Measurement of local forcing on a turbulent boundary layer using PIVY.-S. Park, S.-H. Park, H.J. Sung

Abstract An experimental study was carried out toinvestigate the effect of periodic blowing and suction ona turbulent boundary layer. Particle image velocimetry(PIV) was used to probe the characteristics of the flow.Local forcing was introduced to the boundary layer via asinusoidally-oscillating jet issuing from a thin spanwiseslot. Three forcing frequencies (f +=0.44, 0.66 and 0.88)with a fixed forcing amplitude (A+=0.6) were employed atReh=690. The effect of three different forcing angles(a=60�, 90� and l20�) was investigated under a fixedforcing frequency (f +=0.088). The PIV results showedthat the wall-region velocity decreases on imposition ofthe local forcing. Inspection of the phase-averagedvelocity profiles revealed that spanwise large-scale vorti-ces are generated downstream of the slot and persistfarther downstream. The highest reduction in skin fric-tion was achieved at the highest forcing frequency(f +=0.088) and a forcing angle of a=120�. The spatialfraction of the vortices was examined to analyze the skinfriction reduction.

SymbolsA+ forcing amplitudeCf skin friction coefficient

dx+ vortex diameter

f+ forcing frequency

Reh Reynolds number based on U¥ and hT forcing period (s)t time (s)ur friction velocity (m/s)U¥ freestream velocity (m/s)

Uc+ convection velocity

u¢+ streamwise turbulent intensity

v¢+ wall-normal turbulent intensity

Greek symbolsa forcing angleh boundary layer momentum thickness (m)q density (kg/m3)m kinematic viscosityxz spanwise vorticity

Superscripts– time mean~ phase average¢ random fluctuations+ normalization by wall unit

Subscripts¥ freestream0 no-forcing case

1IntroductionRecent advances in the understanding the coherentstructure of wall-bounded turbulent flow have intensifiedthe interest in reducing skin friction by controlling thecoherent structure. In a turbulent boundary layer, skinfriction is closely associated with the downward sweepmotion induced by the streamwise vortices in the vicinityof the wall (Robinson 1991). Effective control of thestreamwise vortices is very crucial to achieving thereduction of skin friction and consequently the drag.Recent direct numerical simulations by Choi et al. (1994)demonstrated that attenuation of the streamwise vorticesby means of active suction/blowing over the entire wallsignificantly reduces the skin friction. From a practicalviewpoint, however, this active control strategy is difficultto implement because it requires a dense population ofsensors and actuators on the wall.

Many attempts have been made to device a practicalmethod for reducing wall skin friction. These include themodification of the wall surface by installing riblets (Choiet al. 1993), as well as the use of a compliant wall (Choiet al. 1997) or a spanwise oscillating wall (Jung et al. 1992).Among the approaches considered to date, the use of localsuction/blowing to reduce wall skin friction deserves moredetailed study because it provides an efficient and simplemeans for locally actuating the wall-bounded flow. More-over, the strength of the actuation can be controlled withrelative ease by local suction/blowing. Most previous

Experiments in Fluids 34 (2003) 697–707

DOI 10.1007/s00348-003-0604-2

697

Received: 6 August 2002 / Accepted: 27 January 2003Published online: 24 April 2003 Springer-Verlag 2003

Y.-S. Park, S.-H. Park, H.J. Sung (&)Department of Mechanical Engineering,Korea Advanced Institute of Science and Technology,373-1 Kusong-dong, Yusong-ku, 305-701 Taejon, KoreaE-mail: hjsung@kaist.ac.krFax: +82-42-8695027

This work was supported by a grant from the National ResearchLaboratory of the Ministry of Science and Technology, Korea

experimental or numerical studies of local suction/blowinghave focused on steady actuation. Sano and Hirayama(1985) examined the effect of steady suction and blowingthrough a spanwise slot in a turbulent boundary layer.They found that the steady blowing (suction) decreases(increases) skin friction and increases (decreases) turbu-lent intensity behind the spanwise slot. Antonia et al.(1995) demonstrated that intensive local suction relami-narizes the turbulent boundary layer. Park and Choi(1999) conducted a direct numerical simulation to inves-tigate the effects of applying weak steady local suction/blowing. They reported that steady local blowing lifts upthe streamwise vortices, thereby reducing the interactionof the vortices with the wall. In contrast to the previousstudies that considered only steady suction/blowing,Tardu (1998, 1999, 2001) carried out wind tunnel experi-ments in which he compared the behavior of a periodicallyblowing system and a steadily blowing system. He showedthat both types of blowing led to a reduction in drag. Parket al. (2001) and Rhee and Sung (2001) examined the effectof periodic blowing and suction both experimentally andnumerically. They found that local forcing reduces skinfriction and that forcing at a higher frequency is moreeffective.

The majority of previous experimental investigationsused single-point measurement techniques to probe thestatistical properties of the flow. However, the physicalstructures behind the slot and their dynamic behaviorshave been difficult to probe using single-point measure-ment techniques. The difficulty involves the obtainingmultipoint measurements with high spatial resolutionthroughout the flow domain. In contrast, particle imagevelocimetry (PIV) can be used to obtain multidimensionalquantitative measurements of the instantaneous flowstructure and evolving dynamics of turbulent flows. Theprincipal aim of the present study was to investigate thedynamic coherent structure downstream of the localforcing. To achieve this aim, spatio–temporal measure-ments of the velocity and vorticity were acquired usingPIV. Imposition of a periodic disturbance may lead toperiodic variations in the global physical properties of theflow. The technique of phase averaging was used to dis-criminate the periodic velocity component from theinstantaneous velocity. In the experiments reported here,sinusoidal suction and blowing from a spanwise thin slotat the wall was introduced to a flat-plate turbulentboundary layer. The influence of the periodic blowing andsuction was examined by changing the forcing frequency(f +=0.044, 0.066 and 0.088) and the forcing angle (a=60�,90�and 120�). A spanwise vortex was generated by the localforcing, which led to a reduction in the skin friction nearthe wall. The vortex was found to lose momentum in thevicinity of the wall because the vortex flow opposes themainstream in this region. The mechanism behindthe reduction in skin friction was analyzed in terms of thespatial fraction of the vortex near the wall.

2Experimental apparatusMeasurements were performed in a recirculating openwater channel driven by a centrifugal pump. A settling

chamber, a honeycomb and a contraction were placed insequence to secure flow homogeneity. The dimensions ofthe test section were 250 mm (width)·250 mm(depth)·1200 mm (length). A flat plate was installed50 mm above the bottom wall of the test section. Figure 1shows a schematic diagram of the test section and theforcing system. The boundary layer was tripped at theleading edge of the flat plate using a combination of a tripwire of diameter 2 mm and a roughness strip of length100 mm. The roughness strip was installed to reduce thetwo-dimensionality of the wake behind the trip wire. Thiscombination ensured a self-preserving turbulent boundarylayer upstream of the local forcing. The water tunnel couldbe fitted with interchangeable spanwise slotted flat plates.A slot width of 3 mm was employed in all experiments,which corresponded to about 20 in wall units. The slot oflength 150 mm in the spanwise direction, which corre-sponds to 1000 in wall units, was located 600 mmdownstream from the leading edge. The origin of themeasurement coordinate axes was located at the end of thedownstream-edge of the slot, as shown in Fig. 1.

The slot was connected to a cavity in which water wassinusoidally forced by a scotch-yoke system, which gen-erated sinusoidal suction and blowing through the slot.This system was driven by an AC motor with a frequency-controllable inverter (Samsung, MOSCON-G3). The forc-ing frequency was measured by a digital tachometer(Onosokki, HT-5200). Adjustment of the stroke on thewheel at different revolution rates made it possible to varythe frequency of the local forcing while maintaining thesame forcing amplitude. The forcing velocity as measuredby a hot film positioned directly above the slot displayed asinusoidal waveform vf(t)=(vf)maxcos(2pft). The forcingamplitude was defined from the amplitude of the wave-form to be

Aþ ¼ vfð Þmax

U1; ð1Þ

where U¥ is the freestream velocity.In the present experiments, the forcing amplitude was

set to A+=0.6, which is similar to the value of A+=0.4 usedin a previous periodic blowing experiment carried out byTardu (2001). However, there is an inherent distinctionbetween the experiments of Tardu and those describedhere: the periodic blowing in the experiments of Tardu

Fig. 1. Test section and local forcing apparatus

698

caused a nonzero net mass flux into the flow field, whereasthe periodic suction and blowing imposed in the presentwork ensured zero net mass flux. Three forcing frequen-cies were chosen, f +=0.044, 0.066 and 0.088, which cor-responded to four, six and eight integer multiples of thebursting frequency, respectively. Here, f + is defined asf +=fm/us

2. To examine the effect of the forcing angle a, theflow properties were investigated at forcing angles ofa=60�, 90� and 120� while maintaining a fixed forcingfrequency of f +=0.088.

Cross-correlation PIV of single exposure digital imageswas used to acquire instantaneous in-plane velocity fieldmeasurements. A CCD camera (Kodak ES-1.0, 1024 pix-el·1024 pixel CCD array size) coupled to a PC runningimage acquisition software was used to acquire pairs ofsingle exposure images. A 105-mm lens (Nikkon, Micro)was used for imaging, and a field of view of approximately62 mm·62 mm was obtained. Hollow glass beads(q=1.02 kg/m3, d=8~12 lm) were used as tracer particles,and the flow plane of interest was illuminated with a two-head Nd:YAG laser. Each laser (Big Sky Laser, Ultra) wascapable of producing 8-ns, 30-mJ pulses at a repetitionfrequency of 15 Hz. Trigger signals obtained from a digitaltachometer (Onosokki, HT-5200) measuring the revolu-tion rate of the scotch-yoke system were utilized to syn-chronize the camera and the laser. The pulses for the laserand the CCD camera were generated and delayed using apulse delay generator (Berkeley Nucleonics, BNC-555). Anappropriate combination of cylindrical lenses was em-ployed to produce a 150-mm wide by 1-mm thick colli-mated light sheet on the center of the slot. The PIVanalysis of the images acquired by the CCD camera uti-lized a 24 pixel·24 pixel window with a 50% overlap fac-tor. This gave a spatial resolution of 0.75 mm betweenmeasurement points. For vector postprocessing, the localmedian filter criterion (Westerweel 1994) was used.

The imposition of a periodic suction and blowing maylead to periodic variations in the global physical quantitiesof the flow. Hence, it is necessary to represent the velocityfield as a superposition of three components (Hussain andReynolds 1970). An instantaneous quantity S is decomposedinto a time mean component �SS, a phase-averaged coherentcomponent ~ss and a random incoherent component s¢

S ¼ �SSþ ~ssþ s0: ð2ÞThe time average is

�SS ¼ limT!1

1

T

Z T

0

S tð Þdt; ð3Þ

and the phase average is

\S tð Þ >¼ limN!1

1

N

XN

n¼0

S t þ nTð Þ; ð4Þ

where T is a period of the disturbance.When the time mean component �SS was measured using

PIV, there was a small difference between the forcingfrequency measured by the digital tachometer and thefrequency at which images were taken on the CCD camera.Because of this small frequency difference, it took a long

time for these two frequencies to meet at the same phasecompared with the period of forcing T. If many imageswere obtained through this way, the reasonable time meancan be acquired. In this study, 3,000 velocity vectors wereaveraged to obtain the time mean component �SS. Toinvestigate the periodic motion of spanwise coherentstructure, each period T was divided into eight instants.These eight measurement points included the maximumstates of blowing and suction. One thousand images wereacquired at every measurement point and 8,000 imageswere taken at each frequency.

3Results and discussionThe freestream velocity was set to U¥=0.135 m/s and theReynolds number based on the momentum thickness overthe slot without forcing (f +=0) was Reh=690. Figure 2shows the y+ dependence of streamwise mean velocity U+

at four streamwise locations (x+=20, 40, 80 and 160). All ofthe data presented hereafter, including those under forc-ing, are normalized by the wall unit. Care should beexercised when normalizing by wall units, because thefriction velocity changes when a forcing is imposed. In thepresent experiments, the friction velocity (us0) in theabsence of forcing (f +=0) was used for the normalization.This normalization procedure was chosen to enable directcomparison of the near-wall behavior of systems withdifferent forcing frequencies. The distributions in Fig. 2show that the imposition of local forcing leads to areduction in the wall-region velocity in the vicinity of theslot. However, the forcing effect is weaker at y+>30. In the

Fig. 2. Streamwise mean velocity profiles for three forcingfrequencies

699

downstream (x+>80), the retarded flow gradually recoversto resemble the flow without forcing. A similar tendencyto the reduction in the wall-region velocity in the vicinityof the slot was observed not only in the steady blowingcases of Sano and Hirayama (1985), Park and Choi (1999)and Krogstad and Kourakine (2000), but also in the peri-odic forcing of Tardu (2001) and Park et al. (2001). Thissuggests that the blowing effect dominates the meanvelocity profiles under the local forcing conditionsimposed in the present study, even though blowing andsuction are equally applied to the turbulent boundarylayer. This is consistent with the result of Sano andHirayama (1985), who showed that the effect of blowing onthe mean flow field is greater than that of suction with thesame mass flux. The velocity retardation at the wall leadsto a reduction in the local skin friction coefficient (Cf)because of the small friction velocity. This further impliesthat the present forcing causes a net upward shift oflow-speed fluid and therefore an extension of the viscoussublayer. Such an upward shift is typical of a flow withreduced skin friction. As the forcing frequency isincreased, the retardation effect of the mean velocity dueto the forcing becomes stronger and persists fartherdownstream (see Fig. 2), although the quantitative differ-ence is small.

We now consider the behavior of the skin frictioncoefficient Cf, which is defined as

Cf ¼ 2us

U1

� �2

; ð5Þ

where the friction velocity us is calculated under theassumption of viscous sublayer, i.e. U+=y+. The nearestmeasuring position from the wall is y=0.8 mm, whichcorresponds to about y+=6 in wall units. The response ofCf/Cf0 to forcing frequencies of f +=0.044, 0.066 and 0.088 isshown in Fig. 3, where Cf0 represents the skin frictionwithout forcing. As can be seen in Fig. 3, Cf decreasesdownstream of the slot (x+=0) and then the local forcingeffect is gradually attenuated on moving farther down-stream (x+>50). As the flow moves downstream, there is aslight overshoot in Cf/Cf0 (x+>80), after which Cf convergesto Cf0. As the forcing frequency increases, Cf is reduced to

a great extent and the point of minimum Cf moves closerto the slot. At the highest forcing frequency (f +=0.088), themaximum reduction in skin friction is about 75%. Thepresent results are consistent with those of other periodicforcing studies (Park et al. 2001; Rhee and Sung 2001).

Figure 4 shows contour plots of vorticity xz togetherwith phase-averaged velocity vectors for systems subjectedto forcing frequencies of f +=0.044 and f +=0.088. Eightsnapshots taken at intervals of T/8 are shown, where T isthe forcing period. Maximum blowing is imparted at t=0,and maximum suction at t=4T/8. The phase-averagedvelocity vectors were obtained by subtracting the long-time mean velocity from the phase-averaged mean veloc-ity. The velocity vectors in Fig. 4 clearly show the gener-ation of a strong negative spanwise vortical structurecaused by the local forcing, and this vortex convectsdownstream over time. The flow is dominated by thepresence of an unsteady coherent vortex. The incomingfreestream is temporarily blocked by a strong upwardmotion during the blowing phase, and a negative spanwisevortical structure is created at t=7T/8. The pressure in theclose downstream vicinity of the slot is abruptly decreased,leading to the formation of a reverse flow region. Thenegative vortex generated by the local forcing has a max-imum value at t=T/8. At the beginning of the suctionphase, the reverse flow becomes stronger and the prein-jected fluid is then engulfed. This vortex causes a reduc-tion in the wall-region velocity because it generates areverse flow near the wall. Inspection of the flow in thevicinity of the slot reveals the generation of a smallsecondary positive vortex during blowing. This secondaryvortex is weaker than the main vortex and disappearsrapidly in the vicinity of the slot. As the flow movesdownstream, the large-scale vortical structures graduallyshift away from the wall, the vortices increase in size, andfinally the vortical structures disappear. This processbecomes faster as the forcing frequency is increased, as isevident in Fig. 4. Since the vortex size is determined by themass injected during the blowing period, it is inverselyproportional to the forcing frequency. Hence, the vortexsize at f +=0.088 is half that at f +=0.044, and the vorticalstructure decays faster at higher forcing frequency becauseof the smaller size and weaker strength of the vortices.

Fig. 3. Distributions of skin friction forthree forcing frequencies

700

The reduction of skin friction may be related to the roleof the large-scale vortical structure in the vicinity of thewall. In order to characterize the relationship betweenthe vortical structure and skin friction, in Fig. 5 we plotthe trajectories of the vorticity xz at y+=11 (i.e., (xz)max)and <U>)U0 at y+=6, where <U> is the phase-averagedvelocity and U0 is the mean velocity without forcing. Thequantity <U>)U0 provides an estimation of the wall-re-gion velocity because when <U>)U0 is negative (positive),the wall-region velocity is lower (higher). It is interestingthat the trajectories of xz and <U>)U0 are very similar,regardless of the forcing frequency. Downstream of thelocal forcing (x+>0), a strong vortex is clearly detected.The vortex loses momentum in the vicinity of the wall,which reacts as a resistance to the wall. The position of theminimum value of xz is almost the same as that of theminimum <U>)U0. This suggests that the vortex gener-ated by the local forcing reduces the skin friction near thewall. Inspection of the position at which xz�)1 or<U>)U0�0 in Fig. 5 indicates that the vortex center iscloser to the slot (x+>0) for f +=0.088 than for f +=0.044.This observation is closely related to the change in theposition of the minimum value of Cf/Cf0 with changingforcing frequency in Fig. 3.

To elucidate the convective nature of the forcing, con-tour plots of xz and <U>)U0 were calculated. These plots,which are given in Fig. 6, show that as time proceeds theconvection velocity of xz (dashed line) is the same as thatof <U>)U0 for the two forcing frequencies considered.Note that the convection velocities of xz and <U>)U0

change slightly in the downstream region. As seen in

Fig. 6a, the convection velocity is smaller in the region20 £ x+ £ 50 than in the region 50 £ x+ £ 120. As the flowmoves downstream, the spanwise vortex generated by thelocal periodic forcing gradually moves away from the wall.When the vortex shifts away from the wall, its convectionvelocity increases because the convection velocity of thevortex approaches the inherent flow velocity. The con-vection velocities for the two forcing frequencies in Fig. 6are shown to be different on the surface. However, it is

Fig. 5. Trajectories of xz and <U>)U0 at the maximum vorticityphase for f +=0.044 and f +=0.088

Fig. 4. Contour of xz together with phase-averaged velocityvectors for f +=0.044 (a) and f +=0.088 (b)

701

found that the convection velocity of f +=0.044 is equiva-lent to that of f +=0.088. This is because the vortex off +=0.088 is generated more frequently than that off +=0.044. A closer inspection of the contours of xz forf +=0.088 reveals that the pregenerated vortex convectswith a weaker intensity.

Increasing the forcing frequency increases the forcingthe frequency at which the vortex is generated and de-creases the distance between consecutive vortices. In orderto characterize these phenomena, the vortex size and otherparameters were calculated for three forcing frequenciesat the phase when the spanwise vorticity xz has a maxi-mum value. From the vorticity xz contour a large-scalespanwise vortex is obtained by using the local velocitygradient tenser of xz and its corresponding eigenvalues.The procedure used to identify the vortex from xz isdescribed in detail elsewhere (Chong et al. 1990; Dallmanet al. 1991; Jeong and Hussain 1995; Zhou et al. 1996, 1999;Adrian et al. 2000). The vortex obtained using this pro-cedure has the shape of an ellipse with the major axis inthe streamwise direction. As summarized in Table 1, thevortex diameter dx

+ decreases with increasing forcingfrequency, and the span between the vortices D+ decreaseswith increasing forcing frequency. However, the ratio dx

+/D+ increases as the forcing frequency increases. The

ratio dx+/D+ is a relative spatial fraction of the generated

vortices. When dx+/D+ has a higher value, the spanwise

vortices adjacent to the wall occupy a larger portion of theflow over the entire forcing period. Note that the cross-sectional area of the vortex is inversely proportional to theforcing frequency

Area / dþx� �2 / 1

f þ: ð6Þ

This is because the size of vortex is determined by themass injected during the blowing period. Since D+ isproportional to the convection velocity Uc

+ and inverselyproportional to f +, i.e.

Dþ / Uþcf þ

: ð7Þ

The spatial fraction of spanwise vortices obeys the relation

dþxDþ/

ffiffiffiffiffiffif þ

pUþc

: ð8Þ

Given that the convection velocity is almost independentof the forcing frequency (Table 1), the spatial fraction isproportional to the square root of the forcing frequency,dþx�

Dþ /ffiffiffiffiffiffif þ

p. Thus, the spatial fraction increases with

increasing forcing frequency. This suggests that, as theforcing frequency increases, the vortices generated by theforcing affect a wider region of the flow field.

Contour plots of the phase-averaged streamwise tur-bulent intensity (u¢+) together with phase-averagedvelocity vectors are illustrated in Fig. 7 for f +=0.044 andf +=0.088. A strong coherent structure is observed near theslot, and the coherent structure is concentrated on thespanwise vortex indicated by the velocity vectors. The sizeof this coherent structure increases as the flow developsdownstream, with the coherent structure gradually elon-gating to an oblique line in the streamwise direction.Closer inspection reveals that the inclined coherentstructure is lifted away from the wall. On moving furtherdownstream, the structure finally disappears. The inclinedshape of the coherent structure observed in the presentwork is similar to structures reported by Park et al. (2001)and Rhee and Sung (2001). The deformation of thestructure into an inclined shape is caused by the rotatingmotion of the spanwise vortex. A weaker vortex structureis observed above the strong coherent structure. Theweaker structure is engulfed into the wall by the strongvortex, where the front part of the coherent structure isslightly lifted away from the wall. The lifetime of thecoherent structure decreases with increasing forcing fre-quency, which can be attributed to the fact that the moreflow is supplied to the boundary layer at lower forcingfrequencies.

Table 1. Parameters of xz with forcing frequencies

f + (xz)max dx+ D+ Uc

+ dx+/D+

0.044 )2.57 24.5 130.4 5.73 0.1880.066 )2.37 20.3 85.6 5.65 0.2370.088 )2.28 17.9 63.4 5.58 0.282

Fig. 6a,b. Contours of xz and <U>)U0: a f +=0.044. b f +=0.088

702

Contour plots of the phase-averaged wall-normal tur-bulent intensity (v¢+) are shown in Fig. 8 along with phase-averaged velocity vectors. The forcing conditions are thesame as those used in the experiments presented above(i.e., f +=0.044 and f +=0.088). Similar to u¢+, the coherentstructure of v¢+ is concentrated on the spanwise vortexindicated by the velocity vectors. The large-scale structureof v¢+ is gradually lifted away from the wall in the down-stream and the size of the structure increases over time. Incontrast to the inclined shape observed in the distributionof u¢+, the structure of v¢+ has an elliptical shape. As theforcing frequency increases, the lifetime of the coherentstructure decreases.

Next, we consider the effect of forcing angle a on flowcharacteristics, where a is the blowing angle with respectto the horizontal wall (Fig. 1). The slot width was fixed at3 mm, which corresponds to about 20 wall units. Theforcing frequency was fixed at f +=0.088, the value thatgave the greatest reduction of skin friction in the presentstudy. When the forcing angle is a=60�, the blowingvelocity from the slot assists the flow in the streamwisedirection. For the forcing angle of a=120�, however, theblowing velocity opposes the streamwise flow. Figure 9shows the streamwise mean velocity profiles for forcingangles of a=60�, 90� and 120�. The forcing effect ata=120� is so strong that a reverse flow is formed near thewall at x+=20. This negative velocity caused by theblowing in the opposite direction recovers to positivestreamwise flow as the flow moves downstream (x+‡40).

For both a=60� and a=90�, the velocity profiles almostconverge to the undisturbed profile in the regions x+=80and x+=160. However, velocity retardation is still evidentat a=120�. At a=60�, no significant deformation of themean velocity is observed in any region, even at thelocation closest to the slot (x+=20).

The streamwise distributions of Cf/Cf0 are displayed inFig. 10 for the three forcing angles. In agreement with theresults presented above, the effect of using a forcing angleof a=60� is insignificant, with only a slight overshootoccurring after the forcing (x+‡40). On the other hand, theinfluence on Cf of a forcing angle of a=120� is significantover the entire region. A flow is separated in the region8 £ x+ £ 30, where the calculation of Cf is not available.The recovery of the flow from this separation is clearlyobserved in the region (x+ £ 40).

Contour plots of xz along with phase-averaged velocityvectors for a=60� and a=120� are shown in Fig. 11. Asdiscussed above, the vortex size is determined by the massinjected during the blowing period. Since the forcing fre-quency and the forcing amplitude are fixed, the injectedflow rate is the same for a=60� and a=120�. However, thevortex size is different for the two different angles becausethe streamwise velocity of the forcing depends on theforcing angles. During the blowing phase at a=120�, fluidinside the chamber is injected against the flow directionand meets with considerable resistance from the flow.Thus, the flow injected at a=120� takes a longer time tofollow the inherent flow than that injected at a=60�.

Fig. 7. Contour of phase-averaged streamwise turbulent intensityu¢+ together with phase-averaged velocity vectors for f +=0.044(a) and f +=0.088 (b)

703

Although the wall-normal velocity of the injected flow isthe same for a=60� and a=120�, the injected flow at a=120�penetrates farther into the mainstream than that at a=60�.Accordingly, the vortical structure is larger and stronger ata=120� than at a=60�. In the vicinity of the slot, a smallsecondary positive vortex is generated during the blowingperiod. This secondary vortex disappears rapidly becauseit is weaker than the main vortex. Comparison of thesecondary vortices observed at a=60� and a=120� inFig. 11 reveals that the location and lifetime of the sec-ondary vortex are different for the two forcing angles. Ata=60�, the secondary vortex is generated upstream of themain vortex (t=0) and then flows above the main vortex.On the other hand, although the secondary vortex ata=120� is generated in the upstream of the main vortex(t=0), it does not flow down but instead stagnates belowthe main vortex. Because the flow is injected against theflow direction at a=120�, the secondary vortex is accu-mulated in the vicinity of the wall and loses momentumrapidly as a result of interaction with the wall. Thus, thelifetime of the secondary vortex is shorter at a=120� thanat a=60�, with the secondary vortex at a=60� beingmaintained until t=4T/8.

The trajectories of xz and <U>)U0 for forcing angles ofa=60� and a=120� are plotted in Fig. 12, where xz wasmeasured at y+=11 and <U>)U0 at y+=6. Regardless of theforcing angle, the position of minimum xz is the same asthat of minimum <U>)U0. However, the strength of thespanwise vortex generated by the forcing at a=120� islarger than that at a=60�, and the wall-region velocity is Fig. 9. Streamwise mean velocity profiles for three forcing angles

Fig. 8. Contour of phase-averaged wall-normal turbulentintensity v¢+ together with phase-averaged velocity vectors forf +=0.044 (a) and f +=0.088 (b)

704

reduced to a greater extent at a=120� than at a=60�.Contour plots of xz and <U>)U0 for the two forcing an-gles (Fig. 13) were calculated to elucidate the convectivenature of the forcing. As time proceeds, the convectionvelocity of xz is almost the same as that of <U>)U0 forboth forcing angles. Similar to the results presented inFig. 6, the convection velocity changes slightly in thedownstream. However, the convection velocity is larger ata=60� than that at a=120�. Because the vortex generated ata=120� is stronger than that at a=60�, the wall-regionvelocity is reduced to a greater extent at a=120� than ata=60�. Therefore, forcing at a=120� more effectively re-duces skin friction than forcing at a=60�. Closer inspection

of the xz contours reveals that the pregenerated vortex ismore distinctive at a=60� than at a=120�.

The vortex size and other parameters for forcing anglesof a=60� and a=120� are listed in Table 2. These valueswere obtained at the phase when xz has a maximum value.The spatial fraction of the vortices (dx

+/D+) is calculated by

dþxDþ/

ffiffiffiffiffiffif þ

pUþc

: ð9Þ

Since the forcing frequency is fixed at f +=0.088, the spatialfraction is inversely proportional to the convectionvelocity

Fig. 10. Distributions of skin friction forthree forcing angles

Fig. 11. Contour of xz together with phase-averaged velocityvectors for a=60� (a) and a=120� (b)

705

dþxDþ/ 1

Uþc: ð10Þ

As shown in Table 2, the spatial fraction at a=120� is muchlarger than that at a=60�, which leads to a greater reduc-tion of the skin friction at a=120�.

Finally, contour plots of the phase-averaged longitudi-nal turbulent intensity u¢+ together with phase-averagedvelocity vectors are displayed in Fig. 14 for forcing anglesof a=60� and a=120�. For both forcing angles, a strongcoherent structure is observed near the slot, and thestructure is concentrated on the spanwise vortex indicatedby the velocity vectors. The coherent structure at a=120� isstronger and larger than at a=60�, and the vortex gener-ated at a=120� disappears more rapidly than that at a=60�.However, the coherent structure is maintained for a longertime at a=120� than a=60� because the forcing at a=120�opposes the mainstream and leads to the formation of areverse flow. Since the flow injected at a=120� penetratesdeeply into the mainstream, the size of the coherentstructure is larger at this forcing angle.

4ConclusionsThe effect of periodic blowing and suction on a turbulentboundary layer was investigated by means of PIV mea-surements. The local forcing was imposed through aspanwise thin slot. Three forcing frequencies (f +=0.044,0.066 and 0.088) were chosen at a Reynolds number ofReh=690. Spatio–temporal measurements of velocity andvorticity were made using PIV. The results showed a

reduction in the flow velocity in the vicinity of the wallbecause of the local forcing. On moving downstream theflow gradually recovered to the flow without forcing. Theeffect of blowing dominated the mean velocity profilesunder the local forcing conditions considered. Velocityretardation at the wall led to a reduction of skin friction,and the degree to which the skin friction was reducedincreased with increasing forcing frequency. The maxi-mum reduction of skin friction (approximately 75%) wasobserved at f +=0.088. The local forcing generated a strongspanwise vortex that created a reverse flow near the walland consequently reduced the wall-region velocity. As theforcing frequency was increased, vortices were generatedmore frequently, and as a result the distance betweenconsecutive vortices was reduced. The spatial fraction ofthe vortices is proportional to the square root of theforcing frequency. As the forcing frequency was increased,the lifetime of the coherent structures of u¢+ and v¢+decreased. Experiments using forcing angles of a=60�, 90�and a=120� showed that the strength of the spanwisevortex generated by forcing at a=120� is larger than thatgenerated using at a=60�. However, the convectionvelocity at a=60� is larger than that at a=120�. Forcing ata=120� more effectively reduced the skin friction thanforcing at a=60�.

Fig. 12. Trajectories of xz and <U>)U0 at the maximum vorticityphase for a=60� and a=120�

Fig. 13a,b. Contours of xz and <U>)U0: a a=60�. b a=120�

Table 2. Parameters of xz with forcing angles

a (degrees) (xz)max dx+ D+ Uc

+ dx+/D+

60 )2.10 18.7 75.2 6.62 0.24990 )2.28 17.9 63.4 5.58 0.282

120 )3.40 21.4 49.3 4.34 0.434

706

ReferencesAdrian RJ, Christensen K, Liu Z (2000) Analysis and interpretation of

instantaneous turbulent velocity fields. Exp Fluids 29:275–290Antonia RA, Zhu Y, Sokolov M (1995) Effect of concentrated wall

suction on a turbulent boundary layer. Phys Fluids 7:2465–2474Choi H, Moin P, Kim J (1993) Direct numerical simulation of tur-

bulent flow over riblets. J Fluid Mech 255:503–539Choi H, Moin P, Kim J (1994) Active turbulence control for drag

reduction in wall-bounded flows. J Fluid Mech 262:75–110Choi KS, Yang X, Clayton BR, Glover EJ, Atlar M, Semenov BN, Kulik

VM (1997) Turbulent drag reduction using compliant surfaces.Proc R Soc Lond 453:2229–2240

Chong MS, Perry AE, Cantwell BJ (1990) A general classification ofthree-dimensional flow fields. Phys Fluids A 2:765–777

Dallman U, Hilgenstock A, Riedelbanh S, Schulte-Werning B, Voll-mers H (1991) On the footprints of three-dimensional separatedvortex flows around blunt bodies. AGARD CP 494:75–110

Hussain AKMF, Reynolds WC (1970) The mechanics of an organizedwave in turbulent shear flow. J Fluid Mech 41:241–258

Jeong J, Hussain F (1995) On the identification of a vortex. J FluidMech 262:75–110

Jung WJ, Mangiavacci N, Akhavan R (1992) Suppression of turbu-lence in wall-bounded flows by high-frequency spanwise oscilla-tions. Phys Fluids A 4:1605–1607

Krogstad P, Kourakine A (2000) Some effects of localized injection onthe turbulence structure in a boundary layer. Phys Fluids 12:2990–2999

Park J, Choi H (1999) Effects of uniform blowing or suction from aspanwise slot on a turbulent boundary layer flow. Phys Fluids11:3095–3105

Park SH, Lee I, Sung HJ (2001) Effect of local forcing on a turbulentboundary layer. Exp Fluids 31:384–393

Rhee GH, Sung HJ (2001) Numerical prediction of locally-forcedturbulent boundary layer. Int J Heat Fluid Flow 22:624–632

Robinson SK (1991) Coherent motions in the turbulent boundarylayer. Ann Rev Fluid Mech 23:601–639

Sano M, Hirayama N (1985) Turbulent boundary layers with injectionand suction through a slit. Bull JSME 28:807–814

Tardu S (1998) Near wall turbulence control by local time periodicalblowing. Exp Thermal Fluid Sci 16:41–53

Tardu S (1999) Localized unsteady blowing and near wall turbulencecontrol. In: Proc of the 1st international symposium on the tur-bulence and shear flow phenomena, Santa Barbara, CA, 12–15September 1999, pp 399–404

Tardu S (2001) Active control of the near wall turbulence by localoscillation blowing. J Fluid Mech 439:217–253

Westerweel J (1994) Efficient detection of spurious vectors in particleimage velocimetry data. Exp Fluids 16:236–247

Zhou J, Adrian RJ, Balachandar S (1996) Autogeneration of nearwallvortical structures in channel flow. Phys Fluids 8:288–290

Zhou J, Adrian RJ, Balachandar S, Kendall TM (1999) Mechanisms forgeneration of coherent packets of hairpin vortices in channel flow.J Fluid Mech 387:353–359

Fig. 14. Contour of phase-averaged streamwise turbulentintensity u¢+ together with phase-averaged velocity vectors fora=60� (a) and a=120� (b)

707