Contributions of the Transonic Dynamics Tunnel to the Testing of Active Control of Aeroelastic Response

AIAA 2000-1769

Contributions of the Transonic

Dynamics Tunnel to the Testing of

Active Control of Aeroelastic

Response

Boyd Perry, III, Thomas E, Noll, and Robert C, Scott

NASA Langley Research Center, Hampton, VA 23681

AIAA Dynamics Specialists Conference

April 5-6, 2000/Atlanta, GA

For permission to copy or republish, contact the American Institute of Aeronautics and Astronautics 1801 Alexander Bell Drive, Suite 500, Reston, VA 20191-4344

Contributions of the Transonic Dynamics

Tunnel to the Testing of Active Control of

Aeroelastic Response

Boyd Perry, III, _ Thomas E. Noll, t and Robert C. Scott _; NASA Langley Research Center, Hampton, 131 23681 Introduction Y the 1960s, researchers began to investigate the feasibility of using active controls technology (ACT) for increasing the capabilities of military and commercial aircraft. Since then many researchers, too numerous to mention, have investigated and demon- strated the usefuhtess of ACT for favorably modifying the aeroelastic response characteristics of flight vehi- cles. As a result, ACT entered the limelight as a viable tool for answering some very difficult design questions and had the potential for obtaining structural weight reductions, optimizing maneuvering performance, and satisfying the nmltimission requirements being im- Fig. 1 Transonic Dynamics Tunnel.

posed on future military and commercial aircraft de- signs.

the only facility in the world capable of studying a full Over the past 40 years, the NASA Langley Research range of aeroelastic phenomena at transonic speeds.

Center (LaRC) has played a major role in developing The TDT is used by the aircraft industry to aid in ACT in part by its participation in many wind-tunnel clearing new designs for safety front flutter, to eval- programs conducted in the Transonic Dynamics Tun- uate solutions to aeroelastic problems, and to study nel (TDT). These programs were conducted for the aeroelastic phenomena at. transonic speeds. In addi- purposes of: (1) establishing concept, feasibility; (2) tion to flutter clearance studies, the TDT is used t)y demonstrating proof of concept.; and (a) providing researchers to explore flutter trends and aeroelastic data for validating new modeling, analysis, and de- characteristics of new fixed wing and rotorcraft con- sign methods. This paper provides an overview of the cepts, to study the use of active controls, lo determine ACT investigations conducted in the TDT.

the effect, of ground-wind loads on launch vehMes, and For each program discussed herein, the objectives to make steady and unsteady aerodynamic pressure of the effort., the testing techniques, the test results, measurements to support, computational fluid dynamic any significant, findings, and the lessons learned with (CFD) code development..

respect to ACT testing are presented.

The TDT is a closed-circuit, continuous-flow wind tunnel capable of testing at. stagnation pressures from Transonic Dynamics Tunnel near zero to atmospheric conditions and over a Mach The TDT 1 shown in figure 1 became fulls' opera- number range from zero to 1.2. The test. section of tional in 1960 and has served ever since as a "National the TDT is 16 feet square with cropped corners. Con- Facility" dedicated almost exclusively to identifying, trolled variation of pressure in the tunnel simulates understanding, and soMng aeroelastie problems. It. is variations in flight altitude. One feature of the TDT *Assistant Branch Head, Aeroelasticlty Branch, Structures which is particularly useful for aeroelastic testing is a and Materials, AIAA Member.

group of four bypass valves connecting the test. section tBranch Head, Aeroelasticity Branch, Structures and Mate- area (plenum) to the opposite leg of the wind-tnnnel rials, AIAA Associate Fellow.

circuit downstream of the drive fan motor. In the event IResearch Engineer, Aeroelasticity Branch, Structures and Materials, AIAA Senior Member.

of a model instability, such as flutter, these quick- Copyright (_) 2000 by the American Institute of Aeronautics and actuating valves are opened, causing a rapid reduction Astronautics, Inc No copyright is asserted in the United States in the test. section Mach number and dynamic pres- under Title 17, US. Code The US. Government has a royalty- fr_c license to exercise all rights under the copyright claimed herein sure which may result in stabilizing the model. Other for Governmental Purposes. All other rights are reserved by the features that. make the TDT uniquely suited for aeroe- copyright owner 1 OF16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 2000-1769 lactic and ACT testing include: good visibility of the model from the control room; a highly sophisticated computer-controlled data acquisition system; oscillat- ing vanes upstream of the test section that can be used to generate sinusoidal gusts; a variety of model nlotlnt- ing and suspension systems ranging from cantilever sidewall mounts for component models to a 2-cable- suspension system for full-span "free-flying" models; safety screens that protect the tunnel fan blades from debris in case of a model failure; and state-of-the-art instrumentation and test. equipment.

Tests can be performed in the TDT using air o.s the test medium, however, the most distinguishing fea- lure of the tunnel is the use of a heavy gas, presently R-134a refrigerant., as the primary test medium. R- t3.1a is about, four t.imes as dense as air, yet has a Fig. 2 Delta wing semispan model mounted in TDT.

._peed of soimd of about half that of air. These prop- erties of higher density and lower sonic speed have chords of 69.4 inches and 8.8 inches, respectively, beneficial effects on the design, fabrication, and test- a leading-edge sweep angle of 50.5 degrees, and a ing of aeroelastically scaled wind-tunnel models that circular-arc airfoil section with a thickness-to-chord must accurately represent their fuli-scale counterparts: ratio of 0.03. Two high-fineness-ratio bodies were physically larger models may be built, thereby simpli- mounted on the wing lower surface to simulate en- fying the model fabrication process; and the scaled gine nacelles. The model was constructed of a primary natural frequencies of these larger models are lower, load-carrying plate structure covered with balsa wood resuMng in lower flntter frequencies, thereby reduc- that was contoured to the desired airfoil shape. A ing the risk of model destnwt.ion during flutter. Other rigid sidewall mounting block was used to sinmlate a advantages resulting from the use of a heavy gas are fllselage fairing.

a nearly three-fold increase in Reynolds number and lower tunnel drive horsepower. The model w_s equipped with both leading- and trailing-edge control surfaces centered at. 78.5 percent of the wing semispan, each with a span of about 6 Active Control Wind-Tunnel Tests inches. The chord of the trailing-edge surface was This section of the paper contains a summary of the about. 20 percent of the local wing chord; the chord of tests performed in the TDT during the last 30 years the leading-edge surface varied from about. 15 percent that involved active contor] system demonstrations. A of the local wing chord inboard to about. 20 percent.

common thread among these projects is that each has outboard.

has a signifcant impact on the state-of-the art of ACT.

One of the significant contributions of this program Delta Whig Active Flutter Suppression to the then-emerging state of the art was the develop- ment of two miniature eleetrohydraulic vane actuators During/he middle and late 1960s and into the early for mechanizing the active control surfaces. These ac- 1970s there was a growing expectation, that soon tuators were small enough to be mounted immediately turned to a realization, that active controls technology adjacent to the control surfaces and still fit. within the could achieve a variety of aeroelastic benefits. After contours of the airfoil. The importance of this proxim- numerous analytical studies this technology found its ity to the control surface wa_s the elimination of drive way onto a few airplanes (references 2 and 3) and con- shaft wind-up experienced by earlier electromechanical firmed that fatigue life could be increased and that actuators mounted external to the model. The elec- gust loads and fuselage accelerations could be reduced.

trohydraulic actuators provided over 3 foot-pounds of These early successes led to the belief that the much torque output over the frequency range O to 25 Hz with more difficult, and ambitious objective of active flutter a I000 psi supply pressure. The actuators weighed suppression could, indeed, be achieved.

only 2 ounces each and had angular displacement ca- The very frst demonstration of active controls in the pabilities of about +/- 9 degrees.

TDT occurred in 1971 with active flutter suppression of a semispan model of a low-aspect-ratio clipped- Nissim's aerodynamic energy criterion for flutter delta-wing configuration (reference 4). This configu- suppression, reference 5, was employed in the design ration was representative of the then-current Boeing of the flutter-suppression control laws tested on this model. The criterion states that a necessary and sup supersonic transport design, and is shown mounted in ficient condition for the prevention of flutter is that, the TDT test section in figure 2.

for M1 oscillatory motions of an elastic system in an The model had a span of 50 inches, root and tip 2 OF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPEI? 2000-1769 airstream, positive work be done by the system on the surrounding airstreanl. Tile mechanisnl by which this condition is satisfied is tile inclusion of active _No Flutter 500 I "__" "1_/- B "Mod," No Flutter control surfaces whose deflections are related by a con- trol law to the plunging and pitching motions of the 400F Open-Loop J _q¢.t_ A' Flutter wing. Nissim points out. that a suitable configuration is one employing both leading- and trailing-edge con- DY2sam_g,300 I Flutter Boundary trol surfaces, since the two working together provide independent control of lift. and pitching moment..

Three different control laws were designed based on Nissim's method. The first, and second used both psf 200 I leading- and trailing-edge control surfaces; the third used only the trailing-edge surface. In reference 5 these control laws were assigned letters, A, B, and

,o L

I I I I I C. Control law A was implemented and tested as 0.6 0.7 0.8 0.9 1.0 designed. For the other two, modified versions of Mach Number the original designs were implemented and tested and these are referred to as control law B "rood" and con- Fig. 3 Delta wing open- and closed-loop experi- trol law C "rood."

mental results.

The flutter suppression control system consisted of the two control surfaces and their correspondiug actua- t.em. These differences were attributed to tile inability tors, control laws implemented on an analog computer, of potential theory to predict detailed aerodynamic and two accelerometers. The accelerometers were lo- behavior on control surfaces of such relatively small cated along a chordwise strip very near the inboard size compared to the lifting surface. It was decided edge of the control surfaces, one at about. 30 per- that all calculations (not presented in the present pa- cent of the local chord and the other at. about. 70 per) should try to account for these differences in some percent.. Within the analog computer the outputs empirical manner. It. was found that if the ratio of the- of the accelerometers were integrated once to obtain measured-to-the-calculated static control surface hinge velocities, integrated again to obtain deflections, and moments was used as an empirical correction factor on then combined to produce normalized plunge and pitch control surface aerodynamic terms (both steady and displacements and normalized plunge and pitch rates unsteady), the differences in active control systeln ef- along the strip. The control surface deflections com- fectivenesses were greatly reduced. Today this type of manded by the control law were linear combinations empiricism is routinely used in correcting aeroservoe- of these displacements and rates. Control laws A, B lastic analyses.

mod, and C lnod were implenaented by changing po- Another contribution of this program was the iden- t.entiometer settings.

tification of the inertia coupling between the control Figure 3 presents dynamic pressure versus Mach surfaces and the main wing as the mechanism by number and summarizes the experimental results. The which still-air closed-loop instabilities occurred. It. was cross-hatch represents the open-loop flutter boundary, shown experimentally that this instability was driven and clearly indicates a transonic drop in the model by the rate feedback terms in the control law. The flutter speed. The circle, square, and diamond sym- modifications in control laws B and C, referred to bols are closed-loop results and demonstrate various above, were changes to avoid these instabilities. Today increases in dynamic pressure above the open-loop this type of still air instability is a common occurrence.

flutter boundary. The solid circle symbol at. 0.9 Math C-5A Active Load Alleviation System number is the only closed-loop flutter point, on the fig- During the 1970s the TDT played a role ill the de- ure and represents a 12 (7(,increase in flutter dynamic velopment of C-5A Active Lift Distribution Control pressure for control law A. The open square symbol System (ALDCS). The then Lockheed-Georgia Com- and open diamond symbols are no-flutter points and pany was interested in performing a correlation study represent the highest dynamic pressures achieved be- of the C-5A ALDCS flight test results with the re- fore testing was terminated. Control law B "rood" sults from TDT wind-tunnel te_ts of a Froude-scaled demonstrated a 22_, increase in dynamic pressure.

aeroelastic model of the C-5A, also equipped with an Control law C "mod" demonstrated increases in dy- ALDCS. This section of the paper is borrowed heavily namic pressure ranging from 1 ltX, at 0.6 Math number from reference 6.

to 30% at 0.9 Mach number.

The C-5A airplane ALDCS was developed to re- Over the course of conducting this program large duce fatigue damage on the wing due to maneuver, differences had been observed between the predicted and the actual effectivenesses of the active control sys- gust and peak-to-peak ground-air-ground load sources.

3 OF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 2000-17'69 The ALDCS was designed to reduce the inboard wing bendiug moment levels by redistributing the wing loads (so as t.o unload the wing tips) and by sup- pressing the airplane response in the short period and wing first-bending mode during maneuvers and ill at- mospheric turbulence. The systenl utilized wing ac- celerometers to form a symmetric aileron command signal through the existing C-5A Stability Augmen- tat.ion System (SAS). The system also used the ex- isting SAS pitch rate gyro and the autopilot normal accelerometer to command the inboard elevators to suppress short-period and first, wing bending mode gust responses and to provide handling quality com- pensation.

The C-5A airplane ALDCS consisted of wing mounted accelerometers to command aileron deflec- tion which suppressed the first wing bending mode Fig. 4 C-5A model mounted in the TDT.

and reduced wing bending lnoments for maneuver con- independent frequency sweeps of the aileron, horizon- ditions. The accelerometers (two on each wing) were tal stabilizer, and TDT flow oscillation vanes. At mounted at= 89% semi-span and were located on the comparable tunnel and flight conditions, wind-tunnel front and rear beams of the wing. Each pair of wing model data were compared with airplane data, with acceleromelers were summed proportionally front to and without the ALDCS engaged. Figure 5 contains rear (40_ to 60%) to optimize the chordwise location comparisons of frequency response functions (magni- and were summed equally between wings to allow only tudes only) of wing-tip vertical acceleration due to vertical input signals to the ALDCS computer.

aileron deflection and is typical of the many compar- The C-5A wind-tunnel model ALDCS was imple- isons in reference 6. Amplitudes for 1)oth airplane mented on an analog computer. Since the model did and model have been normalized by their respective not have elevators, the horizontal stabilizer was com- ALDCS-off maximum values and the model data has manded in pitch to duplicate the tail lift. change due to been appropriately scaled to airplane frequencies. Fig- inboard elevator ALDCS commands. The ALDCS re- ure 5(a) contains flight-test data; figure 5(b), wind- sponse of the model stabilizer was weighted and sched- tunnel data. The same basic result can be seen in both uled proportionately to the elevator transfer function parts of the figure: for the first mode (approximately requirements. The tail lift. generated by one degree 1 Hz) the ALDCS reduces the wing-tip acceleration by of airplane inboard elevator deflection was approxi- about 60_, and broadens the peak; at all other frequen- mated by 0.35 degrees of stabilizer deflection on the cies the ALDCS produces few differences in wing-tip model. The ailerons and the stabilizer were powered acceleration.

by small hydraulic actuators. Position feed back froln The results of this study validated the use of ac- these control surfaces was programmed using opera- tive control technology for the minimization of aircraft tional amplifiers to give the actuators the appropri- aeroelastic response and showed that scaled aeroelastic ate transfer function. The actuator servo responses wind tunnel models can be used in developing active were measured and compared to calculated aircraft control technology. Both the model and airplane re- response. The total model on-board ALDCS System sults showed thai the desired wing load relief for the weighed 49.61 pounds.

C-5A wing first bending mode was achieved with the The wind-tunnel model was a 1/22 geometrically ALDCS and the model and airplane correlated very scaled, aeroelastic model designed to match the air- well for this mode.

plane Froude number in the TDT heavy-gas test B-52 Active Control Systems medium and at. a density ratio of 2.65. The model was constructed of hollow and solid metal spars located The success of flight investigations in the 1960s to along the elastic axis of each aircraft component with reduce the dynamic response of the XB-70 and the balsa wood fairings to achieve an aerodynamic shape.

B-52E aircraft due to gusts opened the gateway to a The wind-tunnel model was supported using the TDT multitude of active control studies and applications.

two-cable mount syst.em with the forward cable in the In conjunction with the B-52E flight investigation, vertical plane and the aft cable in the horizontal plane.

a 1/30th scale, full-span, free-flying B-52 aeroelastic wind-tunnel model was constructed and tested in the Figure 4 presents a photograph of the model installed in the TDT.

TDT. The model was dynamically scaled to match A variety of wind-tunnel model data was acquired the first nine symmetric elastic vibration modes of the during the 1973 wind-tunnel test. The data included flight vehicle (frequency range from 0 to 25 ttz). The 4 OF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 2000-1769 'L0 _'_ I _ ALDCSofl I 0.8 0,6 Normalized mngti vertical acceder ation 0.4 0.2 I 1 I I 1 I I 1 I 2 3 4 Frequency, Hz a) C-5A airplane response.

I"0 r ,_ -- ALDCSoff Fig. 6 B-52 model mounted on the Transonic Dy- namlcs Tunnel's two-cable mount system.

0.16 - Normalized wing tip verlical

I

0.2 _ d _, ' "- Damping, 0.12: J g 0.08 -Mod 0 1 2 3 4 Frequency, Hz b) C-5A wind-tunnel model Response.

0.04 "-- Aiplane AFS on L- A=planeA_on -- _ I_ Fig. 5 Airplane and wind-tunnel model response --_ [ t 1 I "_,._ I I I to aileron sweeps.

160 180 200 220 240 260 280 Airplane airspeed, m/see model was equipped with active ailerons and eleva- tors and demonstrated, in the wind tunnel, what had Fig. 7 Comparison of B-52 TDT and flight test already been demonstrated in flight: the ability of ac- results, AFS system off and on.

tive controls to alleviate structural dynamic response caused by turbulence.

outboard ailerons, new flap segments, and horizon- In parallel with Controlled Configured Vehicles tal canards. These control surfaces were driven by an electromechanical system consisting of d.c. torque mo- (CCV) flight lest program using the B-52E aircraft, an investigation 7 sponsored by the Air Force Flight tors mounted within the fuselage and crank-pushrod Dynamics Laboratory (AFFDL) with Boeing and in linkages and shafting from the motors to the control cooperation with the NASA LaRC was initiated to: l) surfaces. Figure 6 shows the B-52 wind-tunnel model installed in the TDT on the two-cable mount system.

develop active control concept evaluation techniques through wind-tunnel testing; 2) demonstrate that air- The AFS system consisted of two independent feed- craft, active control wst.ems can be sinmlated with back loops. Signals from accelerometers mounted on wind-tunnel models; 3) obtain experimental data for the ballasted external filel tanks were compensated us- validating analysis results and methods; and 4) ob- ing an analog computer and fed back to the aileron tain data for correlation with the B-52E flight test surfaces, while accelerometer signals located near the results. The impetus behind this investigation was the midwing were compensated and fed back to the flap availability of the previously-tested 1/30th scale B-52 segments. Since both syst.ems were designed to sepa- aeroelastic wind-tunnel model.

rately provide a 30 percent increase in flutter speed, they were considered redundant systems. With the The B-52 wind-tunnel model was modified to dy- exception of scaling differences, the AFS systems on namically match equivalent changes made to the full- the wind-tunnel model and on the B-52E aircraft were scale B-52E as required to produce a flutter condition very similar over the frequency range of interest..

within the vehicle's flight envelope. In addition, to The wind-tunnel test data, scaled up to correspond- properly represent the active flutter suppression (AFS) ing flight-test conditions, are compared with flight-test and the vertical ride control (VRC) systems being results in figure 7. Concentrating first on the wind- evaluated during the flight test program, the wind- tunnel model was further modified to include new tunnel results only (closed symbols), it can be seen 5 OF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 2000-1769 that the AFS-on tests were performed at. velocities only slightly higher than the open-loop flutter velocity of 253 m/s. tIowever at. the highest velocity for AFS- on, the damping in the flutter mode showed a large improvement compared to that for' AFS-off and also showed the potential for a significant increase in flut- ter speed. Comparing, now, the AFS-on wind-tunnel results (closed square symbols) with the AFS-on CCV flight-test results (open square symbols) the damping trends are seen to agree quite well.

The VRC system on the fllll-scale aircraft, was de- signed to reduce, by at. least. 30 percent, the gust- indnced vertical acce]erati0n at the pilot's station.

This system used vertical acceleration, sensed at. the pilot's Station and appropriately shaped wi[ha filter, YF-17 model mounted in the TDT.

Fig. 8 to drive the horizontal canards. The VRC system on the wind-tunnel model was a scaled-down version of ming the model at various tunnel conditions. The wing the actual flight system. The performance of the \'RC had leading- and trailing-edge control surfaces pow- system was equivalent to the performance of the sys- ered by electro-hydraulic actuators. Three different tem on the airplane at the structural mode frequencies external store configurations having widely different.

deemed important at. equivalent test. conditions. The flutter characteristics (flutter frequency, modal cou- VRC systenl reduced the magnitude of the 6th and 8th pling, and flutter-mode violence) were also available.

mode peaks on the model by about 60 percent and 7.5 A photo of the model mounted in the TDT is shown percent respectively, and on the airplane by about ,56 in figure 8..

percent and 73 percent respectively.

The first series of wind-tunnel tests were conducted The most significant finding that resulted from the in three entries: June, August, and December of 1977.

B-52 CCV full-span model program was the knowledge For these tests, a different AFS control law, each em- that dynamically-scaled, actively-controlled wind- ploying a single control surface, was developed for each tunnel models could be extremely useful in study- external store configuration. These tests were quite ing and developing advanced active control concepts.

successful. For the first store configuration, character- From that time forward, wind-tunnel models were des- ized by a lightly damped hump mode, passive flutter tined to play the following important roles in the could not be reached within the limits of the TDT.

dex, elopment of active-control concept.s: increase the With a control law operating that used the trailing- confidence level in these concepts by providing data to edge surface, a significant improvement in structural verify analytical models and methods; and eliminate damping in the critical elastic mode was achieved.

the risks and lower the costs associated with flight- The other two store configurations were character- testing these concepts.

ized by violent flutter onsets, one at. about. 5 Hz and the other at about. 10 Hz. With control laws that YF-17 Wing/Store Active Flutter Suppression Program used the leading-edge surface, flutter suppression was successfully demonstrated: for the configuration with The Northrop Corporation, under contract with the most violent flutter characteristics, the model was the AFFDL and in cooperation with NASA LaBC, tested 18 percent above the unaugmented flutter dy- conducted a long-term wind-tunnel investigation of namic pressure without incurring an instability. Based wing/store AFS. The objective of this program was on damping trends at. this condition, the model was to perform several series of tests to evaluate a mul- projected to be stable up to about 29 percent above titude of control concepts based on different design the unaugmented flutter condition, s This program philosophies. These concepts began with simple non- demonstrated that active mlppression of wing/store adaptive analog controllers and evolved into digital flutter was feasible. The program also demonstrated adaptive controllers. For this program, a 30-percent- for the first, time that leading-edge surfaces acting scale, semispan, aeroelastic model of the YF-17 air- alone are viable AFS surfaces.

craft was designed for testing in the TDT. The wind- tunnel model consisted of a wing, a fuselage, and a For the second series of tests, conducted during Oc- horizontal tail. The sidewall naounted model was very tober 1979, the use of multiple loops with multiple con- unique in that it. used cables and a set of bars and link- trol surfaces acting simultaneously was the approach for obtaining further increases in flutter speed. 9 For ages to simulate rigid-body pitch and plunge degrees- of-freedom. The horizontal tail, attached to an electric this series of tests only the wing/store configuration with the most violent flutter mode was used. To reduce motor located within the fuselage, was used for trim- 6 OF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 9000-1769 Mach Number = 0.8 the risk of losing the model during flutter, tile store 8O was modified to include an interual electro-mechanical [2] T = 10 msec I system that served as a flutter stppper. This system m Damping 60 t T = 15 msec I would passively suppress flutter by moving an internal mass very rapidly, thereby changing the wing/store -X --"_ _ Extrapolated structural frequencies in such a manner as to decou- Paramete40 I

O u n;,oop \

pie the critical elastic modes. Whenever an insta- bility was encountered tile syst.em could be triggered either automatically or manually. Research organiza- I / I '1',/ I -A "-_ I tions from three European countries were invited to 60 80 100 120 140 160 participate in the second series of tests through the Dynamic Pressure, q (psf) auspices of USAF data exchange agreements or in- formation exchange programs. These organizations Fig. 9 YF-17 flutter suppression perfortnance for included British Aerospace and the Royal Aeronau- three controller sampling times.

tical Establishmen! from the United Kingdom, the Office National d'Etudes et de [{echerches Aerospa- the digital control laws and the improvement in flutter tiales from France, and Messerschnfitt-Bolkow-Blotma speed demonstrated was comparable to the metrics ob- GmbH from West Germany. Control laws from each of tained during the analog control test demonstrations.

the organizations varied greatly in their design philoso- Although sampling time, T, was identified as a crit- phy, and in the number of sensors and control surfaces ical parameter (figure 9), the control laws performed used. All control laws were highly successful in sup- adequately down to a sampling rate of 100 samples pressing flutter, l° One control law was tested to a per second (T of 10 milliseconds), which was typical dynamic pressure 70 percent above the passive flut- of sampling rates for aircraft digital control systems ter dynamic pressure. Post test evaluation of damping under development in the early 1980s.

trends indicated that this control law could have sta- During the second phase of these tests, a relatively bilized the model up to about 131 percent above the simple adaptive controller was developed and tested passive flutter dynamic pressure boundary.

during April 1982. The first level of adaptation con- Some "firsts" demonstrated in this portion of the sisted of discriminating between possible flutter modes YF-17 semispan model program included the abil- (based on a priori knowledge) and selecting the ap- ity to switch from one control law to another above propriate control law; tile second level consisted of the unaugmented flutter condition and the ability to adapting the control law to changes in flight condition.

switch from a control law that used a trailing-edge This concept was successfully delnonstrated during the surface to one that used a leading-edge control sur- wind-tunnel tests. 11 In addition, the ability of the con- face above the unaugmented flutter condition. The troller to adapt rapidly following a store release was ability to switch control laws above flutter without ex- demonstrated. For this unique demonstration, a wing- periencing any noteworthy transient motions on the tip mounted store was abruptly released transforming the model from a stable condition to a violent, flutter model laid the groundwork for adaptive control. In addition to the test demonstrations, advancements in condition. The adaptive controller recognized the un- test procedures and measurement techniques were also stable behavior, implemented a new control law, and stabilized the model in a small fraction of a second.

accomplished. Procedures to calculate model open- loop characteristics from measured closed-loop trans- DAST Wing fer functions were demonstrated at conditions below and above flutter. Also, techniques developed and In the early 1970s NASA elnbarked on an ambitious high-risk flight-test program 12 whose primary objec- used to extract, system gain and phase margins from tives were to validate analysis and syntliesis methods Nyqnist plot.s provided a measure of model stability for the active control of aeroelastic response and analy- and were useful in identifying ways to improve the con- sis techniques for aerodynamic loads prediction. This trol law performance.

program was called DAST (Drones for Aerodynamic The next step in the logical progression of AFS de- and Structural Testing). It was conceived and imple- velopment in the YF-17 semispan model program was mented at NASA Langley with fligh! tests conducted to transition from analog non-adaptive systems to dig- at NASA Dryden. The flight test vehicle was an nn- ital adaptive controllers. The demonstration of an manned Firebee II target drone whose standard wing AFS digital adaptive controller was accomplished in had been replaced with an aeroelastic research wing two pha.ses. During the first phase, several of the (ARW).

analog control laws tested previously were digitized The first, in a series of these wings, designated ARXV- and implelnented on a digital controller. These con- trol laws were then retested on the YF-17 model in l, was designed to have both symmetric and antisym- metric classical bending-torsion flutter modes within the TDT during November 1981. The performance of 7 oF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 2000-1769 Fig. 11 F-16 wind-tunnel model on the Transonic Dynamics Tunnel's two-cable mount system.

sequence of an inaccurate description of wind-tunnel turbulence, upon which pre-test analyses and pre-tesI control law performance were based. The results of this test emphasized the need for a more accurate de- Fig. 10 DAST wing model mounted in the TDT.

scription of turbulence within the TDT test section.

the flight envelope of the DAST AR_V-I vehicle. Tile F-16 Wing/Store Active Flutter Suppression primary objective of the ARW-1 flight, tests was to Program verify transonic flutter prediction techniques and to The F-16 aircraft, carries many combinations of ex- validate the predicted performance of the vehicle with t.ernal stores. To assist in identifying a large number an active flutter suppression (AFS) system.

of critical wing/store flutter modes for the F-16, a 1/4 As part. of the DAST program, a wind-tunnel model scale, full-span, free-flying flutter model was designed study was undertaken to reduce the technical risks as- and fabricated for testing in the TDT. This model was sociated with implementing an AFS system on the tested successfully many times in the 1970s and 80s to ARW-1. A dynamically-scaled representation of the support, the USAF F-16 flutter clearance program.

ARW-1 wing was designed such that it would flutter Because of the large number of critical flutter modes within the operational limits of the TDT, and a semis- associated with external stores, the USAF and Gen- pan wind-tunnel model of pod-spar construction was eral Dynamics became very interested in all promising built for testing in the TDT. The model was equipped flutter prevention techlfiques, including AFS. To in- with a hydraulically actuated trailing-edge control sur- vestigate the potential of applying AFS to the F-16, face, centered at. 83 percent of the semispan, with a General Dynamics took advantage of their existing span of 13 percent of the semispan, and with a chord 1/4-scale flutter model, by then a mature and reliable of 20 percent, of the local wing chord. A photograph testbed, and fabricated a new set of wings equipped of the model mounted in the TDT is presented in fig- with acceleromete,'s positioned at key locations and ure 10.

flaperon surfaces powered by hydraulic actuators. In Flutter suppression control laws, la based on two addition, ballast, in the fuselage was replaced with different methods, were designed with the objective a hydraulic pump to power the wing servoaetuators.

of demonstrating a 44-percent increase in flutter dy- For the next. eight, years the F-16 model (figure 11) namic pressure over the Mach number range 0.6 to with the new wings became a testbed for evaluat- 0.9. These control laws employed as feedback sensors ing AFS systems that ranged from analog to complex accelerometers located near the control surface. Volt- digital adaptive concepts. This program was carried- ages proportional to acceleration were fed back to an out. by a team of researchers from General Dynamics, analog computer upon which flutter suppression con- the U.S. Air Force Wright Aeronautical Laboratories Irol laws were programmed. In order to demonstrate (AFWAL), and the NASA LaRC and involved three wind-tunnel test entries in the TDT.

the 44-percent increase, the active control system had to operate in the presence of tunnel turbulence and The first, test, conducted in February 1979 for a within the deflection and rate limits of the actuator.

single wing/store configuration, demonstrated the sup- The stated objective of demonstrating a 44-percent pression of an antisymn_etric flutter mode at 8.6 Hz.

Research issues related to AFS that were considered increase in flutter dynamic pressure over the Mach number range 0.6 to 0.9 was not achieved. How- important included: the effects of asymmetry between ever, at 0.95 Mach number, both control laws did left and right wing sensor signals and actuator com- demonstrate the 44-percent increase. The major fac- manded deflections; the sinmltaneous implementation t.or that prevented the 44-percent increase from being of symmetric and ant.isymnletrie control laws; switch- achieved was unexpectedly large (ref. 13 called them ing of control laws above open-loop flutter; and de- "excessive") control-surface peak deflections. These termining if open-loop frequent3" response functions unexpectedly large deflections were, in turn, the con- (FtlF) could be measured accurately enough to pro- 8OF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 2000-1769 2.0 vide useful information. The determination of accu- rate FRF was considered vital to the safety of the tests, ensuring that. the AFS was operating as expected at subcritical speeds, and that the control law was provid- q No Flutter_ ing the correct, gain and phase to suppress flutter. The q"Ref FSS On measurement of tile open-loop FRF with the feedback 1.0 loop physically open and closed were both success- ful. Itowever, with the loop closed, tunnel turbulence caused a distortion of the FRF near the flutter fre- quency, and was most evident near the unaugmented

[:Flutter I

No Flutter I flutter point, hnportant accomplishments from the I I I ! I I first, test included: successful nmdificat.ions to control 0.5 0.6 0.7 0.8 0.9 1.0 1.1 laws (gain/phase changes and sensor changes) dur- Mach ing testing to maximize AFS effectiveness; successful switching of control laws above the unaugmented flut- Fig. 12 F-16 open- and closed-loop flutter bound- ter condition without experiencing any threatening aries, transient motions; and testing closed loop to a dy- namic pressure 100 percent above the unaugmented oping and demonstrating a totally digital adaptive (no flutter dynamic pressure (with flaperon displacements prior knowledge of the aircraft, configuration) system.

never exceeding 0.6 degre6s).

The objectives of this investigation included: demon- strating digital adaptive flutter suppression for three The unaugmented flutter point for the store con- different external store configurations, each having figuration tested in 1979 was deterndned from FRF widely different flutter-mode characteristics; demon- derived from closed-loop system measurements, and as already stated, there was distortion, and therefore st.rating a 30 percent improvement in flntter speed with the AFS operating for each store configuration; uncertainty, present in these FRF. Post-test data re- and demonstrating the suppression of flutter follow- duction and analysis revealed uncertainty in the actual value of unaugmented flutter point. Therefore, in Oc- ing the separation of a store from the wing. These tests were accomplished during December 1986 and tober 1981 this store configuration was retested. The the results are summarized in reference 15. These objectives of this test. were t.o explicitly define the tests demonstrated, for the first, time, the feasibility of unaugmented flutter condition, to determine the accu- racy of lneasured FRF and define approaches for im- using a digital adaptive AFS system having no prior knowledge of the wing/store configuration. Not only proving the accuracy, and t.o investigate the feasibility were significant improvements in flutter speed demon- of suppressing flutter with a single flaperon while simu- st.rated for some wing/store configurations, but the lating a failure in the other. In addition, a second store system performed very well in adapting and stabiliz- configuration that fluttered symmetrically at 10.6 Hz was tested to further demonstrate the usefulness of ing the model following the release of a wing-tip missile that immediately resulted in a post. flutter condition.

flaperons as AFS surfaces. These tests were highly suc- In this unstable condition, the system was able to iden- cessful, satisfying all their objectives. 14 The accuracy of tim measured FRF was det.ermined by actually vary- tify the unstable plant, design a nominal control law, and suppress flutter in less than a second.

ing gain and phase angles and measuring the gain and phase margins up to an unstable condition. Both the Some of the more significant accomplishments of the direct. (actual measurement of open-loop data) and the totally digital adaptive portion of the F-16 full-span indirect methods (extraction of open-loop data from model program included: the use of control laws de- closed-loop responses) of obtaining FRF were found veloped by the adaptive controller as a backup analog to provide reasonable measures of model stability. The safety system; the launching of missiles from a free- AFS was also found to perform satisfactorily with one flying model at. conditions below and above the unaug- flaperon locked out; however, the gain margin was re- mented flutter boundary; and the successful demon- duced by a factor of one-half. Finally, AFS systems stration of adaptive control. For one test run, the employing flaperons performed equally well for both adaptive controller updated the control law over 2500 symmetric and antisymmetrie flutter modes. Some re- times without, losing control of the flutter mode. In sults from the antisymmetric AFS tests are provided addition, the adaptive controller was successful with in figure 12. simulated single actuator failures and with rapidly With successful conventional AFS wind-tunnel test.

changing test conditions.

demonstrations on the F-16 wind-tunnel model and Active Flexible Wing Program successful adaptive AFS test demonstrations on the In the early-1980s Rockwell International Corpora- YF-17 model, the AFWAL, General Dynamics, and tion developed a concept, it. named the active flexible NASA LaRC team became directly involved in devel- 9 oF 16 AMERICAN INSTITUTE OF AERONAI_TICS AND ASTRONAUTICS PAPER 2000-1769 The second two tests involved only Rockwell and NASA and focused on the demonstration of aeroe- lactic control through the application of digital ac- tive controls technology. The results from these tests are reported in a special issue of the AIAA Journal of Aircraft. is For these tests the model was fitted with wing-tip ballast stores to lower the model flut- ter speed into the operational capabilities of the TDT.

The model was sting-mounted utilizing an internal ballbearing arrangement, allowing the model freedom to roll about the sting. A roll degree-of-freedom ])rake was employed for those cases when a fixed-in-roll con- dition was required. The model had two leading-edge and two trailing-edge control surfaces on each wing panel driven by rotary-vane, electrohydraulic actua- tors powered by an onboard hydraulic system. The AFW wind-tunnel model mounted in Fig. 13 model was instrumented with a variety of sensors that TDT.

included accelerometers, strain gages, rotary variable differential transducers, and a roll rate gyro. Ac- wing (AFW) concept, is and ill 1985, in cooperation with the AF\_,AL and the NASA LaRC, Rockwell tive control concepts considered during the second two tests included AFS, rolling maneuver load alleviation undertook a research program to demonstrate this concept. The AFW concept exploits, rather than (RMLA), and a roll rate tracking system (RRTS).

These active control systems were designed to be com- avoids, wing flexibility by employing active leading- patible with each other such that they could be tested and traillng-edge control surfaces, up to and beyond silnuItaneously, even at conditions above the passive control-surface reversal. A high-perfornlance aircraft flutter speed of the wind-tunnel model.

designed using the AFW concept achieves its high roll rates using wing control surfaces only, thereby elimi- The design goal of AFS control laws wa_s to pen- nating the need for a "rolling tail," and, consequently, etrate the open-loop flutter boundary and proceed eliminating the additional structural weight associated to the operating limit of the TDT. Requirements for with a rolling tail.

minimum levels of robustness and acceptable levels of control-surface deflections and rates were speci- In an AFW design an active roll control (ARC) sys- fied. For the wind-tunnel model in the fixed-in-roll tem is required to efficiently manage the rolling of the vehicle. An ARC system monitors flight conditions configuration, symmetric and antisymmetric flutter and, based on those conditions, chooses the most ef- boundaries had to be penetrated to demonstrate any- fective control surfaces to roll the vehicle, and also thing more than a trivial increase in flutter dynamic chooses tho proper sign for control-surface deflections pressure. Four different AFS control laws were de- (one sign if below reversal, the opposite if above). signed, three were tested. A control law designed using a multiple-input/multiple-output (MIMO) con- In an AFW design further weight savings can also strained optimization technique was successful in sup- be achieved by the additional use of active controls.

pressing flutter to a condition 26 percent above the Taken alone or in combination, AFS, gust load allevi- antisymmetric-open-loop flutter dynamic pressure and ation, and maneuver load control all have the potential ]7 percent above the symmetric-open-loop flutter dy- for further reductions in vehicle weight. By taking full namic pressure.

advantage of active controls and the AFW concept, Rockwell predicted that, compared to conventionally- An important goal of the AFW program was designed high-performance vehicles, weight savings of the demonstration of muh iple-input/nmlt_iple- at least 15 percent, of take off gross weight could be output/nmltiple-function control law testing. This achieved for an advanced fighter configuration.

goal was accomplished through the simultaneous operation of AFS and RMLA control laws. The An AFW program grew out of the AFW con- cept. The testbed for the AFW program was the design goal of RMLA control laws was to reduce or aeroelastically-scaled, full-span, wind-tunnel model control wing loads during rolling maneuvers of 90 shown sting mounted in the TDT in figure 13. Tile degrees, hnportant design considerations were to maintain stability, acceptable control-surface deflec- model was designed and built by Rockwell and tested tions and rates, and constant roll performance. These on four different occa_sions (1986, 87, 89, and 91) in the TDT. The first two tests involved Rockwell, tile control laws were implemented with the wind-tunnel Air Force, and NASA and focused on demonstrating model in the free-to-roll configuration, for which only the AFW concept. The results from these tests are one flutter boundary (symmetric) was within the reported in references 16 and 17 tunnel-operating envelope. Four combinations of AFS 10 OF 16 AMERICAN INSTITUTE OF' AERONAUTICS AND ASTRONAUTICS PAPER 2000-1769 3OO 28O q, psf 240 2O0 Symmetric RMLA RRTS open-loop + + flutter FSS FSS AFW multi-function control law perfor- Fig. 14 Fig. 15 PARTI model mounted in the TDT.

mance.

and RMLA control laws were designed and tested.

of an interior composite plate to serve as the main Aggressive (scaled MIL SPEC) roll maneuvers were load carrying structure and an exterior fiberglass shell performed and wing loads were controlled 17 percent to provide the proper aerodynamic contouring. The above the symmetric-open-loop flutter dynamic aerodynamic shell was divided into six sections. Each pressure. This data is summarized in figure 14.

section was attached to the composite plate at two locations to nlinimize the increase in model stiffness Twin-Englne F-16 Derivative attributed to the aerodynamic shell. The composite Wind-tunnel models of a twin-engine F-16 derivative plate consisted of an aluminum honeycomb core with were tested five times in the TDT from 1988 to 1993.

graphite epoxy face sheets. Seventy-two piezoelectric A flutter test program was initiated to characterize actuator patches were distributed on both the upper the symmetric and antisymmetric flutter modes of the and lower surfaces of the composite plate. The actu- aircraft. 19 A stability model and a dynamically sim- ators covered about, two-thirds of the composite plate ilar model were constructed. The full-span, 2/7-scale area and accounted for about seven percent, of the total models were designed and fabricated with a remotely wing weight. Due to the ply orientation of the male- moveable mass in the fuselage to allow for testing stat- rial used in the composite plate and the wing sweep, ically stable and statically unstable configurations on the piezoelectric actuator patches were connected in the TDT cable-mount system. The mass weighed 40 fifteen different groups chosen to affect the bending lbs. and could travel up to 28 inches, which moved the and the torsional responses of the model. During model center of gravity up to 12 percent of the mean the control law development and testing these actu- aerodynamic chord.

ator groups were further combined into supergroups This model is included here because it required the (several groups of piezoelectric actuator patches being successful implementation of active controls in order to activated by the same signal). Ten strain gages and meet the objectives of flutter testing a statically un- four acceleromel.ers were available as feedback sensors stable configuration. A stability augmentation system and for monitoring the models response during the (SAS) employing pitch rate feedback to the elevons al- tests. In addition to the piezoelectric actuators, the lowed the model to be tested at tunnel conditions up model had a trailing-edge aerodynamic control surface to a Mach number of 1.1 and a dynamic pressure of driven by an electric motor located in the wing root 250 psf.

(hidden from the airstream) and an automatic flutter- Piezoelectric Aeroelastie Response Tailoring stopper. Figure 15 shows a picture of the model fully Investigation aossembled and installed in the TDT. Figure 16 shows the model with the external shell removed exposing NASA LaRC, in cooperation with the Massachusetts some of the internal details.

Institute of Technology, conducted an investigation For this investigation two wind-tunnel test entries known as PARTI (Piezoelectric Aeroelastic Response were performed using air as the test medium at at- Tailoring Investigation). The objective of the PARTI mospheric conditions. The first ent.ry 2° conducted project was to demonstrate in the wind tunnel the abil- during March 1994, was used to measure the follow- ity of strain-actuated adaptive wings to control aeroe- ing open-loop (control law off) information: the model lastic response at subcritica] speeds and to prevent.

flutter. For this demonstration, an aeroelastic semis- subcritical (below flutter) response; the basic flutter characteristics of the model in its basic configuration pan model with distributed piezoelectric actuators was and with the model in a flutter-stopper configuration; fabricated for testing in the TDT. The model consisted 11 oF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 2000-1769

,o

PSD 1°2 (_02/HZlg

,o °

70% Reduclotn in FIMS response 'wp 1 | | | | | 1 0 5 10 15 20 25 Frequency Hz Fig. 16 Internal details of PARTI model.

Fig. 17 PARTI turbulence response results, Maeh number = 0.43, dynamic pressure = 60 psf, and time-histories and frequency response functions for each important piezoelectric actuator group. These Benchmark Active Controls Technology Model experimental data were useful in constructing math- The successful design of an active control system ematical models for designing control taws and for for controlling aeroelastic response requires overcom- verifying analytical models and techniques.

ing numerous technical challenges. These challenges The second entry 21 conducted during November include: the current inability to accurately model con- 1994, was used to assess and demonstrate the ca- trol surface effectiveness, especially for spoilers; con- pability of piezoelectric actuators to suppress flutter trol system robustness, reliability, and sensitivity to and to reduce aeroelastic response caused by tunnel failures; and proven analysis packages for safely testing turbulence. Many different control laws, based on dif- and evaluating these systems. The Benchmark Ac- ferent design methodologies, actuator groupings, and tive Controls Technology (BACT) program has been able to make contributions to all of these areas. The feedback sensors, were designed using experimentally determined state-space mathematical models and ac- objectives of the BACT program were to perform tuator transfer functions. Control law design tech- wind-tmmel experiments to obtain benchmark-quality niques included classical frequency domain methods, data to validate CFD and computational-aeroelasticity the/l-Synthesis method, a Linear Quadratic Gaussian codes, to verify the accuracy of current aeroservoelas- (LQG) lnethod with loop shaping, and a sensitivity- tic design and analysis tools, and to provide an active weighted LQG method. Most. of the control law de- controls test.bed for evaluating new and innovative con- signs used strain feedback rather than acceleration trol methodologies.

feedback because of the "cleaner" transfer functions The BACT program employed a rigid semispan wind-tuunel model that could be tested on either a provided by strain gages and the ability of the strain gages to capture the first three elastic modes of the flexible or a rigid mount. The model is a rectangu- model (lst bending, 2nd bending, and 1st torsion).

lar wing with an NACA 0012 airfoil, a chord of 16 Twenty-eight of these control laws were tested in the inches and a semispan of a2 inches. The model was TDT and evaluated. The complexity of the control built with a conventional trailing edge (TE) control laws varied from single-input/single-output to nmlti- surface and one upper- (US) and one lower-surface input/nmlti-output controllers having five sensors and spoiler (LS). The model was extensively instrumented nine actuator groups. The most successful control law with pressure transducers and accelerometers to mea- was a single-input/single-output LQG design that used sure surface pressures and model dynamic responses.

one strain gage for feedback and alI fifteen actuator The BACT model was the last model of NASA Laug- groups. [;sing this control law, an increase in flutter ley's Benchmark Models Program. 2_ 25 dynamic pressure of twelve percent was demonstrated.

Each control surface on the BACT model had a span In addition, at. dynamic pressures well below flutter, of 30 percent of the model semispan and was centered within the power spectral density fimction of micro- about, the 60 percent semispan station. These three strain due to tunnel turbulence, the peak value at the control surfaces could be actuated independently of frequency of the first, flexible mode was reduced by eacli other using miniature hydraulic actuators and seventy-five percent (figure 17).

were, therefore, suitable for use a.s active control sur- faces. The TE control surface had a chord of 25 The significant contributions of this program to the state of the art were the demonstrations of flutter percent of the model chord; the spoilers each had a suppression and aeroelastic response control by dis- chord of 15 percent of the model chord, hinged at the 60 percent chord station. The actuators allowed tributed piezoelectric actuators on a large-scale aeroe- lastic wind-tulmel model.

static control surface displacements or dynamic con- 12 OF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 2000-1769 trol theory (tI-,x, and p-synthesis) to flutter suppres- sion. Two-input-l.wo-output flutter suppression con- trol laws, one using H-,a:, and one using /t-synthesis design methods 2r were designed, implemented, and successfully tested. These designs were obtained with weighting fulict i01is that put emphasis on keeping con- trol activity limited to the frequencies near flutter The BACT model offered the opportunity to use neural network based control syslems to suppress flut- ter. Three neural network ba_sed control systenls were developed and tested as part. of the Adaptive Neural Control of Aeroelast.ic tlesponse (ANCAtt) program.

ANCAR was a joint research and development effort conducted by the NASA LattC and The Boeing Con> pany (forlnerly the McDonnell Douglas Corporation) under a Space Act. Agreenmnt.

Fig. 18 BACT wind-tunnel model mounted in the TDT. Phase I of the ANCAB program was the devel- opment and denlonstration of a neural network gain trol surface oscillations about, a mean angle. The TE scheduled flutter suppression system. 2s hi this appli- control surface deflection Was mechanically limited to cation a neural network was used to schedule control 15 degrees either up or down. Each spoiler could be laws as a function of Mach number and dynamic pres- deployed from its stowed position (zero degree deflec- sure. The controller was tested along with a robust tion) to any angle up to 45 degrees.

fixed-gain control law. The neural network scheduled During 1993, the BACT model was tested on both system had better performance than the fixed gain mount systenls, flexible and rigid, and both mounts controller.

required the use of a large splitter plate. Tile flexible Under Phase II of the ANCAR program, two adap- mount was the Langley Pitch and Plunge Apparatus, tive neural network based control systems were devel- or PAPA, 26 a mechanism allowing model motion in oped and denmnstrated. One of these systems was those two degrees of freedom. Figure 18 shows the an implenlentation of model predictive control where BACT model attached to the PAPA and shows the the network was trained using experimental data to model and splitter plate mounted in the wind tun- serve as the plant model. 29 The other ss,st.em was nel. The advantage of the PAPA mount is that the an application of inverse modeling control where the combination of rigid-wing-plus-flexible-mount results network was trained using experimental data to model in an aeroelastic configuration that has a flutter speed.

an inverse of the plant. 30 Both systelns could adapt, to The PAPA mount was used to investigate instabilities, plant, changes by retraining the neural network using obtain frequency response fuuctions, and evaluate con- new plant input/output data. All three control sys- trol laws. The rigid mount consisted of a rigid strut tems tested tinder the ANCAR program successfully attached to the tunnel sidewall turntable and a five suppressed flutter to the limits of the testing appara- degree-of-freedom balance. Results from tests on the tus, and represent the first, experimental applications rigid mount will not be addressed in this paper.

of neural networks to flutter suppression. Figure 19 The active controls BACT wind-tunnel tests were shows conditions above and below the BACT open- performed in 1995 and 1996. Of the many accom- loop flutter boundary where data was acquired for the plishments achieved within the BACT program, the inverse model control system.

following are the most. significant contributions to the Finally, the BACT model offered the opportunity to state of the art in active controls technology.

develop and demonstate neural network based adap- The BACT model offered the first opportunity to tive control. This work is described in reference 31.

suppress flutter with a spoiler. 27 A single-input-single- ttere, a model predictive control approach was taken output, flutter-suppression control law was designed and a. linear plant model was employed in the wind- by classical techniques, implemented, and successfully tunnel demonstration. This system suppressed flutter tested. The control law was designed to maximize ro- to the limits of the testing apparatus.

bustness over a range of dynamic pressures using a SST Active Controls Testbed single, fixed dynamic compensator element and using fixed blending of signals from two aecelerometers, one As part of NASAs High Speed Research (HSR) pro- located inboard near the leading edge, the other lo- gram, a 1970s Boeing-built SST model was refurbished cated inboard near the trailing edge. and readied for testing on the TDT cable mount sys- The BACT model offered the first successful ex- tem. This model was a 1/20 scale, low-speed, full- span, dynamically-scaled model equipped with active perimental applications of nmltivariable robust con- 13 oF 16 AMERICAN INSTITUTE OF AERONAUTICS AN[) ASTRONAUTICS PAPER 2000-1769 20O Dynamic t50 Pressure, psf 130 I-I El Inverse Model Control, System A J Inverse Model Control, System B Training Data A A O_au.er8oun_ary J Training Data B I I l 1 I I J I 0.60 0.65 0.70 0.75 0.80 0,85 090 0.95 1.00 Mach Number Fig. 19 BACT open-loop flutter boundary with conditions where closed-loop neural network con- trol was implemented.

horizontal tails and active ailerons. It was selected as a testbed for developing control laws, test procedures, and analytical tools needed for an HSR wind-tunnel models program.

Fig. 20 SST model mounted in the TDT.

This model was tested in the TDT in earls; 1995.

Two stability atlgmentation control laws were success- using an existing 1/6-scale, sting-mounted F-18 model fulls' tested closed-loop with the model on the cable (figure 21). The first, series of tests a2 were part of the mount system. These control laws featured inner and ACROBAT (Actively Controlled Response of Buffet outer loops and demonstrated that additional damping Affected Tails) project.. The objectives of the AC- could be added to the pitch and plunge flying modes ROBAT project were twofold: first., to apply active and to the model first flexible mode (fuselage bending).

controls technology using a variety of force produc- Each of the inner loop laws, as well as the inner/outer ers to alleviate buffeting on twin vertical tails; and, combination, exhibited good stability robustness to er- second, to determine the spatial relationships of the rors at. the plant input., errors at. the plant output, and differential pressures at various angle of attack con- to additive plant error. Unfortunately, a third con- ditions with the buffeting alleviation (BA) system off trol law was unstable and caused the model to enter and on. Five new vertical tails were fabricated for a cable-mount instabilhy from which recovery was im- these tests. Two of the tails were rigid surfaces for possible. As a result., the model was damaged beyond measuring pressures. The other three tails were flexi- repair. This model is shown mounted on the cables in ble surfaces equipped with different control devices: a the TDT test section in figure 20. The thick umbilical rudder surface; a tip vane configuration containing a beneath the model contains instrumentation wires.

slotted cylinder or an embedded slotted cylinder; and Actively Controlled Response of Buffet Affected piezoelectric actuators. All three flexible tails were Tails instrumented with a root strain gage aligned to mea- Buffeting is an aeroelastie phenomenon which sure bending moment and with two tip accelerometers plagues high performance aircraft, especially those located near the leading and trailing edges. The re- mainder of the model, namely the fuselage, the wings, with twin vertical tails. For aircraft of this type at and the leading edge extensions, was rigid.

high angles of attack, vortices emanating from the wing/fuselage leading edge extensions burst, immers- The ACROBAT wind-tunnel tests were performed ing the vertical tails in their wake. The resulting buffet with the model angle of attack varying from 20 to 37 loads cause large oscillatory stresses to be applied to degrees. Data were measured for several cases: open the vertical tails with a consequent loss of fatigue life.

loop (no actuator commands); actuator commanded There are two important parameters that determine by a linear sweep; actuator commanded by constant the stress distribution of the tail in flight. The first frequency sinusoidal motion; and closed loop (control is the angle of attack and the second is the dynamic law on). It was determined that control systems us- pressure. If these stresses could be reduced by 10_ ing either the rudder or the piezoelectric actuators the fatigue life associated with the twin vertical tails were best. for suppressing the buffeting. One time- could be doubled.

invariant, fixed-parameter, single-input/single-output A series of wind-tunnel tests were performed in the (SISO) BA control law worked well to alleviate the buf- TDT beginning in 1995 and continuing into late 1999 feting for all flight conditions tested. This control law 14 OF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPEr_ 2000-1769 Compensated Bandwidth By: I_--Rudder_ _-_ Piezos --- Closed Loop Normalized Fin Tip -- Open Loop Acceleration PBD,

r

Units2/Hz 0 2O 4O 60 80 100 Frequency, Hz Fig. 21 ACROBAT wind-tunnel model mounted in the TDT. Fig. 22 ACROBAT tip accelerations, system on and off, Mach number -- 0.10, dynamic pressure ---- 14 psf, angle of attack -- 26 degrees.

was not optimized for any particular flight condition and it is thought that its performance would be im- ing the SIDEKIC tests. Boeing, who participated proved considerably using an optimal controller or an in the wind-tunnel tests through a NASA Space Act adaptive controller. With this simple control law, the Agreement, designed and tested shunt circuits and peak of the power spectral density function of the root neural predictive controllers (NPC). Also. a variety of bending moment at the frequency of the first bend- modern state-space controllers for the blended system ing mode was reduced by as much as 60% for certain were designed by LaRC and tested. With the blended angles of attack, using gains well below the physical BA system operating, the tip accelerations and root limits of the actuator being investigated. At angles of bending moments (root mean square values) could be attack up to about 30 degrees, both the rudder surface reduced by 25 percent. At 26 degrees angle of attack and the piezoelectric actuator control laws were nearly and at a Mach number of 0.1, the rudder reduced buf- equally effective in alleviating buffeting. However at feting in the first bending mode, around 16 Hz, while higher angles of attack, the rudder effectiveness was the piezoelectric actuators reduced buffeting in the limited by degrading flow field conditions due to the first torsion mode, around 58 Hz (figure 22). Similar separated flow around the tail while the piezoelectric results were obtained for the all-piezoelectric system actuators maintained their effectiveness regardless of on the port tail. An assessment of the NPC controllers flight condition.

indicated that this concept performed very similar to During 1998, a second series of tests 33 were per- the MIMO controller systems, but the "piezoelectric formed in the TDT using the F-18 model test bed.

shunting" concept provided negligible reductions in This project was referred to as SIDEKIC (Scaling In- the buffeting of the vertical tails.

fluences Derived from Experimentally-Known Impact of Controls). New vertical tails were fabricated for Concluding Remarks this project. These tails differed from those used pre- Over its forty-year history, more than 500 tests have viously in that continuous skin construction techniques been conducted in the Transonic Dynamics Tunnel were used and an effort was made to match the layout (TDT) and, of these, about 35 have involved the active of the piezoelectric actuators used during a full-scale control of aeroelastic response. Flutter-suppression, F-18 ground test at the Australian Aeronautical and load-alleviation (maneuver, gust, and buffeting), and Maritime Research Laboratory (AMRL). 34 In addi- stability-augmentation active control systems have tion, the type of amplifiers used in the BA system been successfully demonstrated in the TDT. The TDT were different. For this test, one fin employed both has contributed to the state of the art in a number of an active rudder for controlling responses in the first significant ways, including the following list of firsts: bending mode, around 16 Hz, and active piezoelectric first wind-tunnel demonstration of flutter suppression actuators for controlling the responses in the first tor- on a large scale model; first wind-tunnel demonstra- sion mode, around 50 Hz. This configuration of control tion of buffeting alleviation on a large scale model; effectors was referred to as a blended system because first use of piezoelectric devices for flutter suppression two actuator technologies were combined to provide on a large scale model; first wind-tunnel demonstra- a compromise between the use of an existing control tion of MIL SPEC rolling maneuvers above the flutter surface and a reduced number piezoelectric actuators.

boundary; first use of spoilers for flutter suppression; A variety of control schemes were investigated dur- and first use of neural networks for flutter suppression.

15 oF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 2000-1769 References Tailoring Investigation," 37th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, t Cole, S. and Garcia, J., "Past, Current, and Future Capa- Salt Lake City, UT, April 1996.

bilities of the Transonic Dynamics Tunnel from an Aeroelasticity 22Rivera, Jr, J. A., Dansberry, B. E., Farmer, M. G., Eck- Perspective," Tech. rep.

strom, C. V., Seldel, D. A., and Bennett, R. M., "Experimental 2Dempster, J. B. and Arnold, J. I., "Flight Test Evaluation Flutter Results With Steady and Unsteady Pressure Measure- of an Advanced Stability Augmentation System for the B-52 ments of a Rigid Wing on a Flexible Mount System," AIAA Aircraft," No. 68-1068, Oct. 1968.

91-1010, April 1991.

3Wykes, J. H. and Kordes, E. E., "Analytical Design and 23Rivera, Jr, J. A., Dansberry, B. E., Bennett, R. M., Flight Tests of a Modal Suppression System on the XB-70 Air- [)urham, M. I-/., and Silva, W. A., "NACA 0012 Benchmark plane," Aeroelastic Effects from a Flight Mechanics Standpoint, Model Experimental Flutter Results With Unsteady Pressure No. 46, 1970, pp. 23-1 and 23-18.

Distributions," NASA TM 107581, March 1992.

4Sandford, M. C., Abel, I., and Gray, D. L., "Development _4Durham, M. H., Keller, D. F., Bennett, R. M., and Wiese- and Demonstration of a Flutter-Suppression System Using Ac- man, C. D., "A Status Report on a Model for Benchmark Active tive Controls," NASA TR R-450, 1975.

Controls Testing," AIAA 91-1011, April 1991.

SNissim, E., "Flutter Suppression Using Active Controls 2SSc0tt, R. C., Hoadley, S. T., Wieseman, C. D., and Based on the Concept of Aerodynamic Energy," NASA TN D- Durham, M. H., "The Benchmark Active Controls Techmology 6199, 1971.

Model Aerodynamic Data," Proceedings of the 35th Aerospace 8McWhirter, H. D., Hollenbeck, W. W., and Grosset, W. F., Sciences Meeting and Exhibit, No. 97-0829, Reno, CA, Jan.

"Correlation of C-5A Active Lift Distribution Control System 1997.

(ALDCS) Aeroelastic Model and Airpland Flight Test Results," 2¢Farmer, M. G., "A Two-Degree-of-Freedom Flutter Mount NASA CR 144903, 1976.

System with Low Damping for Testing Rigid Wings at Different ¢Sevart, F., "Development of Active Flutter Suppression Angles of Attack," NASA TM 83302, 1982.

Wind-Tunnel Testing Technology," AFFDL-TR 74-126, Jan.

2TWaszak, M. R., "Robust Multivariable Flutter Suppression 1975.

for the Benchmark Active Control Technology (BACT) Wind- s Hwang, C., Winter, B., Noll, T., and Farmer, M., "Demon- Tunnel Model," Prodeedings of the Eleventh Symposium on stration of Aircraft Wing/Store Flutter Suppression Systems," Structural Dynamics and Control, Blacksburg, VA, May 1997.

AGARD 668, April 1978.

2s Lichtenwalner, P., Little, G., and Scott, R., "Adaptive Neu- 9Hwang, C., Johnson, E., Mills, G., and Pi, W., "Additional ral Control of Aeroelastic Response," SPIE 1906 Symposium on Demonstration of Active Wing/Store Flutter Suppression Sys- Smart Structures and Materials, San Diego, CA, Feb. 1996.

tems," AFWAL-TR 80-3093, 1980.

2SLichtenwalner, P., Little, G., Pado, L., and Scott, R., l°Hwang, C., Johnson, E., Mills, G., Noll, T., and Farmer, "Adaptive Neural Control for Active Flutter Suppression," M., "Wind-Tunnel Test of a Fighter Aircraft Wing/Store Flut- SP]E 1997 Symposium on Smart Structures and Materials, San ter Suppression System, An International Effort," AGARD 689, Diego, CA, March 1997.

April 1980.

3°Scott, R. C., "Active Control of Wind-Tunnel Model 11Hwang, C. and Johnson, E., "Test Demonstration of Dig- Aeroelastlc Response Using Neural Networks," SPIE's Seventh ital Adaptive Control of Wing/Store Flutter, Part I - Demon- Annual International Symposium on Smart Structures and Ma- stration of Digital Control." AFWAL-TR 82-3044, July 1982.

terials, Newport Beach, CA, March 2000.

12Murrow, H. N. and Eckstrom, C. V., "Drones for Aerody- sI I-Ialey, P. and Soloway, D., "Generalized Predictive Control namic and Structural Testing (DAST) A Status Report," AIAA for Active Flutter Suppression," IEEE Control Systems Maga- Journal of Aircraft, Vol. 16, No. 8, Aug. 1979, pp. 521-526.

zine, Vol. 17, No. 4, Aug. 1997, pp. 64-70.

13Newsom, J. R., Abel, I., and Dunn, J. H., "Application of 32Moses, R. W., "Active Vertical Tail Buffeting Alleviation Two Design Methods for Active Flutter Suppression and Wind- on a Twin-Tail Fighter Configuration in a Wind Tunnel," CEAS Tunnel Test Results," NASA TP 1653, May 1980. International Forum on Aeroelasticity and Structural Dynam- ics, Rome, Italy, June 1997.

_4Peloubet, R. and Hailer, R., "Wlnd-Tunnel Demonstration of an Active Flutter Suppression Using F-16 Modelwith Stores," 33Moses, R. W., "Contributions to Active Buffeting Alle- afwal-tr 8.3-3046, April 1983. viation Programs by the NASA Langley Research Center," _Oth AIAA/ASME/ASCE/AHS/ASC Structures, Structural lSPeloubet, R., Bolding, R., and Penning, K., "Adap- Dynamics, and Materials Conference, St. Louis, MO, April tive Flutter Suppressslon Wind-Tunnel Test Demonstration," 1999.

AFWAS-TR 87-3053, Oct. 1987.

34Ryall, T., Moses, R., Hopkins, M., Henderson, D., Zim- 16 Miller, G., "Active Flexible Wing (AFW) Technology," cik, D., and Nitzsche, F., "Buffet Load Alleviation," $nd Aus- AFWAL-TR 87-3096, Jan. 1975.

tralasian Congress on Applied Mechanics, Canberra, Australia, lrPerry, III, B., Dunn, H., and Sandford, M., "Control Feb. 1999.

Law Parameterization for an Aeroelastlc Wind-Tunnel Model Equipped with an Active Roll Control System and Comparison with Experiment," 29th AIAA/ASME/ASCE/AHS Structures, Structural Dynamics, and Materials Conference, April 1989.

IS,,Speclal Section on the Actiave Flexible Wing," AIAA Journal of Aircraft, Vol. 32, No. 1, Jam-Feb. 1995, pp. 9-76 and 205-207.

19Hailer, R. L., "Design, Fabrication, and Test of the IDF Flutter Model, Monthly Progress Reports 16-26," General dy- namics corporation contract, !989.

2°Heeg, J., McGowan, A., Crawley, E., and Lin, C., "The Piezoelectric Aeroelastic Response Tailoring Investigation: Analysis and Open-Loop Testing," CEAS International Forum on Aeroelasticity and Structural Dynamics, Manchester, United Kingdom, June 1995.

_1McGowan, A., Heeg, J., and Lake, R., "Results from Wind- Tunnel Testing from the Piezoelectric Aeroelastic Response 16 OF 16 AMERICAN INSTITUTE OF AERONAUTICS AND ASTRONAUTICS PAPER 2000-1769