Aerospace Science and Technology 10 (2006) 181191

www.elsevier.com/locate/aescte

Control of a decelerating boundary layer.

Part 1: Optimization of passive vortex generators
G. Godard 1 , M. Stanislas
Laboratoire de Mcanique de Lille, UMR CNRS 8107 France, Bv Paul Langevin, cit Scientifique, 59655 Villeneuve dAscq, France
Received 30 September 2005; received in revised form 21 November 2005; accepted 22 November 2005
Available online 6 January 2006

Abstract
The control of boundary layer separation on the suction side of an airfoil at high angle of attack has been renewed by the possibilities of active
control. Nevertheless, such an active control needs a deep understanding of the flow to manipulate and of the actuating flow, both being 3D and
unsteady. For that purpose, a model experiment has been designed in the frame of a coordinated European project called AEROMEMS, with a
simpler (2D) geometry and with a dilatation of the scales in order to be able to characterize the actuation flow. This model is a bump in a boundary
layer wind tunnel, which mimics the adverse pressure gradient on the suction side of an airfoil at the verge of separation. The present contribution
describes preliminary tests done to optimize standard passive devices before testing active systems. The optimization was done with hot film shear
stress probes, the characterization with hot wire anemometry and PIV. The results show quantitatively the improvement brought by the passive
devices in terms of skin friction. They also show the mechanism which is at the origin of this improvement. The next step of the project is to
replace passive devices by synthetic jets.
2006 Elsevier SAS. All rights reserved.
Keywords: Flow control; APG boundary layer; Vortex generators; PIV

1. Introduction
In the last decades, significant progress has been made in
both numerical and experimental fluid dynamics. New tools
are available which allow a deeper insight in the physics of
complex flows, including turbulence. Beside this, the strong development of Micro Electro-Mechanical Systems (MEMS) is
opening a new field of flow actuation at a small scale and at
low energetic cost [1]. The main advantage of such devices is
that they could be integrated smoothly in the skin of the airfoil, acting only when needed. The possibility of reducing skin
friction in cruise flight with such devices have already been addressed, both from the theoretical [2,3] and experimental [4]
points of view and will not be discussed here. Besides, the
control of the flow around an airfoil at high angle of attack is
also of strong interest. Such a flow is encountered in several
flight phases (take-off, landing, manoeuvre. . .) and can lead to
* Corresponding author. Tel.: +33 03 20 33 71 70, fax: +33 03 20 33 71 69.

E-mail address: stanislas@ec-lille.fr (M. Stanislas).

1 Presently: CORIA, UMR CNRS 6614, Rouen, France.

1270-9638/$ see front matter 2006 Elsevier SAS. All rights reserved.
doi:10.1016/j.ast.2005.11.007

partial or large separation, with the well known consequences

(vibration, drag increase and loss of control). To delay and even
suppress this separation, it is necessary to bring momentum
into the boundary layer (BL) in order to allow it to sustain the
strong adverse pressure gradient (APG). Various attempts have
been made in the past with suction, blowing and passive devices [511] at a macro scale. In order to use MEMS devices,
the actuation must be performed at micro scales and with a
limited amount of energy available for actuation. Thus, natural
non-linear mechanisms of the flow should be used in order to
amplify the actuation effect up to the macro scales. In the case
of a separating boundary layer, due to the strong convection,
this means that the actuation has to be done far enough upstream of the separation point so that the non linear mechanisms
have time enough to develop. The idea is thus to enhance the exchange of momentum between the boundary layer and the outer
flow with the help of optimal coherent structures, triggered
by an optimal actuator. Of course, these optimal coherent
structures should have something to do with the natural coherent structures of the boundary layer. Lumley [12] estimates at
80% the contribution of these natural coherent structures to the

182

G. Godard, M. Stanislas / Aerospace Science and Technology 10 (2006) 181191

Nomenclature
Latin symbols
Cd
Cf

drag coefficient
skin friction coefficient, Cf =

Cl
Cp

lift coefficient
pressure coefficient, Cp =

Cq
dP /dx
fp
H
h
H12
H32
H
U, V

p
1
2
U
e
2

P
1
2
2 Ue

total flow coefficient

pressure gradient . . . . . . . . . . . . . . . . . . . . . . . . Pa/m
pulsating frequency . . . . . . . . . . . . . . . . . . . . . . . . Hz
height of the bump . . . . . . . . . . . . . . . . . . . . . . . . . . m
vortex generator height . . . . . . . . . . . . . . . . . . . . . . m

shape factor, H12 =

shape factor, H32 =

Schlichting shape factor
velocity components tangential and normal to the
surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m/s
Ue
external velocity . . . . . . . . . . . . . . . . . . . . . . . . . . m/s
Uinf , U freestream velocity . . . . . . . . . . . . . . . . . . . . . . m/s
U
friction velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . m/s
x, y, z longitudinal, vertical and transverse coordinates m
Xd
abscissa of the passive lines . . . . . . . . . . . . . . . . . . m
Greek symbols
pd

skew angle of the passive devices . . . . . . . . . . . . . .

boundary layer thickness . . . . . . . . . . . . . . . . . . . . . m
displacement thickness . . . . . . . . . . . . . . . . . . . . . . m

total turbulent energy of the boundary layer. As can be immediately inferred, an optimal control with optimal structures
means thus a deep physical understanding of the physics of the
turbulent boundary layer (which is still a subject of strong controversy). However, the situation is not so pessimistic. A lot of
basic work has been done in the last twenty years on the structure of the turbulent boundary layer [13], both experimentally
and (more recently) by Direct Numerical Simulation [14,16].
Although the full mechanism is not well understood, a general
agreement is coming out on a certain number of coherent structures such as ejections, sweeps, streamwise vortices. . . An optimal control structure should a priori be one of these clearly
identified structures. Among those, Jimenez and Pinelli [15]
and Orlandi and Jimenez [16] did show by numerical simulation the important role played by streamwise (or more precisely
quasi streamwise) vortices in the near wall region of the BL.
An experimental study with passive devices performed by Lin
et al. [811] confirms that streamwise vortices tend to transfer
momentum towards the wall and to increase the skin friction.
Thus, this structure was selected in the present study as a good
candidate for being an optimal structure.
Beside this basic physical problem of selecting a control
mechanism, two important technical problems appear quite
rapidly. The first one, which is the selection of an active actuating device, will not be discussed here as the results presented
are preliminary and obtained with passive devices. The sec-

0

P
X
XVG
z

momentum thickness . . . . . . . . . . . . . . . . . . . . . . . . m
energy thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . m
coordinate normal to the bump . . . . . . . . . . . . . . . m
dimensionless coordinate normal to the bump, + =
U
v

distance between two passive devices . . . . . . . . . m

dimensionless distance between two passive devices, + = Uv
wall shear stress . . . . . . . . . . . . . . . . . . . . . . . kg/m2 s
wall shear stress without actuation . . . . . . kg/m2 s
0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . kg/m2 s
pressure difference . . . . . . . . . . . . . . . . . . . . . . . . . . Pa
position of PIV plane with respect to actuation
plane, X = XPIV XVG
streamwise distance between the devices and the
minimum skin friction line . . . . . . . . . . . . . . . . . . . m
lateral displacement . . . . . . . . . . . . . . . . . . . . . . . . . m

Abbreviations
BL
APGBL
ZPGBL
CoR
CtR
VG
PIV

boundary layer
adverse pressure gradient boundary layer
zero pressure gradient boundary layer
co-rotating
counter-rotating
vortex generator
particle image velocimetry

ond one is a size problem. As real boundary layers are quite

thin and MEMS quite small, an experiment at real scale appears quite disappointing on the physics point of view. As the
spatial resolution of available measurement techniques is very
poor at such scale, an experimental approach would be reduced
to a trial and error method with some global measurement of
the effect of the device downstream. No hope to understand the
mechanisms underlying the success or failure of such device.
The concept of optimality would then be reduced to its simplest
form. In order to overcome this problem, as will be described
downstream, it was decided to enlarge significantly the scale of
the phenomenon by simulating it at a macro scale. This has the
main advantage to make it accessible to efficient measurement
techniques such as Hot Wire Anemometry or Particle Image
Velocimetry, allowing a quantitative assessment of both the incident and actuated flow.
The ideas of passive control of flow separation and of vortex
generators is quite old and is already used in several practical
situations. The most recent contributions to the subject are due
to Rao and Kariya [6] and Lin et al. [811]. A good review of
previous works can be found in these contributions.
The work of Rao and Kariya, which is dated 1988, was devoted to submerged vortex generators. The idea was to reduce
the size of the devices, as compared to previous studies. The
flow was a separated flat plate boundary layer. Different configurations were tested with a height of the order of 60% of . The

G. Godard, M. Stanislas / Aerospace Science and Technology 10 (2006) 181191

conclusion was that concave slot devices arranged in V were

the best devices and should be investigated at a smaller size.
In the contribution of Lin et al., three papers between 1989
and 1991 [810] are based on the same basic experiment. The
last contribution [11] is an application to a realistic airfoil in
high lift configuration. In the first three papers, a wide variety of passive devices were studied, with more or less success.
This led the authors progressively to the conclusion that the
most efficient mechanism to reduce flow separation is based on
the exchange of momentum between the near wall region and
the outer flow. They found that this was done quite efficiently
by vortex generators producing longitudinal counter-rotating
vortices. They also showed that such devices could be much
smaller than the boundary layer thickness (of the order of 10%
of ) and still fairly efficient.
In the last paper from Lin [11], micro vortex generators
VGs were studied in detail. These generators were thin plates
of triangular or trapezoidal shape, placed normal to the surface
and at a lateral angle to the flow. They were staggered in either
co-rotating (parallel) or counter-rotating (V shape) configurations. The conclusion was that VGs of h/ = 0.2 placed in a
counter-rotating configuration at 5h to 10h upstream of the separation line were the most effective. Increasing this height parameter was increasing drag without improving significantly the
pressure recovery. Reducing h/ was reducing the efficiency of
the VGs.
Recently, Betterton et al. [17] did a detailed study of the
wake of different vortex generators inspired from the studies of
Lin et al. [811]. This was performed in ZPGBL and APGBL.
Their main conclusion was that the spaced counter rotating
vanes of the type used by Lin [11] were the most efficient of
the set tested to delay separation.
Flow separation control, by means of passive devises, is today the less expensive and the quickest solution to implement.
They can be used to control from low-speed separated flows
in adverse pressure gradient to transonic shock-induced separation. The usual and most effective devices are the vane-type
vortex generators (VGs). They consist of an array of plates fixed
perpendicularly to the surface and skewed at an angle pd to
the main flow direction. They act as streamwise vortex generators (VGs). Fig. 1 shows two VGs configurations that produce
co-rotating and counter-rotating longitudinal vortices arrays re-

183

spectively. Streamwise vortices develop downstream of these

devices and induce momentum transfer between the free-stream
and the near wall region, this mechanism is well describe by
Angele [18], see also Rao [6], Pauley [7], Lin [811], Betterton
[17] and Jenkins [19].
The co-rotating array transports low momentum fluid upward (away from the wall) and high momentum fluid downward between two adjacent streamwise vortices (Fig. 1(a)). For
small values of the spanwise spacing, these opposite phenomena reduce the vortices effectiveness and persistence. Besides, a
co-rotating array of vortices induces its own spanwise displacement by self induction while developing downstream. The major advantage of the co-rotating system is that, usually, the vortices do not move away from the surface. The counter-rotating
system dissociates the upward and downward momentum transport (Fig. 1(b)). The high momentum is transported downward
to the wall around each VGs plane of symmetry. The low momentum is transported upward to the free stream between two
different VGs. In spite of higher efficiency in momentum transfer, some authors (Pauley et al. [6]; Betterton et al. [17]) have
observed the vortices motion away from the wall region. This
appeared for vortices array embedded in a zero pressure gradient boundary layer and for an equidistant spanwise spacing
between the vortices. Results of recent studies (Betterton et al.
[17]; Angele et al. [18]) show that, in adverse pressure gradient
condition, the counter-rotating array of vortices stays near the
wall, even far downstream. Lin [11] estimated the vortices life
distance to be around 100 devices height (h). The maximum
streamwise distance between the devices and the separation
line (XVGs ) reported by Lin for backward-facing ramp tests
was 10h. Recent studies (Betterton et al. [17]; Angele et al.
[18]) have concluded to the effectiveness of submerged VGs
for higher distances (17 < XVGs / h < 52).
Because of drag penalty induced by these passive devices,
many optimization studies have led, in 20 years, to a significant
reduction of the vortex generator height (h) from the order of
the boundary layer thickness () to only a fraction of it.
Despite the numerous results available at the time of the
present study, it was not possible to estimate precisely the skin
friction improvement in the present configuration from the literature data. In fact, very few skin friction measurements are
available. It was thus of interest to perform these measurements

Fig. 1. (a) Co-rotating passive device configuration, (b) counter rotating passive device configuration.

184

G. Godard, M. Stanislas / Aerospace Science and Technology 10 (2006) 181191

Fig. 2. Sketch of the bump in the wind tunnel.

Fig. 3. Lins passive vortex generators in the wind tunnel.

in order to have a documented reference case for comparison

with active devices. Besides, it was of interest to perform an optimization study, as the optimal configuration was not straightforward from the literature survey. Finally, it was useful to identify the mechanisms at the origin of the skin friction increase in
order to compare them with those driving the active devices.
2. Experimental set-up
2.1. Wind tunnel
For the present experiments, the same boundary layer wind
tunnel as in Bernard et al. [20] was used (see this reference for
a detailed description). The test section is 1 2 m2 and 20 m
long. The last 5 m are transparent on all sides to allow the use of
optical methods. The free stream velocity can be varied continuously from 1 to 10 m/s. The wind tunnel can be used in closed
loop with temperature regulation or opened to the outside to allow the use of smoke. In the present study, it was used in the
closed loop configuration for PIV measurements and for HWA.
In this configuration, the wind tunnel is computer controlled by
a PC, both for velocity (0.5%) and temperature (0.2 C). In
open loop, only the velocity is regulated. The boundary layer
under study develops on the lower wall. It is tripped at the entrance of the tunnel by a grid laid on the floor. The origin of the
coordinate system is placed in the middle of the lower wall at
the entrance of the tunnel. The x axis is parallel to the wall and
to the flow, the y axis is normal to the wall, the reference frame
is direct.
2.2. Bump
In order to obtain a significant 2D adverse pressure gradient
in the wind tunnel, a bump was designed and manufactured in

the AEROMEMS project. The shape was computed by Dassault

Aviation, using a two-dimensional NavierStokes solver with a
k turbulence model. The objective was to approach separation without reaching it in order to prevent the flow to become
three-dimensional. The 2D bump is attached on the wind tunnel floor. It was designed in three main parts (see Fig. 2). The
first part is converging, realized with a 2 mm thick steel sheet,
it is sliding on the main part of the model. The main part is
made of wood covered with a thin steel sheet and connected to
an aluminum plate in the downstream divergent part. In this flat
aluminum plate, a window was opened to fit actuators and other
measurement devices (sensors, glass window. . . ). The last part
is a flexible sheet of PVC which is shaped to obtain a good continuity with the wind tunnel floor. The bump surface is painted
in black and is equipped with some pressure taps along the longitudinal median plane, in order to measure the longitudinal
pressure gradient. The main characteristics of the flow around
this bump are detailed in Bernard et al. [20].
2.3. Actuators
Based on the literature survey, only one type of actuator
was tested in the present study. This is the thin plate vortex
generator as suggested by Lin [10] and as illustrated by the
photographs of Fig. 3. They were manufactured from thin steel
plates (e = 0.5 mm) which were cut and folded. Series of actuators were manufactured with different heights, lengths and
shapes (rectangular or triangular). They were glued with double sided tape along a span wise line, all across the wind tunnel.
It was thus easy to vary their configuration (co- or counterrotating), position, spacing and skew angle. Fig. 4 gives the
geometrical parameters which were used to characterize the two
configurations tested.

G. Godard, M. Stanislas / Aerospace Science and Technology 10 (2006) 181191

185

Fig. 4. Geometry of (a) co-rotating passive devices, (b) counter-rotating passive devices.

3. Measurement techniques
Different measurement techniques were used to characterize
the effect of the devices on the flow.
3.1. Hot wire anemometry
The velocity profiles in the wake of the actuators were measured using a single Hot-Wire Anemometer. The wires are
2.5 m in diameter and 0.5 mm in length. The anemometers
used are of constant temperature type: AN 1003 manufactured
by AALab Systems. Details can be found in Bernard et al. [20].
3.2. Skin friction measurements with hot film sensors
In order to obtain a quantitative information on the efficiency of each configuration tested, the wall shear stress was
measured with hot film probes. These measurements were performed mainly at the location of minimum shear stress of the
smooth configuration.
Senflex SF9902 hot film probes were used. They are 1.5 mm
long and they are deposited on a polyamyde substrate with a
thickness less than 0.2 mm. They can be glued directly on the
surface with 60 m double-sided tape. The experience showed
that it is necessary to drill a hole of 2 mm in diameter and
0.5 mm in depth under the sensor to minimize heat losses to
the substrate. The probes were connected to the 4 channels AN1003 AAlab anemometer used also for HWA. The signal from
the anemometer was low pass filtered at 5.5 kHz and digitized
by a PC computer equipped with an A/D converter board and a
sample and hold circuit. The acquisition frequency was 11 kHz
and 1.1 million samples were recorded for each measurement.
Calibration of wall shear stress probes is always a delicate
problem [21]. A specific device was realized for this calibration. A pipe with an inner diameter of 0.19 m and a length of
20 m was connected to a pump and settled in by-pass to the
wind tunnel. The temperature was thus regulated to less than
0.2 C. The pressure gradient was found constant in the last
8 m of the pipe. The probes were mounted at 4 m from the pipe
outlet and calibrated using the constant pressure gradient in the
fully developed part of the flow. The calibration was done at the
same temperature as the measurements. The repeatability of the
measurements was carefully tested. It was better than 5% on the
skin friction for measurements performed on the same day, but

it could reach 20% from one day to the other. Thus, only comparative measurements on the same day were performed.
3.3. Particle image velocimetry
A standard stereo PIV set-up was used in this experiment.
It basically consists of two Nd:Yag laser cavities with amplifiers, each of them producing about 500 mJ per pulse at 12.5
Hz (nominal pulse frequency). The pulse duration is 5 ns. The
light sheet optics consists of two lenses: one spherical to adjust
the light sheet thickness (about 1 mm) and one cylindrical lens
to fix the light sheet width (about 400 mm). As the main flow is
going through the light sheet, a small separation, of the order of
0.5 mm is set between the first and the second pulse of the laser
system, to allow a larger dynamic range.
To record the images, two PCO SensiCam cameras from
Lavision were set on both sides of the wind tunnel in a
sheimpflug configuration. They provide 1280 1024 pix2 image pairs with a 12 bits dynamic range. For the present experiments, Nikon lenses of focal length f = 100 mm were used.
The magnification was around M = 0.14. With an aperture of
f # = 4 the diffraction spot size was of the order of 20 m.
The Davis software and hardware from Lavision were used
for recording. The calibration was performed by recording images of a plane target at 3 different positions around the light
sheet.
Measurements were performed at three locations along the
bump: Xd = 17.67 m, 18.09 m, 18.57 m. The images from both
cameras were processed with a standard multigrid algorithm
with discrete window shifting and Gaussian peak fitting. The
final interrogation window size was 32 32 pixels with 50% of
overlap. This gives 5336, 4292 and 4275 vectors respectively
in the three planes investigated. The Soloff method with 3 calibration planes [22] was used to reconstruct the three velocity
components in the plane of measurement. This was done using
a home made software.
4. Basic flow main characteristics
The flow without control was previously characterized in detail by Bernard et al. [20]. These results will not be recalled
in detail here, the reader is addressed to this paper for a comprehensive characterization of the APG boundary layer flow.
Fig. 5 gives the pressure gradient along the bump which shows

186

G. Godard, M. Stanislas / Aerospace Science and Technology 10 (2006) 181191

5. Optimization results

Fig. 5. Pressure gradient along the bump (from Bernard et al. [20]).

Fig. 6. Skin friction distribution along the bump (from Bernard et al. [20]).

that the flow is strongly accelerating in the first half of the

bump with a rapid variation of the pressure gradient magnitude. This pressure gradient changes sign at x = 17 m and
then varies more progressively in order to avoid separation.
This is fairly representative of what happens on the suction side
of an airfoil at moderately high angle of attack. Fig. 6 gives
the skin friction distribution in the decelerating part from both
2D NavierStokes computation by Dassault Aviation and experiments (see Bernard et al. (2003) for details). As expected
the skin friction decreases rapidly and reaches a minimum
around XCfmin = 18.58 m. This point will serve as a reference for the skin friction variations in the present optimization
study.

In boundary-layer flow separation studies, previous authors

usually used oil-flow visualization, streamwise pressure distributions and velocityprofile measurements to compare the VGs
effectiveness. The first part of the present study did consist in
testing and optimizing passive devices by means of skin friction
measurement at the streamwise location where the Cf reaches a
minimum on the smooth wall. As these measurements are very
difficult to make absolute, they were performed relatively to the
skin friction without actuation. Both measurements (with and
without actuation) were always performed in the same test at a
short time interval in order to minimize the effect of calibration
and temperature variation. Each test was repeated at least twice
to check the repeatability of the results. This procedure allowed
to get a quantitative assessment of the skin friction variation
due to the actuation. The error on this variation was of the order
of 5%. The optimal configuration obtained with these passive
devices will be used afterwards as a reference for the active
devices. The passive device characteristic parameters are presented in Fig. 4(a) for co-rotating devices and 4(b) for counterrotating ones. As can be observed, there are several geometrical
parameters, with most of them common to both configurations.
In both cases, the devices are triangular plates of height h and
length 1 placed normal to the surface. The skew angle is pd
and the leading edge of the triangles are aligned transversely
at the abscissa Xd . The skin friction measurements are performed at XCfmin = 18.58 m. The distance between the trailing
edges of the devices and the measurement station is XVG . The
transverse spacing of the devices is . For the counter rotating
system an extra length is needed. The distance L between the
trailing edges of the two plates of one system was chosen. Table 1 gives the most effective values of these parameters found
in previous studies. The counter rotating configuration was extensively studied with different geometries, while few data are
available for the co-rotating case. The agreement is quite good
on h/, which is not the case for the other parameters. Table 2
shows the different parameters tested in the present study and
the range of values investigated. Based on the literature survey
summarized in Table 1, the selected starting parameters for the
optimization procedure are given in Tables 3 and 4 respectively

Table 1
Effective parameters used in the previous passive device studies
Model used
Lin 1991
Lin 1999
Lin 1999
Bernard 2000
Betterton 2000
Jenkins 2002

APGBL (Ramp)
APGBL (Ramp)
High-lift Airfoil
APGBL (Bump)
APGBL (Bump)
APGBL (Ramp)

CtR
CtR
CtR
CtR
CtR
CoR

VGs

h/

Xvg / h

l/ h

L/ h

/ h

pd ( )

Wishbones
Rectangular vanes
Trapezoid vanes
Delta wing shape a = 45
Triangular plates
Trapezoid vanes

0.20
0.20

0.12
0.30
0.20

[510]
[510]
[510]
52
52
19

3.4
2.0
7.0
1.0
10.0
4.0

2.7
2.0
2.2
1.5
6.0

4
4
4
5
12
4

23
15
23
45
14
23

Table 2
Range of values of passive devices parameters tested
Model used

VGs

h/

XVG / h

l/ h

L/ h

/ h

pd ( )

APGBL (Bump)

CtR CoR

0.200.46

4565

13

1.44

412

1328

G. Godard, M. Stanislas / Aerospace Science and Technology 10 (2006) 181191

187

Table 3
Counter-rotating configuration starting parameters
h/

XVG / h

l/ h

L/ h

/ h

pd ( )

0.20

45

23

Table 4
Co-rotating configuration starting parameters
h/

XVG / h

l/ h

/ h

pd ( )

0.37

57

18

Fig. 8. Influence of the actuators position XVGs for counter rotating devices.

Fig. 7. Influence of the skew angle pd for different counter rotating device
spacing.
Fig. 9. Influence of the height h/ for counter rotating devices.

for the counter and co-rotating devices. Both configurations are

not far from Lins [10] optimal sets.
In the counter rotating configuration, skin friction measurements were performed at XCfmin = 18.58 m and at two spanwise positions: in the plane of symmetry of the device and at
mid distance between two devices (see Fig. 4(b)). Fig. 7 shows
the effect of the skew angle (pd ) and device spacing (/ h).
The circles give the results on the axis of a counter rotating
system (z/ = 0) and the triangles halfway between two devices (z/ = 0.5). The two lines referenced AEROMEMS I
correspond to a preliminary passive device tested in the first
AEROMEMS project. The results show a significant variation
of /0 in the spanwise direction. In each case the higher
values are located on the axis of the device (z/ = 0). For
/ h = 12, in spite of good results at z/ = 0, the device spacing
is too high to produce a significant improvement at z/ = 0.5.
On the contrary, for small spacing (/ h = 4), /0 rises
at mid distance and decreases along the VG axis. The intermediate value (/ h = 8) seems to be a good compromise. It
preserves the maximum variation (+180%) on the system axis
and increases /0 by +70% at halfway. A higher skew angle sensitivity of the / h = 4 configuration can be noticed. The
skin friction increases when pd decrease from 28 . The optimal value is around 18 for / h = 8 and 12 but can be lower for
/ h = 4. Obviously, this parameter is not too critical as soon
as / h is large enough: 17 < pd < 25 for / h
8. Values
of pd = 18 and / h = 8 were retained for further investigation.
Fig. 8 shows the influence of the streamwise position of the
actuators. In this test the hypothesis was made that, when the
VGs height increases above 0.2, the device drag increases

without a significant increase of /0 . Thus, the physical

height of the VGs was kept constant. Fig. 8 gives the skin
friction variations obtained for the same passive devices array located successively at various streamwise position with
respect to the Cfmin location (XVGs / h). There is no significant variation of /0 (10%) at z/ = 0 and only a 20%
variation at z/ = 0.5 for the range investigated. Hopefully,
this parameter is not too critical. The position corresponding
to XVGs / h = 65 was retained for further investigations. The
optimal value of the skew angle pd = 18 was confirmed by a
complementary test at this position.
Fig. 9 gives the skin friction variation /0 obtained for
different VGs height. At z/ = 0.5 the results show an increase
of /0 up to h/ = 0.28. For z/ = 0, /0 increases
monotonously in the range of h/ investigated (0.20.46). This
result is somehow in contradiction with the existing literature.
One can of course expect the curve to level off for higher values
of h, but obviously for h/ larger than 0.5.
In Fig. 10, the influence of L/ h for l/ h = 1 and l/ h = 2
is given. The two devices have the same behavior and a nearly
constant /0 difference (+30%). The higher /0 values
are obtained for l/ h = 2. At z/ = 0.5, /0 is maximal for
L/ h > 3. At z/ = 0 the optimal value is for L/ h = 2.5. This
parameter appears relatively sensitive at z/ = 0. A value of
L/ h = 2.5 was chosen for the remaining tests.
Fig. 11 gives, at this station (XVGs / h = 65) and for L/ h =
2.5, the influence of the device spacing / h. This parameter
appears fairly sensitive between two devices. The optimal value
of /0 is found for / h = 6 at this location. On the system

188

G. Godard, M. Stanislas / Aerospace Science and Technology 10 (2006) 181191

Fig. 10. Influence of the VGs width L/ h for counter rotating devices.

Fig. 13. Influence of the spanwise separation / h for co-rotating devices.

The counter-rotating configuration being optimized, some

complementary tests where performed on the co-rotating system in order to assess the difference between both for an equivalent geometry. Table 4 gives the starting values of the parameters of Fig. 4(a). In this case, only the device spacing in
the spanwise direction was varied. Fig. 13 shows the effect of
this parameter on the spanwise skin friction variation. For that
purpose, skin friction measurements were performed at several
spanwise locations between z/ = 0 and z/ = 1. For / h = 4
the value of /0 is homogeneous in the spanwise direction.
Higher skin friction improvements are obtain locally when / h
increases, but at the expense of significant spanwise variations.
The spanwise mean value of /0 is nearly the same in all
cases. A value of / h = 6 was retained as the best compromise
in that case.
As a conclusion, Table 5 summarizes the optimal parameters
obtained for both the counter (CtR) and the co-rotating (CoR)
passive devices. Table 5 includes the spanwise minimum and
maximum values of the skin friction variation. It clearly comes
out that for equivalent geometries, the counter-rotating configuration is twice as efficient as the co-rotating one, which is
already quite efficient. These results are obtained with fairly
small actuators. For a 1 mm boundary layer thickness, they
would be less than 0.4 mm in height. The main problem would
then be their fragility.

Fig. 11. Influence of the spanwise spacing / h for counter rotating devices.

Fig. 12. Influence of the VGs aspect ratio l/ h for counter rotating devices.

6. PIV results
axis, /0 increases significantly from / h = 4 to / h = 8
and remains constant for / h > 8. A value of / h = 6 was
retained as the best compromise.
To conclude the procedure, the influence of the device length
was investigated. Fig. 12 shows the results obtained. The only
significant evolution of /0 (+20%) is observed between
l/ h = 1 and l/ h = 2 on the axis of the device. A value of
l/ h = 2 was thus retained for the final configuration.

Using the PIV set-up described in Section 3.3, measurements were performed at 3 different stations: X = 17.67 m,
18.09 m, 18.57 m corresponding to X/ h = 22, 38, 57
with respect to the point of minimum skin friction (XCfmin =
18.58 m). The optimal counter-rotating array was located at
Xd = 17.10 m, which corresponds to XVG / h = 57. At each
measurement station, 200 instantaneous velocity fields were

Table 5
Optimal parameters for both configurations of passive devices
VGs
CtR
CoR

Triangular vanes
Triangular vanes

h/
0.37
0.37

XVG / h
57
57

l/ h
2
2

L/ h
2.5

/ h
6
6

pd ( )
18
18

/0 (%)
min

max

110
55

200
105

G. Godard, M. Stanislas / Aerospace Science and Technology 10 (2006) 181191

189

Fig. 14. Mean velocity maps at X/ h = 22, for the smooth wall and the counter rotating actuators. The vectors show the in plane velocity components. The grey
levels the out of plane (streamwise) component.

Fig. 15. Mean velocity maps at X/ h = 38, for the smooth wall and the counter rotating actuators. Same representation as Fig. 14.

recorded, corresponding to 800 images of 1280 1024 pix2

(a total of 2 Gbytes). Based on the analysis parameters described in 3.3, each velocity map was 92 58, 74 58 and
75 57 vectors respectively for X/ h = 22, 38, 57. Starting
from these instantaneous maps, a mean velocity map, giving the
three components of the velocity vector in the plane of investigation, was computed. Such maps were obtained at each station
for both the smooth wall and for the passive devices.
Figs. 14 to 16 give the results obtained at the three measurement stations. Each figure compares the mean velocity map
for the smooth wall to the corresponding map with the actuators. The vectors give the velocity in the plane of measurement,
while the contours give the (main) out of plane component.
In Fig. 14, at X/ h = 22, one can see that the flow without actuation is 2D and uniform in the spanwise direction. The
strong gradient region near the wall, which is still quite thin at
this station, is clearly visible from the rapid contour variation
at the bottom part of the map. At this station, which is quite
near to the actuators, the vortices are clearly detectable. They
are centered at |z/| = 0.25 and y/ = 0.12. The perturbed region reaches y/ = 0.35. A wide region of strong streamwise
velocity is evidenced near the wall, between the two counterrotating vortices. Besides, a smaller region of low velocity is
observed under each vortex and between two sets of actuators.

The flow shows a clear periodicity of / = 0.8. At this station,

the boundary layer thickness without actuation is = 190 mm.
The next station (X/ h = 38), is presented in Fig. 15. The
smooth wall boundary layer has clearly increased in thickness,
while keeping its 2D nature. The streamwise vortices are still
identifiable they are now centered at |z/| = 0.12 and y/ =
0.12 but = 257 mm at this station.
They have decreased in intensity and increased in size. The
regions of high and low streamwise momentum are still identifiable but less intense. The region of low momentum has spread
significantly away from the wall.
At X/ h = 57 (Fig. 16), the smooth wall BL has again
increased in thickness ( = 455 mm). The slight transverse
asymmetry should be attributed to the PIV accuracy which was
a bit less in this configuration, due to some difficulties to adjust the set-up. For the counter-rotating passive devices, the
results show an homogenization, with a streamwise velocity in
the range of 0.5 to 0.7 Ue , all down to the wall. The vortices are
hardly detectable, although a downwash is still visible in the
central part of the field.
To confirm these PIV results, the mean velocity profiles was
measured with a single hot wire at XCfmin = 18.58 m, in the
plane of symmetry (z = 0) with and without device. The results
are shown in Fig. 17. The local external velocity does not show

190

G. Godard, M. Stanislas / Aerospace Science and Technology 10 (2006) 181191

Fig. 16. Mean velocity maps at X/ h = 57, for the smooth wall and the counter rotating actuators. Same representation as Fig. 14.

In Table 6, significant reduction of is observed, whereas

remains nearly unchanged. There is consequently an improvement of the H12 shape factor (22.8%). The increase of the
Schlichting shape factor H from 0.85 to 0.98 indicates that the
counter-rotating devices reduce significantly the adverse pressure gradient effect on the boundary layer.
7. Conclusion

Fig. 17. Comparisons between longitudinal velocity profiles measured at

XCfmin in the plane of symmetry (z = 0) of the counter-rotating system with
and without devices.

significant differences with or without device in contradiction

to the Lin [11] results. The S-shaped profile is in good agreement with the results of Angele et al. [18], using higher and
rectangular devices (h/ = 0.5), at about the same streamwise
distance (XVGs / h = 50). It is also in good agreement with
the PIV data of Fig. 16. In particular, the peak of velocity is at
the same wall distance. Table 6 gives the global parameters deduced from the velocity profiles of Fig. 17. H is the Schlichting
shape factor defined as:


H23
H = 0.5442.H23
H23 0.5049
and known as being a reliable separation criterion. The range
0.723 < H < 0.761 is characteristic of velocity profiles prone
to separate and the value H = 1 is the limit between favorable
and adverse pressure gradient.

In the present contribution, a detailed parametric study of

passive vortex generators was performed with the aim to have a
reference configuration for further studies with active devices.
Both co- and counter-rotating configurations were tested. Starting from characteristics given in the literature (mostly Lin [11]),
an optimization study was performed, based on the skin friction
improvement at the location of Cfmin along the bump. This test
campaign led to optimal configurations which are described in
Table 5 for both devices. The configurations obtained are near
to what can be found in the literature (see Table 2) and they
both show a significant improvement of the wall skin friction.
Nevertheless, the counter-rotating device appears more effective with a ratio of two in skin friction increase as compared to
the co-rotating actuators.
Based on this optimization study, the following conclusions
can be drawn on the effect of the different characteristic parameters for a counter-rotating system:
The first test performed highlighted the importance of
the actuator shape. Triangular actuators produce a significant improvement (+20%) compared to rectangular ones.
Moreover the triangular shape is better in term of drag
penalty.

Table 6
Global parameters deduced from the velocity profiles
Actuator

Ue (m/s)

/0 (%)

(m)

(m)

(m)

H12

Re

Re

Without CtR
Pas. Dev.
Variation

10.34
10.32

0
200

0.455
0.455

0.0668
0.0513
23.1%

0.0401
0.0399
0.5%

1.667
1.288
22.8%

0.848
0.980
15.6%

27610
27431
0.6%

46026
35318
23.3%

G. Godard, M. Stanislas / Aerospace Science and Technology 10 (2006) 181191

In agreement with previous studies, the co-rotating configuration is less effective than the counter-rotating one. The
average difference is about 100%.
The results show a sensitivity to the skew angle. The skin
friction increases when pd decrease. The optimal value
of pd is around 18 . It is consistent with the results of
Pauley et al. [7], who observe a linear increase of the vortices strength up to a skew angle of 18 .
The VGs aspect ratio (l/ h) is not very sensitive in the range
tested. A minimum value of 2 is recommended and the optimal value of 2.5 is the one predicted by Pearcey [5].
The transverse aspect ratio /L converges towards 2.5.
This value is slightly higher than the one used for the studies referenced in Table 1. It is lower than the value of 4
suggested by Pearcey [5] in order to avoid the vortices ejection away from the wall. In spite of this lower value, no
indication of ejection was found in the present tests. This
value is coherent with the results of Betterton et al. [17]
and Angele et al. [18] in adverse pressure gradient boundary layer.
The optimal streamwise position was found at Xd =
17.10 m were the boundary layer thickness is = 0.07 m.
The optimal distance between the VGs trailing edge and
the Cfmin line (57h) is equivalent to Betterton et al. [17].
Nevertheless, this parameter is not very sensitive.
The height h of the actuator appears to behave differently
from what is indicated in the literature. A monotonic increase is observed with h in the plane of symmetry for
h/d > 0.2. The high momentum transfer toward the wall
seems to be much more sensitive to the device height than
the low momentum transfer outward.
Having optimized the actuators, it was of interest to investigate in some details the flow physics associated to the best one.
For that purpose, a stereo-PIV experiment was settled. The PIV
results in the three planes normal to the flow, downstream of the
actuators, clearly explain the flow structure. The mechanism of
momentum transfer between the near wall region and the outer
flow is based on counter-rotating vortices which organize themselves in a very stable way. They grow rapidly in size as they
progress downstream but stay coherent and attached to the wall.
A high streamwise momentum area rapidly appears around the
plane of symmetry of each actuator, while a low momentum region is initially attached to the wall between two devices and
takes more time to spread outward. At the last station, although
the vortices become less visible, the streamwise momentum is
improved in the whole region of interest as compared to the basic flow.
Acknowledgements
The research reported here was undertaken as part of the
AEROMEMS II project (Advanced Aerodynamic Flow Control Using MEMS, Contract No G4RD-CT-2002-00748). The
AEROMEMS II project is a collaboration between BAE SYSTEMS, Dassault, Airbus Deutschland GmbH, EADS-Military,
Snecma, ONERA, DLR, LPMO, Manchester University, LML,

191

Warwick University, TUB, Cranfield University, NTUA, and

Auxitrol. The project is funded by the European Union and the
project partners.
References
[1] M. Gad-El-Hak, Introduction to flow control, in: M. Gad-El-Hak, A. Pollard, J.-P. Bonnet (Eds.), Flow Control-Fundamental and Practices, in:
LNPm, vol. 53, Springer, Berlin, 1998, pp. 1107.
[2] J. Jimnez, On the structure and control of near wall turbulence, Phys.
Fluids 6 (1994) 944953.
[3] K.S. Choi, J.R. Debisschop, B.R. Clayton, Turbulent boundary layer control by means of spanwise wall oscillation, AIAA J. 36 (7) (1998) 1157
1163.
[4] L.G. Pack, R.D. Joslin, Overview of active flow control at NASA Langley Research Center, in: SPIEs 5th Annual International Symposium on
Smart Structures and Materials, San Diego, 1998.
[5] H.H. Pearcey, Shock-induced separation and its prevention by design and
boundary-layer control, in: G.V. Lochmann (Ed.), Boundary Layer and
Flow Control, its Principle and Application, vol. 2, Pergamon Press, Oxford, 1961.
[6] D.M. Rao, T.T. Kariya, Boundary-layer submerged vortex generators for
separation control an exploratory study, Space Programs and Technologies (1988) 839846.
[7] W.R. Pauley, J.K. Eaton, Experimental study of the development of longitudinal vortex pairs embedded in a turbulent boundary layer, AIAA
J. 26 (7) (1988) 816823.
[8] J.C. Lin, F.G. Howard, Turbulent flow separation control through passive
techniques, in: AIAA 2nd Shear Flow Conference, March 1316 1989,
Tempe AZ, AIAA Paper 89-0976, 1989.
[9] J.C. Lin, F.G. Howard, D.M. Bushnell, Investigation of several passive and
active methods for turbulent flow separation control, in: AIAA 21st Fluid
Dynamics, Plasma Dynamics and Laser Conference, June 1820 1990,
Seattle, WA, AIAA Paper 90-1598.
[10] J.C. Lin, G.V. Selby, F.G. Howard, Exploratory study of vortex-generating
devices for turbulent flow separation control, in: 29th Aerospace Sciences
Meeting, January 710 1991, Reno, Nevada, AIAA Paper 91-0042.
[11] J.C. Lin, Control of turbulent boundary layer separation using microvortex generators, AIAA Paper 99-3404, 1999.
[12] J. Lumley, Control of turbulence, Lecture in Research of the American
Institute of Aeronautics and Astronautics, AIAA Paper 96-0001, 1996.
[13] S.J. Kline, S.K. Robinson, Turbulent boundary layer structure: Progress,
status and challenges, in: Structure of Turbulence and Drag Reduction,
IUTAM Symp. 1989, Springer, Berlin, 1990.
[14] S.K. Robinson, Coherent motion in the turbulent boundary layer, Ann.
Rev. Fluid Mech. 23 (1991) 601639.
[15] J. Jimnez, A. Pinelli, The autonomous cycle of near wall turbulence, Fluid
Mech. 225 (1999) 211240.
[16] P. Orlandi, J. Jimnez, On the generation of turbulent wall friction, Phys.
Fluids 6 (2) (1994) 634641.
[17] J.G. Betterton, K.C. Hackett, P.R. Ashill, M.J. Wilson, I.J. Woodcock, C.P.
Tilman, K.J. Langan, Laser Doppler anemometry investigation on subboundary layer vortex generators for flow control, in: 10th Intl. Symp. on
Appl. of Laser Tech. to Fluid Mech., Lisbon, July 1013 2000.
[18] K. Angele, Experimental studies of turbulent boundary layer separation
and control, Thesis, Dept. Mechanics, Royal Institute of Technology,
Stockholm, 2003.
[19] L. Jenkins, S.A. Gorton, S. Anders, Flow control device evaluation for an
internal flow with an adverse pressure gradient, AIAA Paper 2002-0266,
40th AIAA Aerospace Sciences Meeting and Exhibit, Reno 2002.
[20] A. Bernard, P. Dupont, J.M. Foucaut, M. Stanislas, Decelerating boundary
layer: a new scaling and mixing length model, AIAA J. 41 (2) (2003) 248
255.
[21] R. Houdeville, J.C. Julien, J. Cousteix, Skin friction measurements with
hot element gauges, La Recherche Arospatiale 1 (1984) 6779.
[22] S.M. Soloff, R.J. Adrian, Z.C. Liu, Distortion compensation for generalized stereoscopic particle image velocimetry, Meas. Sci. Technol. 8 (1997)
14411454.