Experiments in Fluids 32 (2002) 44±54 Ó Springer-Verlag 2002

DOI 10.1007/s003480100340

Control of backward-facing step flow using a flapping foil

J. C. S. Lai, J. Yue, M. F. Platzer

44
Abstract The effect of oscillating a small foil in plunge on x streamwise direction measured from the corner of the
the reattachment of a separated shear layer in a two-di- step (Fig. 1)
mensional backward-facing step ¯ow has been studied xa streamwise distance of the leading edge of the foil
using ¯ow visualization and single component laser from the step (Fig. 1)
Doppler velocimetry (LDV) measurements. It has been y transverse direction measured from the corner of the
shown that a jet instead of a wake is generated by the step (Fig. 1)
¯apping action of the foil. Results indicate that this action ya transverse distance of the leading edge of the foil from
induces strong mixing and entrainment when the foil is the wall (Fig. 1)
located within the recirculation ¯ow region, thereby re- z spanwise direction in a Cartesian coordinate system
ducing the reattachment length by as much as 70%. Fur- h momentum thickness
thermore, it has been shown that the ¯apping foil is most
effective in reducing the size of the separation zone when
1
placed close to the wall and to the step. It is least effective
Introduction
when placed outside the separated shear layer or down-
Backward-facing step ¯ows have received considerable
stream of the reattachment zone.
attention in the past primarily because of two reasons.
First, separated and reattached ¯ows are encountered in
List of symbols
most practical engineering applications such as ¯ows in
ap amplitude of ¯apping
diffusers, combustors, turbomachinery, and ¯ows around
c chord
airfoils. Second, backward-facing step ¯ows represent the
f frequency of ¯apping (Hz)
simplest ¯ow for studying the reattachment process and
h step height
for providing data for turbulence modeling and CFD val-
k reduced frequency parameter (k ˆ 2pfc/Uo)
idation. A critical review of turbulent ¯ows over backward-
Reh Reynolds number based on step height
facing steps was given by Eaton and Johnston (1981).
Sth Strouhal number based on step height
Earlier studies of backward-facing step ¯ows were re-
Sth Strouhal number based on momentum thickness h
stricted by the instrumentation available and were limited to
U mean streamwise velocity
¯ow visualization (for example, Abbot and Kline 1962),
u¢ streamwise turbulence intensity
pressure and hot-wire measurements (such as Bradshaw
Uo free-stream velocity
and Wong 1972; Kim et al. 1980; Chandrsuda and Bradshaw
1981). Although hot-wire measurements provide detailed
Received: 26 August 1999 / Accepted: 29 May 2001 velocity information on the ¯ow downstream of the reat-
tachment zone, data obtained in the recirculation and
J. C. S. Lai (&) reattachment zones are subject to severe errors due to the
School of Aerospace and Mechanical Engineering inability of traditional hot-wire techniques to discriminate
University College against reversed ¯ow and to resolve high turbulence levels.
The University of New South Wales With the advent of non-intrusive optical techniques the past
Australian Defence Force Academy decade has seen numerous studies of backward-facing step
Canberra, ACT 2600, Australia ¯ows using LDV such as Stevenson et al. (1984), Driver and
e-mail: [email protected] Seegmiller (1985), Adams and Johnston (1988), and Otugen
Tel.: +61-2-62688272; Fax: +61-2-62688276
(1991), and particle-tracking velocimetry (Kasagi and
J. Yue, M. F. Platzer Matsunaga 1995).
Department of Aeronautics and Astronautics It has been shown by Eaton and Johnston (1981) and
Naval Postgraduate School Adams and Johnston (1988), among others, that the reat-
Monterey, CA 93943±5106, USA tachment length is greatly in¯uenced by the state of the
separating boundary layer. Generally, at low Reynolds
This work has been supported by the Of®ce of Naval Research number based on step height (h) (Reh < 1,000), the ¯ow is
(project monitor Dr. E. Rood). Partial support from the Naval
Postgraduate School provided under the National Research laminar at separation and at reattachment, and the reat-
Council Research Associateship program to the ®rst author tachment length increases with Reynolds number. As the
(JCSL) while on leave from the Australian Defence Force Reynolds number increases further, transition to turbu-
Academy is gratefully acknowledged. lence occurs for the separated free shear layer and the
reattachment length decreases with increasing Reynolds of 2% of turbulence intensity near the wall at the separa-
number. At higher Reynolds numbers (Reh > 100,000), the tion point caused a reduction of 2h in the reattachment
reattachment length becomes almost independent of length. By inserting a cylinder of diameter 0.4h at three
Reynolds number. The increase in reattachment length different locations (0.6h, 0.8h), (0.6h, h), and (0.6h, 1.5h)
from the upstream laminar condition to the turbulent for Reh ˆ 700, 1,000, and 1,400, Suzuki et al. (1991)
condition can be as much as over 30%. The reattachment showed that the velocity characteristics in the backward-
length decreases with increasing boundary layer thickness facing step ¯ow could be changed signi®cantly, resulting in
at separation (Adams and Johnston 1988), but the trend is increased heat transfer.
weak. Eaton and Johnston (1981) and Isomoto and Ho- It was ®rst recognized by Knoller (1909) and Betz
nami (1989) have found that the reattachment length (1912) that a ¯apping foil generates thrust. Here a
shortens as the free-stream turbulence level increases. It ¯apping foil refers to one being oscillated in plunge as
45
can be inferred from the experimental results documented opposed to being in pitch. This Knoller±Betz effect was
by Eaton and Johnston (1981) and Adams and Johnston ®rst con®rmed experimentally by Katzmayr (1922). Re-
(1988) that for an expansion ratio less than 1.8, the reat- cently, Dohring et al. (1996) showed both experimentally
tachment length increases with expansion ratio. Kuehn and computationally that when a foil is ¯apped with an
(1980), Driver and Seegmiller (1985) and Ra and Chang appropriate combination of frequency and amplitude, a
(1990) have all shown that the reattachment length in- jet (instead of a wake) is produced downstream of the
creases with increasing adverse streamwise pressure gra- trailing edge. Subsequently Lai and Platzer (1999)
dient. By using ¯ow visualization, Roos and Kegelman quanti®ed the jet characteristics of a ¯apping NACA
(1986) showed that the vortex roll-up and merging process 0012 airfoil for a range of frequencies and amplitudes of
is also present in a reattaching laminar shear layer. By ¯apping at various free-stream velocities. The charac-
obtaining the power spectra of streamwise ¯uctuating ve- teristics of a ¯apping foil at zero free-stream velocity
locity from hot-wire signals, the dominant fundamental have also been recently documented by Lai and Platzer
frequencies of vortex roll-up have been identi®ed by Roos (2001). The ¯apping foil is essentially a two-dimensional
and Kegelman (1986) to be Sth (Strouhal number based on propeller that energizes the ¯ow by entraining the
step height) 0.4 while the results of Bhattacharjee et al. upstream ¯ow into the propeller and imparting the
(1986) show a range of 0.2±0.6. Subsequently, Hasan downstream ¯ow additional momentum. This effect has
(1992) showed that the laminar reattaching shear layer has been demonstrated by Jones et al. (1998). Thus the en-
two distinct modes of instability, namely, the ``shear layer trainment induced by the jet structures of a ¯apping foil
mode'' of instability at Sth  0.012; and the ``step mode'' of may provide a mechanism whereby the reattachment of
instability at Sth  0.185. The shear layer instability fre- a separated shear layer may be controlled.
quency reduces to the step mode via one or more stages of The objective of this paper is to present the results of a
the vortex merging process. study designed to explore whether the recirculation ¯ow
With improved understanding of the ¯ow, the objective region in a two-dimensional backward-facing step ¯ow
of research often becomes one of studying means whereby can be controlled by a small ¯apping foil. In addition, the
the ¯ow can be controlled to suit individual applications. effects of the frequency and amplitude of oscillation and
The controlled excitation may be implemented using the location of the ¯apping foil on the recirculation ¯ow
various means among which mechanical, acoustic or region will be discussed.
¯uidic devices are more commonly used. By introducing
global acoustic excitation at Sth ˆ 0.2±0.4 and a constant 2
forcing amplitude of 92 dB at the step edge, Bhattacharjee Experimental set-up
et al. (1986) found that the spreading rate in the separated
shear layer was increased, resulting in a 10±15% reduction 2.1
in reattachment length. Their results seem to be relatively Experimental rig and flow condition
independent of Reynolds number Reh over the range All the experiments reported here were conducted in the
26,000±76,000. Roos and Kegelman (1986) applied a con- water tunnel facility at the Department of Aeronautics and
stant oscillating ¯ap amplitude of 1% of the boundary Astronautics, Naval Postgraduate School. The water tunnel
layer thickness (d) at the ¯ap trailing edge at Sth ˆ 0.29. is a closed circuit continuous ¯ow facility with a contraction
They found that the reattachment length was reduced by at ratio of 6:1. The test section is 380 mm wide, 1,500 mm long,
least 1h for a laminar separating shear layer and by at least and 510 mm high. The side walls and bottom of the test
2h for a fully turbulent separating shear layer. Controlled section as well as the end wall are made of glass to provide
perturbation was introduced by Hasan (1992) at optical access for ¯ow visualization and LDV measure-
Sth ˆ 0.218, 0.49, and 0.845 with an amplitude of up/Uo ˆ ments. The top of the test section is open to the atmosphere.
0.015 at x ˆ 0 and y ˆ y0.9 (where U/Uo ˆ 0.9) for The ¯ow velocity can be adjusted in a range from 0 to about
Reh ˆ 11,000. His results indicate a reduction in reat- 0.5 m/s. Despite the free surface in the water tunnel, Lai and
tachment length of 18, 12.5, and 12% respectively. Other Platzer (2001) have not observed any signi®cant free surface
attempts in modifying the ¯ow dynamics of backward- effects on the ¯ow characteristics of the ¯apping foil.
facing step ¯ows are mainly based on increasing the local Figure 1 shows the set-up of the backward-facing step
turbulence intensity. By placing a rod upstream of the and the foil. The backward-facing step structure was made
separation point to increase the turbulence intensity, of 4.76 mm thick aluminum plates. The step height (h)
Isomoto and Honami (1989) showed that only an increase is 30 mm and the step width is 380 mm, thus giving an
 3
Basic two-dimensional backward-facing step flow
The mean streamwise velocity pro®les at x/h ˆ ) 0.067 for
z/h ˆ ) 2, 0 and 2 in Fig. 2 show that they agree with each
other to within 5% and follow the 1/7 power law, thus
indicating that the separating boundary layer is two-di-
mensional and fully turbulent. Contours of mean stream-
wise velocity and turbulence intensity are displayed in
Fig. 3a and b. The recirculation ¯ow region can be iden-
ti®ed from the negative mean streamwise velocity in
Fig. 3a and high streamwise turbulence intensity occurs in
46
the shear layer near the dividing streamline (Fig. 3b). By
integrating the mean streamwise velocity pro®les in Fig. 3a
Fig. 1. Schematic showing experimental set-up for backward- with respect to the transverse distance (y) from the wall,
facing step ¯ow contours of the non-dimensional mean streamline are
plotted in Fig. 4 from which the reattachment length has
been determined to be 5.2h. While this value seems to be
aspect ratio of 12.33, larger than the value of 10:1 recom- slightly lower than the normally quoted value of 6h for
mended by de Brederode and Bradshaw (1972) for two-dimensional backward-facing step ¯ow, it compares
avoiding signi®cant three-dimensional effects near the well with the value of 4.9h determined by Etheridge and
center plane. The section downstream of the step is Kemp (1978) using LDV under similar conditions in a
700 mm (L2) long. The plate upstream of the step is water tunnel with a free surface. It should be pointed out
910 mm (L1) long. The free water surface was 272 mm that while most backward-facing step data were obtained
(Ho) and 242 mm (Hi) above the plate and the step in wind tunnels with the test section bounded on four
respectively, thus giving an expansion ratio of 1.12. The sides by walls, our data and those of Etheridge and Kemp
free-stream velocity was 0.32 m/s, giving a Reynolds (1978) were obtained with a free top surface. Figure 4 also
number Reh (based on step height h) of 12,700. reveals the existence of a corner eddy within x/h ˆ 1 that
Experiments were conducted using a NACA 0012 is due to a secondary separation and has been observed by
airfoil with chord (c) of 10 mm and span of 355 mm. A Abbot and Kline (1962) and Kasagi and Matsunaga (1995).
model 113 Elektro-seis shaker from APS Dynamics Inc. At x/h ˆ 5.5, which is just downstream of the reattach-
driven by an APS model 114 ampli®er was mounted on ment point, the mean streamwise velocity pro®les at
top of the test section to oscillate the foil sinusoidally in z/h ˆ ) 1.33, 0, and 1.33 in Fig. 2 indicate some slight
plunge. The signal from a HP200CD frequency generator three-dimensional effects that are to be expected in the
was varied from 5 to 20 Hz. The amplitude of oscilla- reattachment zone. Further downstream, the ¯ow is
tion, measured with a Lucas model No. DC-E500 DC- expected to recover to two-dimensionality as found by
operated linear variable differential transformer (LVDT), Armaly et al. (1983).
was varied from 1.24 to 4.94 mm.

2.2
Instrumentation
A TSI single component LDV was used to measure the
mean streamwise velocity and streamwise turbulence in-
tensity distributions up to 10h downstream of the step.
The ®ber optic probe head could be traversed with an
accuracy of 0.1 mm in the streamwise (x), transverse (y),
and spanwise (z) directions using a Lintech model 41583
traverse table driven by Applied Motion Products 1618
power supply and interface. The light source was an
Omnichrome model 543±300A Argon ion laser with a rated
output of 300 mW at 8.8 A. The beam separation was
50 mm and the focal length was 350 mm, giving a fringe
spacing of 3.427 lm for the blue beam. The Doppler signal
was processed with a TSI IFA 550 intelligent ¯ow proces-
sor. The data acquisition was automated using a 386 PC
and TSI ¯ow information display (FIND) software version
4.5. At least 1,000 data samples were used for each mea-
surement point. Although the maximum differences be-
tween corrected and uncorrected data were less than 2%,
velocity correction based on time between data was ap-
plied. Uncertainty estimates indicate that the uncertainty
in the mean velocity measurements is within 5 mm/s. Fig. 2. Mean streamwise velocity pro®les
 47

Fig. 3a. Contours of mean

streamwise velocity (U/Uo) for
backward-facing step ¯ow;
b Contours of streamwise tur-
bulence intensity (u¢/Uo) for
backward-facing step ¯ow

Fig. 4. Mean non-dimensional

streamline for backward-
facing step ¯ow

4 indicate that a ¯apping foil when suitably placed in a

Controlled excitation with a flapping foil backward-facing step ¯ow may induce signi®cant entrain-
ment to reduce the size of the recirculation zone.
4.1
Characteristics of a flapping foil 4.2
While the wake patterns behind a foil undergoing small LDV survey close to the wall
amplitude pitching oscillations have been studied quite Before detailed LDV measurements were made, ¯ow vi-
extensively (see, for example, Koochesfahani 1989), wakes sualization using dye injection at x/h ˆ 3.5, 4, 4.5, 5, 5.5,
behind ¯apping foils have not received as much attention and 6 was conducted to observe the effects of ¯apping a
until recently. Freymuth (1988) demonstrated thrust pro- small foil at various frequencies and amplitudes of oscil-
duction for ¯apping foils. By using ¯ow visualization, Lai lation. Three different frequencies were used, namely, 5,
and Platzer (1999) showed that when the ¯apping velocity 10, and 20 Hz with an amplitude (ap) of either 0.247c or
increases, the familiar Karman vortex street indicative of 0.123c. Various locations (xa, ya) of the foil were tested,
drag (Fig. 5a) changes to a vortex street indicative of thrust ranging from upstream of the step, in the recirculation
(Fig. 5b). As shown in Fig. 6, while the mean streamwise ¯ow region, outside the shear layer to downstream of the
velocity pro®le at 0.41c downstream of the trailing edge of a reattachment point. The ¯ow visualization results indicate
stationary foil is a wake (indicative of drag), that for the that there is very little in¯uence on the reattachment
same foil ¯apping at 10 Hz with an amplitude of 0.247c ex- length by ¯apping a small foil outside the recirculation
hibits the characteristics of a jet. It should be noted that the ¯ow region or outside the shear layer.
maximum jet velocity at this location is almost four times Consequently, the following locations were selected for
that of the free-stream velocity of 0.327 m/s. These results surveying the downstream variation of mean streamwise
48

Fig. 5a. Wake structures behind a stationary foil indicative of

drag; b Wake structures behind a plunging foil indicative of
thrust

Fig. 6. Mean streamwise velocity pro®les at x/c ˆ 0.41 for a foil

with and without ¯apping

velocity (U/Uo) close to the wall surface (y/h ˆ 0.066)

using LDV. Three streamwise stations, where the leading
edge of the foil was located at xa/h ˆ 1.83, 2.5, and 4, were
chosen. For each streamwise station, the foil was posi- Fig. 7a±c. Variation of U/Uo at y/h ˆ 0.067 with x/h for various
locations of a stationary foil. a xa/h ˆ 1.83; b xa/h ˆ 2.5;
tioned at four different transverse locations, namely c xa/h ˆ 4
ya/h ˆ 0.2 (corresponding to the point close to the wall
that still allows the foil to ¯ap with a suf®ciently large also be seen that the variation of U/Uo (at y/h ˆ 0.066) for
amplitude), ya/h ˆ 0.6 (corresponding to the center of the a stationary foil is almost independent of the location of
standing vortex at xa/h ˆ 1.83 for the backward-facing the foil except when it is located within the shear layer
step ¯ow in Fig. 4), ya/h ˆ 0.9 (corresponding to a loca- near the dividing streamline. When the stationary foil is
tion within the shear layer for xa/h ˆ 1.83, and 2.5), and located near the inner edge of the separated shear layer
ya/h ˆ 1.3 (corresponding to a location outside the shear at (1.83h, 0.9h) (Fig. 7a), the mean streamwise velocity
layer). (U/Uo) becomes positive much closer to the step, thus
As shown in Fig. 7, for all locations of a stationary foil, indicating earlier reattachment. As the stationary foil is
the mean streamwise velocity (U/Uo) at y/h ˆ 0.066 ini- moved further out from the inner edge of the separated
tially is close to zero or negative, decreases with x/h until it shear layer to (2.5h, 0.9h) in Fig. 7b and (4h, 0.9h) in
reaches a minimum, then increases and turns positive as Fig. 7c, the mean streamwise velocity turns positive fur-
the ¯ow proceeds beyond the reattachment point. It can ther downstream compared with Fig. 7a, thus indicating
 not as dramatic as observed in Fig. 8a. Results at all other
locations far from the wall indicate virtually no effect on
the reattachment length by ¯apping a foil at xa/c ˆ 2.5.
Figure 8c shows that if the ¯apping foil is located close to
the reattachment zone, its ability to reduce the reattach-
ment length is greatly diminished.

4.3
Detailed LDV measurements
Based on the ¯ow visualization studies and LDV surveys
close to the wall as described in Sect. 4.2, detailed LDV
49
measurements of mean streamwise velocity and turbulence
intensity distributions were made for the foil positioned at
(1.833h, 0.2h), (1.833h, 0.6h), (1.833h, 1h), and (1.833h,
1.3h), with and without ¯apping.
For a stationary foil at (1.83h, 0.2h), the mean non-
dimensional streamlines in Fig. 9a do not seem to be
signi®cantly altered by the presence of the foil when
compared with that without a foil shown in Fig. 4.
Although the size of the recirculation ¯ow region is slightly
reduced especially near the edge of the step, the reat-
tachment length appears to be virtually unaffected and
remains at 5.2h. Furthermore, the corner eddy is observed
in the ¯ow with a stationary foil. These results serve to
support the conclusion deduced from Fig. 7 that when a
stationary foil is located far from the dividing streamline,
its effect on the recirculation ¯ow region is very small.
With the foil ¯apping at a frequency of 20 Hz and an
amplitude of 0.123c, the mean non-dimensional streamline
in Fig. 9b shows that the size of the recirculation ¯ow
region has been substantially reduced and the reattach-
ment length has been reduced by over 70% to 1.5h. The
mean streamwise velocity (U/Uo) contours in Fig. 10a
show that (U/Uo) is large and positive downstream of the
foil, which is due to the strong jet generated by the ¯ap-
ping action of the foil as has already been displayed in
Fig. 6. Figure 10b shows that while a local maximum in
streamwise turbulence intensity can still be identi®ed in
the shear layer near the dividing streamline as observed
for the original backward-facing step ¯ow in Fig. 3b, an
even larger peak in streamwise turbulence intensity is now
located downstream of the ¯apping foil near the wall. This
region of intense streamwise turbulence intensity indicates
strong mixing as a result of the ¯apping action of the foil.
According to Lai and Platzer (1999), the strength of the jet
produced by a ¯apping foil in a uniform ¯ow is proportional
Fig. 8a±c. Variation of U/Uo at y/h ˆ 0.067 with x/h for various to k(ap/c). It may therefore be expected that by increasing
locations of a ¯apping foil (f ˆ 10 Hz, ap ˆ 0.247c). a xa/h ˆ 1.83;
b xa/h ˆ 2.5; c xa/h ˆ 4 either k or ap or both, the strength of the jet and hence the
entrainment induced may be increased, thereby enhancing
the ability of a ¯apping foil to reduce size of the recirculation
¯ow region. With the foil located at (1.833h, 1h) and ¯apping
the diminishing effect of the foil. Figure 8 shows the cor- at 10 Hz, the mean non-dimensional streamlines in Fig. 11a
responding (U/Uo) at y/h ˆ 0.066 for the foil ¯apping at show that when the amplitude of oscillation is set to 0.247c,
10 Hz with an amplitude 0.247c. Figure 8a indicates that the size of the recirculation ¯ow region has been signi®-
based on the distribution of positive streamwise velocity, cantly reduced when compared with the unexcited back-
the ¯apping foil located closest to the wall (ya/h ˆ 0.2) has ward-facing step ¯ow in Fig. 4. However, the recirculation
the most dramatic effect on reducing the reattachment ¯ow region now consists of a recirculation zone where the
length while the bene®t of a ¯apping foil located at other ¯ow reattaches at about x/h ˆ 3.3 and a very small recir-
locations is marginal. When the ¯apping foil is moved culation zone betweenx/h ˆ 3.2 and 5.1. As the amplitude of
downstream to xa/c ˆ 2.5 (Fig. 8b), the effect of the ¯ap- ¯apping is increased from 0.247c to 0.494c, the size of the
ping foil located at ya/h ˆ 0.2 is still very signi®cant but recirculation ¯ow region has been further substantially
50

Fig. 9a, b. Mean non-dimen-

sional streamline for a foil lo-
cated at (1.83h, 0.2h).
a stationary foil; b foil ¯apping
at 20 Hz with amplitude 0.123c

Fig. 10a, b. Velocity ®eld for

an foil located at (1.83h, 0.2h)
¯apping at 20 Hz with ampli-
tude 0.123c. a mean stream-
wise velocity (U/Uo);
b streamwise turbulence
intensity (u¢/Uo)

reduced as shown in Figure 11b with the reattachment the size of the recirculation ¯ow region reduces as the am-
length being about 2.7h. Furthermore, the small recircula- plitude of oscillation increases. With the foil located at
tion zone in Fig. 11a is completely eliminated. These limited (1.833h, 0.2h) and ¯apping with an amplitude of 0.247c,
results tend to suggest that for a given frequency of ¯apping, Figure 12a shows that when the frequency of oscillation is
 51

Fig. 11a, b. Effect of ampli-

tude of ¯apping on the mean
non-dimensional streamlines
for a ¯apping foil at (1.83h, 1h)
with f ˆ 10 Hz. a 0.247c;
b 0.494c

Fig. 12a, b. Effect of frequency

of ¯apping on the mean non-
dimensional streamlines for a
¯apping foil at (1.83h, 0.2h)
with ap ˆ 0.247c. a 5 Hz;
b 10 Hz

set to 5 Hz, the size of the recirculation ¯ow region has been ¯ow. As the frequency of ¯apping is increased from 5 to
reduced with the reattachment length being about 4.5h 10 Hz, Fig. 12b shows that the size of the recirculation ¯ow
compared with 5.2h for the unexcited backward-facing step region has been substantially reduced. These limited results
 tend to suggest that for a given amplitude of ¯apping, the Although the corner eddy can still be identi®ed, the orig-
size of the recirculation ¯ow region reduces as the frequency inal large standing vortex is now broken up into a pair of
of oscillation increases. vortices due to the ¯apping action of the foil. With the
The effects of the location of a ¯apping foil on the re- ¯apping foil located at (1.83h, 1h), which is within the
circulation ¯ow region are illustrated by the mean non- shear layer and in the vicinity of the dividing streamline,
dimensional streamlines in Figs. 12b and 13a±c for four the original recirculation ¯ow region has been signi®cantly
different foil locations of (1.833h, 0.2h), (1.833h, 0.6h), reduced in size. Although the separated shear layer ap-
(1.833h, 1h), and (1.833h, 1.3h) respectively. The foil was pears to reattach at x/h ˆ 3, there is a small weak recir-
oscillated in plunge at 10 Hz with an amplitude of 0.247c. culation zone between x/h ˆ 3.5 and 5 (Fig. 13b). When
The ¯ow with a ¯apping foil installed at (1.83h, 0.2h) is the ¯apping foil is moved to (1.83h, 1.3h), the mean non-
rather complex. Figure 12b seems to indicate that the dimensional streamline in Fig. 13(c) shows that the shear
52
shear layer reattaches at x/h ˆ 2 but then separates again layer reattaches at x/h ˆ 4.7.
at 2.9h before ®nally reattaching to the wall at 4.5h. The The mean non-dimensional streamline results in
original recirculation ¯ow region shown in Fig. 4 is sig- Figs. 12b and 13, substantiating the observations made
ni®cantly reduced in size and is broken up into three small from Figs. 7 and 8, therefore, show that the size of the
recirculation zones with very low strength. When the foil is recirculation ¯ow region is most reduced when the foil is
positioned at (1.83h, 0.6h), which is approximately the close to the wall (ya/h ˆ 0.2) followed by that located
standing vortex center of the original ¯ow, the size of the within the separated shear layer in the vicinity of the di-
original recirculation zone is reduced and the reattach- viding streamline (ya/h ˆ 1.0), at the standing vortex
ment length has been reduced by 27% to 3.8h (Fig. 13a). center in the absence of the foil (ya/h ˆ 0.6) and outside

Fig. 13a±c. Mean non-dimen-

sional streamline for various
locations of a foil ¯apping at
10 Hz with amplitude 0.247c. a
xa/h ˆ 1.83, ya/h ˆ 0.6; b xa/
h ˆ 1.83, ya/h ˆ 1.0; c xa/h ˆ
1.83, ya/h ˆ 1.3
the separated shear layer (ya/h ˆ 1.3). These results may References
be explained as follows. LDV measurements made by Lai Abbott DE; Kline SJ (1962) Experimental investigation of sub-
and Platzer (1999) show that at a given distance down- sonic turbulent ¯ow over single and double backward facing
stream of a ¯apping foil in a uniform ¯ow, the maximum step. J Basic Eng 84: 317±325
Adams EW; Johnston JP (1988) Effects of the separating shear
streamwise velocity is linearly proportional to k(ap/c). In layer on the reattachment ¯ow structure, Part 2. Reattachment
the control of the backward-facing step ¯ow, the foil is length and wall shear stress. Exp Fluids 6: 493±499
located in a recirculating ¯ow region where the local ve- Armaly BF; Durst F; Pereira JCF; Schonung B (1983) Experi-
locity is not uniform and an effective local velocity will be mental and theoretical investigation of backward-facing step
a more appropriate velocity scale to use. Since the effective ¯ow. J Fluid Mech 127: 473±496
local velocity is lower near the wall than away from the Betz A (1912) Ein Beitrag zur Erklaerung des Segel¯uges.
wall, for a given frequency and amplitude of ¯apping, the Z Flugtech Motorluftschiffahrt 3: 269±272
Bhattacharjee S; Scheelke B; Troutt TR (1986) Modi®cation of 53
foil will be operating at a higher reduced frequency k when vortex interactions in a reattaching separated ¯ow. AIAA J 24:
it is closer to the wall. Furthermore, LDV measurements, 623±629
Navier±Stokes and inviscid unsteady panel code calcula- Bradshaw P; Wong FYF (1972) The reattachment and relaxation
tions made by Dohring et al. (1996) con®rm that the of a turbulent shear layer. J Fluid Mech 52: 113±135
strength of a jet produced by a ¯apping foil is augmented de Brederode V; Bradshaw P (1972) Three dimensional ¯ow in
when the foil is in close proximity to a wall. Thus, due to nominally two-dimensional separation bubbles. I. Flow behind
the combined effects of the higher reduced frequency and a rearward-facing step. Imperial College, Aeronautical Report
72±19
wall proximity, it is more effective to control the recircu- Chandrsuda C; Bradshaw P (1981) Turbulent structure of a
lating ¯ow region by ¯apping a foil close to the wall than reattaching mixing layer. J Fluid Mech 110: 171±194
far from the wall. Dohring CM; Platzer MF; Jones KD; Tuncer IH (1996) Compu-
Results presented here demonstrate that the entrain- tational and experimental investigation of the wakes shed from
ment induced by a ¯apping foil appropriately located is ¯apping airfoils and their wake interference/impingement
suf®cient to impart additional momentum to the ¯ow to characteristics. In 78th AGARD Fluid Dynamics Panel Sym-
promote reattachment. A US patent to exploit the ¯apping posium, 20±23 May 1996, Trondheim, Norway, Paper 33: 1±9.
AGARD, Neuilly-sur-Seine, France
foil for ¯ow control has been granted to Platzer (1999) and Driver DM; Seegmiller HL (1985) Features of a reattaching tur-
it is envisaged that a ¯apping foil can be used to control bulent shear layer in divergent channel ¯ow. AIAA J 23: 163±
separation in both internal and external ¯ows. The pro- 171
pulsive ef®ciency of a ¯apping foil has been calculated by Eaton JK; Johnston JP (1981) A review of research on subsonic
Tuncer and Platzer (1996) for a range of reduced fre- turbulent ¯ow reattachment. AIAA J 19: 1093±1100
quency parameter k. However, the energy bene®t resulting Etheridge DW; Kemp PH (1978) Measurements of turbulent ¯ow
from ¯ow control using a ¯apping foil cannot be reason- downstream of a rearward-facing step. J Fluid Mech 86: 545±
566
ably estimated from the experimental data here and is Freymuth P (1988) Propulsive vortical signatures of plunging and
being studied by Platzer and Jones (2001). pitching airfoils. AIAA J 26: 881±883
Hasan MAZ (1992) The ¯ow over a backward-facing step under
5 controlled perturbation: laminar separation. J Fluid Mech 238:
Conclusions 73±96
In this paper, the effects of a small foil oscillated in plunge Isomoto K; Honami S (1989) The effect of inlet turbulence in-
on a backward-facing step ¯ow have been explored using tensity on the reattachment process over a backward-facing
step. J Fluids Eng 111: 87±91
single component LDV measurements for a Reynolds Jones KD; Dohring CM; Platzer MF (1998) Experimental and
number of 12,700 and a step aspect ratio of 12.33. Results computational investigation of the Knoller±Betz effect. AIAA
of mean streamwise velocity and turbulence intensity for a J 36: 1240±1246
backward-facing step ¯ow with two-dimensional upstream Kasagi N; Matsunaga A (1995) Three-dimensional particle-
¯ow condition are consistent with those reported in the tracking velocimetry measurement of turbulence statistics and
literature. The ¯ow characteristics for various locations of energy budget in a backward-facing step ¯ow. Int J Heat Fluid
a ¯apping foil, and various frequencies of oscillation and Flow 16: 477±485
Katzmayr R (1922) Effect of periodic changes of angle of attack
amplitude have been studied. Results indicate that the on behavior of airfoils. NACA TM 147
global ¯ow behavior in terms of the size of the recircula- Kim J; Kline SJ; Johnston JP (1980) Investigation of a reattaching
tion ¯ow region is not signi®cantly modi®ed when the foil turbulent shear layer: ¯ow over a backward-facing step. J Fluids
is not ¯apping unless it is located within the separated Eng 102: 302±308
shear layer close to the dividing streamline. The most ef- Knoller R (1909) Die Gesetze des Luftwiderstandes. Flug Motor-
fective location for the ¯apping foil is to be as close as tech (Wien) 3(21): 1±7
possible to the step and to the wall, and the next is one Koochesfahani MM (1989) Vortical patterns in the wake of an
oscillating airfoil. AIAA J 27: 1200±1205
located within the separated shear layer near the original Kuehn DM (1980) Effects of adverse pressure gradient on the
dividing streamline. The least effective location is outside incompressible reattaching ¯ow over a rearward-facing step.
the shear layer or downstream of the reattachment zone. AIAA J 18: 343±344
The effectiveness of a ¯apping foil increases with fre- Lai JCS; Platzer MF (1999) The jet characteristics of a plunging
quency and amplitude of oscillation. The reattachment airfoil. AIAA J 37: 1529±1537
length has been shown to be reduced by as much as over Lai JCS; Platzer MF (2001) Characteristics of a plunging airfoil at
70% with a 10 mm NACA 0012 airfoil ¯apping at 20 Hz zero free stream velocity. AIAA J 39: 531±534
with an amplitude of 0.247c.
 Otugen MV (1991) Expansion ratio effects on the separated shear Roos FW; Kegelman JT (1986) Control of coherent structures in
layer and reattachment downstream of a backward-facing step. reattaching laminar and turbulent shear layers. AIAA J 24: 1956
Exp Fluids 10: 273±280 Stevenson WH; Thompson HD; Craig RR (1984) Laser velocim-
Platzer MF (1999) Integrated propulsion/lift/control system for eter measurements in highly turbulent recirculating ¯ows.
aircraft and ship applications. US Patent 5,975,462 J Fluids Eng 106: 173±180
Platzer MF; Jones KD (2001) The unsteady aerodynamics of Suzuki H; Kida S; Nakamae T; Suzuki K (1991) Flow and heat
¯apping-foil propellers. In Proceedings of the 9th International transfer over a backward-facing step with a cylinder mounted
Symposium on Unsteady Aerodynamics, Aeroacoustics and near its top corner. Int J Heat Fluid Flow 12: 353±359
Aeroelasticity of Turbomachines, Ecole Centrale de Lyon, Lyon, Tuncer IH; Platzer MF (1996) Thrust generation due to airfoil
France, 4±8 September 2000. Springer, Heidelberg ¯apping. AIAA J 34: 324±331
Ra SH; Chang PK (1990) Effects of pressure gradient on reat-
taching ¯ow downstream of a rearward-facing step. J Aircraft
54 27: 93±95