flight control using synthetic jets on a cessna 182 model

7/31/2019 Flight Control Using Synthetic Jets on a Cessna 182 Model

1/12

Flight Control Using Synthetic Jetson a Cessna 182 Model

Marcus Ciuryla

Rensselaer Polytechnic Institute, Troy, New York 12180Yong Liu

Ohio University, Athens, Ohio 45701

John Farnsworth

Rensselaer Polytechnic Institute, Troy, New York 12180

Chiman Kwan

Intelligent Automation, Inc., Rockville, Maryland 20855

and

Michael Amitay

Rensselaer Polytechnic Institute, Troy, New York 12180

DOI: 10.2514/1.24961

The application of activeflow control via synthetic jet actuators for separation and roll control on a scaledCessna

182 model was investigated experimentally in a low-speed wind tunnel. The model was instrumented with eitherailerons or syntheticjets embedded withinthe outer portion of thewings span, in lieu of ailerons.Forceand moment

measurements were performed for various aileron deflections and synthetic jet momentum coefficients (on either

both wingtips or only on one). It was found that the effectiveness of the synthetic jets is comparable to that of

conventional ailerons at moderate deflection angles. The model was also instrumented with a hot-film shear stress

sensor downstream from the synthetic jet exit. The sensors rms output was monitored in real time by a computer.

When the rms reached a predetermined threshold value, the computer automatically turned on the synthetic jet

actuators. Using theappropriatethreshold value resulted in complete avoidanceof wingtip stall at theangleof attack

where separation would have occurred. In addition, the shear stress sensor and wind tunnel force data were used to

identify the system dynamics. A computerized dynamic model of an RC version of the Cessna 182 showed that at

moderate to high angles of attack, synthetic jets alone could be used to control the roll of the aircraft.

Nomenclatureb = wing spanCD = vehicle drag coefficientCL = vehicle lift coefficientCm = vehicle pitching moment (about quarter-chord)Cm = derivative ofCm with respect to the angle of

attackCn = vehicle yaw moment coefficient (about quarter-

chord and center plane)Cr = vehicle rolling moment coefficientC = synthetic jets momentum coefficient,

njjU2

j hLj=12

1U21 cb

c = mean chordfrV, fnV = coef ficients listed in Eqs. (4) and (5)g = acceleration due to gravityh = synthetic jet orifice width

Lj = synthetic jet orifi

ce lengthm = massnj = number of active synthetic jetsRe = Reynolds number, U1 c=T = reattachment time constantTf = time offlight, c=U1Tj = synthetic jet periodt = timeUj = synthetic jet orifice velocity (during the

outstroke), 1Tj

R

0 ujt dtU1 = freestream velocityub = x-component of inertial velocity in airplane

body frame FBujt = phase-averaged velocity at the jet exit planeV = input voltage to the synthetic jets (before

amplification)wb = z-component of inertial velocity in airplane

body frame FBxsj = streamwise location of the synthetic jet actuators = angle of attackCr = change in rolling moment coefficient with

respect to aileron deflection of 3 degCrsj = change in rolling moment coefficient from

synthetic jet configuration (jets off)a = aileron deflectionE = elevator commandR = rudder commandSV = synthetic jet commandTh = engine command = closed-loop control system characteristics rootj = synthetic jet density1 = freestreamfluid density = synthetic jet blowing time, Tj=2

Received 4 May 2006; revision received 1 August 2006; accepted forpublication 11 August 2006. Copyright 2006 by the American Institute ofAeronautics and Astronautics, Inc. All rights reserved. Copies of this papermaybe made forpersonal or internal use, on condition that the copier pay the$10.00 per-copy fee to the Copyright Clearance Center, Inc., 222 RosewoodDrive, Danvers, MA 01923; include the code 0021-8669/07 $10.00 incorrespondence with the CCC.

Graduate student, Mechanical, Aerospace and Nuclear EngineeringDepartment.

Graduate student, School of Electrical Engineering and ComputerScience. Student Member AIAA.

Undergraduate student, Mechanical, Aerospace and Nuclear EngineeringDepartment. Student Member AIAA.

Vice President, 15400 Calhoun Drive, Suite 400.Assistant Professor, Mechanical, Aerospace and Nuclear EngineeringDepartment; [email protected] (corresponding author). Senior MemberAIAA.

JOURNAL OF AIRCRAFTVol. 44, No. 2, MarchApril 2007

642
http://dx.doi.org/10.2514/1.24961http://-/?-http://-/?-http://-/?-http://-/?-http://dx.doi.org/10.2514/1.24961


2/12

Introduction

O PTIMUM aerodynamic performance that avoids flowseparation on wing surfaces has been traditionally achievedby appropriate aerodynamic design of airfoil sections. However,when the wing design is driven by nonaerodynamic constraints(stealth, payload, etc.), the forces and moments of the resultingunconventional airfoil shape may be much smaller than aconventional airfoil. Therefore, either active or passive flow controlcan be used to maintain aerodynamic performance throughout the

normal flight envelope. Although passive control devices such asvortex generators have proven, under some conditions, to be quiteeffective in delaying flow separation, they offer no proportionalcontrol and introduce a drag penalty when theflow does not separate(or when they are not needed) [1].

In contrast, active control enables coupling of the control input toflow instabilities that are associated with flow separation and thusmay enable substantial control authority at low actuation levels.Furthermore, active actuation is largely innocuous except whenactivated and has the potential for delivering variable power. Inprevious studies, active control efforts have employed a variety oftechniques including external and internal acoustic excitation [24],vibrating ribbons orflaps [5], and steady or unsteady blowing [6].

Over the last decade, the synthetic jet has emerged as a versatileactuator for active flow control. The formation and evolution ofsynthetic jets are described in detail in the work of Smith and Glezer[7], Amitay and Glezer [8], and Amitay and Cannelle [9]. Theeffectiveness offluidic actuators based on synthetic jets is derivedfrom theinteraction of these jets with theflowneartheflow boundarythat can lead to the formation of a quasi-closed recirculating flowregion, resulting in a virtual modification in the shape of the surface.Past research work has focused on the use of open-loop actuationstrategies to generate the required modulated input signals to jetarrays [6,1012], which is highly dependent on the availability ofaccurate and comprehensive wind-tunnel-validated flow models.However, the underlying flow mechanisms and interactions of jetarrays are usually very complicated and highly nonlinear. Moreover,it is extremely difficult to accurately model the changes in the systemdynamics due to varied flight conditions, inevitable external

disturbances, measurement noise, actuator anomalies, and failures.Therefore, the development of closed-loop nonlinear adaptive flowcontrol techniques that can automatically compensate for modelingerrors and adapt to changes in the system dynamics, are particularlyattractive to realize the full potential of synthetic jet technology.

Recently, the effectof synthetic jets ona scaled model of a Piper J-3 Cub was examined [13] at Reynolds numbers from 150,000 to280,000. Increases in totalvehicle lift of up to16% were observed. Adynamic analysis on the Piper showed that synthetic jets had only aminor influence on stability characteristics, and that the effect wasgreater at high angles of attack. The previous work also showed thatsynthetic jets on the main wing of an aircraft could be used to controlthe pitch of the aircraft to a modest extent.

The present work has three objectives: 1) to explore,experimentally, the feasibility of using active flow control, via

synthetic jet actuators, as a roll control mechanism instead ofconventional ailerons for the small-scale Cessna 182 model; 2) todevelop a closed-loop stall suppression system on a wind tunnelmodel of the Cessna 182; and 3) to develop an integrated flightcontroller (body rate controller) for the RC Cessna flying model,instrumented with synthetic jet wingtips.

Experimental Setup

The experiments were conducted in a closed-return low speedwind tunnel, having a test section measuring 608 by 608 mm, whichis equipped with a six-component force sting balance to measure theaerodynamic forces and moments. All forces and moments reportedin the paper are about the quarter-chord of the wing along the modelcenterline. The resolution of the sting balance was such that themoments andforceswere measured with an accuracyof less than 5%.The experiments were conducted at 15 and 30 m=s, with acorresponding Reynolds number of 67,300 and 134,600,

respectively, and (wing) angles of attack from 0 to 16 deg. Theuncertainty in the velocity is less than 1% whereas the angle of attackalignment is within 0.1 deg.

A 1=24th Cessna scaled model (457 mm span) was constructedfrom stereolithography, where the main wing consists of a NACA2412section,and a NACA 0005was used for boththe horizontal andvertical tails. The main wing was designed such that differentwingtips could be used (Fig. 1) where aileron deflections from 0 to18 deg in 3 deg increments could be achieved. Note that each

ailerons streamwiselengthis 20% ofthe rootchord ontheouter spanportions (25% of the span for each aileron) of the wing.In addition to the aileron deflection wingtips, several wingtips

with embedded synthetic jets were also designed (Fig. 1b). Asynthetic jet is a jet that is synthesized by a train of two-dimensionalvortexpairs. The vortices areformed atthe edgeof an actuator orificeby the motion of a small diaphragm that is driven by a piezoceramicdisc andmounted at thebottom of a sealedshallow cavity. Duringtheforward motionof thediaphragm,fluidis ejectedfrom thecavity, andthe ensuingflowseparatesat the sharp edgeof the orifice forming twovortex sheets that roll into a vortex pair that begins to move awayfrom the orifice under its own self-induced velocity. When thediaphragm begins to move away from the orifice, the vortex issufficiently removed and is thus unaffected by the ambientfluid thatis drawn into the cavity. Therefore, during each cycle the net mass

flux out of the cavity is zero whereas the hydrodynamic impulse ofeach vortex is nonzero. The diaphragm is driven at its resonance andthus the electrical power input to the actuator is typically small.Furthermore, the synthetic jet is very simple mechanically and does

Fig. 1 Cessna wind tunnel model.

CIURYLA ET AL. 643
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


3/12

notrequire compressed air or tubing. For a detailed description of thesynthetic jet see the review by Amitay and Glezer [8]. In the presentwork the synthetic jet was mounted at xsj= c 0:25 (near theseparation point for high angles of attack,Fig. 2), and its synthetic jetperformance was measured using the momentum coefficient, C,which was varied between 1:5 104 and 8:7 103 (when bothwingtips are actuated) and between 7:4 105 and 4:3 103 (whena single wingtip is actuated).

The baseline Cessna has a small taper over the aileron portion ofthe span. However, to avoid any three-dimensional taper effects, thesynthetic jet wingtips were designed without any taper

( C 67 mm), but with the same planform area and overall span asthe aileronwingtips. The synthetic jet-instrumented wingtips have an18% thick Clark-Y airfoil section (the effect of flow control, viasynthetic jet actuators, on a 2-D Clark-Y airfoil was previouslyinvestigatedby Amitayet al.[14]). A thicker airfoil(thanthe baselinewingtip) was chosen to provide enough room for the synthetic jetcavities. Moreover, a stall fence was added to the synthetic jetwingtips to prevent anytip effectsfrom imposing on the inner span ofthe wing. Each wingtip consists of 3 synthetic jets (at the 25% chordlocation and7.5 mm apart, Fig. 3) each havinga 0.5 mmwideslit andextending 25.4 mm along the span. Each synthetic jet actuator is

driven at a frequency of 750 Hz using a piezoceramic disk. Inaddition, a hot-film shear stress sensor was placed downstream thesynthetic jet exit (at 35% chord, Fig. 3).

Results

The implementation of synthetic jet actuators for separationcontrol and as ailerons for roll control on a scaled Cessna 182 modelhas been investigated experimentally. First, the baseline configura-tion is discussed, where ailerons are used for roll. Following, theeffectiveness of synthetic-jets-based active flow control instead ofthe ailerons is presented, and finally, the development of anintegrated flight controller for an RC Cessnaflying model, which isinstrumented with synthetic jets wingtips, is shown.

Aileron Configuration

First, the aerodynamic performance of the baseline Cessna model(i.e., with conventional ailerons) is presented. Figures 4a and 4b

presentthe variationof thevehicles lift coefficient,CL, with angle ofattack and the lift-drag polar, respectively. The data are at Re 67; 300 and 134,600 and four aileron deflection angles. The liftcoefficient increases linearly for < 8 degand reachesits maximumvalue, CL max, of 1.2 at 11 deg (Fig. 4a). As expected, the liftcoefficient is not noticeably affected by aileron deflection. However,the drag coefficient increases slightly with increasing ailerondeflection (Fig. 4b). Furthermore, there is no noticeable effect due toReynolds number at this range of Reynolds numbers.

Thepitchingmoment characteristics for thebaselineconfigurationwere also measured and the change of the pitching moment (aboutthe wings quarter-chord) with the angle of attack is presented inFig. 5. As expected, the moment on the entire vehicle is negative (e.g., stable vehicle), where the moment coefficient increases (i.e.,becomes more negative) linearly with angle of attack (the derivativeofCm with respect to the angle of attack, Cm , is0:037 per degree).

xsjc

Piezo disk

Cavity

xsjc

Piezo disk

Cavity

Fig. 2 Cross section schematics of the wingtip instrumented with

synthetic jet actuators.

Synthetic jets

Shear stress sensor

Synthetic jets

Shear stress sensor

Fig. 3 Cessna model with synthetic-jets-instrumented wingtip.

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 4 8 12 16

0 134,600

6 134,600

12 134,600

18 134,600

0 67,300

6 67,300

12 67,300

18 67,300

CL

a Re

0

0.1

0.2

0.3

0.4

0 0.2 0.4 0.6 0.8 1 1.2 1.4

CL

CD

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 4 8 12 16

0 134,600

6 134,600

12 134,600

18 134,600

0 67,300

6 67,300

12 67,300

18 67,300

CL

a Re

0

0.1

0.2

0.3

0.4

0 0.2 0.4 0.6 0.8 1 1.2 1.4

CL

CD

0

0.1

0.2

0.3

0.4

0 0.2 0.4 0.6 0.8 1 1.2 1.4

CL

CD

a)

b)

Fig. 4 a)Lift coefficient vsangleof attack; b)drag polarfor thebaselineCessna.

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0

0 2 4 6 8 10 12 14 16

0 134,600

6 134,600

12 134,600

18 134,600

0 67,300

6 67,300

12 67,300

18 67,300

Cm

a Re

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0

0 2 4 6 8 10 12 14 16

0 134,600

6 134,600

12 134,600

18 134,600

0 67,300

6 67,300

12 67,300

18 67,300

Cm

a Re

Fig. 5 Pitching moments vs angle of attack at different ailerondeflections.

644 CIURYLA ET AL.
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


4/12

Similar to the earlier results, neither the Reynolds number nor theaileron deflection angle has a meaningful effect on Cm.

Next,thechangeintherollcoefficient with theaileron deflection ispresented in Figs. 6a and 6b for Re 67; 300 and 134,600,

respectively. Here, the roll coefficient at an aileron deflection angleof 3 deg (for each angle of attack) is subtracted from the rollcoefficient such that ata 3 deg, Cr 0 for all angles of attack.This representation makes it easy to display the effect of ailerondeflection on Cr for different . The effects are similar at bothReynolds numbers tested; for a given , as the aileron deflectionincreases, the rolling moment also increases. For low to moderateangles of attack ( < 9 deg) the increase ofCr with a is the largest(and it is linear). At > 9 deg the flow over the wing separates,resulting in a decrease in aileron effectiveness.

To properly formulate a mathematical dynamic model of thevehicle, the relationship between the rolling moment and the ailerondeflection is needed. Therefore, the data for angles of attack from 0 to9 deg are replotted and a linearfit found to be the bestfit for the datawhere dC

r= d

a 0:025 (Fig. 7). Note that these data are used to

develop the dynamic model of the vehicle.

Synthetic Jet Configuration

Next, the effectofflow control, via synthetic jet actuators, onflowseparation and roll control was investigated and is presented in thissection. Figure 8ashows the effect of the momentum coefficient ofthe synthetic jets, C, on the lift coefficient at different angles ofattack and for Re 67; 300. For low angles of attack ( < 3 deg)there is hardly any effect offlow control (the flow is fully attachedover the wingtips, whereas for higher angles of attack the activationof thesynthetic jets results in an increase in thelift coefficient, even atC aslowas 5:8 10

4).Notethat for3 < < 6 deg theflowonthewingtips is partially separated; thus, the effect of the control is thesame regardless of the momentum coefficient. However, at moderateto high anglesof attack, where theflow is massivelyseparated(on theentire wing) theeffect of the momentum coefficient is clearlyvisible:as C increases the lift coefficient increases, suggesting that

proportional control can be achieved. This is very important if

closed-loop control is to be implemented.As mentioned before, the airfoil used for the synthetic jet wingtips

is 18% thick Clark-Y, which does not have the same aerodynamicperformance as the baseline airfoil (NACA 2412). Thus, it isexpected that the lift coefficient of the Cessna will be smaller whenthe synthetic-jet-instrumented wingtips are used (without flowcontrol) as shown in Fig. 8b. Figure 8b presents a comparison of thelift coefficient (for different angles of attack) for the baseline(aileron) configuration, and the synthetic jet configuration (with andwithout activation of the synthetic jets). Without activation of the

0

0.15

0.3

0.45

0 4 8 12 16

Cr

a

0

0.15

0.3

0.45

0 4 8 12 16

0

5

8

9

10

16

Cr

a

0

0.15

0.3

0.45

0 4 8 12 16

Cr

a

0

0.15

0.3

0.45

0 4 8 12 16

Cr

a

0

0.15

0.3

0.45

0 4 8 12 16

0

5

8

9

10

16

Cr

a

0

0.15

0.3

0.45

0 4 8 12 16

0

5

8

9

10

16

Cr

a

a)

b)

Fig. 6 Change of the roll moment (with respect to the value at

a 3 deg) vs the deflection angle at different angles of attacks;a) Re 67; 300 and b) 134,600.

0

0.15

0.3

0.45

0 4 8 12 16

a

Cr

0671.0025.0 = arC

0

0.15

0.3

0.45

0 4 8 12 16

a

Cr

0671.0025.0 =

arC

Fig. 7 Change of the roll moment (with respect to the value at

a 3 deg) vs the deflection angle for 5 9 deg.

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 2 4 6 8 10 12 14 16

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 2 4 6 8 10 12 14 16

CL

C= 5.8x10-4

C= 1.61x10-3

C= 3.48x10-3

C= 8.6x10-3

C= 0

x

CL

C= 8.6x10-3

C= 0

Baseline (ailerons)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 2 4 6 8 10 12 14 16

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 2 4 6 8 10 12 14 16

CL

C= 5.8x10-4

C= 1.61x10-3

C= 3.48x10-3

C= 8.6x10-3

C= 0

x

CL

C= 8.6x10-3

C= 0

Baseline (ailerons)

a)

b)

Fig. 8 Lift coefficient vs angle of attack; a) synthetic jets wingtips andb) comparison with the baseline Cessna. Re 67; 300.

CIURYLA ET AL. 645
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


5/12

jets, the lift coefficient of the synthetic jet configuration has worseperformance than thebaselineconfiguration for > 2 deg. When the

synthetic jets are activated (with C 8:6 103

) the lift coeffi

cientis recovered (to the baseline values) for < 6 deg whereas for > 6 deg the lift coefficient surpasses the baseline values by up to15%. Furthermore, the stall angle is increased by 2 deg.

The effect offlow control on the pitching moment coefficient wasalso measured as a function of the angle of attack and is presented inFig. 9 for various synthetic jet momentum coefficients. There is adefinite trend that as the momentum coefficient is increased, thevehicle moment coefficient is more negative (i.e., more stablevehicle). Moreover, proportional pitching control is obtained for > 6 deg.

In the data presented in Figs. 8 and 9 the synthetic jets on bothwingtips were activated; however, to obtain a rolling moment thesynthetic jets on either the left or the right wingtip were activatedseparately. Figure 10 shows the change in the roll coefficient (with

respect to the corresponding values when the synthetic jets are notactivated) with the synthetic jet momentum coefficient. The dashedlines correspond to the starboard synthetic jets activated, and thesolid lines represent the port synthetic jets activated. At 0 deg avery small roll moment is obtained and it is similar for all themomentum coefficients that were tested. When the angle of attack isincreased to 6 deg proportional control is obtained, where as Cincreases Cr decreases. Note that similar trendswere obtained at all

angles between 5 and 9 deg, and the 6 deg is shown here as arepresentative case. At higher angles of attack, where the flow iscompletely separated over the wingtips (withoutflow control), theeffect of the momentum coefficient is different. At 10 deg, forlow C, increase in C yields a higher rolling moment whereas forC > 1:73 10

3, as C increases Cr decreases. At angles ofattack past thestall angle, low momentum coefficient isnot sufficientto reattach the flow over the wingtip and thus low rolling moment isachieved. However, high momentum coefficient jets can reattach the

flow and create a rolling moment. It is noteworthy that themagnitudes of the rolling moments obtained with synthetic jetactuators is similar to those obtained with ailerons at deflectionangles of 12 deg and smaller.

Closed-Loop Active Stall Suppression

In this section a simple closed-loop control is used to suppressseparation over the wingtips using a dynamic shear stress sensor todetect the separation and synthetic jet actuators to reattach the flow.The separation was observed qualitatively, using tuft flowvisualization, and detected quantitatively, using the shear stresssensor. Figures 11aand 11b present the time trace of the shear stresssensor output voltage, and the tuft visualization over the wingtip,

respectively, at 0 deg. At this low angle of attack thefl

ow iscompletely attached over the wingtip as indicated by the tufts and thevery low rms shear stress sensor output. Note that the shear stresssensor was not calibrated and is used here only to detect separationbased on the rms of its output signal.

Asthe angle ofattackis increasedto 8:5 deg theflowoverthewingtips is separated as indicated by the high rms values from theshear stress (Fig. 12a)andconfirmed by thetufts,which clearlyshowreversed flow over parts of the wingtip (Fig. 12b). The time trace ofthe shear stress sensor exhibits two main frequencies of 10 and100 Hz, which correspond to themodel vibrationand thesheddingfrequency of the separated flow over the wingtip, respectively. Notethat visual examination of the wingtips at these conditions confirmedthe buffeting of the wingtips, withoutflow control.

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0

0 2 4 6 8 10 12 14 16

Cm

C= 5.8x10-4

C= 1.61x10-3

C= 3.48x10-3

C= 8.6x10-3

C= 0

x

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0

0 2 4 6 8 10 12 14 16

Cm

C= 5.8x10-4

C= 1.61x10-3

C= 3.48x10-3

C= 8.6x10-3

C= 0

x

Fig. 9 Pitching moments vs angle of attack at different momentumcoefficients.

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0 0.002 0.004 0.006 0.008 0.01

06

1014

06

1014

C

Cr

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0 0.002 0.004 0.006 0.008 0.01

06

1014

06

1014

C

Cr

Fig. 10 Change of the roll moment (with respect to the unforced case)vs the momentum coefficient at different angles of attack. The dashed

lines correspond to the starboard synthetic jets activated whereas thesolid lines represent the port synthetic jets activated.

-11

-9

-7

-5

-3

-1

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

t(s)

ShearStressSensorOutput(V)

-11

-9

-7

-5

-3

-1

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

t(s)


a)

b)

Fig. 11 a) Time trace of the shear stress sensor output; b) tufts flowvisualization over the wingtip. 0 deg, synthetic jets off.

646 CIURYLA ET AL.
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


6/12

When the synthetic jets are activated with C 8:6 103 the

flow is completely reattachedas indicatedby thelow rmslevels of theshear stress sensor (Fig. 13a) and the tufts (Fig. 13b). Note thepresence of thelow-amplitude periodic oscillationsin theshear stresssensor signal at the synthetic jets actuation frequency, due to its

proximity to the synthetic jet actuators. Moreover, the actuation ofthe synthetic jet eliminated completely the low frequency modelvibrations (as was visually confirmed). Also, in the anemometrysystem used in these experiments, decrease in the output voltage ofthe sensor corresponds to increase in the shear stress; thus, when thesynthetic jets are activated the dc level of the shear stress sensor ismuch lower than that of the uncontrolled case (12a), which isindicative of increase of the shear, as expected when the flow isreattached. From the shear stress data it was concluded that either thesensors dc level or its rms (or both) can be used to detect separation.In the present work we used the rms as the separation indicator.

From the data in Figs. 1113 it is clear that a stall detection systemcan be implemented simply by continuously monitoring the rms ofthe shear stress sensor reading. Thus, a simple closed-loop stallsuppression system was implemented by choosing a threshold rms

value which,if reached, will trigger theactivation of thesynthetic jetsto reattach the flow. In what follows, the angle of attack was slowlyincreased until stall occurred. Then, based on the rms thresholdcriterion, thesynthetic jets wereautomatically activated through a D/A board, using a LabVIEW code.

Figures 14a14c present the time trace of the shear stress sensoroutput as the angle of attack increases monotonically for differentrms thresholds. Note that in these figures the time axis has beenarbitrarily set to zerowhen theflowis onthe verge of separationfromthe airfoils upper surface. In the first experiment, the rms thresholdwas set to 0.5 V (Fig. 14a). As time progresses (i.e., as the angle ofattack increases) the shear stress output voltage increases(corresponding to a decrease in the shear stress). When the angleof attack is 8 deg (t 0) the flow separates, which is indicated bythe increase in the rms as well as by the jump in shear stress reading(marked by the arrow). It takes about 2.5 s before the control istriggered and the synthetic jets are activated, resulting in a completeflow reattachment, as indicated by the significant decrease in the

shear stress sensor output voltage. The figure also includes a sketchof theroll moment(from thestingbalance reading), which shows thatwhen theflow is separated a roll moment is generated (due to a slightasymmetryof themodelthat yieldsan asymmetricflow separationonthe wings). When the synthetic jets are activated the roll moment

diminishes due to flow reattachment on both wings, resulting in asymmetric flow over the wingtips.

Next, the rms threshold was reduced to 0.25 V (Fig. 14b). Again,the flow separates and it takes 0:5 s before the synthetic jets areactivated (compared to 2.5 s for the 0.5 V rms threshold case). Whenthe rms threshold is reduced to 0.1 V (Fig. 14c), the threshold wasreached before the wingtips ever stalled. This suggests that throughthe selection of an appropriate shear stress rms threshold, syntheticjets can not only recover a wing from stall, but avoid it altogether.Furthermore, a roll moment is not generated.

Dynamic Model

As part of the development of closed-loop control system, thedynamic response of the model following the activation of the

synthetic jet is required. Here, the analysis was conducted on an RCflying Cessna 182 model (1.65 m span and 1.27 m total length) basedon the scaled wind tunnel model data. Ideally, the dynamic responseis found by measuringthe aerodynamic force andmoment responses.However, this requires dynamic force/torque sensors, which werenot available for these tests. Instead, the shear stress sensor outputwasused,whichindicates thestateof theflow andis closely relatedtoaerodynamic forces andmoments.Thus, thedynamic response of theshear stress sensor was used to characterize the aerodynamicresponse of the entire vehicle. Experiments were conducted todetermine the dynamic response of the shear stress when a step inputto the synthetic jets was applied. The signal was on for 0.5 s and offfor 0.5 s and the transient response of the shear stress following theactivation and termination of the synthetic jets was measured byphase locking the data acquisition to the onset of the synthetic jetsinput signal.

A typical response of theshear stress sensorto the step input signalis presented in Figs. 15aand 15b (the onset and termination times of

-11

-9

-7

-5

-3

-1

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7t(s)

Shea

rStressSensorOutput(V)

-11

-9

-7

-5

-3

-1

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7t(s)

Shea

rStressSensorOutput(V)

a)

b)

Fig. 12 a) Time trace of the shear stress sensor output; b) tufts flowvisualization over the wingtip. 8:5 deg, synthetic jets off.

-11

-9

-7

-5

-3

-1

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

t(s)


-11

-9

-7

-5

-3

-1

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

t(s)


a)

b)

Fig. 13 a) Time trace of the shear stress sensor output; b) tufts flowvisualization over the wingtip. 8:5 deg, synthetic jets on(C 8:6 10

3).

CIURYLA ET AL. 647
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


7/12

the synthetic jets are marked on the trigger signal). In theseexperiments the angle of attack is 12 deg and the Reynolds number is134,600. At this angle of attack the flow over the wingtips isseparated and the activation of the control results in a decrease in theshear stress sensor output voltage (which corresponds to an increaseinthe shear). It takes about 0.3 s for theflow to completely reattach tothe surface (following the activation of the synthetic jets) and aboutthe same time for the flow to reseparate once the synthetic jets areturned off. Note that the response of the shear stress sensor resemblesthat of afirst-order system. Furthermore, the sensor response has acomponent of the synthetic jet driving signal frequency (750 Hz) andits harmonics. Therefore, in the analysis conducted here, a low passfilter with cutoff frequency at 400 Hz was implemented (Fig. 15b).Note that the time-averaged shear stress varies for different windtunnel speeds, angles of attack, and different input signals; however,its transitory behavior in all experiments conducted is very similar.The response of the shear stress to the activation of the synthetic jetscan be described as afirst-order system as follows:

1

TS 1(1)

where T is the time constant.From the experimental data, the time constantT is approximated

by the time it takes the shear stress sensor reading to reach 63% of itsfinal value, which, in the present experiments, is T 0:04 s. Note

that this time definition is commonly used in dynamic stabilityanalysis. Furthermore, the characteristic time of the flow (i.e., thetime of flight, Tf c=U1) is 0.026 and 0.052 s for Reynoldsnumbers of 67,300 and 134,600, respectively. Therefore, the timeconstant for the reattachment and reseparation (using synthetic jetactuators) can be approximated by the characteristic time of theflow.

From thewind tunnelexperiments, thesynthetic jetwingtip modelcan be described as

Cr frV1

TS 1(2)

Cn fnV1

TS 1(3)

where frV and fnV are functions presented in Eqs. (4) and (5),respectively, and V is the input voltage to the port wings syntheticjets.

Cr frV 0:016V 0:0102V2 0:0027V3 (4)

Cn fnV 0:2164 103V 0:294 103V2

0:2349 103V3 (5)

From theprecedingapproximations a highfidelity dynamic modelfor both the baseline Cessna and the Cessna with synthetic jetwingtips was developed and implemented in Simulink. Note that thevalues are scaled up to a 1.65 m wing span RC Cessna model, whichwill be used in the future as the testbed for the technology presentedin this paper. The structure of the simulation model is illustrated inFig. 16. The aerodynamic coefficients that were not measured in thewind tunnel experiments were approximated by a mathematicalmodel of a Cessna 182 [15], and they will be replaced in the futurewith empirical data when additional experiments are conducted.

The 6-DOF Cessna dynamic model (Fig. 16) is described byEqs. (615) for the rotational and translational dynamics,respectively.

0 2 4

-1 0

-8

-6

-4

-2

0

2

4

6

8

1 0

0 2 4- 1 0

-8

-6

-4

-2

0

2

4

6

8

1 0

0 2 4- 1 0

-8

-6

-4

-2

0

2

4

6

8

1 0

Roll torque

reading

Onset of

Separation

Activation of SJ

Complete

reattachment

Roll torque

Reading

Roll torque

reading

t(s)


0 2 4

-1 0

-8

-6

-4

-2

0

2

4

6

8

1 0

0 2 4- 1 0

-8

-6

-4

-2

0

2

4

6

8

1 0

0 2 4- 1 0

-8

-6

-4

-2

0

2

4

6

8

1 0

Roll torque

reading

Onset of

Separation

Activation of SJ

Complete

reattachment

Roll torque

Reading

Roll torque

reading

t(s)


a)

b)

c)

Fig. 14 Changeof theshear stresssensor output with time (asthe angleof attack increases). a) rms 0:5, b) 0.25, and c) 0.1 V.

0 0.2 0.4 0.6 0.8 1-6

-4

-2

0

2

4

6

t(s)


Trigger

signal

On Off

-6

-4

-2

0

2

4

6

0 0.2 0.4 0.6 0.8 1-6

-4

-2

0

2

4

6

t(s)


Trigger

signal

On Off

-6

-4

-2

0

2

4

6

Trigger

signal

On Off

-6

-4

-2

0

2

4

6

a)

b)

Fig. 15 Response of the shear stress sensor output following theactivation and termination of the synthetic jets. 12 deg,C 8:6 10

3, Re 134; 600. a) Unfiltered; b) filtered.

648 CIURYLA ET AL.
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


8/12

_ p q sin tan r cos tan (6)

_ q cos r sin (7)

_ sin sec r cos sec (8)

_p Ippqpq Ipqrqr g

pl Tl g

pn Tn (9)

_q Iqppp2 Iqrrr2 I

qprpr g

qmTm (10)

_r Irpqpq Irqrqr g

rl Tl g

rnTn (11)

_u

1

m Fx mgS qw rv (12)

_v 1

mFy mgCS ru pw (13)

_w 1m Fz mgCC pv qu (14)

_x_y

_z

24

35

CC SSS CS CSC SSCS SSS CC CSS SCS SC CC

24

35

uv

w

24

35

(15)

Figure 17 presents the diagram of the Simulink simulation model.The actuator commands refer to engine Th, aileron a, elevator E,and rudder R commands for the baseline configuration. For thesynthetic jets configuration, the aileron command is replaced by aninput voltage SV, where SV is defined as

6DOF

Dynamic MotionEquation

Propulsion System Model

Aerodynamic Model

Synthetic Jets

Actuator Model

Control SurfaceActuator Model

Fig. 16 Cessnas 6-DOF dynamic model.

WIND &TURBULENCE

0

Turbulence(1/0)

0

Constant wind field(1/0)

Control Cmd

Wind

body rate

Euler Angle

Body Speed

Position

Aerodynamic

Cessna 182 Simulation

ActuatorDynamic

Actuator Dynamic

1

Actuator Command

WIND &TURBULENCE

0

Turbulence(1/0)

0


Control Cmd

Wind

body rate

Euler Angle

Body Speed

Position

Aerodynamic


ActuatorDynamic

Actuator Dynamic

1

Actuator Command

Fig. 17 Diagram of the Simulink simulation model for the RC Cessna.

CIURYLA ET AL. 649


9/12

SV 2 5; 5 V (16)

If SV > 0 the starboard wingtip jets are activated with an inputvoltage, and the port wingtip jets are off. IfSV < 0 the port wingtipjets are switched on with an input voltage ofSV, and the starboardwingtip jets are off.

In this simulation system, at high angles of attack, the aileroncoefficients for the baseline case are represented by spline functions,obtained fromthe experiments. Thepropulsion model is described asThrust Th mg, where Th is the unitless number representing thethrottle setting for the engine.

The dynamics of actuators are approximated by independentfirst-

order systems with time constants, whereas aileron, elevator, andrudder time constants are estimated from servo motor properties andare presented in Table 1. The synthetic jet wingtips time constant is

scaled up from the experimental results to the flying model scale(having 1.5 m wing span).

Integrated Flight Control with Flow Control System

Based on the Cessna dynamic model, an integrated flightcontroller (body rate controller) was designed for the Cessna that isinstrumentedwith syntheticjet wingtips. A body ratecontroller is thebasic building block for a future autonomous flight control system.The designed body rate controller can follow a given command byemploying the synthetic-jets-controlled wingtips, rudder, andelevator.

The structure of the control system is illustrated in Fig. 18. Thenominal controller is based on feedback linearization. It is

augmented by an adaptive neural network controller to compensatefor themodel uncertainty. In the currentsimulation model, it is foundthat the nominal controller by itself is very robust, thus the neural

Table 1 Actuator simulation time constants

Time constants, s

Aileron 0.06Elevator 0.06Rudder 0.06Synthetic jet wingtips 0.075Engine 0.6

Nominal Controller

Reference

ModelLinear

Compensator

Dynamic

Inverse

Control

Allocation

6DOF

Cessna

Model

command

Command derivative

Neural

Network

v0

p, q, r

p, q, r, v0

vk

vad

Nominal Controller

Reference

ModelLinear

Compensator

Dynamic

Inverse

Control

Allocation

6DOF

Cessna

Model

command

Command derivative

Neural

Network

v0

p, q, r

p, q, r, v0

vk

vad

Fig. 18 Structure of the closed-loop control system.

Table 2 Controller parameters

Command filter and controllerclosed-loop characteristicbandwidth

Proportional gain(Kp)

Integral Gain(KI)

P channel 2 2.828 4q channel 1 1.414 1r channel 1 1.414 1

Position

Body speed

Wind &Turbulence

0

Turbulence(1/0)

Aerodynamic

To Workspace4

Position

To Workspace3

Body speed

To Workspace2

Euler angle

To Workspace1

Body rate

To Workspace

phi

Va

Command Derivative

Command

References

K*u

Gain1

K*u

Gain

Euler Angle

0


Control Command

Wind

Body rate

Euler Angle

Body Speed

Position

Aerodynamic


Command Derivative

Body rate Command

Body rate measurement

Aerodynamic

Contr. outputs

Controller

ActuatorDynamic

Actuator Dynamic

Position

0

Turbulence(1/0)

Aerodynamic

To Workspace4

Position

To Workspace3

Body speed

To Workspace2

Euler angle

To Workspace1

Body rate

To Workspace

Va

Command Derivative

Command

K*u

Gain1

K*u

Gain

Euler Angle

0


Control Command

Wind

Body rate

Euler Angle

Body Speed

Position

Aerodynamic


Command Derivative

Body rate Command

Body rate measurement

Aerodynamic

Contr. outputs

ActuatorDynamic

Actuator Dynamic

Fig. 19 Simulink simulation of the integrated flight control using synthetic jets.

650 CIURYLA ET AL.


10/12

network adaptive controller was not engaged. It is also worth notingthat although the controller is only for body rate, the full 6-DOFdynamics is simulated to evaluate the controller performance.

The Cessna body rate dynamics can be rewritten as

_p Ippqpq Ipqrqr g

pl Tlaero Tlctrl g

pnTnaero Tnctrl (17)

_q Iqppp2 Iqrrr2 I

qprpr g

qmTmaero Tmctrl (18)

_r Irpqpq Irqrqr g

rl Tlaero Tlctrl g

rnTnaero Tnctrl (19)

where the applied torques are divided into two parts: theaerodynamictorques on the wing-body, Tlaero , Tmaero ,and Tnaero , whichare not related to actuator control input and engine thrust generatedtorque, and the actuator generated torques, Tlctrl , Tmctrl , and Tnctrl . Thenominal controller design can be described by the following twoparts.

-2

0

2

rollrate

command

actual signal

-0.2

0

0.2

pitchrate

0 2 4 6 8 10 12 14 16 18 20-5

0

5

10

time (s)

yawrate

-1

0

1

2

LeftWingtip

Voltage(V)

0 2 4 6 8 10 12 14 16 18 20-1

0

1

2

RightWingtip

Voltage(V)

-2

0

2

S

yntheticJets

Voltage

-20

-10

0

tailplance

0 2 4 6 8 10 12 14 16 18 20-20

0

20

rudder

time

(a)

(b)

(c)

-2

0

2

rollrate

command

actual signal

-0.2

0

0.2

pitchrate

0 2 4 6 8 10 12 14 16 18 20-5

0

5

10

time (s

yawrate

-1

0

1

2

LeftWingtip

Voltage(V)

0 2 4 6 8 10 12 14 16 18 20-1

0

1

2

RightWingtip

Voltage(V)

-2

0

2

S

yntheticJets

Voltage

-20

-10

0

tailplance

0 2 4 6 8 10 12 14 16 18 20-20

0

20

rudder

time

a)

b)

c)

(deg/s)

(deg/s)

(deg/s)

(V)

(deg)

(deg)

(s)

time (s)

Fig. 20 Simulation results of a coordinated turn duringthe approach.a) Body rate command andresponse,b) actuator commands to thesyntheticjets,elevator, and rudder, and c) input voltage to each of the wingtips.

CIURYLA ET AL. 651


11/12

Feedback Linearization Based Control Design

Tlctrl

Tmctrl

Tnctrl

264

375

gpl 0 gpn

0 gqm 0

grl 0 grn

2

64

3

75

IppqpqIpqrqr _pckp ~pkpl ~pI

Iqppp2Iqrrr2I

qprpr _qckp ~qkqI ~qI

IrpqpqIrqrqr _rckc ~rkrI~rI

2

664

3

775

Tlaero

Tmaero

Tnaero

264

375 (20)

where pc, qc, and rc are body rate commands, ~p p pc,~q q qc, ~r r rc are tracking errors, ~pI

Rt

0~p, ~qI

Rt

0~q,

~rIR

t0 ~r are integral tracking errors, kp, kpI, kq, kqI, kr, and krI are

feedback gains. Therefore, the body rate closed-loop controller hasthe following characteristics equations:

2 kp kpI

2

kq kqI2 kr krI

24

35

0

00

24

35

(21)

Controller Allocation

From therotation dynamic equations, onecan derivethe followingequation:

TlctrlTmctrlTnctrl

24

35 CA

SVTR

24

35 (22)

where CA is the controller allocation matrix obtained from theactuator model. Thus, the actuator command can be calculated as

SVTR

24

35 C1A

TlctrlTmctrlTnctrl

24

35 (23)

In the controller design process, the coefficients for the syntheticjet configuration [Eqs. (4) and (5)] are simplified as follows:

Cr 0:0098SV (24)

Cn 0 (25)

The controller parameters are shown in Table 2. Note that bycomparing the controller closed-loop bandwidth, the actuator timeconstant (shown in Table 2) is fast enough. The bandwidth of theactuator is more than three times faster than the desired controllerclosed-loop bandwidth. Thus, the actuator dynamics can be ignoredin the controller design process. The model error induced by thissimplification during controller design process can be compensatedfor by the closed-loop control, which is verified in the simulationresults. The Simulink simulation of integrated flight control usingsynthetic jets is shown in Fig. 19.

Simulation Results of Integrated Flight Control and Flow Control

The simulation of integrated flight control and flow control hasbeen tested for a scenario in which the Cessna is approaching therunway for landing at an angle of attack of 10 deg. Using the trimprogram of the simulation model, the trim condition ise 0:288 rad, Th 0:162, ub 11:67 m=s, wb 1:15 m=s.The command in a coordinated turn during approach is given by

rcommand g tan

Va(26)

Figure 20 shows thesimulation resultsof a coordinated turn duringthe approach. Figure 20a shows the body rate command andresponse, Fig. 20b presents an actuator command to the syntheticjets, elevator, and rudder, whereas Fig. 20c shows theinputvoltage toeachof the wingtips. These simulation resultsshow that the syntheticjet wingtips are able to replace traditional ailerons and provide

enough control authority for roll control (at least for the examplemaneuver presented here). It is worth noting that during thecontroller design, the synthetic jet configurations coefficients werereplaced by a very simplified model and that the closed-loop flightcontrol is robust enough to the handle the error introduced by thecontroller design process.

Summary

The application of activeflow control, via synthetic jets actuators,for separation control and roll control has been tested in wind tunnelexperiments. Using synthetic jets, mounted in the wingtips, delaysseparation by 2 deg whereas the maximum lift coefficient isincreased by up to 15%. It was shown that synthetic jets can providesimilar control authority as conventional ailerons, where propor-

tionalcontrol in pitch androll wasachievedfor anglesof attack largerthan 6 deg.

The rms output of a shear stress sensor was used to detectseparation on the wingtips of a scaled Cessna 182 model. Thisdetection technique was integrated with a computer and synthetic jetactuators to create a simple closed-loop stall suppression system. Tothe authors knowledge this is the first time that such a system everdeveloped for a plane model.

Using the wind tunneldata, an analysis of thedynamic response ofan RC scale model of a Cessna 182 was performed and an integratedflight control with flow control was developed. The next logicalstep is to instrument an RC model with synthetic jets along thewingtips and attempt roll control in flight using only synthetic jetactuators.

Acknowledgments

This work was supported as part of SBIR phase I, ContractNumber FA8650-05-M-3539, monitored by John Casey and JamesMyatt. The authors would also like to thank Eric White from RPI,whoassistedin running allof thewind tunnelexperiments,and Frank(Xiaodong) Zhang fromIAI for his contribution in the early stages ofthe work.

References

[1] Wentz, W. H., Jr., Effectiveness of Spoilers on the GA(W)-1 Airfoilwith a High Performance Fowler Flap,. NASA CR-2538, 1975.

[2] Ahuja, K. K., and Burrin, R. H., Control of Flow Separation bySound, AIAA Paper 1984-2298, 1984.

[3] Zaman, K. B. M. Q., Bar-Sever, A., and Mangalam, S. M., Effect ofAcoustic Excitation on the Flow over a Low-Re Airfoil, Journal ofFluid Mechanics, Vol. 182, Sept. 1987, pp. 127148.

[4] Hsioa, F. B., Liu, C. F., and Shyu, J-Y. Control of Wall-SeparatedFlow by Internal Acoustic Excitation, AIAA Journal, Vol. 28, No. 8,1990, pp. 14401446.

[5] Neuberger, D., and Wygnanski, I., The Use of a Vibrating Ribbon toDelay Separation on Two Dimensional Airfoils, Proceedings of AirForce Academy Workshop on Unsteady Separated Flow, Rept. TR-88-000, 1987.

[6] Seifert, A., Darabi, A., and Wygnanski, I., Delay of Airfoil Stall byPeriodic Excitation,Journal of Aircraft, Vol.33,No.4, 1996, pp. 691698.

[7] Smith, B. L., and Glezer, A., The Formation and Evolution ofSynthetic Jets,Physics of Fluids, Vol.10,No.9, 1998, pp. 22812297.

[8] Amitay, M., and Glezer, A., Controlled Transients of Flow

Reattachment over Stalled Airfoils,International Journal of Heat andFluid Flow, Vol. 23, No. 5, 2002, pp. 690699.[9] Amitay, M.,and Cannelle,F., Evolutionof Finite SpanSyntheticJets,

Physics of Fluids, Vol. 18, No. 054101, 2006.

652 CIURYLA ET AL.
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


12/12

[10] Amitay, M., Smith, B. L., and Glezer, A., Aerodynamic FlowControl Using Synthetic Jet Technology, AIAA Paper 1998-0208,1998.

[11] Amitay, M., Smith, D. R., Kibens, V., Parekh, D. E., and Glezer, A.,Modification of Lifting Body Aerodynamics Using Synthetic JetActuators, AIAA Journal, Vol. 39, No. 3, 2001, pp. 361370.

[12] Amitay, M., Pitt, D., and Glezer, A., Separation Control in DuctFlows, Journal of Aircraft, Vol. 39, No. 4, 2002, pp. 616620.

[13] Morel-Fatio, S., Pines, D. J., Kiddy, J., UAV PerformanceEnhancements with Piezoelectric Synthetic Jet Actuators, AIAAPaper 2003-394, 2003.

[14] Amitay, M., Horvath, M., Michaux, M., and Glezer, A., VirtualAerodynamic Shape Modification at Low Angles Attack UsingSynthetic Jet Actuators, AIAA Paper 2001-2975, 2001.

[15] Roskam, J., Airplane Flight Dynamics and Automatic Flight Controls,DAR Corporation, Lawrence, KS, 1995.

CIURYLA ET AL. 653

flight control using synthetic jets on a cessna 182 model

Documents