynamic stall experiments on oscillating airfoils. AIAA Journal 14, 5763.
Ohmi, K., Coutanceau, M., Phuoc Loc, T., Dulieu, A., 1991. Further experiments on vortex formation around an oscillating and
translating airfoil at large incidences. Journal of Fluid Mechanics 225, 607630.
Raffel, M., Kompenhans, J., Wernet, P., 1995. Investigation of the unsteady ow velocity eld above an airfoil pitching under deep
dynamic stall conditions. Experiments in Fluids 19, 103111.
Tuncer, I.H., Wu, J.C., Wang, C.M., 1990. Theoretical and numerical studies of oscillating airfoils. AIAA Journal 28, 16151624.

Simulation of Dynamic Stall For A NACA 0012 Airfoil Using A Vortex Method

Uploaded by

Copyright:

Available Formats

Simulation of Dynamic Stall For A NACA 0012 Airfoil Using A Vortex Method

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Simulation of Dynamic Stall For A NACA 0012 Airfoil Using A Vortex Method

Uploaded by

Copyright:

Available Formats

Journal of Fluids and Structures 17 (2003) 855874

Simulation of dynamic stall for a NACA 0012 airfoil

4. Results and discussion

4.1. Static airfoil

4.2. A flow visualization comparison

4.3.2. Effect of Reynolds number

4.3.3. Effect of mean incidence

4.3.4. Effect of pitch axis location

You might also like