Flow Turbulence Combust (2007) 78:383–407

DOI 10.1007/s10494-007-9068-4

Active Control of a Circular Cylinder Flow

at Transitional Reynolds Numbers

A. Naim & D. Greenblatt & A. Seifert & I. Wygnanski

Received: 16 May 2006 / Accepted: 15 January 2007 /

Published online: 23 March 2007
# Springer Science + Business Media B.V. 2007

Abstract Active and passive control of flow around a circular cylinder, at transitional
Reynolds numbers was investigated experimentally by measuring cylinder surface
pressures and wake velocity profiles. Two- and three-dimensional passive boundary layer
tripping was considered and periodic active control using piezo-fluidic actuators was
introduced from a two-dimensional slot that was nearly tangential to the cylinder surface.
The slot location was varied circumferentially by rotating the cylinder and this facilitated
either upstream- or downstream-directed actuation using sinusoidal or modulated wave-
forms. Separation was controlled by two distinct methods, namely: by forcing laminar-
turbulent transition when applied at relatively small angles (30–60°) from the forward
stagnation point; and by directly forcing the separated shear-layer at larger angles. In the
latter case, actuation produced the largest load changes when it was introduced at
approximately 90° from the forward stagnation point. When the forcing frequency was
close to the natural vortex-shedding frequency, the two frequencies “locked-in” creating
clear and persistent structures. These were examined and categorized. The “lock-in” effect
lowered the base pressure and increased the form-drag whereas delaying separation from
the cylinder did the opposite.

Keywords Flow control . Cylinder . Boundary layer . Transition . Separation . Shedding

Nomenclature
b slot width
Cd drag coefficient, d=q1 D
Cdp form drag coefficient, dp=q1 D

A. Naim : A. Seifert (*) : I. Wygnanski

School of Mechanical Engineering, Faculty of Engineering, Tel-Aviv University, Tel-Aviv 69978, Israel
e-mail: [email protected]

D. Greenblatt
Institute of Fluid Dynamics and Engineering Acoustics,
Berlin University of Technology, Berlin, Germany
e-mail: [email protected]
384 Flow Turbulence Combust (2007) 78:383–407

Cl lift coefficient, l=q1 D

Cp pressure coefficient, ðp  p1 Þq1
Cpb base pressure coefficient
Cm momentum coefficient, ð2b=DÞðUrms =U 1Þ2
D cylinder diameter
D′ distance between opposite streamlines
DC duty cycle
h distance between opposite sets of trailing vortices
k velocity ratio Us =U1
f frequency
fe excitation frequency
fm modulation frequency
fr piezo-element resonance frequency
F+ dimensionless forcing frequency, fe D=U1 , fr D=U1 or fm D=U1
p pressure
q1 free-stream dynamic pressure
Re Reynolds numbers U1 D=ν
Re* wake Reynolds number, UsD′/3
Req momentum thickness Reynolds number, U1 θ=ν
St Cylinder Strouhal number, fD=U1
St* wake Strouhal number, fD′/Us
T period
Umax maximum of slot velocity
Us velocity at separation
Urms rms of slot-exit velocity
U1 free-stream velocity
x streamwise coordinate measured from the cylinder forward stagnation point
y cross-stream coordinate measured from the cylinder forward stagnation point
z spanwise coordinate cylinder center-span
a angle on the cylinder measured from the forward stagnation point
as separation angle measured from forward stagnation point
3 kinematic viscosity
θ momentum thickness
sep separation location
(Subscript)

1 Introduction

1.1 The baseline flow

The cross-flow over a nominally two-dimensional circular cylinder is a prototypical bluff-

body problem, with widespread direct and indirect applications to fixed- and rotary- wing
aircraft, wind-exposed buildings and underwater structures. The flow patterns around a
cylinder vary considerably for 0<Re<1000, but for the approximate range 1,000<Re<
200,000, the basic flow characteristics are similar with little movement of the separation
line (70°<as <85°) and vortex shedding frequency (St∼0.2). When a critical Reynolds
number is exceeded, typically around 300,000 depending upon the free-stream turbulence
level, surface roughness, etc., transition is triggered in the developing boundary layer
Flow Turbulence Combust (2007) 78:383–407 385

resulting in turbulent boundary layer separation further downstream (120°<as <140°) with
a narrower wake and accompanying reduction in drag (from Cd≈1.1 to Cd≈0.6; Achenbach
[1]). The boundary layer can also be tripped by distributed roughness, trip-wires or
upstream turbulence-generating grids [30].

1.2 Active control of cylinder flows

Control of flow separation from a cylinder poses substantial challenges because the
separation line is not precisely (geometrically) fixed and multiple instabilities are present
simultaneously: Viscous (Tollmein-Schlichting) instability of the laminar boundary layers,
centrifugal (Görtler) instabilities due to curvature (see Saric [24]); inflectional (Kelvin-
Helmholtz) instability of the separated shear layer and the ubiquitous global instability that
drives vortex shedding [12]. The effect of excitation was initially demonstrated using sound
waves, emanating from the wind tunnel walls, as a means of controlling boundary layer
separation (e.g. Peterka and Richardson [21]), but was later demonstrated by Williams et al.
[31] to be related to fluidic and not acoustic excitation. Later studies focused on
hydrodynamic excitation, for example from a slot (e.g. Hsiao et al. [11]; Williams et al.
[31]; Schewe [25]; Pal and Sinha [18]; Heine et al. [10]; Amitay et al. [2]; Liu and Brodie
[14]; Béra et al. [4]), producing profound effects on transition and separation, with
consequences for lift, drag and vortex shedding. For example, Hsiao et al. [11] showed that
a suction peak could be generated on one side resulting in Cl =0.6 with the excitation
location at 100° when F þ  fe D=U1  1; similar observations were made by Amitay et al.
[2] for F+ =1.5 who also showed that the separation point could be moved by approximately
60°, resulting in a 25% reduction in form-drag.
The Kelvin–Helmholtz instability produces coherent vortices and forcing affects both the
vortex strength as well as the downstream distance over which the vortices roll up (cf.
Roshko [23]). While exciting the flow at a frequency corresponding to F+ =2, Liu and
Brodie [14] observed a shortening of the vortex-shedding-formation region that was
associated with separation control. For the above-mentioned studies, 5≤F+/St≤10, exci-
tation was most effective when introduced at 90°≤a ≤100°, with perturbations normal to the
surface. In some cases, excitation affects the upstream surface pressure distribution in spite
of the fact that pressures downstream of the actuator are not affected [33]. When
perturbations are introduced via a downstream blowing slot, centrifugal instability intro-
duced by the control jet may compete with the inflectional instability traditionally exploited
by two-dimensional excitation and limit the spanwise coherence of the eddies [20].
Perturbations can also be introduced by oscillating the cylinder, either in a cross-stream
(transverse) mode [5, 6, 8, 9, 13], a streamwise (in-line) mode [9] or a rotational mode [7,
27, 29]. These investigations demonstrated the phenomenon of “lock-in,” in which vortices
are shed at the excitation frequency and stationary phase is found between the excitation
and the flow response at all locations in the near wake. Harmonic forcing can increase
vortex strength by up to 29% with simultaneous base pressure and form-drag changes of
−22% and +16%, respectively [9]. Achieving lock-in over a range of frequencies using
fluidic excitation control at relatively large Reynolds numbers (Re>104), has not been
demonstrated yet, to the best of our knowledge.

1.3 Semi-empirical considerations

Roshko [22] constructed a semi-empirical model for analyzing the wakes of different bluff-
bodies. This was achieved by extending Kirchhoff’s free-streamline theory, where the
386 Flow Turbulence Combust (2007) 78:383–407

separating free-shear layers are idealized by streamlines across which the velocity is
discontinuous. In Kirchhoff’s formulation, the flow is divided into a wake and an outer
potential field where the velocity on the free-streamline is U1 . In Roshko’s extension, the
surface velocity at separation is expressed as Us ¼ kU1 and determined empirically. Since
the pressure is approximately constant aft of the separation point (Ub ≈Us), the base
pressure coefficient can be approximated as

Cpb ¼ 1  k 2 ð1Þ

using Bernoulli’s equation. The model facilitates the calculation of CD and (D′/D) for different
cylinder shapes as a function of k only, where D′ is the distance between free streamlines.
Using von-Kármán’s relations, Roshko [23] further established equations for CD(D/h), also as
a function k and assumed that the width of the vortex sheet is equal to the spacing between
the vortices (D′=h). The theory asserts that the shedding frequency is proportional to Us/D′.
Consequently, a “universal” or wake Strouhal number can be defined as

St* ¼ fD0 =Us ð2Þ

where St*≈0.16 for 10,000<Re*<44,000, irrespective of the bluff-body shape. It is related to

the conventional Strouhal number by
U1 D0 St D0
St* ¼ St ¼ ð3Þ
Us D k D

and is assumed to depend on a wake Reynolds number

Us D0 D0
Re* ¼ ¼ Re k ð4Þ
ν D

Application of the method is straight forward: The shedding frequency is measured and St is
calculated. The ratio kD/D′ is then determined from (3), the corresponding k can be found
iteratively from the extended streamline theory and hence Cd can be calculated. Application
of the above model to instances where active flow control is employed using fluidic actuators
has not been attempted, to the best of our knowledge.

1.4 Objectives and scope

This paper reports on the results of a parametric study, where active zero-mass-flux
perturbations were used to control separation on a circular cylinder. The parameter space
included Reynolds number (Re), reduced frequency (F+), momentum coefficient (Cm ), slot
location (a) and orientation (either forward facing or backward facing), as well as the
waveforms of the control perturbations. Considerable difficulty arises in attempting to
distinguish between transition control (or active boundary layer tripping) and direct
separation control, particularly at sub-critical Reynolds numbers where the separating
boundary layers are laminar. This work addresses the issue of drawing a distinction between
the two phenomena by considering the effect of passive and active perturbations at various
sub-critical Reynolds numbers. Specifically, two- and three-dimensional tripping was
considered in addition to active periodic excitation. Primary measurements included steady
and unsteady surface pressures and wake surveys. Integration of surface pressures was used
to estimate loads and the phenomenon of vortex lock-in was investigated.
Flow Turbulence Combust (2007) 78:383–407 387

Section 2 contains a description of the experimental setup, including measurement

techniques, actuator-slot calibration, control input definitions and experimental uncertain-
ties. Section 3 presents the main results of the parametric study, including the effects of Re,
passive tripping, pure-tone and modulated waveform excitation and vortex-shedding lock-
in. Pertinent conclusions are presented in Section 4.

2 Experiment

Experiments were conducted on a circular cylinder of diameter 76.2 mm and span 609 mm,
with 46 mid-span surface static-pressure ports. The cylinder incorporated internally
mounted piezo-fluidic actuators, across it’s entire span, which communicated with the
external flow through a 0.5 mm wide two-dimensional slot. The slot introduced almost
tangential zero-mass-flux periodic perturbations into the boundary-layer (Fig. 1). Rotating
the cylinder counterclockwise by more than 70° from its position shown in Fig. 1, directed
the excitation in the direction opposite to the boundary layer velocity. This allowed
assessment of the sensitivity to excitation direction relative to the oncoming boundary layer
flow. A sensitive unsteady pressure transducer (136 mV/Psi) was installed inside the
actuator’s cavity to monitor its operation and provide real-time information on the cavity
pressure fluctuations. This allowed real-time monitoring of the in situ pressure fluctuations
present in the flow as well as an estimate of the oscillatory momentum introduced at all
conditions investigated, without the presence of a hot-wire at the slot exit. It also served to
measure the frequency (and to a lesser extent the magnitude) of the vortices shed from the
cylinder.
A comprehensive bench-top calibration was performed in order to determine the
frequency and amplitude response of the actuators and to verify the quality of the
(spanwise, z-direction) two-dimensionality of the excitation. A typical frequency response
of the actuator is shown in Fig. 2a. Data were obtained using a single one-component hot-
wire positioned in the exit region of the slot. The peak slot-exit velocity was considered in
order to avoid the need to de-rectify the velocity signal. Nevertheless, de-rectification and
analysis of the signals revealed small higher-harmonic content and hence confirmed the

Fig. 1 Schematics of the cylin-

der (diameter D=76.2 mm, slot
width h=0.5 mm at exit) with
taped slot and roughness strip
388 Flow Turbulence Combust (2007) 78:383–407

Umax Umax
[m/s] [m/s] 20 VAC 26 VAC
30.0 primary freq. @ 1090Hz 35.0
Z=100mm
Z=100mm
Z=200mm
30.0 Z=200mm
25.0 Z=300mm
Z=400mm Z=300mm
Z=500mm 25.0 Z=400mm
20.0 Z=500mm
secondary freq. @ 680Hz
20.0
15.0
15.0

10.0
10.0

5.0
5.0

0.0 0.0
500 600 700 800 900 1000 1100 1200 1300 1400 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
a Frequency [Hz] b Amplitude [Vac]
(arbitrary)

UMAX
[m/s] Pure sine wave input voltage
30

y = 0.0342x
25
R2 = 0.9974 Measured slot velocity

20 d
Amplitude Modulation (AM) input voltage
15

Tm=1/fm
10
Amplitude Modulation (AM) measured slot velocity
Cal Data
5
Linear (Cal Data) e

0 Burst Mode (BMn) input voltage

0 200 400 600 800
c P' cav [Pa]
Tm=1/Fm n=No. of cycles
Burst Mode (BMn) measured slot velocity

f
Time

Fig. 2 Slot calibration data showing actuator output velocities: a optimum Piezo frequency (resonance); b
two-dimensionality along the slot span; c Umax vs. cavity pressure fluctuations (for Cm assessment); d Pure
sine excitation; e amplitude modulation; and f burst Mode

validity of this approach. For all data acquired, two spectral peaks were identified, at
680 Hz and at 1,090 Hz. It was noted that the maximum scatter of the amplitude at these
resonance frequencies was less than ±5%, indicating adequate two-dimensionality of the
excitation. The amplitude-response of the actuators, measured at five different locations
along the span of the slot, when excited at 1,090 Hz, is shown in Fig. 2b. The scatter in the
peak velocity was less than ±5% stemming mostly from the uncertainty associated with hot-
wire location in the exit region of the slots. The calibration curve of the internal pressure
Flow Turbulence Combust (2007) 78:383–407 389

transducer discussed above is shown in Fig. 2c, indicating that the peak exit velocity
fluctuations are linearly proportional to the pressure oscillations magnitude (rms) within the
actuator’s cavity.
The actuators were operated in three different modes, namely (1) pure sine-wave; (2)
sinusoidal amplitude modulation; and (3) pulsed modulation (see Wiltse and Glezer [32]
and Amitay and Glezer [3]). Figure 2d–f show examples of the three modes, respectively,
where the function generator (input voltage) signal is shown together with the de-rectified
hot-wire signal for all modes. Mode (1) is self-explanatory. For sinusoidal amplitude
modulation (AM; mode 2), the amplitude of the carrier signal which has a resonance
frequency fr =1/Tr is modulated by the modulating signal which has a frequency fm =1/Tm,
and is an integer division thereof. For burst mode (BM; mode 3), the carrier frequency is
modulated by a step, that varies between 0 and 1, and whose duration for n cycles is nTr.
The square wave modulating frequency is also denoted fm. Therefore, in addition to the above
definitions we can also define a duty cycle of the modulating step function, DC≡nTr/Tm
(alternatively or nfm/fr), which expresses the fraction of the modulating period during which
the actuators are active. Consequently, in BM, the momentum coefficient, Cm , is also a
function of DC (and not only of the peak excitation velocity, see below). Note that for the
latter two modes, the reduced excitation frequency F+ is based on fm. A previous investigation
indicated that the momentum produced from the slot is not significantly altered by the
external stream [17].
Following bench-top calibration, the cylinder was installed in the 1500 mm by 609 mm
test section of the low-speed, low-turbulence Meadow-Knapp wind tunnel, described by
Seifert et al. [26]. A rake of total pressure probes was located 1150 mm (x/D=15.1; the
furthest possible downstream location) downstream of the cylinder and was also equipped
with an unsteady pressure sensor. The cylinder surface pressures and wake velocities were
acquired by a computer-controlled pressure scanner with a full scale of 2,600 Pa and a
resolution of 0.1% of full-scale. The maximum test section speed (approximately 60 m/s),
with the cylinder installed, limited the investigated Reynolds number range to Re≤280,000.
The temperature in the test section was maintained constant to within ±0.5°C and Re was
determined to within ±2%. The pressure port and slot locations were accurate to within ±1°
and the unsteady pressures were measured with ±3% uncertainty. The uncertainties
associated with lift, drag and pressure coefficients were ±0.02 and total drag uncertainty
was ±5%.

3 Results

3.1 Scope of control methods considered

Several parameters that reflect upon the condition of the flow around the cylinder were
examined, namely the integral aerodynamic coefficients Cl, Cd, Cdp and the pressure
distributions (Cp) over the cylinder. Spectral analysis of the pressure fluctuations inside the
actuator’s cavity and the far-wake total-pressure fluctuations were also considered. The
flow over the cylinder was subjected to either passive tripping (Sections 3.2 and 3.3) or
two-dimensional (2D) active control (Sections 3.4–3.6) at a variety of Reynolds numbers
(Re). Two-dimensional passive tripping was achieved by means of the excitation slot,
facing either towards or away from the free-stream, where the slot was either taped over or
left open but inactive and 3D tripping was achieved by means of distributed roughness.
Active control included different 2D excitation modes (see Sections 3.4 and 3.5) applied
390 Flow Turbulence Combust (2007) 78:383–407

over a wide range of reduced frequencies (F+) from being significantly below the natural
vortex-shedding frequency to being approximately 50 times higher than the shedding
frequency. The tests also included various oscillatory momentum input levels quantified by
Cm . In all the experiments, the effects of the slot-location (a-scans) and slot orientation
(facing upstream or downstream as indicated by the sign of a, where negative a indicated a
slot facing upstream or counter-flow), were also considered. Finally, the phenomenon of
vortex shedding “lock-in” to the excitation frequency was examined (Section 3.6).

3.2 Effects of Re and 2D passive boundary layer tripping

A comprehensive set of data was acquired where the cylinder boundary layer was subjected
to two-dimensional passive tripping resulting from the slot; a sample of these data is
presented in Fig. 3. These experiments were conducted on two configurations. In the first
configuration, the slot was left open, thereby causing a relatively large (but nevertheless
passive) perturbation to the laminar boundary layer that might have been enhanced by the
interaction between the cavity and the external flow (these data are not shown in Fig. 3 but
will be shown later). In the second configuration, the severity of the perturbation was
reduced by taping-over the slot with a thin (0.1 mm), 13 mm wide tape. Previous
experimental investigations, as well as the present one, indicate that the maximum attainable
Reynolds number (Re  3  105 ) in the present setup was insufficient to generate natural
transition to turbulence on the smooth surface of the cylinder. The form-drag (Cdp),
however, decreased from about 1.25 at Re=100,000 to approximately 1.0 at Re=280,000
(see Fig. 3b for slot locations close to a =0°, where the taped slot is expected to have no
measurable effect on transition), indicating the flow is on the verge of becoming
supercritical. For all Reynolds numbers exceeding 105, the upstream-facing (negative a),
taped-slot affects the drag as it causes a disturbance sufficient to delay the separation
location on the cylinder. The form-drag plotted in Fig. 3b decreases substantially as the
taped upstream-facing slot is positioned in the region of flow acceleration (a ≈−45°,

0.2 1.4

Cl Cdp
0.0 1.2

-0.2 1.0

Re40k
Re100k
-0.4 Re150k 0.8
Re200k
Re220k
Re280k
-0.6 0.6
0 -30 -60 -90 -120 0 -30 -60 -90 -120
a Slot location, α [deg] b Slot location, α [deg]
Fig. 3 Cylinder data measured with taped slot at various Re for negative slot angles (i.e. slot facing
upstream): a Cl vs. a; b Form drag (Cdp) vs. a (same legend for a and b)
Flow Turbulence Combust (2007) 78:383–407 391

measured from a =0° in Fig. 1). The effect of this two-dimensional discontinuity is still
measurable when the slot is moved farther downstream, up to a =−85°. The effect of
surface imperfections is less pronounced at a <−90°, since the slot is located within a
separated flow region.
The range of slot-location angles, a, where the flow is sensitive to the two-dimensional
surface discontinuity decreases with decreasing Re (Fig. 3a and b). Since the taped-over slot
is located on one side of the cylinder, the asymmetry in the separation location contributes
to lift generation (Fig. 3a, note that only negative, upstream-facing, slot locations are
presented, due to higher sensitivity of the flow to the upstream-facing slot). Since most of
the contribution to lift is associated with the forward part of the cylinder, where the flow is
relatively steady and is less affected by vortex shedding, the lift provides a clearer
indication of the state of the flow and is therefore a convenient diagnostic tool for the
expected effects of active flow control. Note that the abrupt generation of lift corresponds to
the slot being closer to the forward stagnation point of the cylinder with increasing Re, due
to transition being triggered further upstream. Furthermore, the lift generated at Re >
1:5  105 (Fig. 3a), has two maxima; for example at Re ¼ 2:8  105 , one maximum is
located at a ≈−45° and the other at a ≈−85°. It is hypothesized that they are associated with
two different transition phenomena: closer to the forward stagnation point, transition to
turbulence in the attached boundary layer renders the resulting turbulent boundary layer less
susceptible to separation; near and slightly downstream of the mean separation location (i.e.
at a ≈−85°), the periodic vortices generated by the surface discontinuity induce more rapid
transition in the separating shear-layer and thereby delay separation. At some intermediate
angles, around a =−60°, the tape-generated step is less effective in delaying separation as
presumably it does not result in bringing about transition to turbulence. Most of the active
separation control data, presented below, are for Re < 1:5  105, where only one Cl
maximum (Fig. 3a) and one Cdp minimum (Fig. 3b) are generated. Note than in the present
case the slot is taped over, but the same findings hold for an open slot, as will be
subsequently shown. The two lift peaks form a single zone of effective influence around
−50°>a >−80°, while the edges of this zone relate to the two phenomena discussed above.

3.3 The effect of 3D distributed roughness

In order to effectively induce transition, three-dimensional roughness (grit #36, creating

RMS surface roughness of 1.1 mm, and mounted on a double-sided tape of thickness
0.1 mm, covering an a range of about 17.5° of the cylinder’s circumference, see Fig. 1) was
placed on the cylinder, upstream of the taped slot. Figure 4a and b present the lift and form-
drag coefficients for the taped slot with and without the 3D roughness strip, respectively.
Note that currently, only positive slot-location angles are presented. The 3D perturbations
are more effective in triggering transition that occurs now at the smaller a ≈30° (the
roughness strip extends from the forward stagnation region to a ≈15°). Note that surface
roughness triggers transition at a ≈40° less than the taped slot alone (see Fig. 4a and b) and
its effect decreases readily with increasing a. Three-dimensional roughness does not
produce the required two-dimensional disturbances to generate spanwise coherent vortices
that control separation at a >80°.
Two pressure distributions measured on the surface of the cylinder, that generate the
maximum lift in each case of those presented in Fig. 4a and b, are compared in Fig. 4c at
Re=200,000: one is for the slot covered with tape at a =60° and the other with the
roughness at a =35°. In the case without roughness, the taped-slot affects the range 52.5°<
a <67.5°. Whereas with the roughness strip added at a =35°, affects the range 2.5<a <20°.
392 Flow Turbulence Combust (2007) 78:383–407

1.0 1.4
Cl Taped
Cdp
0.8 Taped Grit #36
1.2
0.6

0.4 1.0

0.2
0.8
0.0

-0.2 0.6
0 50 100 150 0 50 100 150
Slot location, α [deg] Slot location, α [deg]
a b

-3.0

CP

-2.0

-1.0

Taped
0.0
Taped Grit #36

1.0
0 0.25 0.5 0.75 1
c x/D

Fig. 4 Cylinder slot taped over with and without roughness strip centered at 15° upstream of the slot,
covering about 17.5° of the cylinder circumference: a Cl vs. a; b Cdp vs. a (Re=200,000). Legend applies to
both figures. c Pressure distributions corresponding to the maximum lift conditions shown in Fig. 4c
(cylinder slot taped over with and without roughness). Re=200,000. Note that for the “Taped” data a =60°
while for the “taped Grit no. 36” data a=35°

Roughness almost reinstates the ideal (potential flow) pressure distribution over the front-
upper-part of the cylinder (Cp =−2.8 at x/D=0.38) and delays separation more effectively
than the taped slot, resulting in higher lift. The base pressures in both cases are almost
identical (Cpb =−0.9, see Fig. 4c), resulting in a similar form-drag (Fig. 4b).
To summarize Sections 3.2 and 3.3, passive 2D perturbations have a significantly
different effects on the separation control over a cylinder depending upon location, a, at
150,000≤Re≤280,000. At lower a, attached boundary layer transition is promoted; at larger
a, transition is promoted in the separating shear layer. Three-dimensional roughness, on the
other hand, produces its greatest effect in the attached boundary layer, at lower a, but is
ineffective at promoting transition in the separating shear-layer.
Flow Turbulence Combust (2007) 78:383–407 393

3.4 Pure tone excitation at the actuator’s resonance frequency

Experiments involving pure-tone excitation were conducted with the actuators operating at
their resonance frequency of 1,090 Hz, i.e. without modulation. The reduced excitation
frequency employed here was based on the cylinder diameter, i.e. F þ  fr D=U1 and this
results in values that are of a similar order of magnitude to those that are based on the
length of the separated regions over airfoils and flaps [16, 26]. It is not clear what the
appropriate length-scale should be when transition promotion has also to be considered.
Another length scale to be considered is the width of the near wake (D′ or h) as was
suggested by Roshko [23], because it relates natural vortex-shedding frequency to the
excitation frequency. The two current definitions nearly coincide.
Experience shows [16, 26] that F+ ≈O(1) is effective for separation delay in the absence
of curvature when the separation point is fixed or takes place near the introduction of the
excitation. On a circular cylinder, at sub-critical Re this is not the case because the location
of the separation point moves cyclically with time and it is also sensitive to Re. Although
the instantaneous natural separation location is three-dimensional, the 2D excitation is
assumed to reduce its spanwise variations. Data acquired at Re=105 using F+ =4.2 and
Cm ¼ 1:7% are shown in Fig. 5a and b. We hypothesized that transition promotion via
periodic excitation will be most effective when the perturbations emanate from the surface
in a direction that is opposite to the external flow direction, i.e., a <0. One may also assume
that during the blowing part of the cycle, temporary flow reversal may lead to an absolute
rather than convective instability. The lift generated by the excitation is compared to the lift
generated by the open but passive slot (henceforth referred to as the baseline case), in

1.2 -3.0
Cl CP
0.8

0.4 -2.0

0.0

-1.0
-0.4

-0.8

a 0.0
-1.2 Baseline (α =90˚)
1.4
Baseline (α =-90˚)
Cdp Slot @ 90˚ facing downstream
Slot @ -90˚ facing upstream
1.2 1.0
0 0.25 0.5 0.75 1
1.0
c
x/D

0.8

baseline (open slot)

0.6 1090Hz, Cµ ≈1.7%
1090Hz, Cµ ≈1.7% NEG
NEG baseline (open slot)

b 0.4-120 -90 -60 -30 0 30 60 90 120

Slot location, α [deg]

Fig. 5 Pure sine-wave excitation for the cylinder using F+ =4.2 (1090 Hz) and Cm  1:7%: a Cl vs. a b Cdp
vs. a (Re=100,000). Legend applies for both figures. c Cp vs. x/D for pure sine excitation at F+ =4.2
(1,090 Hz) and Cm  1:7%, a =±90°; Re=100,000
394 Flow Turbulence Combust (2007) 78:383–407

Fig. 5a, for −120°<a <120°. In the absence of excitation Cl ≈0, except when the slot trips
the boundary layer thereby generating a one-sided transition to turbulence and this occurs in
the approximate range of slot locations, 60°<|a|<85° on both sides of the cylinder, but with
significantly stronger magnitude at negative slot locations. It is also evident that the passive
drag reduction is larger at a <0 (Fig. 5b).
When the periodic excitation is activated and it opposes the flow direction (i.e., for a <
0), the velocity perturbations near the slot are larger and so is the generated lift. In fact,
actuation in the direction opposing the stream generates significant lift even at a =−10°. The
largest Cl generated at small negative a is 0.95 and it decreases to 0.6 at a =−60° reaching
Cl =1 at a =−90°. When the actuation is in the direction of the flow, i.e., slot facing
downstream indicated by positive a, Cl is relatively small at small a, but increases at a >
45°. The dependence of Cl on α, for 60°<a <90°, is almost identical to its dependence
when the actuation opposes the flow direction. Similar trends are observed with respect to
the form-drag, Cdp (see Fig. 5b). The optimum slot location for the excitation in the
downstream direction using Cm ¼ 1:7% and F+ =4.2 is around a =102°, whereas excitation
that is introduced further downstream requires a higher Cm to be effective (not shown).
Pressure distributions corresponding to a =±90° (at Re=100,000) are shown in Fig. 5c,
as an example. The baseline flow separates from the cylinder around x/D ≈0.4
(corresponding to a ≈80°, taking the constant-pressure region as an indicator for
separation), the flow is symmetrical in the mean about the x-axis (the analog to the chord
of an airfoil) and the base pressure is Cpb ≈−1.37. Due to the delay of boundary-layer
separation from the upper surface of the cylinder, circulation is created and the flow on the
excited side of the cylinder accelerates while it decelerates over the opposite side. The base
pressure becomes more positive because of the delayed separation on the controlled side
(the flow seems to separate only at a ≈120° in the controlled case, corresponding to x/D=
0.75). Due to the deceleration on the opposite side and the reduced base pressure, the drag
is reduced. Excitation increases the base pressure from Cpb ≈−1.37 to Cpb ≈−0.9 regardless
of the slot orientation relative to the oncoming stream. The separation line on the excited
side is almost the same, |a|=120° (x/D=0.75) for both directions of excitation. There seems
to be a slight drop in Cp for the a =−90° case near the cylinder base probably due to regular
shedding of vortices at the excitation frequency. Upward and downward distortions of the
wake were observed for control at a =−90° and 90°, respectively (not shown), along with a
significant reduction of its momentum deficit. Active perturbations at both injection angles
(i.e., a =±90°) affects the overall circulation generated by the cylinder and therefore, also
the pressure distributions upstream of the slot.
The effects of excitation at different positive slot locations are shown in Fig. 6. The
comparison is made at F+ =4.2 and Cm ¼ 1:1% and at angles that provide almost identical
mean pressure distribution over the cylinder, i.e. at a =55° and a =95° that are indicated in
Fig. 6a and b are compared in Fig. 6c. The pressure distributions over the cylinder (Fig. 6c)
suggest that when the slot is located at a =55° there is a discontinuity in the pressure
distribution that is not as obvious at a =95°. One might presume, given the similarity in the
pressure distributions, that separation occurs at similar locations, and that the mean wake
profiles and characteristic frequencies within the wakes would be comparable. This is not
the case (Fig. 6d and e) where a considerable difference in the total drag is measured at a =
55° and a =95°, corresponding to Cd ≈0.85 and 0.5, respectively (the baseline drag
coefficient for both cases was Cd ≈1.1). The baseline total pressure spectra measured in the
wake at the y-position indicated by the dashed line shown on Fig. 6d, are compared in
Fig. 6e, for the two slot locations These spectra are found to be almost identical having a
peak corresponding to the well-known Strouhal number associated with the von-Kármán
Flow Turbulence Combust (2007) 78:383–407 395

0.9

0.6
CL=0.554
CL=0.513
Cl

0.3

0.0
a 1.5
Baseline
1090Hz

1.2

Cdp
0.9
CDp=0.875 CDp=0.857

0.6
b 0 50 100 150

Slot Location, α [deg]

Fig. 6 a Cl versus a; b Cdp versus a. Legend and abscissa apply for both figures. Cylinder surface pressure
and wake data at Re=100,000 for pure sine wave perturbations at F+ ≈4.2 (1,090 Hz), Cm  1:1%: c Cp
variation on the cylinder surface; d Wake surveys at x/D=15.1 (the dashed line indicates y-position of wake
unsteady pressure sensor). Cylinder data, Re=100,000, Pure Sine at 1,090 Hz (F+ ≈4.2), Cm  1:1%: e Wake
pressure spectra; f Actuator’s cavity pressure spectra. Legend applied to e and f, note the difference in Cp
scale

vortex shedding fD=U1  0:2ð f ¼ 50 HzÞ. The similar peaks in both baselines spectra
also implies that the drag in both baseline cases should be comparable as is indeed the case.
Periodic excitation emanating from the slot located at a =55° increased the shedding
frequency to 64 Hz (i.e. by a ratio of 1.28) that reflects on the reduction in drag from Cd ≈
1.1–0.8. The percentage reduction in Cd corresponds approximately to a similar
proportional reduction in the momentum thickness. Therefore a similar percentage rise in
shedding frequency is seen because it scales with the momentum thickness. One would
anticipate that excitation at a =95° would generate a peak in the spectrum at f≈110 Hz that
would correspond to the measured Cd ≈0.5 but this was not the case. Instead, one observes a
396 Flow Turbulence Combust (2007) 78:383–407

-2.5 300
CP slot @ y Baseline α =55˚
-2.0 α=95˚ [mm] Baseline α =95˚
α =55˚
slot @ 500 α =95˚
-1.5
α =55˚

-1.0
700
-0.5

0.0
Baseline α =55˚ 900
Baseline α =95˚
0.5 α =55˚
α =95˚
1.0 1100
0 0.5 1 0.7 0.8 0.9 1.0
c x/D d u/U∞

0.1 10
50Hz f R =1090Hz
CP CP
64Hz 1

f R =1090Hz 50Hz
1090±365Hz 2180Hz
0.1
64Hz
0.01
60Hz
0.01

Baseline α =55˚
Baseline α =95˚ 0.001
α =55˚
α =95˚
0.001 0.0001
e 10 100 1000 10000 f 10 100 1000 10000
f [Hz] f [Hz]
Fig. 6 Continued

weak peak in the spectrum around f=136 Hz that represents a third sub-harmonic of the
forcing frequency of 1,090 Hz, while the dominant shedding frequency is at 60 Hz, but an
order of magnitude weaker than the baseline magnitude.
The vortex-shedding frequencies measured in the upper region of the wake (Fig. 6e) are
also detected by measuring the pressure oscillations inside the cavity that houses the
actuators (Fig. 6f), along with the excitation frequency of 1090 Hz. The magnitudes of the
Cp fluctuations measured in the cavity are high and are more clearly distinguishable than
the wake oscillations. In addition, one may find in these spectra the second harmonic of the
excitation and some additional peaks at frequencies of 1,090±365 Hz. These sidebands can
Flow Turbulence Combust (2007) 78:383–407 397

be attributed to the highly non-linear interaction between the cavity excitation and the
external flow. The natural vortex shedding frequency (VSF) is clearly identified also in the
cavity-measured Cp fluctuations. In both controlled cases, the VSF was increased and its
magnitude decreased, especially for the a =95° slot location. The increased VSF is
attributed to the narrower wake due to separation delay from the upper side of the cylinder.
While monitoring the excitation signal represents an experimental necessity for the cavity
installed actuators, sensing the frequency and magnitude of vortex shedding represents a
welcome fringe benefit because the sensitivity of the internal pressure transducer suggests
that it can be used to measure the flow characteristics in the vicinity of the body. This
enables its use as a flow-state sensor for feedback-control applications, where data
downstream of the body are not readily available.
One may conclude that actuation at a frequency much higher than the VSF in the
direction of streaming is effective in promoting transition at moderate a and delaying
separation at higher a.

3.4.1 Effects of the excitation magnitude

The effect of Cm on Cl at several representative slot locations and the two different slot
orientations is shown in Fig. 7a and b. At a slot angle of α=60°, an increase in Cm from 1.0
to 2.5%, resulted in a decrease in Cl, whereas for Cm > 2:5%, the lift increased conti-
nuously with Cm (Fig. 7a, and Fig. 8a). The available data does not really provide an
explanation to this complex Cl  Cm relationship. At a =80°, an increase in Cm resulted in a
concomitant gradual weak increase in Cl, whereas at a =100° the dCl dCμ was  large
 for
Cm < 6% but became more moderate at Cm > 6%. At a =110° the slope dCl dCμ was
generally equivalent to that measured at much smaller

slot-angles. In fact, forcing
 at  a =110°

required a much higher amplitude Cm > 8% to increase the slope of dCl dCμ , to
overcome the baseline separation that occurs significantly upstream in the absence of
excitation. At both slot locations (i.e., a =100° and 110°), the actuation emanates down-
stream of the baseline separation location, but due to excitation the separation region was
significantly delayed to approximately 122° and 132°, respectively (see, Fig. 8c and
Table 1). Periodic excitation reduces the pressure upstream of the actuator (see Amitay et al.
[2]; Béra et al. [4]; Taubert et al., 2002 [28]) enabling the flow to remain attached, but it
requires a stronger actuation magnitude if the location of the slot is farther downstream of
the baseline separation location. The maximum lift achieved for the latter two slot locations
was Cl =1.2 corresponding to Cm ¼ 9% (Fig. 7a, corresponding pressure coefficients are
shown in Fig. 8c). Changing the direction of the actuation to counter-flow generates lift on
the actuated side, provided Cm is low and the location of the slot is upstream of the baseline
separated region (Figs. 7b, 8b). The maximum negative Cl achieved by this type of actuation
was −0.5 at Cm ¼ 2:5%. At higher Cm or at larger a, the negative Cl not only decreases in
magnitude but also changes sign (Fig. 7b). The pressure coefficient over the unexcited
(upper) side of the cylinder at a =−80° (Fig. 7c) remains close to that of the baseline, but the
excited side experiences a decrease in pressure as the oscillatory Cm increases. This is because
momentum is injected in a direction that opposes the incoming stream, slows the boundary-
layer flow and eventually causes boundary-layer separation. This effect was seen by Amitay
et al. ([2], Fig. 2e-f), at a ≥100° when introducing excitation normal to the cylinder surface at
Cm  0:06% and F+ =2.6. (Their Cμ should be increased by a factor of four to match the
current definition, and Re was different as well). Hence, the data presented in Fig. 7a and b,
together with that of Amitay et al. [2], show the profound effect of excitation orientation and
location, particularly on bluff bodies such as circular cylinders at sub-critical Re numbers.
398 Flow Turbulence Combust (2007) 78:383–407

1.2 1.2
Cl baseline baseline
Cl
α =60˚ α =-60˚
0.9 0.9
α =80˚ α =-80˚
α =100˚ α =-100˚
0.6 α =110˚ 0.6 α =-110˚

0.3 0.3

0.0 0.0

-0.3 -0.3

-0.6 -0.6
0 2 4 6 8 10 0 2 4 6 8 10
a Cµ [%] b Cµ [%]

-2.5
CP
-2.0

-1.5

-1.0

-0.5

0.0 Baseline α=-80˚

Cµ≈ 1.2%
Cµ≈ 2.5%
0.5 Cµ ≈ 4.1%
Cµ≈ 5.8%
1.0
c 0 0.5 1
x/D
Fig. 7 Cl vs. Cm for pure sine excitation and F+ =10.6 (1,090 Hz): a Slot facing downstream; b Slot facing
upstream; c Cp vs. x/D at a =−80° (Re=40,000)

3.4.2 On the relationship between excitation and vortex shedding

As previously mentioned, when excitation is applied to bluff-bodies experiencing natural

orderly vortex shedding, the excitation can affect and manipulate the shed vortices. These
effects may include the change in the shedding frequency or strength of the vortices [8], and
may result in delaying the rollup of the mixing-layer further downstream [28]. In the current
work, the vortex-shedding frequency was determined from the spectra of the unsteady
pressure measured by the sensor installed inside the actuator’s cavity as well as the
unsteady pressure sensor positioned near the upper boundary of the wake, located at x/D≈
15. Several typical power spectral distributions for various excitation levels and actuation
locations are shown in Fig. 9a and corresponding pressure coefficients in Fig. 10. The
Flow Turbulence Combust (2007) 78:383–407 399

-2.5 -2.5
CP CP
-2.0 -2.0

-1.5 -1.5

-1.0 -1.0

-0.5 -0.5
Baseline α =60˚ Baseline α=-60˚
0.0 Cµ≈1.2% 0.0 Cµ ≈1.2%
Cµ≈2.5% Cµ≈2.5%
0.5 Cµ≈4.1% 0.5 Cµ≈4.1%

1.0 1.0
0 0.5 1 0 0.5 1
a b
x/D x/D
-3.5
CP
-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0 Baseline α=100˚

Baseline α=110˚
0.5 α=100˚
α=110˚
1.0
0 0.5 x/D 1
c x/D
Fig. 8 Pressure distributions for the data shown in Fig. 7a and b, CP vs. x/D for pure sine excitation on the
cylinder at F+ =10.6 (1,090 Hz): a a =60°; b a=−60°; c max % Cm for a =100, 110° (Re=40,000)

spectral data indicate that the frequency of the peak in the pressure oscillations increases
when separation is delayed due to the increase in Cm . The resulting Strouhal number based
on the cylinder’s diameter, D, and the free stream-velocity, U1 , is therefore not constant as
mentioned previously. According to Roshko’s [23] model, the increase in St implies that the
distance between the free-streamlines or vortices (D′,h) should reduce as the control
authority increases. In the present experiment the distance h was not measured and thus two
surrogates were used, namely θ and 2ysep, where ysep is the vertical distance corresponding
to the separation points from the cylinder. In the former case, θ was estimated based on the
400 Flow Turbulence Combust (2007) 78:383–407

Table 1 Cp values and separa-

tion location for the data shown a Cμ (%) Cpb Cp, min asep
in Fig. 8c
baseline 0 −1.14 −1.26 82°
100° 9.2 −0.68 −3.33 122°
110° 9.2 −0.82 −3.29 132°

assumption that Cdp ≈Cd [1]. Although h was not used, Strouhal numbers based on θ and
2ysep indicate that this scaling is appropriate for the shedding frequency, even for controlled
wakes where the excitation frequency is far from the shedding frequency (Fig. 9b), because
in the latter case the control directly altered the separation points. It should be noted that in
Roshko’s [23] Strouhal number (St*≈0.16) the distance between the vortices is somewhat
less than 2ysep but approximately twice as large as θ.
Drag predictions using Roshko’s ([22, 23]; see Section 3.1) model and experimental data
are compared on the basis of
ΔCd ¼ Cd  Cd0 ð5Þ

in Fig. 9c, where the subscript 0 refers to the cases without active control. A similar
expression is used to calculate ΔCdp from experimental data. The theory shows the correct
trend, but the measured drag and form-drag reductions are significantly greater than the
model predictions (Fig. 9c). There are a number of incompatibilities between the theory and
the experiment that are responsible for these discrepancies. From the theory, it is assumed
that separation will only occur on the forward part of the cylinder. This condition is violated
for all control amplitudes used (Fig. 10). In addition, a symmetric wake is assumed and this
is clearly not the case presently (Fig. 10). From the experimental viewpoint, partial
attachment of the separated shear-layer is also expected to have a significant effect on drag
reduction. In addition, momentum issuing from the control slot is not accounted for. This is
shown in Fig. 9c by also plotting Cd þ Cm  Cd0 and Cdp þ Cm  Cdp0 versus frequency
(filled symbols). Although this difference is significant, it is evident that the effect of the
partially attached shear-layer and the limitation associated with the theory produce the large
discrepancies.

3.5 The effect of frequency and modulating waveform

Two additional modes of excitation, discussed in Section 2, were applied: sinusoidal

amplitude modulation mode (AM) and burst (or pulsed) modulation, where F þ ¼ fm D=U1
for modulated waveforms. Burst modulation was achieved using a single cycle,
corresponding to DC=5% and denoted as BM1 (Cμ was based on the calibration of a
single cycle). A comparison of the lift generated by the cylinder when the three excitation
modes were used at approximately the same Cm and F+ is shown in Fig. 11, where fm ≈VSF.
This data shows that for the same momentum input, using BM1 affects lift more profoundly
than either pure harmonic excitation or AM modes. At Re=40,000 BM1 generates Cl,max =
0.7 at 85°<|a|<90°, whereas the pure-sine wave and AM generate Cl,max =0.25 at 60°<|a|<
85°. When considering that both AM and BM1 excitation signals are at the same
modulating frequency, it is clear that the difference in effectiveness is not due to the change
of the fundamental frequency but rather as a result of some other wave feature. There are
two possible reasons for this: (a) the different peak velocities Umax =U1 which are larger for
BM1; and (b) the different spectral content of the forcing.
Flow Turbulence Combust (2007) 78:383–407 401

Fig. 9 Cm effect on VSF and the momentum thickness, θ, for pure sine excitation at F=10.6 (1,090 Hz): a
Power spectrum of actuator cavity Cp, measured at Cl,max ; b St based on approximated wake width at
creation of vortices (Re=40,000). c Drag reductions using the theory of Roshko [22, 23] compared with
experimental data on the basis of Cd −Cd0 and Cdp −Cdp0. Filled symbols account for momentum coefficient
as Cd þ Cm  Cd0 and Cdp þ Cm  Cdp0

When applying short bursts, as is the case of BM1, the momentum and power
coefficients are considerably smaller than its AM and pure sine counterparts, for the same
peak input voltage. One may compensate for this and achieve the same value of Cm for the
BM case through either voltage or duty cycle increase. However, the current experiments
show that the two methods of increasing Cm do not have the same effect on the flow, even
when the same Cm (based on the calibration velocity fluctuations) is maintained. Thus, Cm is
not necessarily the relevant parameter to quantify the input excitation when intermittent
waves are used. The peak voltage change (and therefore the ratio Umax =U1 ) had a more
402 Flow Turbulence Combust (2007) 78:383–407

Fig. 10 Surface pressure distri-

butions, CP , vs. x/D for the data
-4
shown in Fig. 9 for pure sine
excitation at F+ =10.6 (1,090 Hz, CP
Re=40,000)
-3

-2

-1

baseline, α =87˚
0 Cµ≈1.2%, α = 90˚
Cµ≈4.1%, α = 95˚
Cµ≈5.8%, α = 100˚
Cµ≈9.2%, α = 105˚
1
0 0.25 0.5 0.75 1
x/D

pronounced effect on lift and drag than the duty cycle, indicating a higher sensitivity to the
peak amplitude of the fluctuation rather than to its duty cycle or duration.
Regarding spectral content, the difference may depend upon the transient response to
actuation, which is gradual for pure-sine and AM but impulsive (i.e. generating a broad
band of spectral peaks) for BM1. With a wider spectral content (BM1) the flow can

Fig. 11 Cylinder lift vs. slot

location at approximately con- 1.0
stant Cμ , pure sine-wave, AM Cl
and BM1 excitation. Sine wave-
form at 1,090 Hz, AM and BM1
(DC=0.05) at F+ =0.42 (43.6 Hz), 0.5
Cμ ¼ 0:12% (Re=40,000)

0.0

Baseline
-0.5 F+≈10.6, sine
F+≈0.42, AM
F+≈0.42, BM1

-1.0
-120 -60 0 60 120
Slot Location, α [deg]
Flow Turbulence Combust (2007) 78:383–407 403

“choose” the most-unstable frequency and amplify it. Furthermore, the presence of a wide
band of finite-amplitude waves allows the flow to undergo a non-linear interaction between
several of these waves and alter the mean flow as well. The effectiveness of actuation with
a sharp change rather than a gradual change in time was also observed by Amitay et al. [2],
Parekh and Glezer [19] on cylinders, while Margalit et al. [15] used BM forcing to
effectively excite the flow over a delta-wing.

3.6 Effects of vortex shedding lock-in to the excitation frequency

When modulated wave excitation was applied at frequencies that were close to the natural
vortex-shedding frequency (VSF), a lock-in phenomenon was observed (see discussion of
this phenomenon in Section 1.2). Results of measurements carried out at Re=100,000 with
a single burst of excitation (BM1; DC=5%) at frequencies below and approximately 2.5
times higher than the VSF (F+ =0.16 and 0.52, respectively, Fig. 12a and b). The data in
Fig. 12b indicates that the form-drag can be either increased or decreased by altering the
modulating frequency. Lock-in, by means of the low-frequency burst-modulated excitation
at F+ =0.16, increased the pressure drag at a ≈55° by 32% and at a ≈−40° by 17%.

1.0
1.6
Cl Cdp
0.5 1.4

1.2
0.0

1.0
Baseline
-0.5 F+≈0.52, Cµ=0.026%
F+≈0.16, Cµ=0.026% 0.8
F+≈0.52, Cµ=0.12%

-1.0 0.6
-120 -60 0 60 120 -120 -60 0 60 120
a Slot Location, α [deg] b Slot Location, α [deg]

-2.5
CP
-2.0

-1.5

-1.0

-0.5

0.0 Baseline
+≈0.52, C =0.026%
FF+ 0.26, Cµ
µ 0.026%
+≈0.16, C =0.026%
FF+
0.5 0.08, Cµ
µ 0.026%
+≈0.52, C =0.12%
FF+ 0.26, Cµ
µ 0.12%

1.0
0 0.25 0.5 0.75 1
c x/D

Fig. 12 Drag manipulation using frequency locking at frequencies close the shedding frequency and
separation control frequencies a Cl versus slot location; and b Cdp versus slot location (BM1 excitation,
Re=100,000). Legend applies for both figures. Drag manipulation using frequency locking: c CP vs. x/D ;
a=55°, BM1 excitation, Re=100,000. Drag manipulation using frequency locking: d Actuator’s cavity
pressure spectra; e Wake pressure spectra (a =55°, BM1 excitation, Re=100,000)
404 Flow Turbulence Combust (2007) 78:383–407

1
Baseline
CP F+≈0.52, Cµ=0.026%
F+≈0.16, Cµ=0.026%
0.1 F+≈0.52, Cµ=0.12%

0.01

0.001

0.0001
d 1

CP Baseline
F+≈0.52, Cµ=0.026%
F+≈0.16, Cµ=0.026%
0.1 F+≈0.52, Cµ=0.12%

0.01

0.001
e 10 100 1000
f [Hz]
Fig. 12 Continued

Roshko’s [22, 23] model (see section 1.3) predicts a 16% pressure drag rise due to lock-in
at this frequency. The theoretical under-prediction in the former case is discussed further
below. At the higher frequency (F+ =0.52) at the same Cm and a, the form-drag decreases
by 16% and decreases even further at higher a (Fig. 12b). This higher frequency is outside
the range of possible “lock-in” frequencies, though it might be possible to increase the
upper bound of the “lock-in” region by significantly increasing the amplitude of the
excitation [9]. In Roshko’s model, these frequencies produce k values that are outside of
the bounds considered. Thus the observed drag reduction in this instance is due to control
of boundary layer separation over some fraction of the cylinder’s surface.
Considering the base pressure variation between the different cases presented in
Fig. 12c, the dramatic change that the flow undergoes when it is excited at F+ =0.16
becomes evident. The base pressure is as negative as the suction peak and its influence is
recognizable along the entire lee-side of the cylinder (Fig. 12c). This figure also indicates
that a main assumption associated with Roshko’s theory, namely that Cpb is equal to that at
separation, is not always valid. In the uncontrolled case Cpb ≈−1.25, and it has a slightly
negative gradient in the lee portion of the cylinder that is associated with the proximity of
the vortex formation region to the surface. The closer to the cylinder the vortices form, the
higher their induced velocity field over the cylinder is, resulting in lower leeward pressures
Flow Turbulence Combust (2007) 78:383–407 405

Table 2 Vortex-shedding
frequencies and pressure Excitation Cavity Wake
amplitudes for Fig. 12d–e, as
measured by the actuator cavity Cμ F+ f Dominant Amp. Dominant Amp.
and wake dynamic sensors [%] [Hz] f [Hz] Cp f [Hz] Cp
(a =55°, BM1 excitation,
Re=100,000) Baseline 0 0 50 0.12 50 0.053
0.026 0.16 43.60 44 0.61 44 0.113
0.026 0.52 136.25 54 0.062 45 0.050
0.12 0.52 136.25 60 0.048 61 0.029

(cf. Roshko [23]). It is also assumed that the farther downstream the vortices are formed,
the weaker is their effect on the cavity pressure fluctuations that are indicative of the surface
pressure fluctuations, thus showing lower amplitude at the peak frequency spectrum
(Fig. 12d). It was observed that high-frequency excitation induced pressure fluctuations that
were weaker than those induced by the baseline vortices. As for the wake, a significantly
lower level of pressure fluctuations was measured when F+ <0.52 and Cm ¼ 0:026%
applied at a =55° than for the baseline flow (Fig. 12e). This suggests that a high-frequency
excitation (e.g., F+ =0.52) either generates vortices farther downstream from the cylinder
than frequencies comparable to the natural VSF or generates weaker vortices. When a
frequency closer to the natural shedding frequency is used (e.g., F+ <0.16), the vortex-
shedding frequency “locks-on” to it and thus the natural shedding frequency vanishes. The
excitation frequency energizes the vortices that are assumed to be formed closer to the
cylinder, intensifying their strength by an order of magnitude (as shown in Table 2).
The above arguments suggest that the high-frequency vortices are formed further
downstream, consistent with the passive control experiments of Roshko [23]. It is also
possible, that the higher-frequency is associated with separation control and hence resulting
in smaller and weaker wake structures.
Unlike most published experimental results, where the excitation involves the entire
cylinder (cross-stream transverse oscillating cylinder, Davies [8]; oscillatory rotating
cylinder, Taneda [27]; among others) the current excitation is applied asymmetrically to one
side of the cylinder. Consequently, the vortices emanating from that side are stronger and
they lower the pressure on that side. Since the mean symmetry of the baseline flow is
broken, the effects of the excitation on the vortices separating from the opposite side are
unclear. The near wake is curved and it is able to withstand a pressure difference across it.

4 Conclusions

The effects of passive and active boundary-layer transition and separation control over a
circular cylinder at transitional Reynolds numbers were investigated experimentally. It was
found than even the passive presence of the slot, especially when it was oriented upstream
and taped over, had a pronounced effect on transition and separation. Nevertheless,
nominally two-dimensional active periodic forcing offered greater flexibility than passive
methods for altering the lift and drag on the cylinder. It was found that the universal
Strouhal law also holds for active control on cylinders, as long as the excitation frequency
is significantly higher than the natural vortex-shedding frequency. This extension of the
universal Strouhal law is attributed to separation control, that narrows the wake, thereby
increasing the shedding frequency and decreasing the drag. However, drag reduction
406 Flow Turbulence Combust (2007) 78:383–407

predictions based on Roshko’s model are too small, mainly because partial attachment of
the separated-shear layer is not accounted for.
Excitation introduced in the direction opposing the oncoming flow resulted in different
effects depending on the perturbation amplitude. At low momentum coefficients, it was more
effective than co-flow excitation in promoting transition. At high momentum coefficients it
caused boundary layer separation on the fore-body. Exciting the flow using pulsed modulation
of the high resonance frequency of the piezo-fluidic actuators facilitated excitation of the flow
at frequencies comparable to the natural VSF. When the low modulation frequency is close to
the natural VSF, the shed vortices “lock-on” to the excitation frequency, form closer to the
cylinder and therefore increase the drag. When the excitation frequency was above the upper
range of the lock-in frequencies, it affected the flow in the same way as the pure tone
excitation at higher frequencies. This resulted in the forcing of transition accompanied by
further aft boundary-layer separation and drag reduction.

References

1. Achenbach, E.: Distribution of local pressure and skin friction around a circular cylinder in cross-flow up
to Re ¼ 5  106 . J. Fluid Mech. 34(part 4), 625–639 (1968)
2. Amitay, M., Smith, B.L., Glezer, A.: Aerodynamic flow control using synthetic jet technology. AIAA
Paper 98-0208, 36th Aerospace Sciences Meeting and Exhibit, Reno, NV, Jan. 12–15, 1998
3. Amitay, M., Glezer, A.: Role of actuation frequency in controlled flow reattachment over a stalled airfoil.
AIAA J. 40(2), 209–216 (2002)
4. Béra J.-C., Michard, M., Sunyach, M., Comte-Bellot, G.: Changing lift and drag by jet oscillation:
experiments on a cylinder with turbulent separation. Eur. J. Mech. B, Fluids, 1–21 (2000)
5. Bearman, P.W., Currie, I.G.: Pressure-fluctuation measurements on an oscillating circular cylinder. J.
Fluid Mech. 91(part 4), 661–677 (1979)
6. Bearman, P.W., Obasaju, E.D.: An experimental study of pressure fluctuations on fixed and oscillating
square-section cylinders. J. Fluid Mech. 110, 207–321 (1982)
7. Choi S., Choi, H., Kang, S.: Characteristics of flow over a rotationally oscillating cylinder at low
Reynolds number. Phys. Fluids 14(8), 2767–2776 (2002)
8. Davies M.E.: A comparison of the wake structure of a stationary and oscillating bluff body, using a
conditional averaging technique. J. Fluid Mech. 75(part 2), 209–231 (1976)
9. Griffin, O.M.: Flow similitude and vortex lock-on in bluff body near wakes. Phys. Fluids A1(4), April
(1989)
10. Heine, C., Spiess, M.C. Möser, M., Fiedler, H.E.: The influence of feedback control on the vortex
dynamics in a turbulent cylinder wake. Euromech Colloquium 361 – Active Control of Turbulent Shear
Flows, Berlin, 17–19 March (1997)
11. Hsiao, F.-B., Liu, C.-F., Shyu, J.-Y.: Control of wall-separated flow by internal acoustic excitation. AIAA
J. 28(8), 1440–1446 (1990)
12. Heurre, P., Monkewitz, P.A.: Local and global instabilities in spatially developing flows. Annu. Rev.
Fluid Mech. 22, 473–537 (1990)
13. Krishnamoorthy, S., Price, S.J., Paidoussis, M.P.: Cross flow past an oscillating circular cylinder:
synchronization phenomena in the near wake. J. Fluids Struct. 15, 955–980 (2001)
14. Liu, W.P., Brodie, G.: A demonstration of active turbulence transition with MEMS sensors. Abstact
YC06.06, American Physical Society Centennial Meeting Program, March 20–26, Atlanta, GA, 1999
15. Margalit, S., Greenblatt, D., Seifert, A., Wygnanski, I.: Delta wing stall and roll control using segmented
piezoelectric fluidic actuators. (previously AIAA paper 2002–3270), AIAA J. Aircr., May–June (2004)
16. Nishri, B., Wygnanski, I.: Effects of periodic excitation on turbulent separation from a flap. AIAA J. 36
(4), 547–556 (1998)
17. Pack, L.G., Seifert, A.: Periodic excitation for jet vectoring and enhanced spreading. AIAA J. Aircr 38
(3), 486–495 (2001)
18. Pal, D., Sinha, K.: Controlling an unsteady separating boundary layer on a cylinder with an active
compliant wall. AIAA Paper 97-0212, 35th Aerospace Sciences Meeting, Reno, NV, 1997.
19. Parekh, D.E., Glezer, A.: AVIA – Adaptive virtual aerosurface. AIAA Paper 2000-2474, Fluids 2000
Conference and Exhibit, Denver, CO, June 19–22, 2000
Flow Turbulence Combust (2007) 78:383–407 407

20. Patel, V.C., Sotiropoulos, F.: Longitudinal curvature effects in turbulent boundary layers. PIAS 33(1–2),
1–70 (1997)
21. Peterka, J.A., Richardson, P.D.: Effects of sound on separated flow. J. Fluid Mech. 37, 265–287, June
(1969)
22. Roshko, A.: A new hodograph for free-streamline theory. NACA TN 3168, 1954a.
23. Roshko, A.: On the drag and shedding frequency of two-dimensional bluff bodies. NACA TM 3169,
1954b.
24. Saric, W.S.: Goertler vortices. Annu. Rev. Fluid Mech. 26, 379–409 (1994)
25. Schewe, G.: On the forced fluctuations acting on a circular cylinder in a cross-flow from sub-critical to
trans-critical Reynolds number. J. Fluid Mech. 133, 265–285 (1983).
26. Seifert, A., Darabi, A., Wygnanski, I.: Delay of airfoil stall by periodic excitation. AIAA J. Aircr. 33(4),
691–698 (1996)
27. Taneda, S.: Visual observations of the flow past a circular cylinder performing a rotary oscillation. J.
Phys. Soc. Jpn. 45, 1038 (1978)
28. Taubert, L., Kjellgren, P., Wygnanski, I.: Generic bluff bodies with undetermined separation location.
AIAA Paper 2002-3068, 1st AIAA Flow Control Conference, St. Louis, MO, 24–26 Jun 2002.
29. Tokumaru, P.T., Dimotakis, P.E.: Rotary oscillation control of a cylinder wake. J. Fluid Mech. 224, 77
(1991)
30. Tresso R., Munoz, D.R.: Homogeneous, isotropic flow in grid generated turbulence. ASME J. Fluids
Eng. 122, 51–56, March (2000)
31. Williams, D., Acharya, M., Bernhardt, J., Yang, P.: The mechanism of flow control on a cylinder with the
unsteady bleed technique. AIAA Paper 91-0039, 29th Aerospace Sciences Meeting and Exhibit, Reno,
NV (1991)
32. Wiltse, J.M., Glezer, A.: Manipulation of free shear flows using piezoelectric actuators. J. Fluid Mech.
249, 261 (1993)
33. Pack Melton, L.G., Yao, C.S., Seifert, A.: Active control of flow separation from the flap of a
supercritical airfoil. AIAA J. 44(1), 34–41 (2006), JAN (previously AIAA Paper 2003-4005).